Analyzing Flood Risk for

Forecast Informed Reservoir

Operations in the Russian River
Watershed using HEC-WAT

February 2018

Approved for Public Release. Distribution Unlimited. PR-100

 REPORT DOCUMENTATION PAGE Form Approved OMB No. 0704-
0188
The public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing
instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of
information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing
this burden, to the Department of Defense, Executive Services and Communications Directorate (0704-0188). Respondents should be aware that
notwithstanding any other provision of law, no person shall be subject to any penalty for failing to comply with a collection of information if it does
not display a currently valid OMB control number.
PLEASE DO NOT RETURN YOUR FORM TO THE ABOVE ORGANIZATION.
1. REPORT DATE (DD-MM-YYYY) 2. REPORT TYPE 3. DATES COVERED (From - To)
February 2018 Project Report
4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER
Analyzing Flood Risk for Forecast Informed Reservoir
5b. GRANT NUMBER
Operations in the Russian River Watershed using HEC-WAT
5c. PROGRAM ELEMENT NUMBER

6. AUTHOR(S) 5d. PROJECT NUMBER

CEIWR-HEC 5e. TASK NUMBER

5F. WORK UNIT NUMBER

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION REPORT

US Army Corps of Engineers NUMBER

Institute for Water Resources PR-100

Hydrologic Engineering Center (HEC)
609 Second Street
Davis, CA 95616-4687
9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/ MONITOR'S ACRONYM(S)

11. SPONSOR/ MONITOR'S REPORT

NUMBER(S)

12. DISTRIBUTION / AVAILABILITY STATEMENT

Approved for public release; distribution is unlimited.
13. SUPPLEMENTARY NOTES

14. ABSTRACT
The purpose of this study was to demonstrate that the U.S. Army Corps of Engineers (USACE) had a tool in
place that would perform a systems-based risk analysis that could be used to perform reservoir reoperation studies
such as those introducing forecasting. Specifically, this report describes a systems-based risk analysis performed
with the USACE, Hydrologic Engineering Center (HEC) software Watershed Analysis Tool (HEC-WAT) and
associated models to investigate the viability of FIRO (Forecast Informed Reservoir Operations) at Lake
Mendocino in the Russian River watershed in the state of California. The study demonstrates practical tools and
methods consistent with USACE guidance regarding evaluation of flood risk, and the results of the analysis are
used to support the PVA (Preliminary Viability Assessment) of the FIRO demonstration project at Lake
Mendocino.
15. SUBJECT TERMS
HEC-WAT, FIRO, USACE, HEC, PVA, systems-based risk, reservoir reoperation studies, forecasting, models,
Lake Mendocino, guidance, evaluation, analysis, hydrologic, reservoirs, hydraulics, consequences, HEC-HMS,
HEC-ResSim, HEC-RAS, HEC-FIA, HEC-DSSVue, Potter Valley Project, Coyote Valley Dam, Warm Springs
Dam, Lake Sonoma, guide curve, flood operations, Dam Safety Action Classification, hydrologic sampling,
performance metrics, storage, conservation, existing conditions, deterministic simulations, flood risk analysis
simulations, FRA, alternatives, procedures, precipitation, evapotranspiration, basin model, subbasin, routing
reach, flow frequency curves, runoff hydrograph, model performance, parameters, inflows, time-series, schematic
elements, boundary conditions, observed, gridded data, structure inventory, time window, reports, life loss,
damage, program order, model linking, analysis period, meteorology data, critical duration, flood season
distribution, frequency, hyetograph, shape sets, soil moisture, output variable, Hopland, Healdsburg, Guerneville,
Sonoma County, Mendocino County, verification, perturbing, spillway, downstream, back-filling
16. SECURITY CLASSIFICATION OF: 17. LIMITATION 18. NUMBER 19a. NAME OF RESPONSIBLE
a. REPORT b. ABSTRACT c. THIS PAGE OF OF PERSON
ABSTRACT PAGES
U U U
UU 332 19b. TELEPHONE NUMBER

Standard Form 298 (Rev. 8/98)

Prescribed by ANSI Std. Z39-18
Analyzing Flood Risk for Forecast
Informed Reservoir Operations in
the Russian River Watershed
using HEC-WAT

February 2018

Prepared by:
US Army Corps of Engineers
Institute for Water Resources
Hydrologic Engineering Center
609 Second Street
Davis, CA 95616

(530) 756-1104
(530) 756-8250 FAX
www.hec.usace.army.mil PR-100
PR-100 Table of Contents

Table of Contents
List of Figures ............................................................................................................................ ix
List of Tables ............................................................................................................................ xv
Abbreviations ...........................................................................................................................xvii
Executive Summary ................................................................................................................. xix

Chapter Page

1.1 Study Purpose ......................................................................................................1

1.2 Scope....................................................................................................................1
1.3 Acknowledgements ...............................................................................................2

2.1 Basis for Study ......................................................................................................3

2.2 Description of Russian River Watershed ...............................................................3
2.2.1 Location and Size ....................................................................................3
2.2.2 Land Cover .............................................................................................4
2.2.3 River Course ...........................................................................................4
2.2.4 Climate and Storms .................................................................................7
2.2.5 Historical Flooding ...................................................................................7
2.3 Water Management Structures .............................................................................9
2.3.1 Potter Valley Project ................................................................................9
2.3.2 Coyote Valley Dam (CVD) .....................................................................12
Physical Description ............................................................................12
Operations ..........................................................................................14
Guide Curve........................................................................................14
Mode of Operation ..............................................................................15
Flood Operations ................................................................................15
Conservation Operations ....................................................................17
Historical Pool Levels for Conservation Flows ..................................... 17
Impact on Historical Flood Peaks ........................................................19
Dam Safety Action Classification (DSAC) Rating ................................ 20
2.3.3 Warm Springs Dam (WSD) ...................................................................21
2.3.4 Russian River Diversion Structure (RDS) and Inflatable Dam ............... 22
2.4 Potential Forecast-Informed Reservoir Operations .............................................. 23
2.4.1 Existing Uses of Forecasts ....................................................................23
2.4.2 Types of Reservoir Operations ..............................................................23
2.4.3 Types of Forecast Information ...............................................................24
2.4.4 Forecast Skill.........................................................................................25

3.1 HEC-WAT ...........................................................................................................27

3.2 HEC Software Integrated into HEC-WAT ............................................................28
3.2.1 Hydrologic – HEC-HMS .........................................................................28
3.2.2 Reservoirs – HEC-ResSim ....................................................................28
3.2.3 Hydraulics – HEC-RAS .........................................................................28
3.2.4 Consequences – HEC-FIA ....................................................................29
3.3 HEC-DSSVue .....................................................................................................29
3.4 Basic HEC-WAT Concepts..................................................................................29
i
Table of Contents PR-100

Table of Contents
Chapter Page

3.5 HEC-WAT Basic Approach .................................................................................30

3.6 Flood Risk Management .....................................................................................32
3.6.1 Hydrologic Sampling .............................................................................32
3.6.2 Performance Metrics .............................................................................32
3.7 HEC-WAT Interface.............................................................................................32
3.7.1 Panes ....................................................................................................33
3.7.2 Tabs ......................................................................................................34
3.7.3 Menu Bar ..............................................................................................34
3.7.4 Toolbar .................................................................................................36
3.7.5 Map Windows .......................................................................................36
3.8 Overview of a Simulation Map Window ...............................................................37

4.1 Study Objectives .................................................................................................39

4.2 Analysis Design ..................................................................................................40
4.3 Deterministic Simulation ......................................................................................40
4.4 Flood Risk Analysis Simulation ...........................................................................41
4.5 Comparison Metrics ............................................................................................43
4.5.1 Storage Available for Conservation Purposes ....................................... 43
4.5.2 Flood Risk .............................................................................................44
4.5.3 Other Considerations ............................................................................44
4.6 Development of the HEC-WAT Existing Conditions Alternative ........................... 45
4.6.1 Hydrologic Modeling ..............................................................................46
4.6.2 Reservoir System Modeling...................................................................48
4.6.3 Hydraulic Modeling ................................................................................49
4.6.4 Consequence Modeling .........................................................................49
4.7 Development of Simulations – HEC-WAT ...........................................................50
4.7.1 Deterministic Simulations ......................................................................50
4.7.2 Flood Risk Analysis Simulations ............................................................52
4.8 Results ................................................................................................................54
4.9 Formulation of Alternatives..................................................................................55
4.9.1 Without Coyote Valley Dam (WO_CVD) ................................................56
4.9.2 Encroachment .......................................................................................56
4.9.3 Combined..............................................................................................56
4.9.4 Encroachment with Imperfect Forecast .................................................57
4.10 Analysis of Alternatives .......................................................................................58
4.11 Simulation Procedures ........................................................................................60

5.1 Meteorological Data Development ......................................................................61

5.1.1 Precipitation ..........................................................................................61
5.1.2 Evapotranspiration ................................................................................64
5.2 Hydrologic Model Development...........................................................................65
5.2.1 Basin Model ..........................................................................................65
5.2.2 Subbasin Parameterization ...................................................................67
5.2.3 Reach Parameterization ........................................................................69
ii
PR-100 Table of Contents

Table of Contents
Chapter Page

5.3 Calibration ...........................................................................................................70

5.3.1 Monthly Volumes ...................................................................................71
5.3.2 Flow Frequency Curves ........................................................................75
5.3.3 Runoff Hydrograph ................................................................................75
5.3.4 Measures of Model Performance...........................................................75
5.3.5 Final Parameters ...................................................................................81
5.4 Modified HEC-HMS Model for Deterministic Simulations .................................... 81

6.1 HEC-ResSim Data Development ........................................................................83

6.1.1 Inflows ...................................................................................................83
6.1.2 Consumptive Withdrawals .....................................................................84
6.1.3 Lookback Data ......................................................................................84
6.1.4 Miscellaneous External Data .................................................................85
6.2 HEC-ResSim Model Development ......................................................................85
6.3 HEC-ResSim Network Design .............................................................................85
6.3.1 HEC-ResSim Model Schematic Elements ............................................. 86
6.3.2 Reach Routing ......................................................................................86
6.4 HEC-ResSim Alternatives ...................................................................................87
6.4.1 Initial States ..........................................................................................87
6.4.2 Time-Series Boundary Conditions .........................................................87
6.5 HEC-ResSim State Variables ..............................................................................87
6.6 Representation of Coyote Valley Dam.................................................................88
6.6.1 Overview of Lake Mendocino Operations ..............................................88
6.6.2 Coyote Valley Dam Model Operation Sets ............................................ 89
6.6.3 Coyote Valley Dam Model Minimum Flow Rules ................................... 91
6.6.4 Coyote Valley Dam Model Rate-of-Change Rules ................................. 92
6.6.5 Coyote Valley Dam Model Maximum Flow Rules .................................. 92
6.6.6 Other Coyote Valley Dam Model Rules .................................................94
6.7 Representation of Warm Springs Dam ................................................................95
6.7.1 Overview of Lake Sonoma Operations ..................................................95
6.7.2 Warm Springs Dam Model Rules ..........................................................95
6.8 HEC-ResSim Simulation Time-Step ....................................................................96
6.9 Validation ............................................................................................................97
6.9.1 Flood Operations...................................................................................97
6.9.2 Conservation Operations .......................................................................98
6.9.3 Existing Conditions versus Observed ....................................................99

7.1 Hydraulic Data Development............................................................................. 103

7.1.1 Cross-Section Elevation Data.............................................................. 103
7.1.2 Manning's n Values ............................................................................. 104
7.1.3 Bridges and Culverts ........................................................................... 105
7.1.4 Storage Areas .....................................................................................105

iii
Table of Contents PR-100

Table of Contents
Chapter Page

7.2 Hydraulic Model Development .......................................................................... 105

7.2.1 Boundary Conditions ........................................................................... 106
7.3 Model Calibration ..............................................................................................106
7.4 HEC-RAS Outputs Used in HEC-WAT Deterministic Simulations ..................... 114
7.4.1 Time-Series .........................................................................................115
7.4.2 Gridded ...............................................................................................115

8.1 Consequence Model Development ................................................................... 117

8.2 HEC-FIA Simulation Modes ..............................................................................118
8.3 HEC-FIA Model Development ........................................................................... 118
8.3.1 Map Layers .........................................................................................118
8.3.2 Watershed...........................................................................................119
8.3.3 Geographic Data .................................................................................121
8.3.4 Inundation Data ...................................................................................121
8.3.5 Inventory Data .....................................................................................123
Structure Inventory Data .................................................................. 123
Structure Occupancy Types ............................................................. 123
8.3.6 Structure Inventory ..............................................................................124
8.3.7 Critical Infrastructure ........................................................................... 126
8.3.8 Impact Response ................................................................................128
8.3.9 Agricultural Grids.................................................................................128
8.3.10 Warning Issuance Scenario................................................................. 128
8.4 Alternatives .......................................................................................................129
8.5 Time Window ....................................................................................................130
8.6 Simulations .......................................................................................................131
8.6.1 HEC-FIA Results .................................................................................131
8.7 Output Files.......................................................................................................132
8.8 Map Output .......................................................................................................133
8.9 Standard Reports ..............................................................................................134
8.9.1 Individual Structure Damage Report.................................................... 134
8.9.2 Aggregated Consequence Report ....................................................... 136
8.9.3 Detailed Life Loss Report .................................................................... 136
8.10 Structure Damage Validation ............................................................................ 137
8.10.1 Initial Structure Inventory Adjustments ................................................ 137
8.10.2 Non-Damaging Flow Adjustments ....................................................... 139
8.11 Historical Flood Damage Flow Validation .......................................................... 141
8.12 Analysis of HEC-FIA Results ............................................................................. 142

9.1 Planning and Execution ....................................................................................143

9.2 Creating the FIRO HEC-WAT Study ................................................................. 144
9.3 Model Imports ...................................................................................................145
9.3.1 HEC-ResSim .......................................................................................145
9.3.2 HEC-HMS Model Alternatives ............................................................. 147
9.3.3 HEC-RAS and HEC-FIA Model Alternatives ........................................ 148
iv
PR-100 Table of Contents

Table of Contents
Chapter Page

9.3.4 Other Model Entries ............................................................................ 148

Time Window Interval Generator ...................................................... 148
Time Window Modifier...................................................................... 148
9.4 Create a Program Order....................................................................................150
9.5 Create Analysis Periods ....................................................................................151
9.6 Creating Simulations .........................................................................................151
9.7 Model Linking ....................................................................................................152
9.8 Running an HEC-WAT Deterministic Simulation ............................................... 155
9.9 Viewing Deterministic Results in HEC-WAT ...................................................... 156
9.10 HEC-WAT Study Directory Structure ................................................................. 158
9.11 DSS Results......................................................................................................159
9.12 Event Folders ....................................................................................................160
9.13 HEC-WAT Deterministic Simulation HEC-FIA Analysis Tool ............................. 160
9.14 Creating Alternative Simulations ....................................................................... 161
9.15 Running Multiple Simulations with the Compute Manager ................................. 163

10.1 Meteorologic Data Development ....................................................................... 166

10.1.1 Critical Duration...................................................................................166
10.1.2 Flood Season Distribution ................................................................... 172
10.1.3 Precipitation Frequency....................................................................... 172
10.1.4 Hyetograph Locations ......................................................................... 175
10.1.5 Precipitation Shape Sets ..................................................................... 176
10.2 Hydrologic Modeling Uncertainty Analysis ......................................................... 178
10.2.1 Basin and Meteorologic Model Changes ............................................. 178
10.2.2 Parameter Sampling for Initial Soil Moisture ........................................ 179
10.2.3 Selected HEC-HMS Outputs ............................................................... 181
10.3 Reservoir Modeling ...........................................................................................181
10.3.1 Network Changes for Efficient Flood Operation Modeling ................... 182
10.3.2 Sampling of Starting Reservoir Levels ................................................. 183
10.4 Hydraulic Modeling............................................................................................187
10.5 Consequence Modeling ....................................................................................187
10.6 HEC-WAT FRA Modeling ..................................................................................187
10.6.1 FRA Analysis Period ........................................................................... 187
10.6.2 Defining FRA Simulations ................................................................... 188
10.6.3 FRA Linking ........................................................................................190
10.6.4 Skip Rules ...........................................................................................191
10.6.5 Output Variable Editor ......................................................................... 192
10.7 Running the FRA Simulations ........................................................................... 193
10.7.1 FRA Simulation Results ...................................................................... 195
Results from the HEC-WAT Interface ............................................... 195
Results Stored in Files ..................................................................... 197

v
Table of Contents PR-100

Table of Contents
Chapter Page

11.1 Meteorologic Modeling Observations ................................................................ 201

11.2 Hydrologic Modeling Observations .................................................................... 206
11.3 Reservoir System Modeling Observations ......................................................... 209
11.3.1 Existing Conditions Deterministic Lake Mendocino Levels .................. 209
11.3.2 Existing Conditions Lake Mendocino Elevation Frequency Results ..... 213
11.3.3 Significance of Initial Pool Levels ........................................................ 216
11.4 Hydraulic Modeling Observations ...................................................................... 217
11.4.1 Travel Times and Downstream Residence .......................................... 217
11.4.2 Flow and Stage Frequencies ............................................................... 220
Hopland ...........................................................................................223
Healdsburg ......................................................................................224
Guerneville (Hacienda) .................................................................... 225
11.5 Consequence Modeling Observations ............................................................... 226
11.5.1 Annual Average Damage versus EAD ................................................. 226
11.5.2 Mendocino versus Sonoma County ..................................................... 228
11.5.3 Damage Locations for Extreme Events ............................................... 231
11.5.4 Verification of Time Window Modifier Parameters ............................... 232
11.5.5 Verification of Skip Rule Parameters ................................................... 233
11.6 Integrated Modeling Results..............................................................................234

12.1 Formulation of WO_CVD Alternative ................................................................. 237

12.2 Results ..............................................................................................................237

13.1 Formulation of Encroach Alternative ................................................................. 241

13.2 Typical Operations ............................................................................................244
13.3 Results ..............................................................................................................246
13.3.1 Comparison of Coyote Valley Dam Levels .......................................... 246
13.3.2 Compliance with Hopland Rule ........................................................... 250
13.3.3 Downstream Flood Damage ................................................................ 251

14.1 Formulation of Combined Alternative ................................................................ 255

14.2 Typical Operations ............................................................................................258
14.3 Results ..............................................................................................................261
14.3.1 Comparison of Coyote Valley Dam Levels .......................................... 261
14.3.2 Compliance with Hopland Rule ........................................................... 265
14.3.3 Downstream Flood Damage ................................................................ 266

15.1 Formulation of EncroachWIF Alternative ........................................................... 271

15.1.1 Perturbing Method – Deterministic Simulation ..................................... 272
15.1.2 Perturbing Method- FRA ..................................................................... 272
15.2 Typical Operations ............................................................................................272
15.3 Results ..............................................................................................................275
vi
PR-100 Table of Contents

Table of Contents
Chapter Page

16.1 Conservation Storage .......................................................................................279

16.2 Compliance with Hopland Flow Rule ................................................................. 279
16.3 Flood Damage ..................................................................................................280
16.4 Elevation-Frequency at Lake Mendocino .......................................................... 282
16.4.1 Spillway Activation ..............................................................................283
16.4.2 Flood Frequency at Downstream Locations ........................................ 284
16.5 Cases for Further Analysis ................................................................................285
16.5.1 January 2006 - Minor Consequences from Different Back-Filling
Releases .............................................................................................287
16.5.2 December 1964 – FIRO Alternatives can both Improve and Hinder
Flood Operations.................................................................................289
16.5.3 Event 3 from Lifecycle 69 – Spillway Flow Adds to Downstream
Peaks..................................................................................................290
16.5.4 Event 25 from Lifecycle 64 – Late Season Vulnerability ...................... 292
16.6 Other Consequences ........................................................................................294

17.1 Summary of Observations .................................................................................297

17.2 Suggestions for Future Analyses....................................................................... 299

18.1 References .......................................................................................................301

18.2 Background Information ....................................................................................303

Appendix A – Water Control Diagram.................................................................................. A-1

vii
Table of Contents PR-100

viii
PR-100 List of Figures

List of Figures
Figure
Number Page

1 Russian River Watershed Location ..............................................................................4

2 Land Use within the Russian River Watershed.............................................................5
3 Map of the Russian River Watershed including the Potter Valley Project .....................6
4 Flood of 1915 in Guerneville, California (photo provided by the Sonoma County
Library) ........................................................................................................................8
5 Map of the Potter Valley Project .................................................................................10
6 Lake Mendocino Average Annual Inflow for Periods both Prior to and After the
implementation of the Potter Valley Project FERC License Amendment in the
Fall of 2006 ................................................................................................................11
7 Coyote Valley Dam Area ............................................................................................13
8 Coyote Valley Dam Outlet Area .................................................................................13
9 Lake Mendocino Storage Zones ................................................................................14
10 California SWRCB Decision 1610 - Instream Flows ...................................................18
11 Historical Lake Mendocino Spring Maximum Pools ....................................................19
12 Coyote Valley Dam – 1964 Event ..............................................................................21
13 Russian River Diversion Structure and Inflatable Dam ............................................... 22
14 HEC-WAT Framework ...............................................................................................27
15 HEC-WAT – Model Linking Editor ..............................................................................31
16 HEC-WAT Main Window ............................................................................................33
17 HEC-WAT Main Window – Simulation Map Window ..................................................37
18 HEC-WAT Deterministic Simulation Process .............................................................41
19 HEC-WAT Flood Risk Analysis Simulation Process ...................................................42
20 Subbasin Delineations for the Russian River HEC-HMS Model ................................. 47
21 HEC-WAT – Alternative and Simulation Manager ......................................................51
22 HEC-WAT Simulation Editor - Deterministic Simulation (Base_POR_Full) ................. 51
23 HEC-WAT – Analysis Period Editor............................................................................53
24 HEC-WAT Simulation Editor - FRA Simulation (BASE_FRA) – FRA Tab ................... 54
25 Lake Mendocino Guide Curve Adjustments for Combined Alternative ........................ 57
26 HEC-WAT Study Tree ................................................................................................59
27 PRISM Grid with Russian River Rain Gages and Hydrometeorology Stations ........... 63
28 Example Eight-Point Cross Section used in HEC-HMS. ............................................. 70
29 Calpella Monthly Volume Calibration..........................................................................72
30 Ukiah Monthly Volume Calibration .............................................................................72
31 Hopland Monthly Volume Calibration .........................................................................73
32 Cloverdale Monthly Volume Calibration .....................................................................73
33 Healdsburg Monthly Volume Calibration ....................................................................74
34 Guerneville Monthly Volume Calibration.....................................................................74
35 Annual Flow Frequency Curve at Calpella .................................................................76
36 Annual Flow Frequency Curve at Ukiah .....................................................................76
37 Annual Flow Frequency Curve at Hopland .................................................................77
38 Annual Flow Frequency Curve at Cloverdale .............................................................77
39 Annual Flow Frequency Curve at Healdsburg ............................................................78
40 Annual Flow Frequency Curve at Guerneville ............................................................78
41 Simulated versus Observed Runoff Hydrograph at Healdsburg for the Period of
December 2003 through March 2004 .........................................................................79
42 HEC-ResSim Model Schematic Elements ..................................................................86
ix
List of Figures PR-100

List of Figures
Figure
Number Page

43 HEC-ResSim Alternative Editor - Tabs .......................................................................88

44 Lake Mendocino 2006 Event Observed versus Simulated ......................................... 89
45 HEC-ResSim Reservoir Editor - Operations Set Definition ......................................... 90
46 HEC-ResSim Model - WSC I-1610 Q-TUCP IF_Block Structure ................................ 91
47 Reduced Ability to Fill Pool using IF_Block DROC_April2016 .................................... 93
48 HEC-ResSim Reservoir Editor - Warm Springs Dam Baseline Operations Set .......... 96
49 Simulated Flood Operations during the January 1995 Event ...................................... 98
50 Simulated Conservation Operations during May 2008................................................ 99
51 Validation of 1986 Event ..........................................................................................101
52 HEC-RAS - Example of Cutting a Low-Flow Channel into a Cross-Section .............. 104
53 HEC-RAS Boundary Conditions ...............................................................................107
54 Observed High Water Marks on the Theater at Monte Rio ....................................... 107
55 Steady-Flow Water Surface Profiles with Observed High Water Mark Data, Upper
Half of Russian River ...............................................................................................108
56 Steady-Flow Water Surface Profiles with Observed High Water Mark Data, Lower
Half of Russian River ...............................................................................................109
57 HEC-RAS Computed Flood Extent for the 2006 Event Compared to Observed
Marker .....................................................................................................................109
58 HEC-RAS Computed Flood Extent for the 2006 Event Compared to Observed
Marker at Eastside Road and Wohler Road ............................................................. 110
59 Observed versus Computed Flow at Hopland Gage, 2006 Event ............................. 111
60 Observed versus Computed Flow at Cloverdale Gage, 2006 Event ......................... 111
61 Observed versus Computed Flow at Healdsburg Gage, 2006 Event ........................ 112
62 Observed versus Computed Flow at Guerneville Gage, 2006 Event. ....................... 112
63 Computed Water Surface Profile with Observed Data, Upper Half of the Russian
River, 2006 Event ....................................................................................................113
64 Computed Water Surface Profile with Observed Data, Lower Half of the Russian
River, 2006 Event. ...................................................................................................113
65 HEC-RAS Unsteady Flow Dialog Box – Unsteady Flow Analysis Settings ............... 114
66 HEC-RAS Stage and Flow Output Locations Selector.............................................. 115
67 RAS Mapper Results Maps ......................................................................................116
68 HEC-FIA Interface....................................................................................................119
69 HEC-FIA Cross-Section Mappings Editor ................................................................. 120
70 HEC-FIA Watershed Configuration Editor ................................................................ 120
71 HEC-FIA Impact Area Editor ....................................................................................121
72 HEC-FIA New Inundation Configuration Wizard ....................................................... 122
73 Defining an Event for an HEC-FIA Inundation Configuration .................................... 123
74 HEC-FIA Structure Occupancy Type Editor - Depth-Damage Functions Tab ........... 124
75 HEC-FIA Structure Occupancy Type Editor – General Info/Uncertainty
Parameters Tab .......................................................................................................125
76 Parcel Data ..............................................................................................................125
77 Structure Value Breakdown......................................................................................128
78 HEC-FIA Warning Issuance Editor ........................................................................... 129
79 HEC-FIA Alternative Editor ......................................................................................130
80 HEC-FIA Time Window Dialog Box .......................................................................... 131
81 HEC-FIA Simulation Manager ..................................................................................131
x
PR-100 List of Figures

List of Figures
Figure
Number Page

82 HEC-FIA Output Files per Deterministic Event ......................................................... 132

83 Attributes for GIS Output of Damage at Flooded Structures ..................................... 132
84 HEC-FIA Map Window - Grid Query Tool ................................................................. 133
85 HEC-FIA Map Window - Shapefile Query Tool ......................................................... 134
86 HEC-FIA Main Window - Accessing HEC-FIA Reports............................................. 135
87 HEC-FIA - Individual Structure Damage Report ....................................................... 135
88 HEC-FIA - Aggregated Consequence Report ........................................................... 136
89 HEC-FIA - Detailed Life Loss Report........................................................................ 137
90 Structure Inventory Adjustments ..............................................................................138
91 Johnson's Beach Structure Adjustments .................................................................. 139
92 Structure Relocations East of Guerneville across from Rio Nido .............................. 140
93 Structure Damage in Agricultural Area near Cloverdale ........................................... 141
94 Russian River Damage by River Mile for the 2006 Event ......................................... 142
95 Creating the HEC-WAT Study ..................................................................................144
96 HEC-WAT – Study Details Dialog Box – FIRO Study Plug-Ins ................................. 145
97 HEC-WAT Study after Importing HEC-ResSim Model .............................................. 146
98 HEC-HMS Model Alternative ....................................................................................147
99 HEC-WAT - Create New Time Window Modifier Alternative Dialog Box ................... 149
100 HEC-WAT - Time Window Modifier Alternative Editor for Deterministic
Simulations ..............................................................................................................149
101 HEC-WAT - Program Order Dialog Box ................................................................... 150
102 HEC-WAT - Analysis Period Editor .......................................................................... 151
103 HEC-WAT - Simulation Editor ..................................................................................152
104 HEC-WAT – Model Linking Editor ............................................................................ 152
105 HEC-WAT – Model Linking Editor - HEC-ResSim Model Alternative Linkages ......... 153
106 HEC-WAT – Model Linking Editor - Time Window Modifier Alternative Linkage ....... 154
107 HEC-WAT – Model Linking Editor - HEC-RAS Model Alternative Linkages .............. 154
108 HEC-WAT – Model Linking Editor - HEC-FIA Model Alternative Linkages................ 155
109 HEC-WAT - Compute Progress Dialog Box ............................................................. 155
110 HEC-WAT - HEC-ResSim Standard Plot.................................................................. 156
111 HEC-WAT – Simulation Map Window - Elements .................................................... 157
112 HEC-WAT - Results Menu .......................................................................................157
113 HEC-WAT Study Folder Hierarchy for FIRO Study .................................................. 158
114 HEC-WAT – Simulation Browse Run Folder ............................................................ 158
115 HEC-WAT – HEC-FIA Analysis Tool – Simulation Total Map Display ...................... 160
116 HEC-WAT – HEC-FIA Analysis Tool - Single Event Map Display ............................. 161
117 HEC-WAT – Alternative and Simulation Manager .................................................... 161
118 HEC-WAT - Model Linking Editor - Dynamic Linking ................................................ 162
119 HEC-WAT - Dynamic Linkage Editor........................................................................ 162
120 HEC-WAT - Compute Manager ................................................................................163
121 December 1955 Event .............................................................................................167
122 December 1964 Event .............................................................................................168
123 January 1978 Event .................................................................................................168
124 January 1983 Event .................................................................................................169
125 February 1986 Event................................................................................................169
126 January 1995 Event .................................................................................................170
xi
List of Figures PR-100

List of Figures
Figure
Number Page

127 March 1995 Event ....................................................................................................170

128 January 1997 Event .................................................................................................171
129 February 1998 Event................................................................................................171
130 December 2005 Event .............................................................................................172
131 Peaks Found using HEC-DSSVue Find Peaks Tool ................................................. 173
132 HEC-WAT - Hydrologic Sampling Editor - Event Sampling Tab ............................... 173
133 HEC-WAT - Hydrologic Sampling Editor - Frequency Curve Tab ............................. 174
134 HEC-WAT - Hydrologic Sampling Editor – Hyetograph Location Tab ....................... 175
135 Shape Set Hyetographs for December 1995 Event at Russian 10 and
WF Russian Subbasins ............................................................................................176
136 HEC-WAT - Hydrologic Sampling Editor - Shape Sets tab ....................................... 177
137 HEC-HMS Uncertainty Analysis ...............................................................................178
138 Calibrated Beta Distributions for Computed Soil Moisture for the WF Russian
Subbasin ..................................................................................................................179
139 HEC-HMS - Uncertainty Parameter Specified for the Subbasin WF Russian ........... 180
140 HEC-HMS - Uncertainty Parameter for Subbasin Russian 70 - Specified as a
Linear Regression of the Subbasin WF Russian ...................................................... 180
141 HEC-HMS – Uncertainty Analysis Editor .................................................................. 181
142 HEC-ResSim Output Selection for FRA Simulations ................................................ 182
143 HEC-DSSVue Cyclic Analysis of the Existing Conditions Alternative's Deterministic
Pool Levels ..............................................................................................................183
144 Empirical Distributions for Deterministic Existing Conditions Alternative Lake
Levels ......................................................................................................................184
145 Empirical Distributions for Deterministic Combined Alternative Lake Levels ............. 185
146 HEC-ResSim Random Variable Wizard ................................................................... 186
147 HEC-WAT - Analysis Period for FRA Simulations .................................................... 188
148 Creating a New FRA Simulation – Base FRA........................................................... 188
149 HEC-WAT – Simulation Editor - Initial Seeds for FRA Simulations ........................... 189
150 HEC-WAT – Model Linking Editor - FRA Simulation – HEC-HMS Linking ................ 190
151 HEC-WAT – Model Skip Rules Editor ...................................................................... 191
152 Determination of Non-Damaging Flow – Guerneville Gage ...................................... 192
153 HEC-WAT - Output Variable Editor .......................................................................... 193
154 HEC-WAT - Run FRA Simulation Dialog Box ........................................................... 194
155 HEC-WAT Histogram Viewer ...................................................................................196
156 HEC-WAT – Output Variable Viewer ........................................................................ 196
157 HEC-WAT - Frequency Viewer ................................................................................197
158 FRA Results Folder Structure – Simulation Level..................................................... 198
159 Example Output Variable DSS Records ................................................................... 198
160 Example Lifecycle Results Folder ............................................................................ 199
161 Frequency of Shape Set Selection ........................................................................... 203
162 Frequency of Event Date Selection .......................................................................... 204
163 Frequency of Eight-Day Total Precipitation .............................................................. 204
164 Realization Frequency Comparisons........................................................................ 205
165 HEC-HMS Deterministic Simulation Example Results .............................................. 206
166 Validation of Sampling for SMA Loss Method – EF Russian 20 Subbasin –
February Events.......................................................................................................207
xii
PR-100 List of Figures

List of Figures
Figure
Number Page

167 Validation of Sampling for Deficit and Constant Loss Method – WF Russian
Subbasin - December Events...................................................................................208
168 Annual Peak Flow Frequency Curve – Coyote Valley Dam Inflow – Existing
Conditions ................................................................................................................209
169 Simulated and Observed Lake Mendocino Levels 1950 - 1964 ................................ 210
170 Simulated and Observed Lake Mendocino Levels 1965 - 1979 ................................ 211
171 Simulated and Observed Lake Mendocino Levels 1980 - 1994 ................................ 211
172 Simulated and Observed Lake Mendocino Levels 1995 - 2010 ................................ 212
173 Validation of Sampling for Coyote Valley Dam Initial Pool Level – January
Events – Existing Conditions Alternative .................................................................. 214
174 Annual Peak Pool Elevation Frequency Curve – Lake Mendocino – Existing
Conditions ................................................................................................................215
175 Coyote Valley Dam Results for Largest FRA Event .................................................. 216
176 Lake Mendocino Existing Conditions Deterministic Results 1995 - 1998 .................. 217
177 Downstream Impact of Six-Hour Delay to Stopping Coyote Valley Dam Release..... 218
178 Downstream Impact of One Day Delay to Stopping Coyote Valley Dam Release .... 219
179 Annual Peak Flow Frequency Curve – Hopland – Existing Conditions Alternative ... 220
180 Annual Peak Water Surface Elevation Frequency Curve – Hopland – Existing
Conditions Alternative ..............................................................................................221
181 Annual Peak Flow Frequency Curve – Healdsburg – Existing Conditions
Alternative ................................................................................................................222
182 Annual Peak Water Surface Elevation Frequency Curve – Healdsburg – Existing
Conditions Alternative ..............................................................................................223
183 Annual Peak Flow Frequency Curve – Guerneville (Hacienda) – Existing Conditions
Alternative ................................................................................................................224
184 Annual Peak Water Surface Elevation Frequency Curve – Guerneville
(Hacienda) – Existing Conditions Alternative ............................................................ 225
185 Flood Damage from the Deterministic Simulations ................................................... 226
186 Average Damage Results from Deterministic and Lifecycle Simulations for
Existing Conditions Alternative .................................................................................227
187 Variation of EAD across FRA Lifecycles................................................................... 228
188 Existing Conditions Alternative Inundation Extents for 1964 Event in Mendocino
County .....................................................................................................................229
189 Existing Conditions Alternative Inundation Extents for 1964 Event in Sonoma
County .....................................................................................................................230
190 Proportion of Russian River Structure Damage by River Mile across Events ........... 231
191 Existing Conditions Alternative 1995 Flows .............................................................. 232
192 Relationship between Event Damage and Peak Flow at Selected Locations ........... 233
193 Least-Damage Flows at Selected Locations ............................................................ 233
194 Selected Existing Conditions Alternative Hydrographs from Deterministic
Simulation ................................................................................................................234
195 Selected Existing Conditions Alternative Hydrographs for Storm
Event 2005 - 2006....................................................................................................235
196 Influence of Coyote Valley Dam for the March 1995 Flood Event ............................. 238
197 Difference in Depths for the 1964 Flood Event ......................................................... 239
198 Lake Mendocino Levels (2013) ................................................................................241
xiii
List of Figures PR-100

List of Figures
Figure
Number Page

199 HEC-ResSim - Reservoir Rules for Encroach Alternative ......................................... 242

200 Typical Operations under the Encroach Alternative .................................................. 244
201 Coyote Valley Dam Levels under Encroach Alternative (1950 - 1980) ..................... 247
202 Coyote Valley Dam Levels under Encroach Alternative (1974) ................................ 247
203 Coyote Valley Dam Levels under Encroach Alternative (1980 - 2010) ..................... 248
204 Encroach Alternative Operations for the 1986 Event ................................................ 249
205 Coyote Valley Dam Levels for Encroach Alternative (1995) ..................................... 249
206 Comparison of Hopland Flows over 8,000 cfs for Encroach Alternative ................... 250
207 Encroach Alternative for Event 5 from Lifecycle 80 .................................................. 252
208 Differences between Damage for Existing Conditions and Encroach Alternatives
for each FRA Event ..................................................................................................252
209 HEC-ResSim - Reservoir Rules for the Combined Alternative.................................. 256
210 Coyote Valley Dam Levels under the Combined Alternative (1983) ......................... 258
211 Coyote Valley Dam Flood Operations under Combined Alternative (1983) .............. 259
212 Coyote Valley Dam Levels under the Combined Alternative (1950 - 1980) .............. 262
213 Distribution of Lake Mendocino Storage on 10 May ................................................. 262
214 Coyote Valley Dam Levels under the Combined Alternative (1974) ......................... 263
215 Coyote Valley Dam Levels under the Combined Alternative (1980 - 2010) .............. 264
216 Combined Alternative Operations for the 1986 Event ............................................... 264
217 Coyote Valley Dam Levels under the Combined Alternative (1995) ......................... 265
218 Comparison of Hopland Flows over 8,000 cfs for the Combined Alternative ............ 266
219 Differences between Damage for Existing Conditions and Combined Alternative
for each FRA Event ..................................................................................................267
220 Combined versus Existing Conditions Alternative Pool Levels for Event 8 from
Lifecycle 52 ..............................................................................................................268
221 March 1987 Simulation of the EncroachWIF Alternative ........................................... 273
222 Simulation of 1986 Event under EncroachWIF Alternative ....................................... 274
223 Simulation of 1964 Event under EncroachWIF Alternative ....................................... 275
224 Differences between Damage for EncroachWIF and Encroach Alternatives
for each FRA Event ..................................................................................................276
225 Distributions of Lake Mendocino September 30th Storage ........................................ 277
226 Comparison of Hopland Flows above 8,000 cfs from Deterministic Simulations ....... 280
227 Flood Damage from the Deterministic Simulations ................................................... 281
228 Comparison of Lake Mendocino Annual Peak Frequency ........................................ 282
229 Comparison of Hopland Flow Frequency Curves ..................................................... 285
230 Comparison of Healdsburg Flow Frequency Curves ................................................ 286
231 Comparison of Guerneville (Hacienda) Flow Frequency Curves .............................. 286
232 Outflows for Combined Alternative – 2006 Flood Event ........................................... 288
233 Higher Stage Due to Prior Outflows Still Resident Downstream ............................... 288
234 Comparison of Alternatives for 1964 Event .............................................................. 289
235 Event 3 from Lifecycle 69 Results for the Combined Alternative .............................. 291
236 Comparison of Simulations for Event 25 from Lifecycle 64 ....................................... 293
237 Agricultural Inundation for 1964 Event – Existing Conditions Alternative .................. 295

xiv
PR-100 List of Tables

List of Tables
Table
Number Page

1 Flood Flows on the Upper Russian River Before and After Coyote Valley Dam. ........ 20
2 Hourly Precipitation Gages ........................................................................................62
3 Daily Precipitation Gages ...........................................................................................64
4 Event Adjustment Stations and Periods .....................................................................64
5 Subbasin Canopy Parameters – Simple Canopy Method ........................................... 65
6 Subbasin Surface Parameters - Simple Surface Method............................................ 66
7 Subbasin Surface Parameters – Soil Moisture Accounting Method (SMA) ................. 66
8 Subbasin Loss Parameters - Deficit and Constant Loss Method ................................ 66
9 Subbasin Transform Parameters - Clark Unit Hydrograph Transform Method ............ 66
10 Subbasin Baseflow Parameters - Linear Reservoir Baseflow Method ........................ 66
11 Reach Routing Parameters - Muskingum-Cunge Routing Method ............................. 67
12 Reach Routing Parameters - Modified Puls Routing Method ...................................... 67
13 Hydrologic Soil Group Description and Infiltration Rate Assignments ......................... 68
14 Initial Parameter Estimates for Subbasin EF Russian 20 ........................................... 68
15 Initial Parameter Estimates for Deficit and Constant Subbasins ................................. 69
16 Performance Ratings for Summary Statistics .............................................................80
17 Summary Statistics at Calibration Points ....................................................................81
18 Calibrated Parameters for Subbasin EF Russian 20 ..................................................81
19 Calibrated Parameters for Deficit and Constant Subbasins ........................................ 82
20 Diversions from the Russian River .............................................................................84
21 Russian River Bridges by Location used in the HEC-RAS Model. ............................ 106
22 Water Surface Elevation Results from Steady-Flow Simulation on the Russian
River. .......................................................................................................................108
23 Summary of Structure Inventory ...............................................................................127
24 Flood Damage Validation Events .............................................................................142
25 Historical Event Critical Duration Analysis ................................................................ 167
26 GEV Parameters for Precipitation Frequency Curve ................................................ 174
27 Shape Set Definition ................................................................................................177
28 Hydrologic Sampler Results for Event 38 from Lifecycle 28 ..................................... 202
29 Coyote Valley Dam Reach Length and Flood Wave Travel Time
(reference: Table 14, CVD Regulation Manual) ........................................................ 218
30 Hopland Rule Compliance Comparison ................................................................... 280
31 Flood Damage Comparison for Deterministic and FRA Simulations ......................... 281
32 Comparison of Lake Mendocino Frequencies at Key Elevations .............................. 283
33 Comparison of Spillway Flow Durations for Deterministic Simulations ..................... 283
34 Comparison of Key Frequencies at Hopland, Healdsburg, and Guerneville
Gages ......................................................................................................................284

xv
List of Tables PR-100

xvi
PR-100 Abbreviations

Abbreviations

AAD Annual Average Damage

ACE Annual Chance Exceedance
AFY acre-feet per year
ASOS Automated Surface Observing System
BIOP biological opinion
CCP common computation point
CESPK USACE, Sacramento District
CESPN USACE, San Francisco
cfs cubic feet per second
CNRFC California-Nevada River Forecast Center
CoP Community of Practice
CVD Coyote Valley Dam
CW3E Center for Western Weather & Water Extremes
CWMS Corps Water Management System software
DEM Digital Elevation Mode
DSAC Dam Safety Action Classification
DWR Department of Water Resources, State of California
EAD Expected Annual Damage
ERDC USACE, Engineer Research and Development Center
ESA Endangered Species Act
ESRI Environmental Systems Research Institute
FAO Food and Agriculture Organization
FEMA Federal Emergency Management Agency
FERC Federal Energy Regulatory Commission
FIRO Forecast Informed Reservoir Operations
FRA Flood Risk Analysis
GEV Generalized Extreme Value
GIS Geographic Information System
HEC USACE, Hydrologic Engineering Center
HEC-DSS Data Storage System
HEC-DSSVue Data Storage Systems Visual Utility Engine
HEC-FIA Flood Impact Analysis software
HEC-HMS Hydrologic Modeling System software
HEC-RAS River Analysis System software
HEC-ResSim Reservoir System Simulation software
HEC-WAT Watershed Analysis Tool software
HH&C CoP USACE, Hydraulics, Hydrology, and Coastal Community of Practice
I/O input and output
xvii
Abbreviations PR-100

KSTS Sonoma County Airport

KUKI Ukiah Municipal Airport
LEHD Longitudinal Employer-Household Dynamics
MW megawatts
NCDC National Climatic Data Center
NEPA National Environmental Policy Act
NGVD National Geodetic Vertical Datum
NMFS National Marine Fisheries Service
NOAA National Oceanic & Atmospheric Administration
NRCS Natural Resources Conservation Service
NSE Nash-Sutcliffe Efficiency
NWS National Weather Service
PBIAS percent bias
PG&E Pacific Gas & Electric
PRISM Parameter-elevation Regressions on Independent Slopes
PVA Preliminary Viability Assessment
PVID Potter valley Irrigation District
PVP Potter Valley Project
QPF Quantitative Precipitation Forecast
R Clark Storage Coefficient
R2 coefficient of determination
RDS River Diversion Structure
RMSE root mean square
RSR Observed Standard Deviation Ratio
SCRIPPS SCRIPPS Institution of Oceanography
SCS Soil conservation Service
SCWA Sonoma County Water Agency
SMA Soil Moisture Accounting
SSURGO Soil Survey Geographic
SWRCB State of Water Resources Control Board (California)
Tc time of concentration
TUCP Temporary Urgency Change Petition
TWIG Time Window Interval Generator (HEC-WAT)
URR Upper Russian River
USACE U.S. Army Corps of Engineers
USAF United States Air Force
USBR U.S. Bureau of Reclamation
USDA U.S. Department of Agriculture
USGS U.S. Geological Survey
WCM Water Control Manual
WSD Warm Springs Dam

xviii
PR-100 Executive Summary

Executive Summary
This study shows that Forecast Informed Reservoir Operations (FIRO) at the Coyote Valley Dam
(CVD) located in the Russian River watershed in Northern California, can potentially result in
more water stored in Lake Mendocino and water in the Russian River available for water supply
and environmental uses without significantly increasing flood risk. The analysis was performed
as part of the Lake Mendocino FIRO Preliminary Viability Assessment (PVA).

This analysis complies with U.S. Army Corps of Engineers (USACE) policy requirements for a
system-wide risk-based approach, and uses methods and data typical for USACE studies. The
Hydrologic Engineering Center's (HEC) Watershed Analysis Tool (HEC-WAT) model
facilitated simulations of project operations for a baseline condition and alternative operations,
allowing comparison of the relative differences among the results. The HEC-WAT project
produced results for both a sixty-year continuous simulation (i.e., deterministic simulation) and a
probabilistic Monte Carlo approach (Flood Risk Analysis (FRA) compute option in HEC-WAT).

The Russian River HEC-WAT model used precipitation data as the primary hydrologic input,
and provided that precipitation data to an HEC-HMS (HEC's Hydrologic Modeling System
software) model that computed the resulting streamflow. An HEC-ResSim (HEC's Reservoir
System Simulation software) model simulated operations at Coyote Valley Dam and Warm
Springs Dam. An unsteady HEC-RAS (HEC's River Analysis System software) model routed
the reservoir releases and tributary flows, determining peak water surface elevations and
inundated extents throughout the Russian River watershed. An HEC-FIA (HEC's Flood Impact
Analysis software) model used a parcel-based structure inventory and stage-damage information
to compute flood damage. The models and data for this research analysis were developed with
the intention that the models and data could be readily applied in a subsequent study to propose
and implement changes to the operation of Coyote Valley Dam.

The analysis takes a "bookend" approach to show the potential effectiveness of FIRO. The
operation alternatives modeled in this analysis employ aggressive operational schemes using
perfect forecast information, i.e., observed precipitation and streamflows in place of actual
forecast data. The specific alternatives evaluated in this analysis do not represent serious
proposals, but illustrate well-known types of operations that could be based on forecast
information. An additional alternative randomly perturbs the "forecast" precipitation to
demonstrate an assessment of plan performance considering forecast error.

The Existing Conditions (or Baseline) alternative represents the current reservoir operations. An
alternative that operates Warm Springs Dam in the absence of Coyote Valley Dam provides
context regarding the total amount of flood damage reduced by Lake Mendocino. The Encroach
alternative illustrates a very simple FIRO approach, adding a single rule to the baseline
operations that permits the reservoir to store additional water in the absence of significant
precipitation during the next five days. The Combined alternative demonstrates other
possibilities with an advance release rule, a rule using forecasted local flows to maximize
downstream channel capacity, and a heavily modified guide curve. The Encroach with Imperfect
Forecast alternative demonstrates possible impacts of forecast error by running the Encroach
alternative using perturbed forecast information.
xix
Executive Summary PR-100

Differences in flood risk among alternatives are quantified by frequency analysis of flows and
stages at Lake Mendocino and key locations along the Russian River, and by expected annual
damage (EAD). Other metrics include reservoir storage at the end of flood season, reservoir
storage at the end of water year, and compliance with downstream flow limitations.

The results demonstrate that with the assumptions made within the analysis, FIRO can
significantly increase the reliability of water supplies at Coyote Valley Dam without significantly
increasing flood risk on the Russian River. Some increase in flood risk is possible when
introducing additional information that has the potential for errors, and can lead to
implementation errors as well. The modeling tools developed here are ready to be adapted to
include characterizations of forecast uncertainty and tolerance for flood risk, and used in
subsequent official analyses for implementing FIRO.

Practical implementation of FIRO at Lake Mendocino will require careful design of reservoir
operations to avoid flood releases that increase flood peaks downstream. Additionally, the
operations must consider the possibility of higher pool levels resulting in outflow over the
uncontrolled spillway early enough to increase downstream flood peaks. The operations must
also be designed with awareness of any dam safety considerations regarding greater hydraulic
loading on the structure, or more frequent spillway flow.

xx
PR-100 Chapter 1 - Introduction

Introduction

Study Purpose
The purpose of this study was to demonstrate that the U.S. Army Corps of Engineers (USACE)
had a tool in place that would perform a systems-based risk analysis that could be used to
perform reservoir reoperation studies such as those introducing forecasting. Specifically, this
report describes a systems-based risk analysis performed with the USACE, Hydrologic
Engineering Center (HEC) software Watershed Analysis Tool (HEC-WAT) and associated
models to investigate the viability of FIRO (Forecast Informed Reservoir Operations) at Lake
Mendocino in the Russian River watershed in the State of California. The study demonstrates
practical tools and methods consistent with USACE guidance regarding evaluation of flood risk,
and the results of the analysis are used to support the PVA (Preliminary Viability Assessment) of
the FIRO demonstration project at Lake Mendocino.

The analysis compares modeling results from an Existing Conditions alternative (baseline)
against alternative reservoir operation plans that leverage "perfect" forecast information. The
FIRO alternatives were designed to provide more Lake Mendocino storage and Russian River
flow available for water supply and environmental uses without increasing flood risk.

Scope
This analysis evaluates the change in flood risk for alternative reservoir operations for Coyote
Valley Dam (Lake Mendocino) in the Russian River watershed. Existing hydrologic, reservoir,
hydraulic, and consequence models for the Russian River watershed were leveraged to better
understand how the incorporation of a forecast component into flood operation at Coyote Valley
Dam might affect flood risk along the Russian River, impact dam safety, and benefit water
supply.

The individual models were updated and integrated into an HEC-WAT study that compared
systems-based flood risk results from an Existing Conditions alternative against alternative water
management plans using forecast information. Alternative operations were developed based on
perfect knowledge of future precipitation, soil moisture, reservoir inflow, and downstream
channel capacity. This process allowed identification of the maximum possible gain in water
stored without significant increase in modeled flood damages. An additional alternative was
developed to demonstrate how a forecast with some level of uncertainty or error might impact
flood risk and water supply.

The HEC analysis described in this report represents a research/demonstration study. USACE's
San Francisco District (CESPN) operates Coyote Valley Dam and holds authority for any actual
changes to water management practices. The modeling framework was intended to be suitable
for subsequent analyses leading to implementation of FIRO at Coyote Valley Dam. The analysis
1
Chapter 1 – Introduction PR-100

complies with relevant USACE policy and guidance, including the requirement for a system-
wide risk and uncertainty based approach. Potential future applications include formulating and
screening alternative reservoir operations and different types of forecast information, or
providing technical justification for a major deviation, Water Control Manual update, or other
official USACE study process.

The USACE analysis used methods and data typical for USACE studies. The individual
computational software applications used within the HEC-WAT framework are approved for use
within the USACE Hydraulics, Hydrology, and Coastal Community of Practice (HH&C CoP).
The HEC-WAT software was officially released during the course of this analysis.

This report comprehensively describes the study methods and evaluates the results in great detail,
in order to document findings specific to Lake Mendocino operations and to provide a detailed
example for subsequent or similar FIRO studies. Casual readers may wish to read only the
overview, results, and conclusions in Chapters 3, 4, 16, and 17, which were summarized in the
PVA. Readers interested mainly in the development and performance of the FIRO alternatives
for Lake Mendocino operations should read Chapters 13 through 15. Readers interested mainly
in the study approach and application of the HEC-WAT framework should focus on Chapters 5
through 10.

Acknowledgements
This study was proposed by Mr. Matthew Fleming, Division Chief, Hydrology and Hydraulics
Technology Division, HEC, and executed by a team at HEC led by Mr. Matthew McPherson,
Senior Hydraulic Engineer, Water Management Systems Division, HEC. Dr. Cary Talbot of the
USACE Engineer Research and Development Center (ERDC) served as advisor, and Mr.
Christopher Dunn was the Director of HEC during the completion of this study.

Mr. Thomas Brauer developed the hydrologic model (HEC-HMS) and related parameters, with
assistance from Mr. Fleming. Dr. Leila Ostadrahimi developed the reservoir model and
alternative operations, with assistance from Ms. Joan Klipsch and Dr. Beth Faber. Mr. Cameron
Ackerman prepared the hydraulic model (HEC-RAS) and Mr. McPherson developed the HEC-
FIA model. Mr. McPherson performed most of the HEC-WAT modeling and analysis of results,
with substantial assistance from the previously named team members and also Mr. George
(Chan) Modini, Ms. Lea Adams, Mr. Michael Bartles, and Ms. Penni Baker. Ms. Marchia Bond
of the USACE Sacramento District (CESPK), Mr. Michael Dillabough of the USACE San
Francisco District (CESPN), and Mr. Chris Delaney and Mr. John Mendoza of the Sonoma
County Water Agency (SCWA) provided data and guidance during the course of this study.

2
PR-100 Chapter 2 – Background Information

Background Information

Basis for Study

Lake Mendocino is a dual use reservoir, which is owned and operated for flood control by
USACE and is operated by the SCWA for water supply. Due to recent changes in the operations
of an upstream hydroelectric facility, the Potter Valley Project (PVP), Lake Mendocino has
suffered from water supply reliability issues since 2007. Recent studies completed by SCWA
have found that the water supply reliability of Lake Mendocino is expected to continue to decline
with the current growth projections for the areas that rely on Lake Mendocino for water supply
and the potential changes to the regions hydrology expected with climate variability.

FIRO is a water management strategy that uses data from watershed monitoring programs and
improved weather and hydrologic forecasting to help water managers selectively retain or release
water from reservoirs in a flexible manner that more accurately reflects natural variability of
meteorology and hydrology, while not increasing flood risk (FIRO Steering Committee, 2015).
The strategy offers a potential alternative to improve the current reliability of Lake Mendocino
for meeting water supply and environmental flow objectives. The FIRO Steering Committee
established a multi-agency pilot project to investigate a FIRO application for the Russian River
watershed.

The FIRO Steering Committee (established in 2014), consists of representatives from USACE,
SCWA, SCRIPPS Institution of Oceanography (SCRIPPS), the National Oceanic and
Atmospheric Administration (NOAA), U.S. Geologic Survey (USGS), U.S. Bureau of
Reclamation (USBR), and the State of California, Department of Water Resources (DWR). In
July 2015, the FIRO Steering Committee developed a framework for evaluating whether FIRO is
a viable strategy to improve water supply reliability while not reducing the existing flood
protection capacity of Lake Mendocino. The PVA addressed the question of whether FIRO can
make better use of existing water supplies without increasing flood risk.

This report describes modeling activities performed to investigate the viability of FIRO at Lake
Mendocino in the Russian River watershed, and supports the technical findings of the PVA.

Description of Russian River Watershed

Location and Size

The Russian River drains a 1,485-square-mile watershed from the Coast Ranges in northern
California (Figure 1). The watershed is eighty miles long and 32 miles across at the widest
point. The Russian River Valley is flanked by two coastal ranges; the Mendocino Range

3
Chapter 2 – Background Information PR-100

Figure 1 Russian River Watershed Location

borders the basin to the west and the Mayacamas Mountains to the east. Elevations within the
watershed range from sea level at the ocean outlet to approximately 2,500 feet at the headwaters
and 4,500 feet along the eastern edge of the watershed.

The only two reservoirs in the watershed are operated by USACE for flood protection and water
supply – Lake Mendocino in the headwaters of the main Russian River and Lake Sonoma on Dry
Creek.

Land Cover
Higher elevations within the watershed are characterized by forested areas with some
concentrated areas of rangeland. The lower portions of the watershed along the Russian River
are characterized by agricultural lands with urbanization at concentrated population centers, as
shown in Figure 2.

River Course
The Russian River headwaters (Figure 3) rise from the east and west forks that merge near the
city of Ukiah, California. Coyote Valley Dam impounds Lake Mendocino on the East Fork
Russian River, about a mile above the confluence. The Pacific Gas and Electric (PG&E) Potter
Valley Project diverts water from the Eel River into the East Fork. Flow gages in the West Fork
and downstream at Hopland, California play important roles in Lake Mendocino reservoir
operations.
4
PR-100 Chapter 2 – Background Information

Figure 2 Land Use within the Russian River Watershed

SCWA defines the region above the USGS gage at Healdsburg, California as the Upper Russian
River (URR). Ukiah serves as the Mendocino County seat, and the 16,000 residents make it the
largest city in Mendocino County. Between Ukiah and Healdsburg the land use is dominated by
agriculture (mostly viticulture and orchards), and sheep and cattle grazing, and the river passes
through several small towns, and additional tributaries join the river along the way. Robinson
Creek enters the Russian River from the east just downstream of Ukiah, Feliz Creek enters from
the west near Hopland, Big Sulphur Creek enters from the east near Cloverdale, California, and
Maacama Creek joins the mainstem upstream of Healdsburg. Eight bridges span the Upper
Russian River.

5
Chapter 2 – Background Information PR-100

Figure 3 Map of the Russian River Watershed including the Potter Valley Project

Russian River hydraulics become more complex at Healdsburg in Sonoma County. The lower
Russian River passes through more developed areas and ten bridges. The largest tributary, Dry
Creek, joins the Russian River just below Healdsburg. The SCWA operates an inflatable dam
and fish passage facility at its Mirabel Diversion facility near Forestville. Various drainage
projects, private levees, and transportation embankments affect flood hydraulics in the region
between Healdsburg and Forestville. Gages at Geyserville and Healdsburg are used for Lake
Mendocino operations. Operations at Lake Sonoma use data from gages at Guerneville, from
gages on Dry Creek, and Russian River flows at the confluence with Dry Creek and the Pacific
Ocean are considered.
6
PR-100 Chapter 2 – Background Information

The southeastern portion of the Russian River watershed is heavily developed and populated.
The city of Santa Rosa, California, which is the Sonoma County seat, is the largest city with
168,000 residents in this portion of the watershed. Mark West Creek enters the Russian River
from the east at Mirabel Park near Forestville and drains approximately 254 square miles. The
Laguna de Santa Rosa (170 square miles) empties into the Mark West Creek approximately 2.5
miles upstream from its confluence with the Russian River and is a natural overflow basin for the
Russian River. After flowing past Mark West Creek, the Russian River turns west and flows
past several small towns before entering the Pacific Ocean near Jenner, California. The river
totals approximately 110 miles from the Potter Valley Project to the ocean.

Climate and Storms

The climate is Mediterranean with cool wet, winters and warm, dry summers, but the watershed
transitions from a dry interior portion dominated by hardwood forests, oak savannah, chaparral,
and grasslands, to a fog-influenced portion near the coast characterized by conifer forest.

Approximately 93 percent of the annual precipitation normally falls during the wet season, which
is October to May, with a large percentage of the rainfall typically occurring during three or four
major winter storms. These major storms often come in the form of an Atmospheric River,
which is the horizontal transport of large amounts of water vapor through the atmosphere along a
narrow corridor. Although brief, Atmospheric Rivers can produce thirty to fifty percent of the
region's annual precipitation during a few days (NOAA, 2018). Flood-producing rainfall is
deposited over the basin due, primarily, to orographic action of the mountain barriers combined
with frontal rainfall waves and/ or occluded frontal systems (USACE, 1954).

The Russian River watershed has variable rainfall due to orographic effects with lower rainfall
volumes at the lower elevations and higher rainfall volumes at the higher elevations. The average
annual rainfall for the entire watershed is approximately 46.4 inches, with average rainfall
approximately 32 inches at the lower elevations and as much as 78 inches in the upper elevations
of the Mendocino Coast Mountains.

Historical Flooding
The Russian River has flooded frequently throughout history. The 1959 version of the Coyote
Valley Dam Regulation Manual references historical large floods in 1906, 1909, 1937, 1940,
1942, 1943, and 1945, although damaging events occurred in a number of other years, as
evidenced by Figure 4, which shows part of Guerneville, California, in 1915 being flooded.

After Coyote Valley Dam was authorized in 1950, subsequent floods in 1953, 1955, 1956, and
1958 occurred before the dam was completed in 1959. The floods of 1909 and 1955 provided
the primary design hydrographs for flood storage and operations.

The largest event in the period of record occurred around Christmas of 1964. The storm remains
the only event to cause flow over the spillway at the Coyote Valley Dam, and serves as an
important reference point for the hydrologic analyses presented in this report.

7
Chapter 2 – Background Information PR-100

Figure 4 Flood of 1915 in Guerneville, California (photo provided by the Sonoma County Library)

In addition to the 1964 storm, the FIRO analysis is focused on events in 1955, 1978, 1983, 1986,
1995, 1997, 1998, and 2006 for validation of results from the individual models, and establishing
parameters for the probabilistic events that will be generated during the FIRO analysis. Specific
details about these events are discussed further in this report.

Floods in the Russian River watershed are normally of short duration, lasting three to four days,
developing within 24 to 48 hours after the beginning of a storm, but rapidly receding within two
or three days (USACE, 1984). Normally floods in the basin are flashy, since the times of
concentration on tributaries are short and flows respond rapidly to variations in rainfall (USACE,
1954). Floods typically occur during the rainy season from November through April, and larger
storms can inundate the portions of the alluvial valleys adjacent to the Russian River (USACE,
2003). On rare occasions, storms in October and May caused minor or moderate flooding.
Chapter 5 discusses the hydrologic sampling parameter development and includes a seasonal
analysis of Russian River storms.

The Russian River exceeds the official National Weather Services (NWS) flood stage
designation relatively frequently for the primary gage locations at Hopland, Healdsburg, and
Guerneville. Flood stage may result in little direct flood damage to contemporary structures, but
can still close roads and cause indirect problems related to drainage. The NWS maintains
additional thresholds at these locations for more damaging "moderate" and "major" floods.

The city of Hopland and surrounding areas are some of the most flood prone regions of the upper
Russian River. Flood stage at the USGS gage near Hopland (USGS station 11462500; NWS
station HOPC1) is 21 feet, which corresponds to a flow rate of approximately 15,000 cubic feet
per second (cfs). Since the completion of Coyote Valley Dam in 1959, the maximum flow rate
8
PR-100 Chapter 2 – Background Information

recorded at the Hopland gage is 33,700 cfs in December of 1964, and water levels have reached
flood stage sixteen times (22 percent of the years). Additionally minor flooding begins occurring
in Hopland when stage exceeds the banks of the channel, which can cause flooding, and closure
of the Highway 175 Bridge. According to the CNRFC (California-Nevada River Forecast
Center), this occurs at a stage of fifteen feet and a flow rate of approximately 8,140 cfs (NOAA,
2018a). Since 1959, flows have exceeded 8,140 cfs 124 times for 62 percent of the years.

The City of Healdsburg is also prone to flooding during extreme rainfall events. Flood stage at
the gage near Healdsburg (USGS station 11464000; NWS station HEAC1) is 53,000 cfs
(NOAA, 2018b). Since 1959, the maximum flow rate recorded at the Healdsburg Gage is 69,300
cfs, which occurred in January of 1995, and water levels have reached flood stage four times
(seven percent of the years).

Flood problems historically occur across many small communities along the Russian River
below Healdsburg. Flood warnings for the area are currently referenced to the Johnson's Beach
gage at the city of Guerneville (USGS station 11467002; NWS station GUEC1). The Johnson's
Beach gage came into service in 1996, and exceeded the NWS flood stage in six of the twenty
years on record. The gage is located on the downstream side of the old Highway 116 Bridge
adjacent to the Johnson's Beach recreation area, and is sometimes called the Guerneville Bridge
gage (note that a separate gage exists approximately six miles upstream at the Hacienda Bridge).
The Russian River near the Guerneville gage (USGS station 11467000; NWS station RIOC1)
offers discharge records since 1940, but does not have a designated flood stage. This report
distinguishes the two Guerneville gages as "Johnson's Beach" and "Hacienda". The Hacienda
gage provides the preferred reference regarding flood frequency, while the Johnson's Beach gage
provides the preferred reference regarding consequence analysis.

Water Management Structures

The only two reservoirs in the watershed are operated by USACE for flood protection and water
supply – Lake Mendocino in the headwaters of the main Russian River, and Lake Sonoma on
Dry Creek. Most of the Russian River community relies heavily on Lake Mendocino for flow
augmentation during the dry season. The river faces rising municipal and agricultural demands,
while also supporting instream flow requirements for fish species listed as threatened. The
scarcity has become more critical in recent years due to substantial reductions in the water
diverted to Lake Mendocino from the neighboring Eel River, through the Potter Valley Project.

The Russian River also contains two other significant water projects designed to support
hydropower and water supply: PG&E's Potter Valley Project, and SSCWA's Russian River
Diversion Structure and related facilities.

Potter Valley Project

PG&E's Potter Valley Project was constructed in 1908 for power generation purposes. Water is
collected for storage in Lake Pillsbury, a reservoir created by the Scott Dam on the Eel River.
Natural flows from the Eel River and water released from Lake Pillsbury storage are diverted
twelve miles downstream from Scott Dam at Cape Horn Dam and then are conveyed through a
diversion tunnel and penstocks to the Potter Valley Powerhouse, which is located in the Russian
9
Chapter 2 – Background Information PR-100

River watershed. A map of the facilities of the PVP is provided in Figure 5. Some of the water
discharged from the powerhouse is diverted into canals from which the Potter Valley Irrigation
District (PVID) receives water under a water supply agreement with PG&E and the PVID's own
appropriative water rights license. The PVID can divert up to 50 cfs of flows from the Potter
Valley Project for use by their customers. The remaining water discharged from the powerhouse
not consumptively used by PVID flows down the East Fork Russian River into Lake Mendocino.
The Potter Valley Project has a maximum flow capacity of approximately 300 cfs and a
generation capacity of 9.4 megawatts (MW). Potter Valley Project diversions and operations are
regulated by a license issued to PG&E by the Federal Energy Regulatory Commission (FERC).
The Potter Valley Project serves multiple purposes, including power generation, Potter Valley
agricultural irrigation uses, and minimum instream flow releases into the East Fork Russian
River.

Figure 5 Map of the Potter Valley Project

PG&E manages releases from Lake Pillsbury to meet FERC-required minimum release
requirements in the Eel River and to provide water for diversions to the Potter Valley Project
powerhouse. PG&E does not manage or coordinate the operation of Potter Valley Project with
USACE or SCWA's operations of Lake Mendocino. However, the historical importance of water
from the Potter Valley Project to Lake Mendocino water supplies is demonstrated by the fact that
the California SWRCB's (State Water Resources Control Board) Decision 1610, which adopted
several terms now in the SCWA's water right permits. From the permits, the SWRCB established
10
PR-100 Chapter 2 – Background Information

a hydrologic index for the Russian River and Dry Creek that contains minimum instream flow
that is based on cumulative inflows into Lake Pillsbury.

In 2004, FERC amended PG&E's license to improve conditions for salmon species listed as
threatened species under the Federal Endangered Species Act (ESA). PG&E began operation of
the Potter Valley Project in accordance with the amended FERC license in 2006, and these
revised operations substantially reduced the amounts of Potter Valley Project diversions
compared to historical levels. These reductions are illustrated in Figure 6, which shows historic
average water year inflow into Lake Mendocino and Potter Valley Project releases shown in blue
for two periods:

1) Operations since the construction of Coyote Valley Dam and prior to the implementation
of the amended FERC license, 1959-2006, and

2) Operations after the implementation of the amended FERC license, 2007 - 2015.

Figure 6 Lake Mendocino Average Annual Inflow for Periods both Prior to and After the
Implementation of the Potter Valley Project FERC License Amendment in the Fall of 2006

For example, in Figure 6, the Lake Mendocino inflows (represented by the orange bars), for the
period, 2007 to 2015 have declined from historic inflows, which is largely the result of reduced
Potter Valley Project transfers. Changes in the seasonal timings of Potter Valley Project
diversions have also affected Lake Mendocino water storage reliability. Reduced inflows in the
11
Chapter 2 – Background Information PR-100

spring have contributed to declining water supply reliability of Lake Mendocino through the
summer months (SCWA, 2015). As a result, the SCWA has had to file several Temporary
Urgency Change Petitions (TUCPs) with the California SWRCB. These TUCPs temporarily
reduce the minimum instream flow requirements in the SCWA's water right permits as necessary
to preserve water supply storage in Lake Mendocino for downstream beneficial uses.

Coyote Valley Dam (CVD)

The Coyote Valley Dam impounds Lake Mendocino on the East Fork Russian River, about four
miles northeast of the City of Ukiah in Mendocino County, California. The Coyote Valley Dam
project was authorized by the Flood Control Act of 1950 for purposes of flood control, water
supply, and recreation. Construction was completed by USACE in January 1959, with SCWA
participating as the non-federal local sponsor. Congressionally authorized mitigation for effects
on the endangered fish caused by the dam were added in the mid 1970's.

Physical Description
Coyote Valley Dam is an earth embankment dam approximately 160 feet high with a crest length
of 3,500 feet. Lake Mendocino has a total current storage capacity of 116,500 acre-feet, which
includes a water supply pool of between 68,400 acre-feet and 111,000 acre-feet, depending on
the time of year. Based on reservoir bathymetric surveys (original in 1952 and most recent in
2001) the average sedimentation rate in the reservoir is estimated to be 143 acre-feet per year
(AFY). The invert of the controlled outlet is at an elevation of 637 feet above mean sea level
(USACE, 2003). This level in the reservoir establishes the top of the inactive pool, which,
according to the 1986 Water Control Manual, was estimated to have a capacity of 135 acre-feet.
Based on the historic rate of sedimentation, it is expected that the inactive pool has reached its
capacity to accumulate sediment.

The watershed of the reservoir has an area of approximately 105 square miles, which is
approximately seven percent of the total watershed area of the Russian River Basin. Average
annual inflow into the reservoir since the construction of Coyote Valley Dam is approximately
230,000 AFY, with a peak annual inflow of 443,000 acre-feet in 1983 and a minimum annual
inflow of 60,000 acre-feet in 1977. Inflow into the reservoir consists of unimpaired flows from
the contributing watershed area and a portion of the water diverted though the Potter Valley
Project from the Eel River.

The Coyote Valley Dam and adjacent area, looking north in a "tilted" satellite view, is displayed
in Figure 7. The spillway of Lake Mendocino is located in a low saddle about 0.6 miles
upstream from the southern abutment of the dam (Figure 7). The spillway discharges flows
through Howard Creek until joining the Russian River just north of the Ukiah city limits. The
spillway structure consists of an 800-foot long approach channel and a 200-foot wide rectangular
weir. Since construction of Coyote Valley Dam, the spillway has only been activated once in
December of 1964 when inflows exceeded 14,000 cfs.

Location of the controlled outlets for Lake Mendocino are displayed in Figure 8. Water is
conveyed to the outlet works of Coyote Valley Dam by a single reinforced concrete pipe
approximately 720 feet long and eleven feet and ten inches in diameter. The flow through this

12
PR-100 Chapter 2 – Background Information

Figure 7 Coyote Valley Dam Area

Figure 8 Coyote Valley Dam Outlet Area

tunnel may be directed to the power plant, or passed through a flood control gate. Maximum
release capacity of the controlled outlet is approximately 7,500 cfs when the water surface
elevation is within the Emergency Release Pool (above elevation 773 feet mean sea level). The
powerhouse contains two turbine/generator units with rated power generation capacities of 2,500
and 1,000 kilowatts.

13
Chapter 2 – Background Information PR-100

Operations
USACE determines the schedule and amount of water released from Lake Mendocino during
flood control operations, while SCWA manages releases from the conservation pool. Regulation
for flood control and water supply operations are described in the "Coyote Valley Dam Water
Control Manual - Appendix I", which was originally published by USACE in April 1959 and
revised in August 1986. Exhibit A of the Coyote Valley Dam Water Control Manual (WCM)
was most recently revised in September 2003 to incorporate the most recent bathymetric survey
information (USACE, 2003).

Most of the river relies heavily on Lake Mendocino for flow augmentation during the dry season.
The Russian River is facing rising municipal and agricultural demands, while also supporting
instream flow requirements for fish species that are listed as threatened. The scarcity of water
has become more critical in recent years due to substantial reductions in the water diverted to
Lake Mendocino from the neighboring Eel River, through the Potter Valley Project.

Guide Curve
Water management operations at Coyote Valley Dam manage storage in the reservoir according
to the reservoir guide curve defined in the Coyote Valley Dam Water Control Manual (Figure 9).
The guide curve sets the maximum threshold for storage of conservation water in the reservoir
and the Lake Mendocino storage zones (Figure 9).

Figure 9 Lake Mendocino Storage Zones

The volume of the conservation pool defined in the Coyote Valley Dam Water Control Manual
(USACE, 2003) is 68,400 acre-feet (elevation 737.5 feet NGVD (National Geodetic Vertical
Datum)) from November through February. From 1 March to 10 May, the defined water supply

14
PR-100 Chapter 2 – Background Information

pool linearly increases to a storage level of 111,000 acre-feet (elevation 761.8 feet NGVD).
From 1 through 30 October, the water supply pool linearly decreases from 111,000 acre-feet to
68,400 acre-feet. The storage volume of the conservation pool is reduced at the onset of the
rainy season to maximize flood space capacity and is then increased in the spring, when the
likelihood of large storm events is low, to maximize storage for water supply purposes. The
maximum guide curve storage elevation of 761.8 feet provides three-feet of freeboard from the
spillway crest (764.8 feet NGVD) to limit spillway overflow resulting from variations due to
wind and wave action, diurnal fluctuations in Potter Valley Project releases, or possible minor
surface runoff (USACE, 1954).

The guide curve (Figure 9) that was explained in this section was implemented in the spring of
2007. Prior to 2007 the increase in the conservation pool from 68,400 acre-feet did not begin
until 1 April and reached a maximum level of conservation storage of 86,400 acre-feet (elevation
748 feet NGVD) on 20 April. Prior to 2009, the increase in the conservation pool could start on
1 March, but the SCWA has to provide a written request to USACE annually.

Mode of Operation
The Lake Mendocino power plant and associated facilities operate generally in two modes:
power generation mode and flood control mode. Flood control mode must be in operation when
the flood pool exceeds elevation 755 feet. Power generation mode may occur when the flood
pool is at or below elevation 755 feet. In the power generation mode the gate will be fully
closed, and the turbines and bypass valves will be opened as necessary to pass the required water
release. In the flood operation mode, all outflow typically passes through the gate, and the
turbines and bypass valves are fully closed.

The design of the intake allows flow to the outflow gate or the power plant, but not both
simultaneously. Switching between power generation and flood control modes, requires five to
six hours of transition, so that pressure differentials can be properly managed. The activity
requires a small amount of planning for staff availability and facility readiness. The power plant
has a limited capacity to make flood control releases, providing water managers with some
degree of operational flexibility. Coyote Valley Dam typically transitions in and out of flood
control mode a few times over the winter months.

Flood Operations
The nominal flood control pool consists of the storage between the guide curve and elevation
771.0 feet NGVD (storage level 128,100 acre-feet). This amounts to 59,700 acre-feet of flood
control space during the winter flood season (737.5 feet NGVD; 68,400 acre-feet), and 17,100
acre-feet during the summer (761.8 feet NGVD; 111,000 acre-feet). However, flood control
operations in practice typically regard the upper bound as the spillway crest (764.8 feet NGVD;
116,500 acre-feet), rendering the active flood control space as 48,100 acre-feet in winter and
5,500 acre-feet in summer. The Coyote Valley Dam Water Control Manual (USACE, 2003)
labels these volumes as "flood control reservation".

Lake Mendocino features relatively simple flood operations and under normal circumstances,
Coyote Valley Dam minimizes outflows during large storm events, temporarily storing flow
from the East Fork of the Russian River in reservoir space reserved for flood control, and
15
Chapter 2 – Background Information PR-100

releasing it later as the river recedes downstream. Under unusual circumstances, the chance of
contributing to downstream flood damage may need to be weighed against the risk of insufficient
flood control storage remaining to absorb the next storm. The Coyote Valley Dam Water
Control Manual (USACE, 2003) balances these risks by allowing escalating maximum releases
for three different zones (schedules) within the flood control storage.

Flood control releases from Lake Mendocino are also guided by downstream maximum flow
criteria. Releases from Coyote Valley Dam are constrained by the following outflow limitations:

1. Rate-of-change (also called ramping rates): These values supersede those listed in the
latest Coyote Valley Dam Water Control Manual (USACE, 2003).

a. Increasing: The Biological Opinion report (NMFS, 2008) allow releases to

increase up to 2,000 cfs per hour when outflows from the reservoir exceed 1,000
cfs, and up to 1,000 cfs per hour when less than 1,000 cfs.

b. Decreasing: NMFS requested USACE in April 2016 to implement new ramping

schedule criteria to help protect federally listed salmonids. USACE agreed that in
most circumstances, it was feasible to ramp down flood releases in the following
manner:

• Flood releases from Coyote Valley Dam between 2,500 cfs and 4,000 cfs
ramp down at 250 cfs/hour;

• Flood releases from Coyote Valley Dam below 2,500 cfs ramp down at
100 cfs/hour;

• Between 15 March and 15 May, for releases from Coyote Valley Dam
below 250 cfs, ramp down at 25 cfs/hour not to exceed 50 cfs/day;

• Between 16 May and 14 March, at releases from Coyote Valley Dam

below 250 cfs ramp down at 25 cfs.

2. Flow in West Fork: Releases limited to 25 cfs when the Russian River at Ukiah USGS
gage reaches 2,500 cfs and is rising.

3. Hopland Rule: Insofar as possible, avoid flood releases that contribute to flows greater
than 8,000 cfs at the Hopland gage.

4. Flood Control Schedule Limits (Appendix A): The zones displayed in Figure 9 govern
the releases used to empty the flood control space following a storm.

a. Schedule 1 – up to 4,000 cfs if the pool stayed below 746.0 feet NGVD

b. Schedule 2 – 4,000 cfs if the pool exceeded 746.0 feet NGVD but remained
below 755.0 feet NGVD

16
PR-100 Chapter 2 – Background Information

c. Schedule 3 – up to 6,400 cfs if the pool exceeded 755.0 feet NGVD

d. Emergency Release Schedule - provides guidance for releases made through the
gate when reservoir water levels exceed 771 feet NGVD

The hydraulics of weir flow over the 200-foot uncontrolled spillway quickly increases the
outflow as the lake rises, so that the spillway provides all of the specified Flood Control
Schedule 3 (Appendix A) release and the gate stays closed. In situations with spillway flow, the
outflow from Coyote Valley Dam could become largely outside the control of operators, with
Lake Mendocino no longer providing as much flood protection to downstream locations. Lake
Mendocino has never reached the Emergency Release Schedule (Appendix A).

Conservation Operations
Releases while in power generation mode occur through the power plant, operated by the city of
Ukiah. The plant can release up to 4,375 cfs, and USACE directs the amount of outflow for
flood operations. Flood operations through the power plant are relatively rare. The SCWA
directs the amount of conservation pool releases, which usually occur through the power plant.
A complex set of minimum flow requirements for several downstream locations govern the
releases. The instream flow requirements (SWRCB Decision 1610) are a function of a
hydrologic index based on reservoir storages of Lake Mendocino and Lake Pillsbury, and a
categorization of the water year as normal, dry, or critical. The Biological Opinion of 2008
superseded some of the SWRCB Decision 1610 (Figure 10) minimum flows in the Upper
Russian River.

Historical Pool Levels for Conservation Flows

Operations for Lake Mendocino follow a standard guide curve that allows it to fill during the
spring, so that on 10 May the reservoir reaches its maximum amount of stored water for the year,
corresponding to the minimum storage reserved for floods. Unfortunately, inflows to the
reservoir usually fail to conform to this schedule, and Lake Mendocino typically enters the dry
season lower than desired. Maximizing the storage at the end of flood season makes more water
available during the summer and fall for downstream water supply and environmental needs.

Historically, spring refill of Lake Mendocino followed several different strategies (Figure 11).
During the first six years of operation, Lake Mendocino was drawn down during the winter much
lower than the 737.5 feet NGVD level indicated in the water control manual, and only allowed to
rise between 737.5 and 740.0 feet NGVD. During the next six years, operators drew down the
winter level less severely, and filled the pool during spring to around 743 feet NGVD. The
reasons for these early operations that differed from the 1959 official Coyote Valley Dam Water
Control Manual were unavailable for this analysis. The simplest speculation is that recreation
and water supply demands were still developing, so operators enjoyed flexibility to place a
greater emphasis on flood control.

From 1971 to 2002, reservoir operators filled Lake Mendocino to 748.0 feet NGVD, identified as
the "summer pool elevation" by Note 3 of the 2003 Water Control Diagram:

17
Chapter 2 – Background Information PR-100

Figure 10 California SWRCB Decision 1610 - Instream Flows

18
PR-100 Chapter 2 – Background Information

Figure 11 Historical Lake Mendocino Spring Maximum Pools

"Normally, the summer pool elevation will be kept at 748.0 feet to maximize
recreational opportunities at the lake; however, Sonoma County Water Agency
retains the rights to raise the summer pool to 761.8 feet based on demonstrated
demand and NEPA (National Environmental Policy Act) criteria being met".

The operators were consistently successful in filling the pool to the desired level during those 32
years. Notable outliers occurred due to an April storm in 1974 and the historic drought in 1977.
Reservoir inflows proved inadequate in only five other years during this period, to fill the pool to
748.0 feet NGVD.

The storage that Lake Mendocino carries into the dry season has been much less consistent since
2002. Water supply and environmental demands on the storage led to recognition of 761.8 feet
NGVD as the contemporary summer pool level, although the lake has never achieved it. The
reduction of the Potter Valley Project diversion in 2004 and recent droughts have resulted in
significant variability in the ability to fill Lake Mendocino in the spring. A part of this study
demonstrates how the use of forecast information can allow Coyote Valley Dam operations to
more reliably fill Lake Mendocino to the desired summer pool.

Impact on Historical Flood Peaks

Regulation by Coyote Valley Dam reduced peak flows, increased the lag time between flood
peaks entering and exiting Lake Mendocino, and increased the duration of high flow
downstream. The median of instantaneous peak flows recorded at the Russian River at the
Hopland, Cloverdale, and Healdsburg gages decreased after the Coyote Valley Dam began
operating in 1959, but since the structure only regulates thirteen percent of the watershed above
19
Chapter 2 – Background Information PR-100

Healdsburg, and seven percent of the total watershed, the decreases are minor (Table 1). In
1986, USACE found that the dam reduced flood peaks by 29 percent at Hopland, by 21 percent
at Cloverdale, and by eleven percent at Healdsburg (USACE, 2003). The greatest decreases
occur upstream, closest to the dam and lessen downstream due to greater contributing area and
unregulated tributary inputs.

Table 1 Flood Flows on the Upper Russian River Before and After Coyote Valley Dam
Russian River Russian River Russian River
near Hopland (cfs) near Cloverdale (cfs) near Healdsburg (cfs)
(USGS gage no. 11462500) (USGS gage no. 11463000) (USGS gage no. 11464000)
1937-present 1951-present 1937-present
362 square mile 503 square mile 793 square mile
Date1 drainage area drainage area drainage area
February 1940 34,100 No record 67,000
January 1943 34,000 No record 53,330
January 1954 27,400 33,300 53,700
December 1955 45,000 53,000 65,400
February 1958 32,300 38,100 50,900
Pre dam median 21,250 22,350 33,950
December 1964 41,500 55,200 71,300
January 1974 39,700 51,900 64,700
February 1986 35,600 40,700 71,100
January 1995 27,600 39,400 73,000
December 2005 35,600 50,700 58,900
Post-dam median 14,550 18,200 32,050
1
Before Coyote Valley Dam: pre-1959; Post Coyote Valley Dam: post-1959

If the reservoir pool level rises high enough, flow occurs over the uncontrolled spillway at
Coyote Valley Dam. This is undesirable because the reservoir can no longer completely absorb
upstream flood peaks. Flood protection offered by the reservoir during spillway flow is limited
to attenuation and delay of the flood peak due to routing through the lake storage.

The 1964 event (Figure 12) caused Coyote Valley Dam spillway to discharge for about a day
after a peak inflow was generated with the flow lasting almost two days. Fortunately, the event
had passed, and none of the references consulted for this report attribute any flood damage to the
spillway flow. The rarity of spillway flow itself may also represent an indirect flood risk.
Uncertainty about performance of the structure does not factor into the flood risk calculations in
this analysis. However, the probability of different hydraulic loadings could inform a separate
dam safety analyses.

Dam Safety Action Classification (DSAC) Rating

Coyote Valley Dam was assigned a DSAC Class III. This rating designates dams that have
issues where the dam is significantly inadequate or the combination of life, economic, or
environmental consequences with probability of failure is moderate to high. A 2009 study
showed that the Coyote Valley Dam spillway was unable to pass the Probable Maximum Flood
(PMF) without overtopping the dam, but the specific reasons for the rating were unavailable for
this analysis. The DSAC rating could be very important to the potential application of FIRO to
Lake Mendocino. USACE policy might require a resolution of dam safety concerns at Coyote
Valley Dam before adopting water management practices that lead to any increase in dam safety
risk. Such dam safety considerations would affect the specific design of the FIRO plans. For
20
PR-100 Chapter 2 – Background Information

Figure 12 Coyote Valley Dam – 1964 Event

example, the DSAC rating may preclude formulation of plans that trade gains in captured storage
against the chance of greater hydraulic loading on the dam, or greater likelihood of spillway
flow.

Warm Springs Dam (WSD)

Warm Springs Dam was completed in 1983 on Dry Creek, creating Lake Sonoma. The Warm
Springs Dam was authorized by the Flood Control Act of 1962 for the purposes of flood control,
water supply, environmental stewardship, and recreation. Congressional authorization in the
mid-1970's added mitigation for ESA fish to the mission set for both federal dams on the Russian
River. Warm Springs Dam is a compacted earth fill with an impervious core dam, with a
maximum height above the streambed of 319 feet and a crest length of 3,000 feet. Elevation of
the dam crest is 519.0 feet NGVD. The adjacent uncontrolled spillway has a 100-foot long crest
at elevation 495.0 feet NGVD.

The drainage area above Warm Springs Dam totals approximately 131 square miles, or about 25
percent more area than above Coyote Valley Dam. However, Warm Springs Dam offers
136,000 acre-feet of flood control reservation between the guide curve and spillway crest, which
is 1.8 times the flood control reservation at Lake Mendocino. The outlet works consist of (low
flow water quality outlet) with three five-foot diameter intake tunnels (at elevations of 431, 391,
and 352 feet). The flood control outlets at Warm Springs Dam consist of two 5' x 8' service gates
and two 5' x 8' slide gates. These outlets restrict flows at the Yoakum Bridge near Geyserville
(7,000 cfs), and at Guerneville, where flows cannot exceed 35,000 cfs.

Similar to Coyote Valley Dam, flood control operations at Warm Springs Dam typically require
outflows to be minimized during storms. Lake Sonoma offers much more storage than Lake
Mendocino, so reservoir operators have greater flexibility regarding when to release storage
accumulated during storms. Storm storage retained in Lake Sonoma substantially reduces flood
peaks on Dry Creek and its confluence near Healdsburg, but has a limited effect on flood peaks
along the lower Russian River.
21
Chapter 2 – Background Information PR-100

Similar to Coyote Valley Dam, Warm Springs Dam supports instream flows at locations on Dry
Creek and the lower Russian River. Warm Springs Dam also maintains flow to a fish hatchery
immediately below the dam. Coyote Valley Dam and Warm Springs Dam perform no explicit
system operations. However, Warm Springs Dam releases to provide minimum flows on the
Russian River, take prior releases from Coyote Valley Dam into consideration, representing an
implicit system operation.

Russian River Diversion Structure (RDS) and

Inflatable Dam
SCWA operates an inflatable dam on the Russian River near Mirabel, about halfway between the
Wohler Bridge and confluence with Mark West Creek. The SCWA inflates the dam in late
spring or early summer to divert water into recharge ponds adjacent to the river, supporting six
large collector wells that pump water from a depth of about 100 feet after natural filtration
through sand and gravel. Figure 13 shows the inflated dam. The fish passage facility and
viewing gallery along the right bank (under the concrete and glass) were completed in 2016. The
RDS and associated withdrawals play a very important role in modeling conservation flows in
the Russian River system, but have little relevance to flood flow modeling.

Figure 13 Russian River Diversion Structure and Inflatable Dam

22
PR-100 Chapter 2 – Background Information

Potential Forecast-Informed Reservoir

Operations
Ongoing research into Atmospheric Rivers may offer improvements in the variety and quality of
forecast information available to water managers in the Russian River watershed. Developing
modified operations that effectively leverage the forecast information requires creativity and
cooperation across several scientific and engineering specialties. This analysis considered only
the most obvious and familiar types of forecast information and associated operations.

Existing Uses of Forecasts

Operations leveraging forecast information can act in advance of flood events, reducing the
amount of reservoir storage reserved for flood operations in the absence of a forecasted event,
and thus offering the potential to retain more water during the wet season for use during the dry
season. For example, some long-standing reservoir operation plans shift the guide curve higher
or lower according to basin conditions such as soil saturation, snow water equivalent, or storage
in other upstream reservoirs. Reservoirs operating to protect downstream locations often
consider anticipated flows routing in from streams along the intervening reach.

The Water Control Manuals of the Coyote Valley Dam and Warm Springs Dam both stipulate
use of specific real-time precipitation forecasts. Note 4 from the Coyote Valley Dam and Warm
Springs Dam Water Control Manuals states:

"When the NWS QPF (Quantitative Precipitation Forecast) is one-inch or more for
the next 24 hours, or a half- inch or more for any six-hour period in the next 24
hours, flows in the Russian River will be monitored hourly so that reductions can be
made to ensure dam operations will adhere to all other limitations and operating
criteria."

These existing QPF thresholds represent a commitment to operator readiness, but do not modify
reservoir operations. The forecast-informed reservoir operations analyzed in this report
explicitly determine Coyote Valley Dam outflows based on forecast information.

Types of Reservoir Operations

The Coyote Valley Dam lends itself to two basic water management strategies: releasing
reservoir storage in advance of storms, and storing more water in the absence of storms. These
concepts could be implemented using a variety of specific rules and water management
guidelines able to leverage a wide variety of forecast information.

The advance-release strategy trades off the advantage of more available flood storage against the
possibility of the releases causing problems downstream. Advance release rules can be triggered
by a variety of information, using different outflow rates and schedules under different
circumstances. Reservoir operators routinely make release decisions about how aggressively to
empty flood storage, and often exercise significant discretion in the schedule and amount when
not in an emergency. The operator's judgment is influenced by weather forecasts, or a variety of
23
Chapter 2 – Background Information PR-100

miscellaneous considerations such as impacts on fish habitat, water quality, recreation, or

construction projects. Alternately, if advance releases are expected to cause downstream
damage, the releases may be codified in the regulation plan according to emergency measures.
Conceptually, advance releases could also be used to clear storage out of the conservation zone
and provide flood protection beyond the design of the project. In that case, the trade-off also
includes the chance that anticipated inflows might fail to bring the reservoir pool back to the
guide curve.

Another case, where the storing of more water in the absence of flood trade-offs, includes the
chance that the gain in water against the risk that floods might occur could overwhelm the
available storage. The guide curve may adjust higher or lower at various times, in a prescribed
manner according to some specific parameter, such as an inflow volume forecast or snow-water
equivalent in a well-defined procedure.

When the reservoir rises into the flood pool, water managers must use judgment regarding how
quickly to release the stored water to free up the flood storage space. The Coyote Valley Dam
Water Control Manual offers guidelines, but allows discretion in balancing potential downstream
consequences of higher outflows against the risk of worse problems in facing subsequent storms
with a fuller reservoir. Forecast information traditionally factors into such decisions.

Multiple reservoir operations based on a combination of forecast information types can be

combined in a regulation plan, in order to use the information most helpful for a given season
and type of risk.

Types of Forecast Information

The types of forecasts chosen for use in this analysis reflect common products that are available
for the Russian River watershed, or could be provided with existing technology. For simplicity,
only deterministic forecast information was used. Reservoir operations were formulated as
follows:

a) Basin-average forecasted precipitation in the drainage area above Coyote Valley

Dam - the NWS forecasts precipitation totals for various durations in the future for many
locations and areas, and the forecast information used in this report could easily be
derived from existing NWS forecast products. Precipitation forecast information could
be especially useful for some types of reservoir rules because forecasters can often
predict the absence of rain with great accuracy.

b) Forecasted soil moisture conditions in the drainage area above Coyote Valley Dam -
USACE water management personnel could produce forecasts of the moisture storage or
deficit in the soil using the Corps Water Management System (CWMS). Alternately, the
CNRFC could provide the information as a by-product of its other forecast modeling.
Soil moisture information is sometimes used to adjust reservoir operations to reflect the
immediacy and scale of rainfall-runoff response, such as preserving more flood storage in
a saturated basin.

24
PR-100 Chapter 2 – Background Information

c) Forecasted Lake Mendocino reservoir inflow - USACE water management staff

members could forecast the reservoir inflows using CWMS. Alternately, the CNRFC
provides forecasted volumes of water expected to arrive at certain California reservoirs,
and could accommodate Lake Mendocino. Reservoir inflow is typically the most
important metric affecting operations, and can be very useful for driving a variety of
rules. In addition, this information represents a sort of "master" forecast, since it
incorporates precipitation, soil moisture, and other forecast information.

d) Forecasted unregulated downstream flow at Hopland - USACE water management

personnel could forecast the downstream flow using the CWMS. Alternately, the
CNRFC forecasts streamflow for the Russian River at Hopland, and could provide a
version of the forecast that does not include releases from Coyote Valley Dam. This
information is useful for maximizing Coyote Valley Dam releases while observing the
constraint regarding 8,000 cfs maximum flow at Hopland.

Other types of forecast information could be considered in subsequent analyses. In particular,

ongoing research about Atmospheric River events may yield useful predictors regarding the
imminence or absence of large storms.

Reservoir operators associated with the Russian River have shown great confidence and
familiarity with forecast products of five days or less. In accordance with the desire to use
common and practical forecast information, the longest forecast period considered in this report
is five days. The utility of longer forecasts is of interest to the FIRO project, and under research
in efforts parallel to the study presented in this report. The results of the research could identify
other forecast information of use to potential reservoir operation that offers longer warning
times.

Forecast Skill
In addition to the numerical quantities and timing of forecasted data, some forecasts may include
quantifications of associated "skill", providing an indication of the uncertainty in a forecast. In
principle, reservoir rules could be designed that incorporate the skill, making reservoir operation
explicitly a function of the uncertainty around a forecast, as well as the scale of the forecast.
Ongoing research by CW3E (Center for Western Weather and Water Extremes) into "Go/No Go"
thresholds for Lake Mendocino reservoir inflow forecasts holds promise for operational logic
that considers forecast skill.

However, reservoir operations that directly factor numerical estimates of skill into the release
decisions remain a research topic. For simplicity, the rules employed in this analysis use only a
deterministic approach.

25
Chapter 2 – Background Information PR-100

26
PR-100 Chapter 3 – HEC Software

HEC Software

HEC-WAT
HEC-WAT provides a framework that allows users to conduct water resources studies of
complex riverine systems with an integrated, comprehensive and systems based approach. This
includes a robust analysis of flood risk, including modeling of alternative reservoir operations
based on different types of forecast information. Within HEC-WAT, the HEC suite of software
applications are used to perform many of the necessary hydrologic, hydraulic, and planning
analyses from a single interface. The analysis procedures are consistent with requirements of ER
1105-2-100, ER 1105-2-101, and EM 1100-2-1619.

The Russian River watershed implemented in HEC-WAT includes previously created HEC-
HMS, HEC-ResSim, HEC-RAS, and HEC-FIA models. The integration of the individual pieces
of software (Figure 14) within the HEC-WAT framework is achieved through the concept of a
"plug-in". The plug-in allows the individual pieces of software to integrate without requiring
special code in HEC-WAT to support the individual pieces of software. The individual pieces of
software provide the technical mechanics within the framework (i.e., editing, reporting
capability, computational analyses, etc.) and are not intended to contain any HEC-WAT specific
code. HEC-WAT provides an analysis framework, while the individual pieces of software
provide the analytical computations.

Figure 14 HEC-WAT Framework

27
Chapter 3 – HEC Software PR-100

HEC Software Integrated into HEC-WAT

The primary set of initial software to be used during the analytical process in HEC-WAT is:

• HEC-HMS (Hydrologic Modeling System)

• HEC-RAS (River Analysis System)
• HEC-ResSim (Reservoir System Simulation)
• HEC-FIA (Flood Impact Analysis)

The following sections provide a brief description of the HEC-WAT software applications that
will be used during the FIRO analysis.

Hydrologic - HEC-HMS
The HEC Hydrologic Modeling System software (HEC-HMS) is a standalone software
application that is designed to simulate the precipitation-runoff processes of dendritic watershed
systems. The software is designed to be applicable in a wide range of geographic areas and is a
generalized hydrologic modeling system capable of representing many different watersheds to
simulate either a single flood event or a continuous simulation that spans a couple of weeks or
multiple decades. HEC-HMS in the HEC-WAT compute sequence is a hydrologic precipitation
rainfall-runoff model that generally provides simulated flow at multiple locations within a
watershed. For the FIRO analysis, Version 4.2 of HEC-HMS was used.

Reservoirs - HEC-ResSim
THE HEC Reservoir System Simulation software (HEC-ResSim) is a standalone software
application that is used to model reservoir operations at one or more reservoirs whose rule-based
operations are defined by a variety of operational goals and constraints. The software simulates
reservoir operations for flood management, low flow augmentation and water supply for
planning studies, detailed reservoir regulation plan investigations, and real-time decision support.
HEC-ResSim in the HEC-WAT compute sequence is a reservoir simulation model that provides
regulated flow hydrographs. For the FIRO analysis, Version 3.3 of HEC-ResSim was used.

Hydraulics - HEC-RAS
The HEC River Analysis System software (HEC-RAS) is a standalone software application that
allows the user to perform one-dimensional steady and unsteady flow and two-dimensional
unsteady flow simulations. HEC-RAS contains four one-dimensional river analysis components
for: (1) steady flow water surface profile computations; (2) unsteady flow simulation; (3)
movable boundary sediment transport computations; and (4) water quality analysis. The HEC-
RAS Mapper tool that is available from the HEC-RAS interface, calculates inundation boundary
and water depth maps. HEC-RAS in the HEC-WAT compute sequence is a river hydraulic
model that computes the river stages, water surface profiles, and levee breach information that is
used to compute consequences by HEC-FIA. For the FIRO analysis, Version 5.0.3 of HEC-RAS
was used.

28
PR-100 Chapter 3 – HEC Software

Consequences - HEC-FIA
The HEC Flood Impact Analysis software (HEC-FIA) is a standalone software application that
evaluates consequences using either continuously observed or simulated stage hydrographs
(hydrograph-based) or depth grids (GIS-based). For a specified analysis, the software evaluates
urban and agricultural flood damage, area inundated, number of structures inundated, and
consequences. Consequences include economic and life loss. HEC-FIA also provides
information to flood managers and emergency managers for emergency response activities. For
the FIRO analysis, Version 3.0.1 of HEC-FIA was used.

HEC-DSSVue
HEC-DSSVue (http://www.hec.usace.army.mil/software/hec-dssvue/) graphs and tabulates data
that is stored in a DSS file. Over sixty mathematical manipulation functions are available for
operations on records within a DSS file, as well as data entry functions, and several utility and
database maintenance functions. Data can be displayed from a selection of record names
(pathnames) or from spatially referenced locations with a map background. Common data stored
in a DSS file includes time series data, such as hourly or daily flow, stages, precipitation,
elevation, and storage data; curve data, such as rating tables and frequency curves; gridded data,
such as NexRAD (Next-Generation Radar) data; and a variety of other data types.

Basic HEC-WAT Concepts

An HEC-WAT study is a combination of data, models, and events required to analyze a specific
geographic area. The first step is to define the schematic for the base alternative (Existing
Conditions), and then schematics for other defined alternatives. A schematic is a physical
representation of a watershed boundary, stream alignment, and the flood risk management and
environmental measures to be modeled. The schematic for the base alternative is built usually by
one individual from the study team. A schematic usually includes stream alignment, common
computation points, and existing flood control or restoration measures. The schematic aids in the
placement of element using background maps of gage locations, reservoir locations, sub-basin
delineations, study boundary areas, and other layers can be added to the study.

When an HEC-WAT study is created, the modeler will create a base alternative, which for the
FIRO study is Existing Conditions. The study team will also define other alternatives for the
study. An alternative in HEC-WAT is a way to group a schematic with different operations and
model parameters. Once an alternative is created, the next step is to create analysis periods.
Analysis periods define the time window (events) that could be associated with an alternative.
An event is an occurrence of precipitation/snowmelt that leads to some sort of hydrologic
response normally associated with a specific period (time). The event can be historical or
hypothetical or represent the upper limit of a precipitation runoff response

Now that analysis periods and alternatives have been defined, existing models that have been
identified for the study can be brought into the HEC-WAT study either by importing or by
copying the files into the HEC-WAT directory structure. If models are to be built, HEC-WAT
provides two options. One is the ability to create the models through HEC-WAT by accessing
29
Chapter 3 – HEC Software PR-100

the individual pieces of software directly from within HEC-WAT. The other option is to copy
the files created by HEC-WAT to a computer, and then return the files to the HEC-WAT study
once the models are calibrated and validated. For example, if the HEC-RAS model for a study
needs to be built, within the HEC-WAT directory structure, an HEC-RAS folder exists, which
contains the default HEC-RAS project files. The user can copy these files to another computer,
develop the HEC-RAS model, and then copy the files back to the HEC-WAT study.

From the individual models, HEC-WAT is interested in the individual model alternatives that
have been defined. Each model has their own nomenclature for "alternative":

● HEC-HMS – run, simulation

● HEC-RAS – plan
● HEC-ResSim – alternative
● HEC-FIA - alternative

Once all the model alternatives have been added to the HEC-WAT study, simulations can be
created. A simulation is a combination of an analysis period associated with an alternative. For
each of these combinations, model alternatives need to be selected, a simulation computed, and
results reviewed. The process is then repeated for the other identified alternatives, and after
several HEC-WAT simulations, alternative result comparison can be reviewed.

HEC-WAT Basic Approach

A PDT is conducting a water resources study and is using the following software - HEC-HMS,
HEC-ResSim, HEC-RAS, and HEC-FIA. An HEC-WAT study has been created with a
schematic built by the PDT that will be used by most of the models. The models are brought in to
the HEC-WAT study and all models files are imported or copied into the HEC-WAT data
management system that is based on the HEC-WAT study name.

Now that all the models have been imported, simulations are created. For each simulation, the
individual models are selected for a defined alternative. After the simulations are setup, "linking"
between the models will need to be defined. Linking is the process within the HEC-WAT
framework that allows the individual models to communicate with each other and how data is
handed-off from model to model. The majority of linking in HEC-WAT is accomplished through
DSS mapping and the use of simulation names. Additional linking capabilities include
inundation boundary maps, water depth maps, and other DSS files.

For example, the HEC-ResSim model requires information from an HEC-HMS model. From the
HEC-WAT Model Linking Editor (Figure 15), based on a particular simulation, the modeler will
link the models. From a table that lists the available locations that were defined by the HEC-
ResSim model, the modeler will select HEC-HMS as the model that has the information needed.
Then the modeler will select the appropriate location/results from the HEC-HMS simulation to
pass to the HEC-ResSim model. Once completed, the linkage will be saved for the HEC-ResSim
model. All of the software that was included within the HEC-WAT alternative simulation will
need to be linked. Once all of the linking has been completed, the simulation can be run.

30
PR-100 Chapter 3 – HEC Software

Figure 15 HEC-WAT - Model Linking Editor

The first model is an HEC-HMS model. The model is run and unregulated flows are saved to a
DSS file (simulation.dss). The unregulated flow from HEC-HMS now becomes inflow to the
HEC-ResSim model. Because of the linking that was defined earlier in the process, HEC-HMS
information stored in simulation.dss is used by the HEC-ResSim model during the compute. At
the same time, the HEC-ResSim model is saving peak-regulated flows/stages, flow hydrographs
or stage hydrographs to the simulation.dss file. Once the HEC-ResSim model compute is
complete, the HEC-RAS model begins running using HEC-HMS and HEC-ResSim information
from the simulation.dss file. HEC-RAS is saving river stages, water surface profiles,
hydrographs (unsteady analysis), and levee breach information to the simulation.dss file. HEC-
RAS (through the RAS Mapper tool) is also creating inundation boundary and water depth maps,
which are saved in the HEC-WAT study data management structure.

After the HEC-RAS model compute is complete, the HEC-FIA model compute begins. Based on
the linking, HEC-FIA will retrieve HEC-ResSim and/or HEC-RAS results from the
simulation.dss file and HEC-FIA will also use the inundation boundary maps generated by HEC-
RAS. Once the HEC-FIA model compute is complete, the HEC-WAT simulation is complete
and results can be reviewed from within the HEC-WAT framework.

With the HEC-WAT approach there is flexibility in the performance of water resources studies,
the PDT is able to perform the study in a coordinated fashion, the data management of study files
is centrally located within the HEC-WAT directory structure, alternative analyses can be
performed easily and review of analyses results is from a central location.

31
Chapter 3 – HEC Software PR-100

Flood Risk Management

The HEC-WAT framework also includes a flood risk analysis (FRA) compute option that
provides tools that will help a planner address the issues faced during a USACE planning study,
and specifically the requirements for systems, watershed, and uncertainty analyses. The FRA
compute builds on the HEC-WAT framework to provide a convenient method of implementing a
Monte Carlo analysis across multiple models for flood risk management. With the initial HEC-
WAT framework providing all the initial software that would meet the needs of a performing a
system-wide analysis, the FRA compute option allows a user to perform plan formulation or
system performance analyses while incorporating risk and uncertainty capabilities. Chapter 10
provides further details on the flood risk analysis compute option.

The Monte Carlo approach provides a method for incorporating a probabilistic description of
inputs (i.e., boundary conditions, initial conditions, model parameters) into an analysis while
making use of detailed deterministic models. During a Monte Carlo simulation, the single-
valued inputs (deterministic models) are replaced with randomly sampled values as defined by
probability distributions. In turn, the sampled model information is repeated several iterations to
span the possible range of model inputs to quantify uncertainty in model results. During an FRA
simulation input values are sampled, algorithmic decisions are performed, results are
summarized, and convergence is reached (for Version 1.0 of HEC-WAT automatic convergence
is not a feature, the user must define the total number of events).

Hydrologic Sampling
When an FRA compute is being performed, hydrologic sampling will generate system-wide
flood event samples for a Monte Carlo simulation. The data produced by the plug-in can take the
form of hydrographs or hyetographs spanning the life cycle of the system, usually fifty years (ER
1105-2-100 (USACE, 2006)).

Performance and Economic Metrics

Risk for USACE is the measure of the probability and severity of undesirable consequences (ER
1105-2-101 (USACE, 2017)). From these analyses, the performance of a system's reaction to an
event (flood) must be reported. The performance and economic metrics that can be computed in
HEC-WAT meet these USACE requirements, and are as follows: Assurance describes the
probability that a system will contain a specific annual chance exceedance flood event, at or
below a specific target (e.g., the top of a levee). Annual Exceedance Probability (AEP) and
Long-Term Exceedance Probability (LTEP) are calculated for the full range of events and,
Expected Annual Damage (EAD) is the average or mean of all annual damage. These metrics
assist in communicating risk and the Performance Metrics plug-in generates these metrics. Not
all FRA studies require performance metrics (USACE, 2006).

HEC-WAT Interface
The HEC-WAT main window (Figure 16) displays the framework for HEC-WAT that allows
users to enter data, review data, create alternatives, run simulations, and view results. The Title
32
PR-100 Chapter 3 – HEC Software

Figure 16 HEC-WAT Main Window

Bar (Figure 16) displays the HEC-WAT study title. After a study is opened, the name of the
study will appear in the Title Bar.

Panes
The HEC-WAT main window (Figure 16) is laid out in panes that allow the user to view HEC-
WAT study components at the same time or to display different, yet simultaneous, views of
study components. The HEC-WAT main window has four main panes. The panes are:

Study Pane The Study Pane provides an overview of items that have been defined for an
HEC-WAT study. Based on the view (HEC-WAT User's Manual, Chapter 3,
Section 3.2), the HEC-WAT Study Pane displays information about
alternatives; simulations; analysis periods; models that have been defined for
the study; what map windows are actively open; the available schematics;
stream alignments; measures; common computation points; and, impact areas.

Content Pane This pane (Figure 16) provides details on selected HEC-WAT components in
the Study Pane (Figure 16). For example, in Figure 16, for a study view
(HEC-WAT User's Manual, Chapter 3, Section 3.2), an HEC-WAT
simulation has been selected – Base_FRA. The Content Pane displays the
elements of that simulation. Information includes the alternative; the
analysis period; the program order; the model alternatives that were run; the
time window; the directory where the results from the simulation were
written; and, the status of the simulation.

33
Chapter 3 – HEC Software PR-100

Desktop Pane The Desktop Pane (Figure 16) is where the different map windows that are
available from HEC-WAT will display. For example, in Figure 16, the
Desktop Pane contains a map window of a selected simulation and a map
window of a selected schematic. These two map windows provide the user
with a graphical representation of a simulation and a schematic.

Message Pane The Message Pane (Figure 16) contains messages that the HEC-WAT
software is providing for informational purposes as a simulation progresses.
The information in the Message Pane is a record of HEC-WAT software
activities, records the opening of a study, opening map windows, accessing
DSS, and many other HEC-WAT activities.

Tabs
HEC-WAT tabs (Figure 16) provide different views of the available components of a study and
users can perform certain operations for a study from the tabs. The tabs can change the layout of
the menu bar and available information in the Study and the Content panes can be changed.
There are three tabs - Study, Maps, and Schematic - and all three are described in the HEC-
WAT User's Manual (Chapter 3, Section 3.2).

Menu Bar
The menu bar (Figure 16) of HEC-WAT provides the user with many commands to perform
various activities/functions. Descriptions of the available menu items are described in the
following, and further details are available from the HEC-WAT User's Manual (Chapter 3,
Section 3.3).

The File menu (Figure 16) allows the user to perform study management functions such as
creating, opening, closing, and saving an HEC-WAT study. The File menu also provides
commands that will allow users to provide information about the study team; select the software
applications or tools (plug-ins) that are needed for the study; import a CWMS watershed; import
other HEC software application's model alternatives (HEC-HMS, HEC-ResSim, HEC-RAS,
HEC-FIA); open recently viewed HEC-WAT studies; and, exit the HEC-WAT software. The
File menu is available in all HEC-WAT views.

From the Edit menu (Figure 16) the HEC-WAT user, can create/modify/edit items that are the
building blocks of an HEC-WAT study. A user can manipulate alternatives and simulations;
establish the program order for a study; link the individual models of a study; modify the look
and feel of the common computation point layers (CCPs); provide aliases for the defined stream
segments in the HEC-WAT study; available FRA editors; and, access editors from application
software that are part of a study. The FRA specific editors allow the user to skip models during
an FRA compute, and select variables to save to the simulation DSS file. The above description
is available when in the Study View; the Edit menu in the other views is limited.

The View menu (Figure 16) is where an HEC-WAT user can setup the HEC-WAT main
window. The user can select which toolbars and panes to display and not display; set display
units for plotting and tabulating (English, SI); create and use layouts of the main window;
34
PR-100 Chapter 3 – HEC Software

manage layouts; which common computation point layer to display (Default, Fragility Curve);
and, a report of the common computations point available for the study. The View menu is
available in all HEC-WAT views.

The Maps menu (Figure 16) can be used by an HEC-WAT user to setup map layers and map
windows for an HEC-WAT study. The user can set spatial properties; zoom the active map
window out fully (extents); add and remove map layers; open a new map window; adjust the
active map window's properties; save the contents of the active map window as a graphics file;
and, synchronize the current open map windows. The user can also add an Internet map to the
active map window, and view the map windows in either the watershed coordinate system or the
coordinate system of the Internet map. Other available commands include the ability to print a
copy of the active map window in the Desktop Pane (Figure 16); import various schematic
elements (stream alignment, impact areas, reservoirs, levees, computation points) from a
shapefile; and, export various schematic elements (stream alignment, impacts areas, schematic
elements) to a shapefile. The Maps menu is available in all HEC-WAT views.

The Compute (Figure 16) menu is used to compute individual HEC-WAT simulations; compute
multiple simulations; view a list of simulations for the watershed that have been computed;
compute the active simulation or compute the individual models of the active simulation; and,
view the compute log for the active simulation. For an HEC-WAT, study with multiple FRA
simulations there is an item for simulation failures. The Compute menu is only available in the
Study View.

An HEC-WAT simulation must be selected for the Results menu (Figure 16) to be active. The
Results menu can be different based on the simulation type – deterministic or an FRA
simulation. For a deterministic simulation, the user can view the simulation.dss file and view
results available from the individual models that were part of the simulation. The HEC-WAT
User's Manual (Chapter 3, Section 3.3) provides further details.

From the Tools menu (Figure 16) the HEC-WAT user can access DSS data (HEC-DSSVue);
access the Statistical Software Package (HEC-SSP); provides options for HEC-WAT (HEC-
WAT User's Manual, Appendix F); view the log for HEC-WAT, view logs for DSS files and the
individual models that are part of a simulation; monitor memory usage; what graphic elements
HEC-WAT knows about from the other applications; information on how the different model
alternatives are used for the study; setup for distributed computing; and, if a simulation is
running slowly, setup compute timings to locate where a slowdown might be occurring. The
Tools menu is available in all HEC-WAT views.

The Window menu (Figure 16) is where the HEC-WAT user can control the appearance of the
map windows in the Desktop Pane (Figure 16) of the HEC-WAT main window. The Window
menu is available in all HEC-WAT views.

The Help menu (Figure 16) allows the user access to HEC-WAT documentation; install the
example study; connect to the HEC-WAT web page; connect to the HEC website; and, display
current version information about HEC-WAT.

35
Chapter 3 – HEC Software PR-100

Toolbar
The toolbar area of the HEC-WAT main window (Figure 16) provides quick access to the most
frequently used options from the HEC-WAT menu bar. These toolbars can been toggled on and
off, from the View menu (Figure 16), point to Toolbars, and select the toolbar the user wishes to
toggle on and off. There are six available toolbars from the main toolbar:

Standard There are three tools available from the Standard toolbar (Figure 16). The
tools allow the HEC-WAT user to open a study, close a study, and save a study.
These tools have the same function as items under the File menu (Section
3.7.3).

WAT Tools There are several tools available from the WAT Tools toolbar (Figure 16). The
tools allow the user to set the program order for HEC-WAT; to manipulate
alternatives, analysis periods, and simulations; link the individual model
alternatives; set up rules to skip models during an HEC-WAT compute; and,
select results from the individual models (output variables) that will be saved.

Tools The tools available from the Tools toolbar (Figure 16) are variable depending
on what plug-ins are loaded. Two tools that are always available, is software to
access DSS data (HEC-DSSVue) and HEC-SSP.

Maps There are four tools available from the Maps toolbar (Figure 16). The tools
allow the HEC-WAT user to control the appearance of map windows in the
Desktop Pane (Figure 16) - create a new map window; synchronize active map
windows; tile; and, cascade. These tools have the same function as items under
the Window menu (HEC-WAT User's Manual, Chapter 3, Section 3.7.3).

Programs There are usually four tools available from the Programs toolbar (Figure 16)
that represent the individual software applications available from HEC-WAT:
HEC-HMS, HEC-ResSim, HEC-RAS, and HEC-FIA. These tools allow the
user to open the selected software application's user interface from within HEC-
WAT. The tools available from the Programs toolbar are dependent on which
plug-ins are loaded.

Map Windows
Map windows (Figure 17) are used for graphically displaying stream alignments, schematics,
alternatives, and simulations. From an alternative map window, the user can adjust the stream
alignment. A schematic map window allows the user to edit, modify, delete measures, CCPs,
impact areas, and adjust the stream alignment. The simulation map window (Figure 17) is where
the HEC-WAT user can edit individual model parameters and view results. The HEC-Wat
User's Manual provides further details in Chapter 3 (Section 3.5).

36
PR-100 Chapter 3 – HEC Software

Overview of a Simulation Map Window

For example, in Figure 17, in the Desktop Pane, a Map Window is displaying the selected
HEC-WAT simulation – Base_FRA. From the schematic (Figure 17), the various model
elements involved in the active simulation are displayed. The Stream Alignment (orange
stream centerline, Figure 17) is common to most of the models that are part of the active
simulation and provides a reference point for the HEC-WAT study. The magenta lines and
raindrop symbols visible in Figure 17 represent subbasin elements from the HEC-HMS model
alternative. Part of the cyan polygon (Figure 17) representing the Lake Mendocino reservoir in
the HEC-ResSim model alternative is visible at the top of the schematic. The green lines
displayed in the schematic (Figure 17) correspond to HEC-RAS cross-sections. The HEC-FIA
model alternative used a detailed parcel-based structure inventory and stage-damage information
to simulate flood damage. The blue squares in the schematic (Figure 17) represent the structures
from the parcel-based structure inventory. Simulation results are available by right clicking on
schematic elements, and from the available shortcut menus, the modeler can select results from
the model alternatives that are part of the selected simulation. For example, in Figure 17, the
modeler has selected from a shortcut menu, to view an HEC-RAS cross-section, which is
viewable in an HEC-RAS Cross Section dialog box in the lower left of Figure 17. In the
example (Figure 17), the selected HEC-WAT simulation is an FRA simulation, from the Results
menu, the modeler has selected the Histogram Viewer (Figure 17, visible on the right side), an
HEC-WAT FRA result about Lake Mendocino. Further details from the schematic can be
viewed by zooming into areas on the schematic and revealing further details about the selected
simulation.

Figure 17 HEC-WAT Main Window – Simulation Map Window

37
Chapter 3 – HEC Software PR-100

38
PR-100 Chapter 4 – Overview of the FIRO Analysis

Overview of the FIRO Analysis

Study Objectives
This analysis was designed as a "bookend" analysis in support of the PVA, which estimates the
most conservation storage possible at Lake Mendocino using FIRO without increasing flood risk
on the Russian River. The goal of this analysis was not to design practical FIRO alternatives
suitable for actual implementation by USACE.

The question of whether FRIO could be viable appears straightforward, and the answer seems
obvious to many people with substantial experience in forecasting. USACE guidance for flood
risk management requires a reliable engineering analysis that defines the details of specific
reservoir operations, and evaluates them using a systems perspective that considers interrelated
effects throughout a watershed. USACE policy further requires flood risk studies to consider the
uncertainty of variables that have a significant impact on study conclusions, such as the
probability of different potential events.

Consequently, this analysis was designed to accomplish additional objectives in the course of
determining how much conservation storage could be made available without increasing flood
risk:

1. Demonstrate methodology - This analysis was intended to illustrate a detailed and

comprehensive approach consistent with USACE policy for analyzing the flood risk
associated with FIRO at Lake Mendocino. The analysis should provide an example of
technical assessments for potential proposed water control manual updates, major
deviation analyses, or other official studies likely to be required for implementation of
FIRO at Coyote Valley Dam.

2. Develop tools – During this analysis, the data and models developed should be useful to
other efforts within USACE or FIRO partners. For example, a subsequent FIRO activity
may be needed to perform the deterministic analysis using CNRFC hindcast inflows to
Lake Mendocino. The modeling tools employed must be familiar within USACE and
approved by USACE's HH&C CoP for a variety of applications. Example uses could
include modeling as part of a water control manual update study, conversion to real-time
operations as part of a CWMS analysis, or updating the procedures used to simulate
project benefits regarding flood damage avoided. This report also identifies limitations
of the specific model implementations used in this analysis, and recommends
improvements for future studies.

3. Inform research - The analysis provided practical examples of how reservoir operations
might incorporate forecast information, and help researchers present forecast information
in metrics most usable to reservoir regulation alternatives and models. For example,

39
Chapter 4 – Overview of the FIRO Analysis PR-100

FIRO partners at CW3E may study the ability to characterize the skill of inflow forecasts
as a "Go/No-Go" flag for advance release rules at the Coyote Valley Dam.

4. Understand flood risk drivers related to Lake Mendocino - The analysis of flood risk
illustrated how reservoir operations can affect flood damage along the Russian River. A
better understanding of flood risk drivers related to reservoir operations can help focus
the questions that must be answered by subsequent studies supporting implementation of
FIRO at Lake Mendocino.

Analysis Design
The FIRO analysis compared results from simulations under the Existing Conditions alternative
against results from alternate reservoir operations based on forecast information. The relative
comparison held everything else equal except the reservoir operations at Coyote Valley Dam.
This report refers to the model simulations and results that reflect the current regulation of Lake
Mendocino as the Existing Conditions alternative.

Each FIRO alternative was compared to the Existing Conditions alternative using metrics that
characterize flood risk and water supply benefits. The metrics were defined in coordination with
FIRO study partners, and used in common with a parallel analysis by SCWA and CNRFC. Each
alternative was simulated using HEC-WAT with both a sixty-year deterministic continuous
simulation, and a probabilistic Monte Carlo approach (FRA simulation).

The analysis relied on repeated simulation rather than optimization in order to estimate the
storage that FIRO could make available for water supply. Although an optimization approach
might have provided a more direct estimate of the maximum storage, the simulation approach is
key to other study objectives. The simulation approach provided the best representation of
results from a specific regulation alternative, which will likely be required to evaluate any FIRO
alternatives proposed for actual implementation. The simulation approach also allowed the
models to employ release decision logic that used forecast information available at each time
step. The formulation of the FIRO regulation alternatives involved substantial experimentation
by engineers familiar with the Russian River watershed and experienced reservoir modelers. The
resulting alternatives might not establish the absolute maximum possible gain in reservoir
storage available to water supply without increasing flood risk, but should be close.

Deterministic Simulation
The deterministic sixty-year continuous analysis for the Russian River watershed used historical
data. This provided a familiar context for understanding results, for example, how the 2006
flood event might have played out under a FIRO alternative. Historic precipitation and
streamflow data from 1950 through 2010 was used for the deterministic simulation and
validation. This period was selected as offering the longest period of available precipitation and
streamflow data. This period also includes the flood of record (December 1964) and drought of
record (1977 through 1978).

Antecedent conditions for individual flood events, such as soil moisture or reservoir level at the
start of an event, were created as part of the continuous sixty-year simulations. The continuous
40
PR-100 Chapter 4 – Overview of the FIRO Analysis

simulation using soil moisture methods provided dry season streamflows that allowed modeling
of reservoir operations for minimum flows. The resulting reservoir depletions set the antecedent
pool levels for the eventual winter storms. The process of an HEC-WAT deterministic
simulation is illustrated in Figure 18.

Figure 18 HEC-WAT Deterministic Simulation Process

Three deterministic HEC-WAT alternative are displayed in Figure 18 with the top row
representing the Existing Conditions alternative. The Russian River study program order for the
deterministic simulations included HEC-HMS, HEC-ResSim, HEC-RAS, and HEC-FIA models.
Precipitation data was the primary input to the HEC-HMS model, which simulates the resulting
sixty-year streamflow. An HEC-ResSim model simulates existing operations at Coyote Valley
Dam and Warm Springs Dam. An unsteady flow HEC-RAS model was used to route the
reservoir releases and tributary flows through the Russian River to the ocean, determine water
surface elevations (peak) and inundated extents. An HEC-FIA model used a parcel-based
structure inventory and stage-damage information to simulate flood damage. The HEC-WAT
simulation managed boundary conditions for the models and linked model results to subsequent
models in the program sequence. Over the specified analysis period, 1950 through 2010, the
simulations yielded a time-series of results across all model results.

HEC-WAT alternatives were created by duplicating the Existing Conditions alternative, and then
for each new HEC-WAT alternative, a new HEC-ResSim model alternative that modeled
forecast-informed operations was added. The second and third rows of Figure 18, illustrate the
construction of HEC-WAT alternatives comparing different water management operations with
all else remaining unchanged (e.g., E (Encroach), C (Combined)). Advantages or disadvantages
of the reservoir operation alternatives are defined in terms of differences in simulation results,
such as the storage in Lake Mendocino on 30 September or the peak annual stage at Guerneville.

Flood Risk Analysis Simulation

The FRA simulation for the Russian River system was configured to simulate 5,000 possible
flood events using a Monte Carlo approach. The FRA simulation provides a more robust
consideration of flood risk and uncertainty by including plausible potential precipitation events
that did not occur in the observed record, including very large storms while maintaining their
estimated frequency. The choice of 5,000 events was thought to be adequate for defining the
1% Annual Chance Exceedance (ACE) flow/stage estimates and for computing EAD.
41
Chapter 4 – Overview of the FIRO Analysis PR-100

The FRA simulations for the Russian River system focus solely on short-term flood events to
reduce simulation time. For the FRA simulations, HEC started with the deterministic model
alternatives, and then modified the model alternatives for the FRA simulations. The streamlined
FRA model alternatives avoided unnecessary computations for a flood risk analysis such as
evapotranspiration from HEC-HMS and minimum instream flow rules and diversions from
HEC-ResSim. The event-based FRA simulations also required initial soil moisture and initial
reservoir elevations for each storm. These antecedent conditions were sampled from parameter
distributions develop from the deterministic results. Figure 19 illustrates the process of an HEC-
WAT FRA simulation.

Figure 19 HEC-WAT Flood Risk Analysis Simulation Process

Hydrologic sampling (HEC-WAT Hydrologic Sampler) is the first step in an FRA simulation
and generated a dataset of 5,000 annual maximum precipitation events, based on analyses of
historical precipitation for 1951 through 2010. A critical duration analysis of the ten largest
historical flood events resulted in a selection of eight-days as the duration of the synthetic
storms. A GEV (Generalized Extreme Value) distribution fitted to the annual maximum series of
eight-day basin precipitation totals provided the relationship between probability and size of
event. Flood season analysis is based on 32 of the annual historical storms selected to define the
frequency of events in each month. For each of the 5,000 events, the hydrologic sampling
generated a basin-wide storm total precipitation from the GEV distribution, shaped it according
to one of ten hyetograph patterns (representing the ten largest events between 1951 and 2010),
and then placed it on a sampled calendar date.

Results from the deterministic HEC-HMS model were used to define sampling distributions for
soil moisture for different months of the season. HEC-HMS simulated the resulting streamflows
and provided these as input to the HEC-ResSim model. Results from the deterministic HEC-
ResSim model for each alternative were used to define sampling distributions for the initial
reservoir elevations for Lake Mendocino for various seasons. The HEC-ResSim model
generated reservoir outflows that were input to the HEC-RAS and HEC-FIA models in order to

42
PR-100 Chapter 4 – Overview of the FIRO Analysis

simulate downstream flows, stages, and damage. No parameters were sampled within the
Russian River HEC-RAS and HEC-FIA models.

To allow reproducible results that would be used for comparison of alternatives, the hydrologic
sampling available in HEC-WAT reproduced the same series of randomly generated storm
events in each FRA simulation, and the HEC-HMS model sampled the same starting soil
moisture storage for each event. This modeling option is accomplished through the FRA option,
which allows modelers to maintain the same pseudo-random seeds for each FRA simulation.
This capability also allows the HEC-ResSim model to sample the starting reservoir levels at the
same probability for each FRA event, although the distributions differed according to the
reservoir storage results for each water management plan.

As in the deterministic setup, the HEC-WAT alternatives were created by duplicating the
Existing Conditions alternative, and then for each new HEC-WAT alternative, a new HEC-
ResSim model alternative that modeled forecast-informed operations was added. Advantages or
disadvantages of the reservoir operation alternatives are quantified as differences with respect to
EAD or 1% ACE flow at Guerneville.

Comparison Metrics
Given the need for a consistent set of key statistics for comparing the performance of each
alternative, it was determined that the amount of additional water available for conservation
purposes and the degree of flood risk at selected locations along the Russian River be
characterized by the following statistical metrics.

Storage Available for Conservation Purposes

Several candidate metrics were proposed to account for success at meeting downstream
operational constraints while satisfying instream environmental flow requirements. HEC and
SCWA discussed different minimum flow targets at several river locations at different times
during the summer and autumn that represented municipal water supply needs, or regulatory
requirements for endangered Coho salmon, and threatened steelhead trout and Chinook salmon.
The calculations took the general form of duration analyses, but became complex when trying to
account for storage reserved for later use versus flow targets achieved as the dry season
progressed. While working with an HEC-WAT provisional simulation (Available Water),
SCWA and HEC concluded that calculations reflecting the priorities of operations within
conservation pool provided no useful information. Performance of reservoir operation
alternatives for all of the downstream flow targets was driven simply by the amount of water
held in Lake Mendocino at the beginning of the dry season.

HEC and SCWA agreed that the success of reservoir regulation alternatives in meeting the
various water supply and instream targets was best measured by how well the alternatives
managed the spring refill. Defining even this simpler representative storage metric also required
consideration, since the dry season begins unpredictably sometime between February and May.
Ranking alternatives based on annual maximum pool levels during the spring months proved to
be a flawed metric, since subsequent flood operations after the peak pool level sometimes
determine the storage actually carried into the dry season.
43
Chapter 4 – Overview of the FIRO Analysis PR-100

After these considerations, HEC choose a very simple metric to quantify refill success, and
thereby compare the ability of each reservoir regulation alternative to meet the many water
supply needs in the Russian River watershed. Since the existing guide curve tries to attain the
summer pool level by 10 May each year, the water supply performance of each alternative was
compared using the Lake Mendocino storage that was determined for 10 May. The analysis also
reported the amount of storage in Lake Mendocino on 30 September, which reflects the amount
of water to support autumn flows for fishery management, and aligns with long-standing
reporting practices of SCWA.

Only the deterministic simulations generated these metrics for describing the additional storage
available for conservation purposes. The FRA simulations considered only short duration flood
events, and did not model the entire sequence of operations through 10 May for each event.

Flood Risk
The USACE analysis evaluated flood risk throughout the Russian River and Dry Creek
watersheds. Flood damage was modeled as part of a chain of hydrologic interactions and human
decisions. The decisions for this analysis were represented by reservoir regulation alternatives,
in other words, pre-determined operational guidelines and responses to information about current
and anticipated hydrologic conditions.

The most significant flood risk metric was the difference in EAD between a reservoir regulation
alternative and the Existing Conditions alternative. EAD provides a basis for comparisons
because EAD considers the full range of potential events while incorporating both the economic
consequences and probabilities. EAD was determined using the FRA simulation, while
deterministic results are reported as Annual Average Damage (AAD).

Additional results are presented that provide a better understanding about the flood risk at
various locations in the watershed. These results take the form of flow and stage frequency
curves at key locations in the watershed (e.g., Lake Mendocino, Hopland, Healdsburg, and
Guerneville). Results reported the 1% ACE (i.e., 100-year recurrence interval) water surface
elevation at each location for each alternative.

Other Considerations
Reservoir operations and the ability to satisfy requirements of the water control alternative (plan)
were analyzed. For instance, current operations specify that Coyote Valley Dam operations
should avoid discharges when the streamflow exceeds 8,000 cfs downstream at the Hopland
gage. However, sometimes such releases can occur due to imperfect information available to
operators or the requirements of other higher-priority rules.

HEC simulated a metric for violations of the 8,000 cfs rule for use while formulating FIRO
alternatives. Hopland flows exceeding 8,000 cfs from deterministic reservoir simulations of the
Existing Conditions alternative were compared to results of the FIRO alternative being designed.

44
PR-100 Chapter 4 – Overview of the FIRO Analysis

Differences in the timing of operations resulted in situations where the FIRO alternative avoided
or caused violations of the 8,000 cfs rule, or resulted in a different amount of violation. The
Hopland flow evaluations were not generated for use as a metric in comparing alternatives. The
information was used during the alternative formulation process for the Encroach and Combined
alternatives, where alternative parameters were adjusted until an appropriate balance among
objectives was achieved.

Development of the HEC-WAT Existing

Conditions Alternative
In order to quantify the potential gains in conservation storage available under FIRO, and
determine that no additional flood risk was caused, the analysis had to establish a comparison
dataset. Normally the Existing Conditions alternative reflects current or anticipated conditions in
a watershed, and provides a familiar context for evaluating changes. The simulation results from
modeling Lake Mendocino cannot simply be compared to the observed records of levels and
flows in the Russian River watershed, since the historical water management practices, water
demands, and inter-basin diversions varied substantially over time.

Consequently, the Existing Conditions alternative used in this analysis was developed from
model simulations that represented current conditions in the watershed. The models used for the
Existing Conditions alternative required various data for determining model parameters,
boundary conditions, initialization states, as well as inputs for each simulation interval. The
inputs used for the Existing Conditions alternative were also used in modeling the other FIRO
water management alternatives.

The HEC-WAT study took the approach that performance of FIRO alternatives regarding flood
risk and other metrics can be judged using a relative comparison against existing condition
model results. This assumption was justified by calibrating the Existing Conditions alternative
results against the observed record, or validating that modeled results reflected the expected
outcomes in situations where the observed record did not apply.

Relative comparison can reduce the impact of inaccuracies in data or models if those
shortcomings apply equally to all sets of results, and the results are not especially sensitive to
those inaccuracies. For example, inaccuracies in the total structure damage computations should
be acceptable for relative comparisons, if the damage is distributed across many structures in
different locations for a range of depths. If specific structures, reaches, or depths heavily
influence the total structure damage, then the accuracy of the flood damage computations for
those structures becomes more important.

Models from prior studies of the Russian River watershed served as a starting point for the
models used in the FIRO analysis. However, the prior models were developed for other
purposes, and more than likely will require re-development for use in the FIRO study.
Developing the models and input data for the Existing Conditions alternative constituted the first
major milestone of the FIRO study.

45
Chapter 4 – Overview of the FIRO Analysis PR-100

Hydrologic Modeling
The origin of the HEC-HMS model used for the FIRO study came from an HEC analysis that is
documented in PR-72, "Hydrologic and Hydraulic Modeling of Coyote Valley Dam for With and
Without Failure of the Dam" (HEC, 2009). The main function of the HEC-HMS model
remained the same, to provide streamflows across the watershed, which served as input to the
HEC-RAS and HEC-ResSim models. The requirements of this study led to several major
revisions to the HEC-HMS model.

The previous dam-break analysis required HEC-HMS to produce detailed hydrographs for
critical periods for a limited number of storm events. This study involves repeated continous
sixty-year simulations for deterministic simulations, and many individual storm events under the
FRA simulations. Consequently, the HEC-HMS time step was revised from 15-minute to one
hour. The delineations used for PR-72 (HEC, 2009) were retained because the delineations
represented an appropriate level of detail for this analysis.

The watershed was delineated into eighteen subbasins (Figure 20), for purposes of computational
efficiency. The delineations represented the largest aggregations of area possible, with divisions
according to dissimilar hydrologic characteristics, calibration gages, locations relevant to water
management rules, and requirements of the hydraulic model.

In developing basin model parameters, GIS data such as USGS ten-meter DEM data was used to
provide the basis for catchment delineations, distances, and slopes. USGS data layers also
provided the basis for land use classifications and designation of impervious areas. Soil
properties were estimated using data from the Natural Resources Conservation Service (NRCS)
Soil Survey Geographic (SSURGO) database. Channel routing parameters were initially
estimated using DEM data and calibrated to observed gage flows, then replaced later with
parameters from the HEC-RAS model.

Historic precipitation and streamflow data from published gages were used to drive the
precipitation-runoff response model and its calibration, as described in detail in Chapter 11
(Section 11.1). The availability of this data was the limiting factor that determined the analysis
period for the deterministic simulations as 1950 through 2010. The model was calibrated using
observed flow hydrographs and annual flow-frequency relationships at the key gages on the
Russian River. The model was validated against the historic events listed in Chapter 2 (Section
2.2.5).

In order to analyze dry season conservation operations, and the resulting reservoir depletions that
influence the starting storage for storm events, the HEC-HMS model was modified to perform
continuous simulation across the entire analysis period. The Russian River HEC-HMS model
incorporated the Penman-Monteith evapotranspiration method with the Soil Moisture
Accounting (SMA) method to simulate soil moisture continuously. Penman-Monteith
evapotranspiration inputs for the analysis period were developed from hydrometeorologic
observations at the Ukiah Municipal Airport and Sonoma County Airport.

The HEC-HMS model was calibrated across the sixty-year analysis period according to the
observed runoff hydrograph and seasonal volumes at the key gages on the Russian River. The

46
PR-100 Chapter 4 – Overview of the FIRO Analysis

Figure 20 Subbasin Delineations for the Russian River HEC-HMS Model

modeling approach neglected subsurface flow, which also limited the ability to calibrate against
varying dry season conditions observed over the analysis period. Considerable effort was
required to calibrate a single set of model parameters across such a wide range of hydrologic

47
Chapter 4 – Overview of the FIRO Analysis PR-100

conditions and flood events. The resulting HEC-HMS model provides the best estimate of flows
throughout the watershed resulting from precipitation for any given year.

The calibrated standalone HEC-HMS model was modified slightly during integration into the
deterministic HEC-WAT simulations, as explained in Chapter 11 (Section 11.2). The FRA
simulations employed a third version of the HEC-HMS model, with the evapotranspiration
components removed for faster computation of flood events in the Monte Carlo simulations.
Results of soil moisture storage from the sixty-year deterministic simulations were analyzed to
provide seasonal distributions for sampling initial soil moisture storage for each event of the
FRA simulations.

The hydrologic sampling methodology allowed for modeling a much wider range of potential
events, by introducing sampled flood events in the historic records that were larger than observed
events. The hydrologic sampling generated synthetic precipitation events for the FRA
simulation based on a frequency analysis of historical precipitation records. An analysis of the
largest 32 observed events determined the relative occurrence of events in each month, and
enabled sampling a start date for each event. Analysis of the ten largest historical storms led to
designating eight days as the "critical duration", in other words, the event duration that produces
the greatest flood impact in the basin. The observed annual maximum series of eight-day storm
total precipitation (averaged across the watershed) from 1951 through 2010 was used to estimate
a GEV distribution for precipitation-frequency using the method of linear moments. The
magnitude of each synthetic event was sampled from that relationship between total precipitation
and frequency. Finally, time-series of hourly incremental precipitation throughout the watershed
resulted from scaling the hyetograph patterns from one of the ten largest observed events to the
sampled event magnitude. The HEC-HMS model used the initial soil moisture storage and
synthetic hyetograph to generate streamflows throughout the basin, which were used as data for
the HEC-ResSim and HEC-RAS models.

Reservoir System Modeling

The HEC-ResSim model used in this study originated from a model developed for an HEC
analysis that is documented in PR-85, "Determination of a Hydrologic Index for the Russian
River Watershed using HEC-ResSim" (HEC, 2012). The HEC-ResSim model served as the
vehicle for characterizing and comparing FIRO, and was the centerpiece of the HEC-WAT
simulations. The previous HEC-ResSim model was updated in coordination with CESPK and
SCWA personnel to simulate on a six-hour time-step, include better physical data, more detailed
diversion information and conservation operations, and recent water management changes, as
explained in Chapter 11 (Section 11.3).

The HEC-ResSim model generally maintained the same routing reaches as the HEC-HMS
model, except that the HEC-ResSim model treated Dry Creek as two reaches rather than one, for
modeling a maximum flow rule at Warm Springs Dam. The stand-alone HEC-ResSim model
was validated against the historic events listed in Chapter 2 (Section 2.2.5), and integrated into
the HEC-WAT study. No modifications were necessary for the deterministic simulations. The
conservation operations were removed from the HEC-ResSim model, which reduced the
generation of DSS data, which allowed the FRA simulations to execute faster.

48
PR-100 Chapter 4 – Overview of the FIRO Analysis

During the simulations, the HEC-ResSim model relied on results from the HEC-HMS model for
flows throughout the Russian River watershed. The model applied the same pattern of monthly
evaporation rates for each year. SCWA provided data for consumptive diversions at locations
along the Russian River, diversion amounts from the Eel River through the Potter Valley Project,
and historical values for the existing definition of the hydrologic index.

Hydraulic Modeling
The HEC-RAS model used in this study came from combining HEC-RAS models that were
developed for an HEC analysis that is documented in PR-72 (HEC, 2009) and a river hydraulics
model covering Dry Creek below Warm Springs Dam developed for a CWMS application. The
HEC-RAS model used normal depth at the ocean for a downstream boundary condition. A
provisional simulation of the entire sixty-year period took almost three days to finish, so detailed
hydraulic modeling was performed only for the annual maximum event within each water year
(typically about a two-week duration).

The hydraulic model (HEC-RAS) simulated river stages and flows along the Russian River for
each water management alternative. The RAS Mapper Tool in HEC-RAS determined inundation
extents and water surface depth grids for use in consequence modeling. The combined
HEC-RAS model was re-generated from the same channel geometry and bridge data as the
earlier models. The combined HEC-RAS model was re-calibrated to the 1986 (March), 1995,
1997, and 2006 events, as described in Chapter 11 (Section 11.4).

The HEC-RAS model derived cross-sections and storage areas from ten-meter DEM data, and
assumed trapezoidal channel geometry in the absence of bathymetry data. Although the model
performed well for large events, the model may have been less accurate at lower flows. This
study assumed that the relative comparison of hydraulic model results across a wide range of
flood events provided a valid indication of the difference in flood risk among reservoir operation
alternatives.

The stand-alone HEC-RAS model used for calibration was imported into the HEC-WAT study,
without modification for the deterministic simulations. The HEC-RAS model that was used for
the FRA simulations differed, in that the generation of DSS data was reduced.

Consequence Modeling
An HEC-FIA model for the Russian River watershed had been previously created for software
testing purposes at HEC, and served as a placeholder during the early phases of the FIRO
analysis. The HEC-FIA model was rebuilt using cross-section and storage area geometry from
the provisional HEC-RAS model, in addition to a parcel-based structure inventory data provided
by CESPK. All HEC-FIA results in this analysis were generated using the Grids and Cross-
Sections HEC-FIA simulation option.

The same HEC-FIA model was used to simulate structure damage for each event, using the
gridded maximum depths that were produced by the HEC-RAS model. HEC-FIA determined
the depth at each structure in the structure inventory within the inundated grid cells. The HEC-
FIA model used terrain elevation at a structure, and an assumed foundation height of a structure
49
Chapter 4 – Overview of the FIRO Analysis PR-100

to determine a flooding depth relative to the first floor of each structure. Generalized curves for
each type of structure related the flooding depth to flood damage as a percent of the structure
value. Similar curves related the flooding depth to flood damage for the structure contents.
Aggregating these results across the watershed produced the total flood damage for an event.

The structure inventory did not include data for the number of cars or other property types. The
substantial effort required to integrate such data from other sources was deemed outside the
scope of the FIRO analysis, and the refinements were considered unlikely to affect the results of
the relative comparisons, so flood damage results considered only the value of structures and
contents. Infrastructure damage and indirect losses, such as reduced commerce and employment,
were also not considered.

Validation of the HEC-FIA model started with the simulation of flood damage for an event
known to have caused no damage. The non-damaging HEC-FIA model simulations covered the
same event periods as the HEC-RAS model simulations. Simulated flood damage was validated
as described in Chapter 11 (Section 11.5). The stand-alone HEC-FIA model included life loss
calculations and impact response reports as a by-product of the original model development.
These features were imported into HEC-WAT and used within the deterministic simulations, but
deleted from the model alternative used for the FRA simulations. For the FRA simulations, the
HEC-FIA model data (input) and results were restricted to minimum requirements for calculating
structure damage.

Development of Simulations – HEC-WAT

Deterministic Simulations
The HEC-WAT simulations for the Existing Conditions alternative were created by importing
the stand-alone HEC-HMS, HEC-ResSim, HEC-RAS, and HEC-FIA models into the HEC-WAT
study, and then selecting the appropriate model alternatives for each defined simulation. Three
simulations (described in Section 4.8) were created from the HEC-WAT Alternative and
Simulation Manager (Figure 21). This dialog box allows modelers to create, manage, and
provide a summary of alternatives, analysis periods, and simulations in an HEC-WAT study
(further details are provided in the HEC-WAT User's Manual). The defined simulations are
displayed in the Study Tree of the HEC-WAT main window (Figure 16).

The HEC-WAT Simulation Editor (Figure 22) provides the details for a selected deterministic
simulation, such as the HEC-WAT alternative, HEC-WAT analysis period (sixty-years), and the
HEC-WAT program order (i.e., program sequence). For example, in Figure 22, the selected
simulation includes POR in its name (Figure 22), which for the FIRO study means this is a
deterministic simulation. In the F Part box (Figure 22), the HEC-WAT alternative and analysis
period names have been concatenated to create the F-Part for DSS records being written to the
simulation.dss file. In addition, the HEC-WAT alternative and analysis period names are used
in creating the HEC-WAT directory structure (Run Directory box, Figure 22). For example, in
Figure 22, for the selected simulation, the HEC-ResSim model is named Baseline. The
deterministic simulations for the other water management alternatives were created by
duplicating the Existing Conditions alternative, and simply choosing a different model alternative
for HEC-ResSim.
50
PR-100 Chapter 4 – Overview of the FIRO Analysis

Figure 21 HEC-WAT – Alternative and Simulation Manager

Figure 22 HEC-WAT Simulation Editor - Deterministic Simulation (Base_POR_Full)

51
Chapter 4 – Overview of the FIRO Analysis PR-100

The HEC-WAT Time Window Interval Generator (TWIG) and Time Window Modifier
alternatives (Figure 22) were configured to execute the HEC-RAS and HEC-FIA models only for
a fifteen-day time window that includes the largest event for each water year. Including the
largest event for each water year, provided detailed hydrographs of stages and flows along the
Russian River, as well as flood damage consequences, for use in the analysis of flood events and
computation of frequency statistics based on annual series.

The boundary condition data for deterministic simulations consisted of sixty years of
precipitation, evapo-transpiration, Potter Valley Project flows, and consumptive diversions.
Deterministic results included sixty years of continuous hourly flows from individual subbasins,
and six-hour reservoir inflows, outflows, and pool levels, and flows at certain locations along the
Russian River from hydrological routing. Detailed hydraulic model results and flood damage
consequences were simulated all along the river for the peak event of each water year. Each
sixty-year deterministic simulation required about 2.5 hours to complete.

The continuous simulation results from the deterministic simulations allowed analysis of
performance of reservoir operations across entire flood seasons, and the ability to meet
conservation objectives during the subsequent dry season. In addition, the continuous simulation
illustrated how depletions for some years can carry over into the next year. The event-based
hydraulic details from the deterministic simulations illustrated downstream consequences of
different water management strategies. The results from the Existing Conditions alternative
deterministic simulations allowed modelers to understand how the hydrology and hydraulics
throughout the Russian River watershed interacted during historical periods and events. The
breadth of deterministic results and relatively fast simulation times made the deterministic
simulations most suitable for the iterative modeling employed to formulate water management
alternatives for Lake Mendocino.

Flood Risk Analysis Simulations

The FRA simulations were used to establish existing condition flood frequencies and damage
estimates, based on 5,000 randomly generated events. The FRA simulations consisted of ten
realizations of 500 events each, with each realization representing a different randomly sampled
precipitation-frequency relationship. The choice of 500 events per realization reflects a rule-of-
thumb for reasonable estimates of the 1% flow and stage in each realization. The sampling
method, nested Monte Carlo sampling, allows calculation of performance metrics such as EAD
and the associated confidence bands by separately accounting for knowledge uncertainty and
natural variability. At least 100 realizations (minimally) are required to define the uncertainty in
the simulated results due to boundary conditions, initial conditions, and model parameters.
"Knowledge uncertainty" parameters, such as the precipitation frequency curve, are sampled
once per realization. "Natural variability" parameters, like precipitation magnitude, are sampled
once per event. The number of realizations selected for the FIRO FRA analysis, were not
enough to adequately define the uncertainty around flow and stage frequency curves; however,
the total number of events does provide an adequate estimate of the 1% flow and stage. EAD was
taken as the mean of the 5,000 events.

HEC-WAT groups events into "lifecycles" to support analyses performed for official USACE
planning studies. The analysis period used in FRA simulations defines the length of an
individual project lifecycle. The FIRO analysis did not consider Lake Mendocino water
52
PR-100 Chapter 4 – Overview of the FIRO Analysis

management alternatives from a lifecycle project perspective, and arbitrarily used a fifty-year
analysis period because it represents a common lifecycle found in USACE planning studies. The
100 lifecycles in this analysis each spanned the years 3001 through 3050 (Figure 23). The dates
for this HEC-WAT Analysis Period were arbitrary, and chosen simply to underscore that the
sequences of events were synthetic.

Figure 23 HEC-WAT - Analysis Period Editor

The specialized versions of the individual models (HEC-RAS, HEC-HMS, HEC-ResSim,

HEC-FIA) for the FRA simulations were imported as separate model alternatives into the HEC-
WAT study. An FRA simulation includes more required components than a deterministic
simulation. Random sampling of the hydrology must occur for an FRA simulation, which is
handled by a Hydrologic Sampling alternative. The HEC-WAT hydrologic sampling generates
synthetic storm events that more fully cover the range of potential events. From the HEC-WAT
Simulation Editor (Figure 24), from the FRA tab, the modeler has the capability to control the
"seeds" used for the pseudo-random sampling.

The selected HEC-WAT Hydrologic Sampling alternative determines the subbasin hyetographs
and calendar start and end dates for each flood event. The hyetograph results for each subbasin
were linked to the HEC-HMS model alternative precipitation inputs. The HEC-HMS model
alternative included parameters that allow for sampling of antecedent soil moisture for each
event. The HEC-HMS generated flows were linked as inputs to the HEC-ResSim model
alternative, which was configured to sample initial reservoir levels. The HEC-WAT Skip Flag
Editor was used to simulate the HEC-RAS and HEC-FIA model alternatives when the flow
simulated by HEC-ResSim for Guerneville was high enough to cause flood damage. Since an
FRA simulation separately evaluates each event, the program sequence did not need a TWIG.
To minimize computations even further, a Time Window Modifier alternative was configured to
execute the HEC-RAS and HEC-FIA model alternatives for only ten-day periods, centered on
the time of peak flow at Healdsburg.

The FRA simulations exercised the existing Lake Mendocino water management alternative
(plan) under 5,000 plausible hydrologic scenarios (flood events) that have not occurred in the
historical record, providing insight regarding the performance and limitations of the current
operations, and the downstream impact on the system. The FRA simulations yielded flow-,
stage-, and damage-frequency results along the Russian River, which served as the basis of
comparison regarding the IRO operating alternatives for Lake Mendocino.
53
Chapter 4 – Overview of the FIRO Analysis PR-100

Figure 24 HEC-WAT Simulation Editor - FRA Simulation (BASE_FRA) – FRA Tab

The FIRO simulations in this study were formulated using deterministic simulations limited to
the observed record, and the FRA simulations were used only to stress-test them and compute
more reliable flood risk metrics. In a different application, FRA simulations could be used to
identify key events illustrating certain flood risks, and be integrated into the plan formulation
phase.

Results
The deterministic results for the Existing Conditions alternative consisted of continuous time-
series from 01Oct1950 through 30Sep2010 from the hydrologic and reservoir system models:

• Hourly precipitation for each subbasin (HEC-HMS)

• Hourly moisture deficit in each subbasin (HEC-HMS)
• Hourly flow from each subbasin (HEC-HMS)
• Six-hour storage and pool elevations for Lakes Mendocino and Sonoma (HEC-ResSim)
• Six-hour outflows from Lakes Mendocino and Sonoma (HEC-ResSim)
• Six-hour flows at each gage location (HEC-ResSim)

Deterministic results for the hydraulic and consequence models were for fifteen-day periods
covering the largest event in each water year from 1951 through 2010:

• Hourly flows and stages at each cross-section and storage area (HEC-RAS)
• Annual peak flow and maximum stage at each cross-section and storage area (HEC-RAS)
• Ten meter grid of maximum depth each water year (HEC-RAS)
• Flood damage to structure and contents at each structure (HEC-FIA)
54
PR-100 Chapter 4 – Overview of the FIRO Analysis

These deterministic results were analyzed according to the metrics described in Section 4.5, to
characterize flood risk and potential storage gain. Statistics from the soil moisture and reservoir
level data served as the basis for sampling distributions for the FRA simulations.

The models also produced detailed results that were not used directly in the primary study
results, but used extensively during model development and validation, or for verifying correct
software performance. For example, modelers reviewed results related to subbasin base flow,
candidate reservoir outflows corresponding to each release rule, looped rating curves for
different events, potential loss-of-life at each structure, and incidence of each shape set pattern
sampled for the synthetic events.

Formulation of Alternatives
Including the Existing Conditions alternative, the analysis evaluated five alternatives. The only
difference among the alternatives was the reservoir model (HEC-ResSim). The second
alternative, Without Coyote Valley Dam, was a special limited analysis that was intended simply
to provide context for the total flood damage reduction that Coyote Valley Dam provides to
downstream areas.

The other alternatives employed different HEC-ResSim model operation sets to model outcomes
for different water management alternatives. The Encroachment alternative is based on a very
simple FIRO concept, while the Combined alternative takes a more complex approach. These
two alternatives make use of a simplifying assumption of using "perfect" forecast information,
consistent with the study's "bookend" approach to illustrate the maximum gain in storage without
increasing flood risk. The last alternative, Encroachment with Imperfect Forecast
(EncroachWIF), uses the same simple FIRO concept, but the reservoir simulation model first
randomly "perturbed" the perfect rainfall forecast information to introduce error.

The water management alternatives were formulated with a trial-and-error approach using
deterministic simulations that improved the success of filling Lake Mendocino each spring
without increasing flood peaks downstream. The design of the operations also attempted to
avoid violations of an operational rule specifying maximum flows at Hopland, and to avoid
greater incidence of flow over the spillway. The FRA simulations provided subsequent
confirmation that the alternatives did not significantly increase flood damage, and illustrated
performance of the reservoir operating alternatives under a wider range of hydrologic events than
the deterministic simulations.

This analysis never included the goal to design practical FIRO alternatives suitable for actual
implementation by USACE. The alternatives that received "perfect" forecast information were
able to consider aggressive operations that would capture additional storage, above the guide
curve. Real world operations would likely be less aggressive due to uncertainty in the forecast
information. The types of operations modeled in these FIRO alternatives are based on typical
rules and operations but modified to leverage commonly available types of forecast information.
Actual forecast informed reservoir operations for Lake Mendocino will certainly involve a more
cautious implementation, but some of the operations will likely resemble those explored here.

55
Chapter 4 – Overview of the FIRO Analysis PR-100

The HEC-WAT executed each model alternative in the sequence for the whole analysis period in
a deterministic simulation before the next model was simulated (Section 4.4). Specifically, the
HEC-HMS model alternative was simulated for the whole sixty years before the HEC-ResSim
model alternative began its simulation in a deterministic simulation. Consequently, all HEC-
HMS model alternative results for the entire simulation were available to the HEC-ResSim
model alternative at any time step. This design gave the reservoir simulation model easy access
to "observed" future data, such as precipitation and reservoir inflow. This "perfect forecast" data
was used to drive the water management alternatives based on FIRO.

Without Coyote Valley Dam (WO_CVD)

The Without Coyote Valley Dam (HEC-WAT alternative - WO_CVD) alternative operates the
Warm Springs Dam in the absence of the Coyote Valley Dam. The WO_CVD alternative was
not meant as a proposed operation alternative, and was only used to estimate without-project
flood damage.

Many people have noted the distance of Lake Mendocino from the primary flood damage
centers, and have questioned how much difference the reservoir can actually make. During the
course of this analysis, it became evident that it could be possible to quantify the maximum
impact of Coyote Valley Dam on Russian River flood events by simply modeling an HEC-
ResSim alternative without the reservoir. The flood damage and flow/stage frequency results for
the deterministic simulation are provided for context in understanding the flood risk metrics
associated with other alternatives. Because of this limited purpose, the WO_CVD alternative was
not simulated under an FRA simulation.

Encroachment
The Encroach alternative (with perfect forecast) illustrates a very simple FIRO based approach,
adding a single rule to the existing condition operations during winter/spring that permits the
reservoir to store additional water in the absence of a significant precipitation forecast. The
threshold for significant precipitation was determined using a trial-and-error process that
balanced the amount of storage captured, against added flood risk because resulting operations
increased downstream peak stages, or led to substantially more spillway flow, or resulted in a net
increase of violations of the Hopland 8,000 cubic feet per second (cfs) rule. The selected rule
stated that if at least three inches of total precipitation were forecasted during the next five days,
Lake Mendocino would release as much water as possible to get the reservoir pool back to guide
curve, within the existing condition physical and operational constraints. The amount of
encroachment was capped at 761.8 feet (i.e., the guide curve target for summer pool level). The
rule was in effect from 15 January thru 31 May.

Combined
The Combined alternative (with perfect forecast) demonstrates other potential forecast informed
reservoir operations. Formulation of this alternative started with the preliminary modeling of
individual measures as part of a screening process, in order to identify the most promising
measures. After receiving feedback from others on the FIRO team, HEC refined the forecast-
informed operations to consist of:
56
PR-100 Chapter 4 – Overview of the FIRO Analysis

• An advance release rule that would consider one-, two-, three-, and five-day volumes of
forecasted reservoir inflow.

• A "maximize the Hopland capacity" rule that uses forecasts of unregulated flow at
Hopland to release as much as possible from Lake Mendocino without violating the 8,000
cfs rule.

With the advantage of perfect forecast information driving the rules above, the guide curve was
modified (Figure 25) to capture more storage without increasing flood risk. Modifications
included the following:

1) Change time of pool rise from 01 Mar to 01 Feb (earlier start of fill)
2) Change summer shoulder from 10 May to 10 March (permit faster fill)
3) Change winter shoulder from 31Oct to 30 Nov (slower drawdown)

All of these measures are taken together form the Combined alternative.

Figure 25 Lake Mendocino Guide Curve Adjustments for Combined Alternative

Encroachment with Imperfect Forecast

The EncroachWIF alternative demonstrates the possible impact of known forecast error by
simulating the Encroach alternative using a randomly "perturbed" forecast of five-day total
precipitation instead of observed. The forecast was made imperfect by randomly sampling a
value from a Normal distribution that centers on the correct five-day total precipitation value,
57
Chapter 4 – Overview of the FIRO Analysis PR-100

with a standard deviation equal to some percentage of the correct value, such as fifteen percent,
that represents forecast error or skill. The calculation was performed separately for each day,
allowing greater error for forecasts farther in the future. The forecast was updated every six-
hours in based on the defined time step for the reservoir simulation. The relationship between
successive forecasts was maintained as a serial correlation using an auto-regressive approach.

The results of the Encroach alternative provided context for understanding the significance of a
perfect forecast assumption within the reservoir operation alternatives crafted for Coyote Valley
Dam. The EncroachWIF alternative did not introduce a separate reservoir operation plan for
assessing the viability of forecast informed operations. The EncroachWIF alternative attempts
an initial estimate of the sensitivity of the Encroach alternative to error in the forecast (the
forecast skill).

Characterizations of skill in various types of forecast data, and the approaches used to
incorporate skill into water management alternatives, represent areas of ongoing research and
can be very complex. Proper treatment of these topics lies beyond the scope of this analysis.
The EncroachWIF alternative served only to illustrate an approach to check the sensitivity of one
FIRO alternative to the perfect forecast assumption, and to help shape future strategies for
analyzing the impact of forecast skill.

Analysis of Alternatives
HEC-WAT allows modelers to define simulations comprised of various model alternatives,
analysis periods, and simulation options. In order to provide direct relative comparisons, this
study defined similar simulations for each alternative.

From the HEC-WAT main, window, from the Study Tree (Figure 26), the alternatives and
simulations that have been defined for the FIRO analysis are displayed. Several simulations
have "POR" included in their names (deterministic). Simulations ending in "POR_Short" used a
five-year analysis period from 1960 through 1965, while simulations ending in "POR_Full" used
a sixty-year analysis period. The POR_Short simulations were typically used as a check, to
ensure that all the models, linkages, boundary conditions, and simulation options performed as
expected, before executing the POR_Full simulations.

Results from the Existing Conditions alternative deterministic simulations for the sixty-year
analysis period were closely reviewed to verify that data and modeling produced appropriate
results. The Existing Conditions alternative based simulations provided insight regarding flood
risk drivers in the watershed, and assisted interpretation of differences among results from the
other alternatives. For example, in a few historical years, FIRO alternatives could result in
slightly greater damage than the Existing Conditions alternative for events that happen early in
the flood season. This difference was due to less successful filling of Lake Mendocino during
the previous spring under the Existing Conditions alternative, with deeper depletion of the pool
over the subsequent dry season. This incidental extra storage under the Existing Conditions
alternative made Lake Mendocino more effective in absorbing early flood events.

58
PR-100 Chapter 4 – Overview of the FIRO Analysis

Figure 26 HEC-WAT Study Tree

59
Chapter 4 – Overview of the FIRO Analysis PR-100

The FRA simulations for each alternative produced voluminous results, which were primarily
analyzed in aggregate. The probabilistic results allowed robust frequencies to be associated with
stages, flows, and damage at key locations. Results for the ten largest synthetic events received
closer review, which identified a flood risk not observed in the period of record results.

The HEC-WAT Alternative and Simulation Manager (Figure 21) provides a summary of the
various alternatives, analysis periods, and simulations that were used for the Russian River FIRO
analysis. This feature of HEC-WAT, proved helpful in organizing the analysis and avoiding
errors in the structure of alternatives and simulations. The HEC-WAT allows great flexibility in
defining the study components and structure, and mistakes would be harder to identify by
separately reviewing the definitions of each simulation. For this relatively simple analysis, the
main benefit was that the Alternative and Simulation Manager quickly showed that the
simulations were properly defined. The manager allows the modeler to verify that only the
reservoir model differed between alternatives.

Simulation Procedures
Provisional deterministic simulations were performed throughout the development of the models
and alternatives, with the results for all of the sixty-year simulations being completed overnight.
FRA simulations were performed for this analysis by copying the fully defined HEC-WAT study
to four different personal computers, and executing each simulation in parallel. Each FRA
simulation took about eight hours to complete.

HEC-WAT Version 1.0 includes a Distributed Compute option that automatically manages an
FRA simulation across multiple computers. The Distributed Compute option reduced the
simulation time for one of the Russian River FRA simulations from eighty hours to eight hours,
using ten compute nodes. As with the deterministic simulations, the FRA simulations also
involved provisional simulations. The HEC-WAT allows FRA simulations to resume at
specified lifecycles or realizations, rather than repeat the FRA simulation from the beginning.
This capability allowed effective testing of refinements, corrections to model parameters, or to
HEC-WAT simulation options.

60
PR-100 Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic)

Meteorological and Hydrologic

Model and Data Development
(Deterministic)
The FIRO Existing Conditions alternative is based on models that were developed for previous
studies. The individual models were updated and validated against observed data, with
subsequent refinements after preliminary integration into HEC-WAT and the deterministic
simulations. In the course of formulating the FIRO alternatives, further modifications were
made to the HEC-HMS and HEC-ResSim existing condition model alternatives to support
exchange of forecast information. Chapters 5 through 9 provide information on the data and
model (e.g., HEC-HMS, HEC-ResSim, HEC-RAS, HEC-FIA) development for the HEC-WAT
deterministic simulations. Chapter 10 provides information on what other data and model
development was needed for the HEC-WAT FRA simulations.

This chapter details the meteorological data development and the hydrologic (HEC-HMS) data
and model development for the Existing Conditions alternative and the HEC-WAT deterministic
simulations.

Meteorological Data Development

For the FIRO study, an assumption was made that the Russian River watershed hydrology is
driven entirely by rainfall and neglected any consideration of snowmelt. HEC-HMS contains
methods for simulating snow accumulation and melt; however, the atmosphere is warmer during
historic flood events and precipitation generally falls in liquid form in the Russian River basin.
Hourly precipitation served as the primary input to the HEC-HMS model. The HEC-WAT
deterministic simulations used observed meteorologic data values.

Precipitation
An HEC product, GageInterp, a spatial interpolation utility program, was used to create a
gridded precipitation record by interpolating point precipitation data. GageInterp requires point
precipitation records as input, as well as specifications about the grid, to create the gridded time-
series records. Interpolated values were adjusted in GageInterp using a bias grid. The National
Climatic Data Center (NCDC; 1951-2010, 2001) provided precipitation records for precipitation
gages in or near the Russian River basin. This study utilized twenty-eight hourly precipitation
gages (Table 2). A grid of average annual precipitation from the thirty-year period (1981 through
2010) was obtained from the PRISM (Parameter-elevation Regressions on Independent Slopes
Model; OSU, 2012). The PRISM grid (Figure 27) was applied in GageInterp to improve the

61
Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic) PR-100

Table 2 Hourly Precipitation Gages

Elevation
Station (feet) Begin End
ANGWIN PAC UNION COL 1714.9 10/31/1967 12/6/2013
CLOVERDALE 1 S 319.9 1/31/1973 5/31/2006
CLOVERDALE 11 W 1820.9 6/30/1948 1/31/1973
FINLEY 5 SW 1752.0 10/31/1957 12/27/1963
FINLEY 1341.9 12/31/1955 12/24/1963
HOPLAND 8 NE 2513.1 6/30/1948 9/30/1976
KELSEYVILLE 4 SW 1480.0 12/31/1955 8/9/1962
LAKE MENDOCINO DAM 764.1 5/31/1972 12/6/2013
LAKEPORT SCS 1361.9 12/31/1955 1963/12/63
MAHNKE 2390.1 12/31/1955 12/16/2013
NAVARRO 1 NW 152.9 1/31/1958 12/6/2013
NOVATO 8 WNW 399.9 6/30/1948 11/30/1996
OAKVILLE 4 SW N0 2 1685.0 7/31/1949 4/30/1988
PETALUMA 1 N 29.9 6/30/1948 6/30/1964
POTTER VALLEY 3 NNW 1060.0 7/31/1953 7/31/1964
POTTER VALLEY 3 SE 1100.1 11/30/1952 3/31/1986
POTTER VALLEY PH 1018.0 1/31/1953 4/30/2009
REDWOOD VALLEY 722.1 7/31/1948 5/31/1972
SAINT HELENA 4 WSW 1779.9 6/30/1948 12/6/2013
SAINT HELENA 6 NE 1001.0 6/30/1948 11/30/1974
THE GEYSERS 1740.2 6/30/1948 12/6/2013
UKIAH MASONITE 620.1 11/30/1952 1/31/1958
VENADO 1259.8 6/30/1948 12/6/2013
W. HOWARD FOREST R S 1924.9 6/30/1948 9/30/1989
WARM SPRINGS DAM 224.1 4/30/1973 11/30/1998
WILLITS 1 NE 1353.0 9/30/1989 4/30/2013
YORKVILLE 1100.1 6/30/1948 4/30/2005
CLEARLAKE4 1349.1 10/1/1954 11/1/1985
SEBASTAPOL 67.9 7/1/1948 12/6/2013

interpolation process by spatially distributing precipitation normal to the thirty-year pattern.

GageInterp used an inverse distance squared algorithm to interpolate precipitation values
between gage locations. The interpolation algorithm used a radius of influence of 30,000 meters
for all precipitation gages during the period from 1973 to 2010. From 1950 to 1973, a radius of
influence of 50,000 meters was used because only four gages contained precipitation
measurements. The increased radius of influence provided spatial coverage over the basin during
a time period where fewer stations were available for interpolation.

62
PR-100 Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic)

Figure 27 PRISM Grid with Russian River Basin Rain Gages and Hydrometeorology Stations

In HEC-HMS, a simulation run applied the gridded precipitation records to the Russian River
subbasins. HEC-HMS calculated basin average precipitation for each time step and stored the
resulting hyetograph for each subbasin for use in subsequent simulations.

Meteorologic model accuracy was improved for three large events: December 1964, February
1986, and December 2005; and three dry year events, February 1976: February 1987, and
January 1988; using daily precipitation information. Daily precipitation records were
downloaded from NCDC for six locations in the basin. For the selected events, the interpolated
63
Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic) PR-100

storm precipitation totals (using hourly precipitation gage data) deviated significantly from daily
precipitation observations in the basin. New gridded precipitation records for these events were
created by incorporating daily precipitation observations for gages in the basin with the temporal
pattern of the encompassing subbasin (Table 3). A summary of dates and gages used for each
event adjustment are presented in Table 3. The adjusted event precipitation records (Table 4)
replaced the existing gridded precipitation record.

Table 3 Daily Precipitation Gages

Encompassing
Station Subbasin Begin End
POTTER VALLEY PH EF Russian 20 9/1/1937 3/7/2016
UKIAH Russian 70 1/1/1893 5/24/2013
CLOVERDALE Russian 40 7/22/1950 1/31/2016
HEALDSBURG Russian 30 2/1/1893 8/31/2012
SANTA ROSA Santa Rosa Cr 20 6/1/1902 1/31/2013
GUERNEVILLE Russian 10 11/8/1939 4/28/1971

Table 4 Event Adjustment Stations and Periods

Event Stations Available Begin End
December 1964 POTTER VALLEY PH, UKIAH, 12/20/1964 12/24/1964
CLOVERDALE, HEALDSBURG, SANTA
ROSA, GUERNEVILLE
February 1976 POTTER VALLEY PH, UKIAH, 1/24/1976 3/4/1976
CLOVERDALE, HEALDSBURG, SANTA
ROSA
February 1986 POTTER VALLEY PH, UKIAH, 2/10/1986 2/23/1986
CLOVERDALE, HEALDSBURG
February 1987 POTTER VALLEY PH, UKIAH, 2/8/1987 2/16/1987
CLOVERDALE, HEALDSBURG, SANTA
ROSA
January 1988 POTTER VALLEY PH, UKIAH, 1/1/1987 1/18/1987
CLOVERDALE, HEALDSBURG, SANTA
ROSA
December 2005 UKIAH, CLOVERDALE, HEALDSBURG, 12/25/2005 1/2/2006
SANTA ROSA

Evapotranspiration
In the Russian River HEC-HMS model, the Penman-Monteith evapotranspiration method was
used to extract moisture from the canopy, surface, and soil during periods without rainfall. The
Penman-Monteith method requires shortwave and longwave radiation time series data as input.
Hydrometeorology data sources in the basin provided shortwave radiation directly. The FAO-56
(Food and Agriculture Organization) Penman-Monteith Method (FAO, 1998) was used to
calculate longwave radiation using hydrometeorologic data, maximum/minimum temperature
and moisture, along with location and elevation. The United States Air Force (USAF) 14th
Weather Squadron provided hydrometeorologic data for two stations in the Russian River basin -
64
PR-100 Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic)

Ukiah Municipal Airport (KUKI) and Sonoma County Airport (KSTS). The NCDC provided
Automated Surface Observing System (ASOS) data for stations KUKI and KSTS. The USAF
data included records for solar radiation, air temperature, wind speed, and relative humidity. The
ASOS data included records for air temperature, wind speed, and relative humidity. Complete
hydrometeorologic time series records of solar radiation, air temperature, wind speed, and
relative humidity for the sixty-year simulation were generated from the data sources. USAF data
was used as the primary data source. ASOS data supplemented the USAF 14th Weather
Squadron record for periods with missing data. Climate normals generated from the USAF data
completed the hydrometeorologic time series records when both data sources had missing data.

Hydrologic Model Development

The HEC-HMS model has three main components: basin model, meteorologic model, and
control specifications. The meteorologic model creates the precipitation and evapotranspiration
boundary conditions based on the hydrometeorological input described previously. The control
specification defines the time period and time step required for simulations. A one-hour time
step was chosen as the finest resolution consistent with the input precipitation data and
streamflow calibration data.

HEC-HMS contains many methods for simulating the precipitation-runoff response in a

watershed. For example, nine methods are available for modeling the interception of
precipitation by the canopy and soil surface and infiltration into the soil. These methods include
simple approaches, like the SCS (Soil Conservation Service) curve number method with only
one parameter, to a more complex approach, like the 24 parameter SMA method. HEC-HMS
also contains multiple methods for modeling the transformation of excess precipitation to direct
runoff at the subbasin outlet, baseflow, and the movement of water from an upstream location to
a downstream location (channel routing). A summary of the equations and required parameters
for each method are included in the HEC-HMS Technical Reference Manual (HEC, 2000).

Basin Model
Basin modeling methods were chosen based on data availability, appropriateness for the project
area, and study objectives. Tables 5 through 12 contain the modeling methods used for the
Russian River HEC-HMS model and a list of required parameters. For the loss method, SMA
was used for the subbasin contributing to Coyote Valley Dam. Otherwise, the Deficit and
Constant loss method was used throughout. The SMA loss method was used upstream of Coyote
Valley Dam because of the method's ability to account for soil moisture with high fidelity; this
capability is consistent with study objectives to inform reservoir operations at the dam.

Table 5 Subbasin Canopy Parameters – Simple Canopy Method

Parameter/Method Description
Initial Storage Percentage of the canopy storage that is full of water at the beginning
of the simulation
Maximum Storage Amount of water that can be held in canopy
Crop Method Method for specifying the crop coefficient
Uptake Method Method for extracting water from the soil

65
Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic) PR-100

Table 6 Subbasin Surface Parameters - Simple Surface Method

Parameter/Method Description
Initial Storage Percentage of the surface storage that is full of water at the beginning
of the simulation
Maximum Storage Amount of water that the surface can hold

Table 7 Subbasin Surface Parameters - Soil Moisture Accounting Method (SMA)

Parameter/Method Description
Soil Storage Total water storage in rooting zone
Tension Storage Water held in tension storage in rooting zone
Maximum Infiltration Rate Maximum infiltration rate from surface into the soil
Soil Percolation Percolation rate from soil layer into Groundwater 1 layer
Groundwater 1 Storage Water storage in Groundwater 1 layer
Groundwater 1 Percolation Percolation rate from Groundwater 1 layer into Groundwater 2 layer
Groundwater 1 Coefficient Groundwater 1 linear reservoir coefficient
Groundwater 2 Storage Water storage in Groundwater 2 Layer
Groundwater 2 Percolation Percolation rate from Groundwater 2 Layer into deep aquifer
Groundwater 2 Coefficient Groundwater 2 linear reservoir coefficient
Percent Impervious Area Percent of subbasin directly connected to the stream network

Table 8 Subbasin Loss Parameters - Deficit and Constant Loss Method

Parameter/Method Description
Initial Deficit Initial loss parameter that accounts for the moisture condition in the
watershed at the beginning of the simulation - loss from canopy
interception, surface storage, and infiltration
Maximum Deficit When the soil is at its driest, the maximum deficit is the amount of
moisture/precipitation the soil can hold before runoff begins.
Constant Loss Rate Infiltration rate during saturated soil conditions
Percent Impervious Area Impervious area directly connected to the channel network (no losses
are computed)

Table 9 Subbasin Transform Parameters - Clark Unit Hydrograph Transform Method

Parameter/Method Description
Time of Concentration Travel time along the longest flow path (hour)
Storage Coefficient Linear reservoir coefficient (hour)

Table 10 Subbasin Baseflow Parameters - Linear Reservoir Baseflow Method

Parameter/Method Description
Groundwater 1 coefficient Linear coefficient for computing flow from the first reservoir
Groundwater 1 reservoirs Number of first layer reservoirs used for routing
Groundwater 2 coefficient Linear coefficient for computing flow from the second reservoir
Groundwater 2 reservoirs Number of second layer reservoirs used for routing

66
PR-100 Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic)

Table 11 Reach Routing Parameters - Muskingum-Cunge Routing Method

Parameter/Method Description
Channel Length Length of the channel reach for routing
Slope Energy slope in the channel reach
Channel Shape Representative channel shape (eight-point cross section)
Manning's n values Representative channel roughness

Table 12 Reach Routing Parameters - Modified Puls Routing Method

Parameter/Method Description
Storage-Discharge Relationship between storage and discharge in the entire routing reach
Relationship (computed by HEC-RAS)
Number of Steps Calibration parameter affecting storage (attenuation) in the routing reach

Subbasin Parameterization
A DEM was downloaded from the USGS National Elevation Dataset (USGS, 2013). The HEC-
GeoHMS utility delineated the stream and subbasin boundary network based on the USGS DEM,
and forced subbasins at reservoir and stream gauge locations as calibration points. Subbasin
delineations for the Russian River watershed are shown in Figure 20.

Land use and soils data were used to compute spatially averaged basin parameters. Land use
data was acquired from the USGS based on the year 2000 land usage (USGS, 2005). The land
use data were categorized as agriculture/rangeland, forested, residential/commercial, and water.
The study area is primarily characterized by agriculture/grasslands and forested lands in the
upper and lower basins and residential/commercial near the southern-most region of the basin.
The land use dataset was used to compute the percent impervious area for each subbasin by
assigning a percent impervious percentage to the residential/commercial land use type and
computing subbasin average values.

Soils data were obtained from the SSURGO database (USDA, 2015). SSURGO data were used
to estimate loss parameters for the Deficit and Constant (e.g., Constant Rate, Maximum Deficit)
and SMA (e.g., Maximum Infiltration, Soil Storage) loss methods initially. The SSURGO
database is organized in terms of map units. A soil map unit is a collection of areas defined and
named the same in terms of their soil components. Map units are subdivided into components.
Components consist of vertical layers known as horizons.

Soil properties were estimated by taking area and depth weighted averages of select SSURGO
variables and aggregating them over the subbasin. Tables produced by the USGS, titled "Area-
and Depth-Weighted Averages of Selected SSURGO Variables for the contiguous United States
and District of Columbia" (USGS, 2014), facilitated soil calculations. Constant Rate and
Maximum Infiltration rate estimates were made by assigning infiltration rates for each
hydrologic soil group and calculating a weighted average based on hydrologic soil group content
for each map unit. Infiltration rate assignments for each hydrologic soil group were based on
ranges published in Technical Release 55, "Urban Hydrology for Small Watersheds" (NRCS,
1986) and are presented in Table 13. Maximum Deficit and Soil Storage were estimated by
multiplying the porosity of the soil for each map unit by an assumed active root zone depth of
67
Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic) PR-100

Table 13 Hydrologic Soil Group Description and Infiltration Rate Assignments

Infiltration
Hydrologic Soil Rate
Group Description (in/hr)
Low runoff potential when thoroughly wet. Water is transmitted freely through 0.35
A
the soil (e.g., gravel, sand, loamy sand, or sandy loam).
Moderately low runoff potential when thoroughly wet. Water transmission 0.25
B
through the soil is unimpeded (e.g., silt loam or loam).
Moderately high runoff potential when thoroughly wet. Water transmission 0.15
C
through the soil is somewhat restricted. (e.g., sandy clay loam).
High runoff potential when thoroughly wet. Water movement through the soil is 0.05
D restricted or very restricted (e.g., clay loam, silty clay loam, sandy clay, silty
clay, or clay).
Dual hydrologic soil group. The first letter applies to the drained condition. The 0.05
A/D, B/D ,C/D
second letter applies to the undrained condition.
NA Undefined 0.05

24 inches. Loss parameters were generated for subbasins by taking an area weighted average of
values calculated for each map unit. Soil Storage and Max Infiltration rate initial estimates are
presented in Table 14. Maximum Deficit and Loss Rate estimates are presented in Table 15.

Table 14 Initial Parameter Estimates for Subbasin EF Russian 20

Element Parameter Value
Canopy Crop Coefficient 0.75
Loss Soil Storage (inches) 10.9
Tension Storage (inches) 3.8
Max Infiltration Rate (inches/hour) 0.11
Soil Percolation (inches /hour) 0.2
Groundwater 1 Storage (inches) 3
Groundwater 1 Percolation (inches /hour) 0.1
Groundwater 1 Coefficient (hour) 10
Groundwater 2 Storage (inches) 10
Groundwater 2 Percolation (inches/hour) 0.05
Groundwater 2 Coefficient (hour) 400
Transform Time of Concentration (hour) 6
Storage Coefficient (hour) 5
Baseflow Groundwater 1 Coefficient (hour) 0.5
Groundwater 1 Reservoirs 1
Groundwater 2 Coefficient (hour) 0.5
Groundwater 2 Reservoirs 1

Initial soil water content parameters ('Soil (percent)' for SMA and 'Initial Deficit (inches)' for
Deficit and Constant) were input for the model run but did not affect results because a warm up
period was used for the continuous simulation.
68
PR-100 Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic)

Table 15 Initial Parameter Estimates for Deficit and Constant Subbasins

HMS Maximum Constant Time of Storage GW 1 GW 2
Subbasin Deficit Loss Rate Concentration Coefficient Coefficient Coefficient
Name (in) (in/hr) (hr) (hr) (hr) (hr)
Austin Cr 10.70 0.17 6.7 6.7 30 600
Big Sulphur Cr 10.60 0.11 5.2 5.2 30 600
Dry Creek 10 11.00 0.14 8.8 17.5 30 600
Dry Creek 20 10.40 0.13 2.4 2.4 30 600
Dry Creek 30 10.90 0.13 10.8 10.8 30 600
EF Russian 10 8.80 0.11 1.2 1.2 30 600
Green Valley 10.60 0.15 9.6 9.6 30 600
Laguna 10.70 0.10 8.5 8.5 30 600
Russian 10 11.00 0.18 3.2 3.2 30 600
Russian 20 10.60 0.14 7.0 7.0 30 600
Russian 30 11.10 0.12 4.0 4.0 30 600
Russian 40 10.80 0.13 4.1 4.1 30 600
Russian 50 10.90 0.10 7.5 7.5 30 600
Russian 60 10.80 0.10 8.0 8.0 30 600
Russian 70 10.30 0.11 7.0 7.0 30 600
Santa Rosa Cr 10 10.80 0.11 8.4 8.4 30 600
Santa Rosa Cr 20 10.80 0.10 7.6 7.6 30 600
WF Russian 10.80 0.13 6.0 6.0 30 600

The time of concentration, or the time it takes for runoff to travel from the most distant point in
the watershed to the outlet, was estimated using the TR-55 methodology (NRCS, 1986). The
longest flow path was sub-divided into segments for sheet flow, shallow concentrated flow, and
channel flow. Physical parameters such as length and slope were computed using the DEM.
Cross-sectional data was estimated from field measurements, while additional data required for
the TR-55 method such as the two-year 24-hour rainfall, were entered based on published data
(Weather Bureau, 1961).

The Clark storage coefficient (R) is a calibration parameter that affects the shape of the runoff
hydrograph, meaning this parameter is set using measured precipitation-runoff data. Initial
estimates for R were set using a relationship where time of concentration (Tc) and Storage
Coefficient (R) were equal to one another. Initial Tc and R values for each subbasin are
presented in Tables 14 and 15.

Reach Parameterization
Modified-Puls storage routing was used for the routing method for reaches along the Russian
River. A storage-discharge relationship was developed using a calibrated steady-flow HEC-RAS
model and simulating a range of flows large enough to encompass historical peak flows. The
relationships were then developed for each routing reach and written to a DSS file from HEC-
RAS for input into an HEC-HMS model.
69
Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic) PR-100

Muskingum-Cunge routing was selected for a limited number of routing reaches that were not
included in the Russian River HEC-RAS model. The Muskingum-Cunge routing method was
selected because the routing is based on physical parameters such as channel shape, routing
reach length, and surface roughness (Manning's n value). Muskingum-Cunge routing lends itself
to circumstances where limited observed data is available. Cross sections were represented using
an eight-point description of the land surface representative of the routing reach. Cross sections
were estimated from USGS topographic maps. An example cross section is shown in Figure 28.

Figure 28 Example Eight-Point Cross Section used in HEC-HMS

Calibration
Model calibration is the process of adjusting model parameters to improve agreement between
computed data and observed data. Continuous simulation of sixty years under a single set of
precipitation-runoff model parameters posed a difficult calibration challenge. The challenge was
compounded by calibrating for both annual flow peaks and seasonal volumes, with special
attention to specific peak flows and volumes in historic event hydrographs. Several runoff
signatures were considered during calibration to support FIRO objectives:

1) The average monthly runoff volume was considered, to ensure that the water supply
operations of the reservoir model were representative of the existing conditions and
suitable for use in determining the additional storage possible under FIRO.

2) The annual maximum flow-frequency curve was considered as the goal of this study to
assess flood risk (low probability events) under FIRO alternatives.

3) The runoff hydrograph, including timing of peak flows, and the recession limb of
hydrographs was considered to ensure fine resolution accuracy and establish baseline
forecast for the FIRO analysis.

The model was calibrated starting upstream and proceeding sequentially downstream. The
model was calibrated for the East and West Forks of the Russian River at the Calpella and Ukiah
70
PR-100 Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic)

gages respectively. The calibration then proceeded downstream along the main stem through the
Hopland, Cloverdale, Healdsburg, and Guerneville gages. During calibration, model parameters
were adjusted within the physically reasonable range. For each simulation, model parameters
remained constant and state variables evolved with no interference. At each calibration point, a
monthly runoff volume curve and annual peak flow-frequency curve, generated from model
results, were evaluated against observations. Additionally, summary statistics were calculated
for daily flow quantifying model performance against observations.

Monthly Volumes
The beginning of the wet season is characterized by a dry soil layer that progressively becomes
saturated with each rainfall event. The first major runoff event of the season usually corresponds
with the soil layer reaching maximum capacity. The soil remains at or near a saturated state for
the remainder of the wet season, where minimal to no precipitation is abstracted before surface
runoff begins. The SMA parameters soil depth and soil tension storage and the Deficit and
Constant parameter maximum deficit were adjusted such that the timing of soil saturation
corresponded with the first major runoff event of the wet season. This portion of calibration
focused on average monthly flow volume at the beginning of the wet season corresponding to the
months November and December.

The Penman-Monteith evapotranspiration method is physically based and inputs are based on
measured atmospheric parameters. The method was developed for estimating plot scale crop
evapotranspiration and incorporates a crop coefficient that adjusts the rate of evapotranspiration
for a variety of vegetative surfaces. For the Russian River study, Penman-Monteith
evapotranspiration was applied at the catchment scale for a composite of vegetative surfaces. The
crop coefficient was adjusted to reproduce the monthly runoff distribution of the watershed. This
portion of calibration focused on agreement between computed and observed average monthly
flow volume. Calibrated monthly-average flow signatures for the six gauged locations are
presented in Figures 29 through 34. The average flows for a defined period essentially represent
volume. The averages were computed using HEC-DSSVue, and plotted with the blue lines
representing the model results and the red line showing the observed gage values.

The HEC-HMS flows served as the foundation of the baseline inputs to the reservoir model and
hydraulic model. Calibrating the average monthly volumes helped ensure that the water supply
operations of the reservoir model were representative of the existing condition, and suitable for
use in determining the additional storage possible under FIRO. The water supply operations also
affected flood peaks for some early-season events, since depletions modeled during the dry
months determined the reservoir pool level entering the next flood season.

Flows at Calpella represent the majority of inflow into Lake Mendocino. On average, the HEC-
HMS flows were somewhat higher than observed during the winter, but lower in spring and
summer. Flows into Calpella from the Potter Valley Project likely account for some of the
difference, whereas a majority of the difference is due to availability of representative
precipitation occurring over the drainage area upstream of the Calpella flow.

Considerable effort went into the Calpella calibration, with trade-offs between monthly volume
and flood peaks, and impact on calibrations for downstream locations. The results emphasized

71
Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic) PR-100

Figure 29 Calpella Monthly Volume Calibration

Figure 30 Ukiah Monthly Volume Calibration

72
PR-100 Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic)

Figure 31 Hopland Monthly Volume Calibration

Figure 32 Cloverdale Monthly Volume Calibration

73
Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic) PR-100

Figure 33 Healdsburg Monthly Volume Calibration

Figure 34 Guerneville Monthly Volume Calibration

74
PR-100 Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic)

flood modeling as the most critical study purpose. Minor seasonal differences between the HEC-
HMS model volumes and the observed record were considered acceptable, since the flows were
used as a common input for a relative comparison.

If the inflow volumes from HEC-HMS did not reasonably reflect the actual basin conditions,
then hypothetically the benefits and risks of various FIRO strategies could be distorted.
Calibration results for monthly volumes at Ukiah and below showed an extremely close
correspondence with the observed values.

Flow Frequency Curves

SMA and Deficit and Constant are conceptual loss methods. While some parameters are loosely
based on physically properties, for example, Maximum Infiltration Rate, all parameters are
subject to calibration. For the Russian River study, the Maximum Infiltration Rate and Constant
Rate were adjusted to reproduce observed hydrograph peaks. This portion of calibration focused
on agreement between computed and observed flow-frequency curves. Flow frequency curves
(peak) were computed using annual maximum peak flows from HEC-HMS model results
compared to flow frequency curves (peak) from observations.

The unregulated period from 1950 through 1959 was combined with the regulated period to
capture early events. Calibrated flow frequency curves for the six gaged locations are presented
in Figures 35 through 40. The flow frequency curves compare flow-frequency analyses of
annual maximum peak flow between HEC-HMS computed results and USGS observations.
Results show that the HEC-HMS model reproduces the extreme upper (low annual exceedance
probability) end of the flow frequency curves at the six gage locations. There is discrepancy in
the flow frequency curves at the lower (high annual exceedance probability) end, but no damages
are expected for these low flows. The differences in flow frequency should have no impacts on
the computed damages for the deterministic simulations.

Runoff Hydrograph
Clark unit hydrograph parameters Tc and R were adjusted to reproduce timing of the hydrograph
peak; Groundwater 1 Coefficient and Groundwater 2 Coefficient were adjusted to reproduce the
interflow and baseflow portions (recession limb) of the hydrograph. This portion of calibration
focused on agreement between the computed and observed runoff hydrographs with emphasis
and timing of peaks and the recession limb. Computed versus observed runoff hydrograph at
Healdsburg for December 2003 through March 2004 is displayed in Figure 41.

Measures of Model Performance

For the Russian River study, performance was evaluated for computed versus observed daily
flows, for gauged locations over the sixty-year period of record, at USGS gage locations: East
Fork Russian River near Calpella and the Russian River near Ukiah, Hopland, Cloverdale,
Healdsburg, and Guerneville. Summary statistics were evaluated based on published ranges of
model performance from Moriasi, 2007. Parameter adjustment proceeded until the model
performed within the "very good" range for all performance metrics.

75
Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic) PR-100

Figure 35 Annual Flow Frequency Curve at Calpella

Figure 36 Annual Flow Frequency Curve at Ukiah

76
PR-100 Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic)

Figure 37 Annual Flow Frequency Curve at Hopland

Figure 38 Annual Flow Frequency Curve at Cloverdale

77
Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic) PR-100

Figure 39 Annual Flow Frequency Curve at Healdsburg

Figure 40 Annual Flow Frequency Curve at Guerneville

78
PR-100 Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic)

Figure 41 Simulated versus Observed Runoff Hydrograph at Healdsburg for the Period of December
2003 through March 2004

Statistics were used to quantify model performance compared to observations. Statistics include
Nash-Sutcliffe Efficiency (NSE), Observed Standard Deviation Ratio (RSR), coefficient of
determination (R2), and percent bias (PBIAS) were computed using daily average flow. NSE
measures the relative magnitude of the residual variance compared to the measured data
variance. NSE ranges between -∞ and 1.0, where NSE equal to one is optimal. Values of NSE
less than 0.0 indicate the mean observed value is a better predictor than the simulated value,
meaning that the performance is unacceptable. NSE is computed using Equation 1.
𝑛𝑛
𝛴𝛴𝑖𝑖=1 (𝑌𝑌𝑖𝑖𝑜𝑜𝑜𝑜𝑜𝑜 − 𝑌𝑌𝑖𝑖𝑠𝑠𝑠𝑠𝑠𝑠 )2
𝑁𝑁𝑁𝑁𝑁𝑁 = 1 − � 𝑛𝑛 � (1)
𝛴𝛴𝑖𝑖=1 (𝑌𝑌𝑖𝑖𝑜𝑜𝑜𝑜𝑜𝑜 − 𝑌𝑌� 𝑜𝑜𝑜𝑜𝑜𝑜 )2
where:

n = number of observed values compared to computed over the duration of the

simulation
𝑌𝑌𝑖𝑖𝑜𝑜𝑜𝑜𝑜𝑜 = observed daily flow values
𝑌𝑌𝑖𝑖𝑠𝑠𝑠𝑠𝑠𝑠 = computed daily flow values
𝑌𝑌� 𝑜𝑜𝑜𝑜𝑜𝑜 = average of observed daily flow values

79
Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic) PR-100

RSR normalizes the Root Mean Square Error (RMSE) by using the standard deviation of the
observations, incorporating the benefits of error index statistics so that the resulting statistic can
be applied to various constituents. The RSR is computed using Equation 2.

𝑛𝑛
��𝛴𝛴𝑖𝑖=1 (𝑌𝑌𝑖𝑖𝑜𝑜𝑜𝑜𝑜𝑜 − 𝑌𝑌𝑖𝑖𝑠𝑠𝑠𝑠𝑠𝑠 )2 �
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅
𝑅𝑅𝑅𝑅𝑅𝑅 = = (2)
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑜𝑜𝑜𝑜𝑜𝑜 𝑛𝑛
��𝛴𝛴𝑖𝑖=1 (𝑌𝑌𝑖𝑖𝑜𝑜𝑜𝑜𝑜𝑜 − 𝑌𝑌�𝑖𝑖𝑠𝑠𝑠𝑠𝑠𝑠 )2 �

where: 𝑌𝑌�𝑖𝑖𝑠𝑠𝑠𝑠𝑠𝑠 = average of simulated daily flow values

R2 describes the degree of collinearity between simulated and observed data, and describes the
proportion of the variance in measured data explained by the model. R2 is oversensitive to
outliers and insensitive to additive and proportional differences between model predictions and
measured data. R2 is computed using Equation 3.
2
⎡ ⎤
𝑛𝑛
𝛴𝛴𝑖𝑖=1 (𝑌𝑌𝑖𝑖𝑜𝑜𝑜𝑜𝑜𝑜 − 𝑌𝑌� 𝑜𝑜𝑜𝑜𝑜𝑜 )(𝑌𝑌𝑖𝑖𝑠𝑠𝑠𝑠𝑠𝑠 − 𝑌𝑌� 𝑠𝑠𝑠𝑠𝑠𝑠 )
𝑅𝑅 2 = ⎢ ⎥ (3)
⎢ 𝑛𝑛 ⎥
�𝛴𝛴 (𝑌𝑌 𝑜𝑜𝑜𝑜𝑜𝑜
− 𝑌𝑌� 𝑜𝑜𝑜𝑜𝑜𝑜 )2 �𝛴𝛴𝑖𝑖=1 (𝑌𝑌𝑖𝑖
𝑛𝑛 𝑠𝑠𝑠𝑠𝑠𝑠
− 𝑌𝑌� 𝑠𝑠𝑠𝑠𝑠𝑠 )2
⎣ 𝑖𝑖=1 𝑖𝑖 ⎦

PBIAS measures the average tendency of the simulated data to be larger or smaller than the
observed data. The optimal value for PBIAS is zero (0.0), with low absolute percent bias
indicating accurate model simulation. Positive values mean the model underestimation bias
when compared to the observed, whereas negative values indicate the model overestimation bias.
PBIAS is computed using Equation 4.

𝑛𝑛
𝛴𝛴𝑖𝑖=1 (𝑌𝑌𝑖𝑖𝑜𝑜𝑜𝑜𝑜𝑜 − 𝑌𝑌𝑖𝑖𝑠𝑠𝑠𝑠𝑠𝑠 ) × 100
𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 = � �
𝑛𝑛
𝛴𝛴𝑖𝑖=1 (𝑌𝑌𝑖𝑖𝑜𝑜𝑜𝑜𝑜𝑜 ) (4)

Summary statistic performance ratings based on Moriasi (2007) are presented in Table 16.

Table 16 Performance Ratings for Summary Statistics

Performance
Rating NSE RSR R2 PBIAS
Very Good 0.65<𝑁𝑁𝑆𝑆𝐸𝐸≤1.00 0.00<𝑅𝑅𝑆𝑆𝑅𝑅≤0.60 0.65< 𝑅𝑅2≤1.00 𝑃𝑃𝐵𝐵𝐼𝐼𝐴𝐴𝑆𝑆< ±15
Good 0.55<𝑁𝑁𝑆𝑆𝐸𝐸≤0.65 0.60<𝑅𝑅𝑆𝑆𝑅𝑅≤0.70 0.55<𝑅𝑅2≤0.65 ±15≤𝑃𝑃𝐵𝐵𝐼𝐼𝐴𝐴𝑆𝑆<±20
Satisfactory 0.40<𝑁𝑁𝑆𝑆𝐸𝐸≤0.55 0.70<𝑅𝑅𝑆𝑆𝑅𝑅≤0.80 0.40<𝑅𝑅2≤0.55 ±20≤𝑃𝑃𝐵𝐵𝐼𝐼𝐴𝐴𝑆𝑆<±30
Unsatisfactory 𝑁𝑁𝑆𝑆𝐸𝐸≤0.40 𝑅𝑅𝑆𝑆𝑅𝑅>0.80 𝑅𝑅2≤0.40 𝑃𝑃𝐵𝐵𝐼𝐼𝐴𝐴𝑆𝑆≥±30
NSE = Nash-Sutcliffe efficiency
RSR = Observed Standard Deviation Ratio
R2 = coefficient of determination
PBIAS = percent bias

Summary statistics of model performance for the deterministic calibration simulation at the six
calibration points are presented in Table 17.

80
PR-100 Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic)

Table 17 Summary Statistics at Calibration Points

Location NSEa RSRb R2c PBIASd
Calpella 0.854 0.382 0.892 -1.242
Ukiah 0.880 0.347 0.880 -2.350
Hopland 0.913 0.295 0.919 -1.435
Cloverdale 0.927 0.270 0.931 -2.710
Healdsburg 0.915 0.292 0.916 -2.210
Guerneville 0.938 0.250 0.939 -3.233
NSE = Nash-Sutcliffe efficiency
RSR = Observed Standard Deviation Ratio
R2 = coefficient of determination
PBIAS = percent bias

Final Parameters
The calibration process, that is described in this entire section yields a calibrated model with
final calibrated parameter values for each subbasin. Table 18 summarizes the final calibrated
parameter values for subbasin EF Russian 20. Table 19 summarizes the final calibrated
parameter values for Deficit and Constant subbasins.

Table 18 Calibrated Parameters for Subbasin EF Russian 20

Element Parameter Value
Canopy Crop Coefficient 0.7
Loss Soil Storage (inches) 10
Tension Storage (inches) 7
Max Infiltration Rate (inches/hour) 0.3
Soil Percolation (inches/hour) 0.1
GW 1 Storage (inches) 4
GW 1 Percolation (inches/hour) 0.1
GW 1 Coefficient (hour) 24
GW 2 Storage (inches) 10
GW 2 Percolation (inches/hour) 0.001
GW 2 Coefficient (hour) 150
Transform Time of Concentration (hour) 5.5
Storage Coefficient (hour) 4.5
Baseflow GW 1 Coefficient (hour) 0.5
GW 1 Reservoirs 1
GW 2 Coefficient (hour) 24
GW 2 Reservoirs 1

Modified HEC-HMS Model for Deterministic

Simulations
A modified version of the calibrated HEC-HMS Russian River model was prepared for HEC-
WAT. The modified HEC-HMS model included one simulation run for HEC-WAT deterministic
simulations and one simulation (Uncertainty Analysis) for the FRA simulation. All event
81
Chapter 5 – Meteorological & Hydrologic --Model & Data Development (Deterministic) PR-100

Table 19 Calibrated Parameters for Deficit and Constant Subbasins

HMS Maximum Constant Time of Storage GW 1 GW 2
Subbasin Deficit Loss Rate Concentration Coefficient Coefficient Coefficient
Name (inch) (inch/hour) (hour) (hour) (hour) (hour)
Austin Cr 6.00 0.400 6.7 6.7 24 200
Big Sulphur Cr 10.00 0.300 5.2 5.2 24 400
Dry Creek 10 6.00 0.400 8.8 19.3 24 200
Dry Creek 20 6.00 0.400 2.4 2.4 24 200
Dry Creek 30 6.00 0.400 10.8 10.8 24 200
EF Russian 10 7.00 0.075 1.2 1.2 24 400
Green Valley 6.00 0.400 9.6 20.9 24 200
Laguna 6.00 0.400 8.5 9.4 24 200
Russian 10 6.00 0.400 3.2 3.2 24 200
Russian 20 6.00 0.400 7.0 15.4 24 200
Russian 30 10.00 0.300 4.0 4.0 24 400
Russian 40 10.00 0.300 4.1 4.1 24 400
Russian 50 8.00 0.125 7.5 7.5 24 400
Russian 60 12.00 0.050 8.0 9.0 24 400
Russian 70 12.00 0.050 7.0 8.0 24 400
Santa Rosa Cr 10 6.00 0.400 8.4 18.7 24 200
Santa Rosa Cr 20 6.00 0.400 7.6 12.1 24 200
WF Russian 7.00 0.100 6.0 3.0 24 400

simulations and corresponding basin models, meteorologic models and control specifications
were removed from the HEC-HMS model to reduce the size of the model in computer storage.

The HEC-HMS simulation run for the deterministic analysis included a modified basin model
and the existing meteorologic model. The deterministic and corresponding basin and
meteorologic models were prepared as follows:

• Meteorologic Model. The HEC-HMS project's meteorologic model - Basin Average 60

yr, required no changes.

• Basin Model. The HEC-HMS project's original basin model was Basin Average 60 yr,
and was renamed to Continuous. All source elements were removed from the HEC-HMS
model. The source elements were used to bring in observed reservoir releases from Lake
Mendocino and Lake Sonoma. These source elements were replaced by flows generated
by the HEC-ResSim model alternative. Observed flow was retained for all model
locations co-located with observed flow gages.

• Simulation Run. The HEC-HMS simulation run included the Basin Average 60 yr
meteorologic model and the Continuous basin model. Results from the HEC-HMS model
for the HEC-HMS simulation were documented in Chapter 10 (Section 10.2.3).

82
PR-100 Chapter 6 – Reservoir Simulation - Model & Data Development (Deterministic)

Reservoir Simulation
Model and Data Development
(Deterministic)
The FIRO Existing Conditions alternative is based on models that were developed for previous
studies. The individual models were updated and validated against observed data, with
subsequent refinements after preliminary integration into HEC-WAT and the deterministic
simulations. In the course of formulating the FIRO alternatives, further modifications were
made to the HEC-HMS and HEC-ResSim existing condition model alternatives to support
exchange of forecast information. Chapters 5 through 9 provide information on the data and
models (e.g., HEC-HMS, HEC-ResSim, HEC-RAS, HEC-FIA) development for the HEC-WAT
deterministic simulations. Chapter 10 provides information on what other data and model
development was needs for the HEC-WAT FRA simulations.

This chapter details the reservoir simulation (HEC-ResSim) data and model development for the
Existing Conditions alternative and the HEC-WAT deterministic simulations. Further details
about the reservoir model development and formulation process regarding the alternative water
management plans is available from the Hydrologic Engineering Center (point-of-contact: Mr.
Matthew McPherson) upon request.

HEC-ResSim Data Development

Inflows
The HEC-HMS results provided reservoir supplies and local inflows to HEC-ResSim junctions
along the Russian River and Dry Creek. The HEC-HMS subbasin outflow hydrographs were
applied at the Russian River junctions held in common with the HEC-ResSim model. HEC-
HMS routed flows from the headwaters of Dry Creek to Lake Sonoma and from upper Austin
Creek to the Russian River. The HEC-ResSim model also relied on HEC-HMS routed flows
from the Santa Rosa Creek subbasins to the Russian River.

Since the HEC-ResSim model required lower Dry Creek to be divided into two reaches, the
HEC-HMS flows were apportioned according to the drainage area. At the "Dry Creek near
Geyserville" junction, 51.6 percent of the flows were added, with the rest of the flow being
added at the confluence with the Russian River.

The only other flow into the system was the diversion from the Eel River through the Potter
Valley Project into the headwaters of the East Fork Russian River. The Potter Valley Project
flows for 1950 through 2010 used in this study were modeled by the SCWA to reflect PG&E's
83
Chapter 6 – Reservoir Simulation - Model & Data Development (Deterministic) PR-100

operation of the Potter Valley Hydroelectric Project since 2006 under the amended FERC
license. Details are provided in the Lake Mendocino Water Supply Reliability Evaluation Report
(Section 2.2, Technical Memorandum).

Consumptive Withdrawals
SCWA provided time-series of estimated diversions from the Russian River. The twelve
diversions (Table 20) out of the system used in the HEC-ResSim model represent simplified net
aggregations of specific diversions, return flows, aquifer recharge, and a variety of distributed
losses. Six of the diversions defined in the HEC-ResSim model reflect only municipal and
industrial withdrawals. The six diversions were modeled as repeating annual patterns, and were
based on usage from 2009 through 2013 (sometimes using separate patterns for wet and dry
years). The SCWA diversion near Mirabel is by far the largest withdrawal from the system,
ranging from around 75 cfs in the winter to 125 cfs in the summer. The other five exclusively
municipal and industrial diversions range between zero and ten cfs during the year.

Table 20 Diversions from the Russian River

Diversion Name Location Mapped To
Redwood Valley Div Coyote Valley Dam REDWOOD VALLEY/UPPER RR
SCWA Div SCWA Diversion Dam SCWA DIVERSIONS/LOWER RR
Healdsburg Fitch Div SCWA Diversion Dam HLDSBRG FITCH MTN WELLS/LOWER RR
Windsor Div SCWA Diversion Dam CITY OF WINSDOR/LOWER RR
Lower Rive Ag Div SCWA Diversion Dam HACIENDA AGG/LOWER RR
Calpella Div Calpella Gage Junction CALPELLA/UPPER RR
Hopland Div Hopland Gage Junction HOPLAND/UPPER RR
Cloverdale Div Cloverdale Gage Junction CLOVERDALE/UPPER RR
Healdsburg Div Healdsburg Gage unction HEALDSBURG/UPPER RR
Healdsburg Dry Creek Div Dry Creek Junction HLDSBRG DRY CREEK WELLS/LOWER RR
Dry Creek Div Dry Creek Junction DRY CREEK/DRY CREEK
RRCWD Div Guerneville Gage Junction RR COUNTY WD/LOWER RR

The other six diversions include some component of agricultural demand, where the losses are
functions of the quantity of water available in the river, and were modeled by SCWA using a
water balance approach. Two of these diversions are computed for each month of the year. The
other four use the water balance approach for May through November, but repeat winter values
based on municipal and industrial demands.

The data for the diversions was provided by SCWA. The values were generally derived
according to the procedures described in the Lake Mendocino Water Supply Reliability
Evaluation Report, with some refinements and updates implemented for the FIRO project.

Lookback Data
Time-series data for levels and releases for the two reservoirs were specified in the alternative
definition. The data was developed during provisional reservoir modeling runs, and was only
used in the final simulations during the "warm-up" period for HEC-ResSim. In addition,
historical values for the type of year (normal, dry, or critical) used in conjunction with the

84
PR-100 Chapter 6 – Reservoir Simulation - Model & Data Development (Deterministic)

hydrologic index were provided by SCWA for the rules governing the SWRCB Decision 1610
minimum flows. The values spanned the entire 1950 through 2010 analysis period, plus a few
weeks prior to the start. Providing lookback for the full analysis period served as a convenience
for ad hoc simulations of certain events or periods, allowing modelers to choose any start time.

Miscellaneous External Data

Rules and state variables used in the deterministic simulations required certain inputs to be
specified throughout the analysis period. The Lake Pillsbury Storage amount was used by the
state variable that computed the hydrologic index.

HEC-ResSim Model Development

The HEC-ResSim existing conditions (Baseline) alternative represents current reservoir
operation and boundary condition assumptions, such as extra-basin diversions and evaporation
losses. The results serve as a "baseline" for relative comparisons with alternative reservoir
operation plans. The reservoir systems model was developed using NGVD, for compatibility
with familiar operational criteria and established reference material. Where practical, equivalent
NAVD elevations are provided in results tables and figures.

HEC-ResSim Network Design

The HEC-ResSim model contained three "networks" that reflect different physical or operational
representations for modeling the Russian River system. Two networks were designed for use
within the HEC-WAT deterministic simulation. One of the networks (POR) contained three
alternatives, including the HEC-ResSim Baseline alternative. The HEC-ResSim Baseline
alternative represents the reservoir operations for the Russian River watershed as of 2016. In
addition, the HEC-ResSim (Baseline) alternative was used as a baseline condition for relative
comparison to the other alternatives. The POR network also contained two FIRO alternatives: a
simple FIRO approach that permits encroachment into the flood pool in the absence of
significant precipitation (Encroach), and a more complex FIRO alternative that blended multiple
measures such as guide curve modifications, an advance release rule based on forecast reservoir
inflow, and forecasted downstream channel capacity (Combined).

The second network (WO_CVD) was also designed for HEC-WAT deterministic simulations.
WO_CVD contained a single alternative that operated Warm Springs Dam in the absence of
Coyote Valley Dam. The WO_CVD alternative provided an estimate of the "without project"
flood damage, to illustrate how much flood protection Coyote Valley Dam provides to various
downstream locations.

The third network (FRA) was designed for HEC-WAT FRA simulations. The network contained
versions of the same three model alternatives as the POR network, but was configured for
efficient analysis of flood operations for events generated as part of extensive Monte Carlo
simulations. The model alternatives in the FRA network were named BaselineF, EncroachF, and
Combined.

85
Chapter 6 – Reservoir Simulation - Model & Data Development (Deterministic) PR-100

HEC-ResSim Model Schematic Elements

An overview of the HEC-ResSim model
layout is displayed in Figure 42. The
orange lines represent the stream
alignment, with small green circles
marking the endpoints and junctions. The
blue shapes indicate Lake Mendocino and
Lake Sonoma. The red lines show the
HEC-HMS subbasin delineations for
reference. Red circles (Figure 42)
indicate HEC-ResSim junctions, which
were created to establish locations for
model inputs, outputs, or places of
interest, such as gages providing
observed values for model calibration.
Many of the junctions indicate changes in
streamflow, such as a confluence,
diversion out of the river, or a
contribution from a local subbasin. The
white "halo" around certain junctions
indicates a local flow assignment. Some
junctions also played a role in reservoir
rules. Squares (Figure 42) around a
HEC-ResSim junction indicates that a
reservoir operates to meet a flow
requirement for that location.

The junctions are connected by reaches,

dark blue lines (Figure 42), with small
blue arrows that indicate direction of
flow. On the schematic (Figure 42), the
black arrows represent diversions. Each
HEC-ResSim element carries a label,
although the rendering is scale-dependent
and usually few labels are visible.

Reach Routing
The HEC-ResSim model includes more
junctions than the HEC-HMS model, and Figure 42 HEC-ResSim Model Schematic Elements
consequently includes more reaches. The
same Modified Puls routing parameters used for the HEC-HMS model were used in the HEC-
ResSim model for reaches along the Russian River. In cases where extra junctions divided a
single HEC-HMS reach into two or more HEC-ResSim reaches, the shorter segments were
assigned null routing.

86
PR-100 Chapter 6 – Reservoir Simulation - Model & Data Development (Deterministic)

Reservoir operations were associated with two of the additional junctions in the HEC-ResSim
model. One case involves a minimum flow rule at the "Cloverdale Diversion" junction just
downstream of the confluence with Big Sulphur Creek (Figure 42). Use of null routing for the
very short reach between the junctions for Big Sulphur Creek confluence and the Cloverdale
Diversion was considered acceptable. It was anticipated that there would be no noticeable affect
to Lake Mendocino releases by adding actual routing parameters.

The other case involved a maximum flow rule at Warm Springs Dam operating for Dry Creek
near the Geyserville junction. Accommodations to make the HEC-ResSim routing reaches
maintain similarity to the HEC-HMS model were not appropriate in this situation. Although
HEC-HMS was able to model Dry Creek from Lake Sonoma to the Russian River as a single
reach, the HEC-ResSim model used two reaches that properly represented the routing of Warm
Springs Dam outflows to Dry Creek near Geyserville. The HEC-RAS model provided Modified
Puls parameters for the two reaches in the HEC-ResSim model.

HEC-ResSim Alternatives
Initial States
The initial conditions for the reservoir simulation were defined in the Lookback tab of the HEC-
ResSim Alternative Editor (Figure 43). In addition to establishing the initial reservoir storage
and values for state variables, the lookback period provided a warm-up period for the reservoir
simulation. The data provided during the lookback period allowed the start of the simulation to
proceed with fully defined values in operational rules, such as rate-of-change or moving
averages. The lookback period must also cover the time required to route releases to the farthest
downstream control point. The longest such dependency among all of the HEC-ResSim model
alternatives in this study was 3.5-days, which was used for all the reservoir simulations.

For the HEC-WAT deterministic simulations, constant outflows of 100 cfs were used during the
lookback period for Lake Sonoma, Lake Mendocino, and the SCWA Diversion Dam. Diversions
flows were set to zero, and state variables were set to their default values. The lookback storages
for Lakes Mendocino and Sonoma were taken from preliminary simulation results.

Time-Series Boundary Conditions

The inputs for each interval of the reservoir simulation and downstream routing were defined in
the Time-Series tab of the HEC-ResSim Alternative Editor (Figure 43). The inputs fall into
four categories: inflows to the system, diversions from the system, external inputs for minimum
flow rules, and lookback data.

HEC-ResSim State Variables

The HEC-ResSim model computed the hydrologic index variable in a state variable named
"StorageState". The hydrologic index provided key information to determine instream flow
requirements affecting conservation operations for Lake Mendocino and Lake Sonoma.

87
Chapter 6 – Reservoir Simulation - Model & Data Development (Deterministic) PR-100

Figure 43 HEC-ResSim Alternative Editor - Tabs

Representation of Coyote Valley Dam

The HEC-ResSim model of Coyote Valley Dam includes an uncontrolled spillway, and an outlet
tunnel that may discharge through a gate or through the powerhouse (Chapter 2, Section 2.3.2).
The Redwood Valley Diversion occurs from Lake Mendocino. The Lake Mendocino elevation-
storage-area relationship was taken from the 2003 version of the Coyote Valley Dam Water
Control Manual, Exhibit A. Monthly evaporation losses from Lake Mendocino were provided
by SCWA.

Overview of Lake Mendocino Operations

Lake Mendocino operates according to a simple "rule curve", also known as guide curve in
HEC-ResSim, representing target pool elevations throughout the year. The rule curve separates
storage used for meeting water demands (i.e., the conservation zone below the rule curve) from
storage reserved for handling flood events (the flood control zone above the rule curve). The
Lake Mendocino rule curve is displayed in Figure 44 as the dashed purple line.

88
PR-100 Chapter 6 – Reservoir Simulation - Model & Data Development (Deterministic)

Figure 44 Lake Mendocino 2006 Event Observed versus Simulated

Reservoir levels above the rule curve, cause flood control operations to trigger, with the
objective of releasing water in order to evacuate storage and get back to the rule curve. The
outflows are subject to constraints, which typically result in reducing project releases and storing
water in the flood zone until local flows downstream begin to recede. Coyote Valley Dam flood
operations are subject to rate-of-change limitations, and include a rule against reservoir releases
that result in flows over 8,000 cfs at the Hopland gage. The release constraints become less of a
consideration for very large events that threaten to exceed the dam's storage capacity due to high
inflows and high incremental flows downstream.

Most of the time Lake Mendocino levels are at or below the rule curve and the dam operates in
"generating mode" with releases that meet downstream demands and generate incidental
hydropower. The historical Lake Mendocino elevations (shown as the dotted blue line Figure
44), illustrates a very common situation, where demands during the summer and fall of 2005
significantly drafted the reservoir, so that the level was far below the rule curve at the beginning
of the winter storm season. The lake then absorbed a very large storm on New Year's Day 2006,
rising to within three feet of the spillway before streamflow receded thus allowing project
releases to evacuate stored floodwater and return to the rule curve. The rule curve slopes upward
during the spring, allowing the reservoir to initiate refill as the threat of flooding typically
diminishes. The inflows in spring 2006 proved insufficient to fill the pool to the summer target
level, which is typical for Lake Mendocino (Figure 11).

Coyote Valley Dam Model Operation Sets

HEC-ResSim organizes reservoir-operating criteria into "operation sets", which allowed the
study to use the same reservoir model to compare different water management alternatives
simply by switching to a different operation. An overview of the Baseline operations set is
displayed in Figure 45. The HEC-ResSim Baseline operations set represents the Existing
Conditions alternative for the FIRO study.
89
Chapter 6 – Reservoir Simulation - Model & Data Development (Deterministic) PR-100

Figure 45 HEC-ResSim Reservoir Editor - Operations Set Definition

The left panel of the HEC-ResSim Reservoir Editor, from the Operations tab (Figure 45)
displays the storage zones described in the Coyote Valley Dam Water Control Manual, and the
operating rules that apply to each zone. Selecting a zone displays the definition for that zone
throughout the calendar year. For example, Figure 45 displays the top of the conservation zone
for Coyote Valley Dam. The rules for each zone were arranged in order of priority, with the
highest priority on top. The rules reflected maximum releases, minimum releases, or explicitly
specified releases.

These maximum flow rules generally came into play with Lake Mendocino in the flood control
zone. The main application in the conservation zone was the situation when the Russian River
rose enough to require reduced outflows. Similarly, the minimum flow rules typically only apply
to operations when the reservoir pool was in the Conservation zone, with the main exception
being the requirement to maintain at least 25 cfs outflow regardless of zone. Consequently, for
simplicity the same rule definitions were applied to both the Flood Control and Conservation
zones, even though some definitions were irrelevant at that reservoir level. Including the same
full set of flow rules in both the Conservation and Flood Control zones also helped simulate
times when the reservoir storage was very near to the guide curve, and could easily cross into the
other zone.
90
PR-100 Chapter 6 – Reservoir Simulation - Model & Data Development (Deterministic)

Coyote Valley Dam Model Minimum Flow Rules

There are two rules for the Coyote Valley Dam that represent instream flow requirements,
originating from SWRCB Decision 1610, and were modified by an SCWA TUCP. The upper
rule, Min25-Release (Figure 45), enforces that Coyote Valley Dam must always release 25 cfs.

The subsequent IF_Block, WSC I-1610 Q-TUCP (Figure 45), describes the more complex
downstream flow requirements. An expanded view of the IF_Block structure for WSC I-1610 Q-
TUCP is displayed in Figure 46. The first condition refers to the hydrologic index, which may
be "Normal", "Dry", or "Critical". If the index is "Normal", then a second condition applies,
which reflects the combined storage in Lake Mendocino on the Russian River and Lake Pillsbury
on the Eel River.

Figure 46 HEC-ResSim Model - WSC I-1610 Q-TUCP IF_Block Structure

Each case of the combined conditions specifies minimum flows that need to be maintained
throughout the upper Russian River. The requirements were expressed for four locations in
order to accommodate routing or releases, tributary inflows, and diversions. HEC-ResSim
resolves the different flow requirements at the locations by taking the maximum of the
minimums. The SCWA Reliability Report (SCWA, 2015) describes the flow requirements in
detail. Per recommendation of SCWA, these flow targets were further padded by varying
amounts to reflect "safety buffer" increments used in operations to overcome potential losses
along the river.

91
Chapter 6 – Reservoir Simulation - Model & Data Development (Deterministic) PR-100

Coyote Valley Dam Model Rate-of-Change Rules

The next highest priority rules reflect rate-of-change constraints (Figure 45), the increasing rate
of change (IROC) rule. The IROC_BIOP rule (Figure 45) represents operational requirements
from the biological opinion (BIOP). Releases below 1,000 cfs may increase by only 1,000
cfs/hour, while higher outflows may ramp-up by 2,000 cfs/hour.

Decreasing rate of change (DROC) IF_Block, DROC_April2016 (Figure 45), reflects an

agreement between CESPN and NMFS in April 2016, primarily intended to minimize fish
stranding. The requirement updates the previous rules established under the 2008 Russian River
BIOP, limiting the "ramp-down" of Coyote Valley Dam releases, based on the magnitude of the
discharge and time of year. Outflows greater than 4,000 cfs may be reduced by 1,000 cfs/hour,
but only by 250 cfs/hour when between 4,000 cfs and 2,500 cfs, and only by 100 cfs/hour for
flows less than 2,500 cfs. Releases less than 250 cfs must ramp down no faster than 25 cfs for
most of the year. Between 15 March and 15 May, releases below 250 cfs may also decline by no
more than fifty cfs/day, which was implemented as 2.083 cfs/hr in the HEC-ResSim model.

The initial HEC-ResSim model used DROC limits from the BIOP. Results were compared after
updating to the more restrictive 2016 values. The new DROC IF_Block (DROC_April2016)
often resulted in a higher outflow than the old rule for a short time after a storm. The 50 cfs/day
limit applies during the season for filling the reservoir to the summer pool. If there were no
further guide curve or flood operations that spring, then the lower pool elevation would be
carried through the entire dry season.

Plotting the Coyote Valley Dam pool levels and outflows using DROC_April2016 in red and the
prior rule in green (Figure 47), provides a graphical illustration of the differences. For reference,
the inflows are shown in grey and the guide curve in purple (Figure 47). The reservoir levels
were the same while inflows from 12 through 16 March raised the lake to the guide curve and
triggered a 440 cfs release. The inflows quickly diminished and the HEC-ResSim model
resumed the minimum 25 cfs outflow as soon as the DROC rule allowed. This event turned out
to be the last storm of the year, and the slower ramp-down under the DROC_April2016
(IF_Block) allowed the lake to fill about seven inches less than under the prior DROC rule. In a
few rare situations, the different storage results affected the timing for crossing thresholds for
minimum flow rules. In one case, the DROC_April2016 (IF_Block) caused a different
hydrologic index to apply for all the minimum flow evaluations of the year.

Coyote Valley Dam Model Maximum Flow Rules

The remaining rules in the HEC-ResSim model limit the maximum outflows from the Coyote
Valley Dam (Figure 45). The lesser of the maximum flows from the different rules determines
the highest release. The IF_Block, RR_Ukiah_Flow, represents a special approach developed for
the Russian River HEC-ResSim model to replace two existing rules specified in the Coyote
Valley Dam Water Control Manual:

• 8,000 cfs maximum allowable flow at Hopland USGS gage (USGS 11462500)
• 25 cfs maximum allowable release when the Russian River at Ukiah USGS gage (USGS
11461000), located on the West Fork of the Russian River, is above 2,500 cfs
92
PR-100 Chapter 6 – Reservoir Simulation - Model & Data Development (Deterministic)

Figure 47 Reduced Ability to Fill Pool using IF_Block DROC_April2016

The RR_Ukiah_Flow IF_Block contains rules that were necessary for evaluating the impact of
forecast information on flood control operations. The standard HEC-ResSim rule for maximum
releases computes the outflow based on a downstream limit that performs an approximate
routing from the reservoir to the control point, and incorporates intervening local flows. Such
assumed knowledge of the near future conditions downstream may be appropriate when
simulating release decisions, where experienced operators might consider a variety of
information, including flows at the control point forecasted by the NWS. However, for this
study to explore the impact on release decisions of perfect forecast information for the
downstream local flows, the modeling team attempted a more detailed representation of the
actual decision logic typically employed by the operator for Lake Mendocino.

In practice, the operator did not determine flood operation outflows by routing candidate releases
from Coyote Valley Dam to Hopland, and did not know the local flow hydrographs. The
operator made judgments based on experience and awareness of basin conditions, primarily
considering the level of Lake Mendocino and the flow observed at the nearby gage on the West
Fork Russian River. The RR_Ukiah_Flow IF_Block models the decision process according to
three different situations, where outflows were determined as a function of West Fork flows:

(1) Flow rising on the West Fork gage. This situation is the first condition
(RR_Ukiah_Flow IF_Block) and determined how to reduce releases in order to protect
downstream locations. Preliminary modeling indicated that the original West Fork rule
typically controlled in this situation, instead of the Hopland rule, and the operator
concurred that West Fork flows were normally the key information for deciding whether
to reduce outflows.

93
Chapter 6 – Reservoir Simulation - Model & Data Development (Deterministic) PR-100

(2) Flow declining on the West Fork gage and the reservoir stage is above 755 feet. This
situation is the second condition (RR_Ukiah_Flow IF_Block) where the reservoir stage is
above 755 feet and governs the emptying of the flood pool when the reservoir was high,
and operators felt more concern with the reservoir level than the possibility of exceeding
8,000 cfs at Hopland.

(3) Flow declining on the West Fork gage and the reservoir stage is below 755 feet. This
situation is the third condition (RR_Ukiah_Flow IF_Block) where the reservoir stage is
below 755 feet and governs the emptying of the flood pool when the reservoir was not
high, and greater consideration was given to the Hopland flow.

Outflows were determined according to relationships using the West Fork Russian River flow,
developed in coordination with SCWA. Review with operators confirmed that this substitute
rule acceptably captured the actual operations for the rising and falling limbs of Russian River
flows, without using forecast information.

The rule, MaxReleaseWCM-FC (Figure 45), sets maximum flows at 2,000 cfs, 4,000 cfs, or
6,400 cfs, based on elevation, according to typical operational practices. The 4,000 cfs and
6,400 cfs thresholds represented limits in the Water Control Diagram (Appendix A) for Flood
Control Schedule 2 and Flood Control Schedule 3. For Flood Control Schedule 1, the current
water control manual allows greater operator discretion, which posed a modeling challenge. The
threshold (2,000 cfs) in the MaxReleaseWCM-FC rule represents what was actually specified in
the prior version of the Water Control Diagram. Releases around 2,000 cfs still occur often in
practice, and correspond to the approximate transition point between power generation mode and
flood operations. Operators sometimes maintain 2,000 cfs instead of going to 4,000 cfs because
of ramping rate constraints on the ability to reduce flows quickly later.

Other Coyote Valley Dam Model Rules

The highest-priority rule in the Conservation and Flood Control zones was the specified flow for
the Redwood Valley diversion out of the lake (RVWD Full Diversion). The flow values were
provided in an external time-series as described in Table 20.

The transition operations between power generating mode and flood control operations were
modeled using the Limit Rel thru Pwr Plant rule. This rule ensured that the HEC-ResSim model
simulated using the correct total outflow capacity for all outlets of Coyote Valley Dam,
regardless of whether using the gate and/or powerhouse. Limit Rel thru Pwr Plant zeroes out the
power plant capacity at 755 feet, so that the total controlled outlet capacity reflects only the gate
rating (only applied to the Flood Control and Emergency zones).

The rule, MaxReleaseFlood_Gates (Figure 45), specified releases from Lake Mendocino through
the gate until the pool exceeded 773 feet, per the emergency release schedule of the Water
Control Diagram (applies only in the Emergency Zone).

The Dummy_Pillsbury TS rule allowed the HEC-ResSim model to ingest the Lake Pillsbury
storage as an external time-series DSS file. The Pillsbury storage was used by the StorageState
state variable. Dummy_Pillsbury TS called for a minimum release of zero cfs, so it did not affect
94
PR-100 Chapter 6 – Reservoir Simulation - Model & Data Development (Deterministic)

releases. The rule was applied outside the Conservation and Flood Control zones to avoid
unnecessary evaluations during simulations.

Representation of Warm Springs Dam

The HEC-ResSim model of Warm Springs Dam includes an uncontrolled spillway, a power
plant, and a control structure with a pair of flood control gates. The power plant and gates may
be operated concurrently. There are no diversions from the pool, but a hatchery operates
immediately below the dam. The Lake Sonoma elevation-storage-area relationship was taken
from the Warm Springs Dam Water Control Manual, Exhibit A. Monthly evaporation losses
from Lake Sonoma were provided by SCWA.

Overview of Lake Sonoma Operations

Lake Sonoma operates according to a guide curve, similar to Lake Mendocino, with rules
defined for Conservation, Flood Control, and Emergency storage zones. The operations set
supports instream flow requirements similar to the ones applicable to Lake Mendocino. Warm
Spring Dam flood operations also resemble those of Coyote Valley Dam, in that the operations
set reduces outflows to protect downstream locations until the river recedes.

The HEC-ResSim representation of Lake Sonoma operations was simpler than for Lake
Mendocino, which in turn made the operations at Warm Springs Dam simpler. In addition, Lake
Sonoma has a greater ability to store floodwaters, which lessens the urgency to clear the flood
storage by "backfilling" receding flows downstream. The water supply operations of Lake
Sonoma also face less stress in comparison to Lake Mendocino. Operations at Warm Springs
Dam were modeled in the same manner as for each of the water management alternatives
evaluated at Coyote Valley Dam. An operations set was defined for the Warm Springs Dam
representing the Existing Conditions alternative, and was named Baseline.

Warm Springs Dam Model Rules

The guide curve for Lake Sonoma remained constant at 451.1 feet NGVD throughout the year.
The highest priority rule (Min70-Base Hatchery) in the Conservation and Flood Control zones
was the minimum 70 cfs required for the fish hatchery. Similar to Lake Mendocino, other
minimum flows required at different downstream locations per SWRCB Decision 1610 were
specified according to the hydrologic index. IROC and DROC constraints were also imposed on
releases from Warm Springs Dam by the BIOP – IROC_BIOP and DROC_BIOP (Figure 48).

Lake Sonoma operations included maximum flow rules for two downstream locations, each
resembling the Hopland Rule used at Lake Mendocino. One rule (MaxatGeyserville) ensures
that Dry Creek near Geyserville does not exceed 7,000 cfs. The other rule (MaxatGuernevilee
Gage) requires releases to avoid contributing to flows at the Guerneville Gage above 35,000 cfs.

The IF_Block Rising_5000 (Figure 48), contains the logic governing the releases to be made, in
accordance with Flood Control Schedules 1 through 3 of the Water Control Diagram (Appendix
A). Rising_5000 set maximum outflows based on the reservoir level, unless the reservoir
inflows were over 5,000 cfs and rising. In this case, the outflows were limited to 100 cfs, in
95
Chapter 6 – Reservoir Simulation - Model & Data Development (Deterministic) PR-100

Figure 48 HEC-ResSim Reservoir Editor -Warm Springs Dam Baseline Operations Set

order to reduce the hazard to an emergency water supply line. The least of these conditions
controlled the highest allowable total outflow.

The lone rule in the Emergency zone, MaxReleaseFlood_Gates, (Figure 48) described the
emergency release schedule, which specified how to operate the gates in conjunction with flow
over the spillway.

HEC-ResSim Simulation Time-Step

The reservoir simulation interval in HEC-ResSim established the amount of time between each
release decision. The outflows determined for each time-step were averages for each period, and
the model reported reservoir storages for the end of the periods. HEC-ResSim averaged the one-
hour HEC-HMS inputs to the reservoir simulation time-step, prior to performing routing
calculations. The daily or monthly diversion flows were also converted to the reservoir
simulation time-step before combining with other flows.

Provisional analyses compared results from one-, two-, three-, four-, and six-hour simulation
time steps. The choice of reservoir simulation interval proved complicated, due to the interplay
of time step, routing methods and parameters, reservoir maximum or minimum outflow rules
96
PR-100 Chapter 6 – Reservoir Simulation - Model & Data Development (Deterministic)

based on downstream control, and rate of change rules. The flow values played out differently
across the thresholds used in rules, and when evaluating rising versus falling criteria. This
typically caused small differences in outflows and reservoir storage among the different time-
steps, which persisted until the model tracked back into guide curve operations. Depending on
the timing of when thresholds were crossed, differences of up to six inches were observed. In
cases where no further guide curve or flood control operations occurred, differences in lake
levels among the results for the various reservoir simulation intervals carried on through the
summer.

After careful consideration, the modelers chose a six-hour reservoir simulation interval. The six-
hour time-step offered much faster computation speeds, which was of great value to the iterative
alternative formulation process. Operational advisors also considered a six-hour simulation
time-step as a reasonable decision interval for Lake Mendocino water management operations,
and noted that forecast information from the CNRFC was typically updated every six hours.
Detailed analyses of the provisional simulations showed that the finer intervals required more
computation time but did not provide substantially superior results. Timing differences both
slowed and hurried the crossing of thresholds and triggering rules among each of the intervals
modeled, without an obvious bias. The ability of simulations with shorter intervals to more
quickly activate or deactivate rules was not generally considered a realistic improvement for
modeling actual operations, except for certain rate-of-change situations.

Validation
Reservoir operations modeled under the Existing Conditions alternative were validated by
reviewing simulated release decisions for Coyote Valley Dam and Warm Springs Dam for
compliance with current operating requirements. The sixty-year analysis period was reviewed,
with special attention to the flood events of 1955, 1964, 1978, 1983, 1986, 1995, 1997, 1998, and
2006, with simulation results being compared to observed data whenever appropriate.

Flood Operations
A typical flood operations simulated at Coyote Valley for the Existing Conditions (i.e., Baseline)
alternative is provided in Figure 49. Prior to 8 January, the pool was below the guide curve and
the reservoir was releasing only the minimum 25 cfs.

Inflows from the first storm drove the pool above guide curve on 8 January, but outflows
remained at 25 cfs due to the maximum rule for rising river conditions at West Fork. Early, on 9
January, the West Fork gage was falling, and the pool remained below 755 feet, so the Coyote
Valley Dam release was based on West Fork flows according to the relationship specified in the
Coyote Valley Dam Model Maximum Flow Rules. According to that relationship, outflows
remained limited to 25 cfs flow while the West Fork flows exceeded 2,500 cfs. On 10 January,
the West Fork flows fell below 2,500 cfs, and the relationship allowed releases ranging between
1,000 to 2,000 cfs, until another storm occurred on 12 January. The reservoir attempted to limit
outflows to 25 cfs according to the rule for rising flows at Ukiah, but required a day to ramp
down because the DROC requirement which limited the change to 100 cfs/hour (i.e., 600 cfs per
simulation time-step).
97
Chapter 6 – Reservoir Simulation - Model & Data Development (Deterministic) PR-100

Figure 49 Simulated Flood Operations during the January 1995 Event

The storm ended and West Fork flows began receding on 14 January, and at this point the pool
exceeded 755 feet. Therefore, the Coyote Valle Dam release was based on West Fork flows
according to the relationship specified in the maximum flow rules for the conditions of declining
flow at the West Fork gage with a reservoir stage above 755 feet. Once the reservoir fell below
755 feet, the RR_Ukiah_Flow IF_Block no longer applied because the 4,000 cfs rule for Flood
Control Schedule 2 became the lesser of the maximum flow limits. When the pool fell below
746 feet, Flood Control Schedule 1 became the control, and limited the outflows to 2,000 cfs.

None of the historical validation events exercised the emergency release schedule. Per the
recommendation of a CESPK water manager, the modeling team performed an event simulation
with the 1964 flood shifted to May, assuming a full pool. This example caused Lake Mendocino
to reach a peak elevation of 773.25 feet NGVD, and demonstrated that the model released the
expected increments of flow through the gates while the pool remained above elevation 771 feet.

Conservation Operations
Conservation releases were generally determined according to the highest flows of the minimum
flow rules. HEC-ResSim iteratively performed linear routing of candidate releases to evaluate
which of the downstream minimums controlled. The calculations included local flows from
downstream tributaries and diversions out of the river represented at Hopland and Cloverdale. A
typical series of release decisions while the reservoir remained in the conservation pool (i.e.,
below guide curve), and the hydrologic index was "normal" is displayed in Figure 50.

SWRCB Decision 1610 requires that the Coyote Valley Dam releases ensure 185 cfs throughout
the Upper Russian River during the month of April, and the model rules included an additional
20 cfs buffer for locations from Hopland through Healdsburg, making the effective minimum
205 cfs. From 28 through 29 April 2008, the local flows at downstream locations slightly
98
PR-100 Chapter 6 – Reservoir Simulation - Model & Data Development (Deterministic)

Figure 50 Simulated Conservation Operations during May 2008

exceeded the diversion amounts, and the reservoir needed to release around 203 cfs to maintain
the desired flow at Hopland. The releases increased slightly to compensate for diminishing local
flows.

On 1 May, diversion amounts increased, and routing effects briefly caused the Cloverdale
location to control for minimum flows. In addition, on 1 May, the BIOP minimum flows
between Hopland and Healdsburg changed to 125 cfs, with a 9 cfs buffer, for an effective
requirement of 134 cfs. The releases ramped down by about 12 cfs per time-step, in compliance
with the DROC rule. The minimum flow at the confluence with the West Fork Russian River
controlled briefly on 2 May as the previous flows routed through the system, then was controlled
at the Hopland location.

Existing Conditions versus Observed

The Existing Conditions (i.e., Baseline) alternative represented current operations and assumed
that current operations persisted during the whole analysis period (i.e., the future condition is
equivalent to the current condition). The primary differences from historical operations,
especially the early years of Lake Mendocino, involve:

1) The Existing Conditions alternative assumed the existence of current Potter Valley
Project flows. The flows diverted from the Eel River are greatly diminished from
historical amounts. The Potter Valley Project flows represent much of the summer
supply to Lake Mendocino and allow higher flows along the Russian River. The Existing
Conditions simulations in the FIRO analysis reflect a drier system than what occurred in
the observed record.

2) The Existing Conditions alternative reflects higher agriculture and municipal demands,
and requirements for instream environmental flows. Reservoir storage depletions
simulated (Baseline) often exceed historical amounts.

3) Reservoir operating practices of the past differ from the Existing Conditions simulation
rules. Some standard practices have evolved, such as earlier reservoir filling dates in
recent years. In other situations, the historical operations differ from the expected plans
99
Chapter 6 – Reservoir Simulation - Model & Data Development (Deterministic) PR-100

for unknown reasons. The Existing Conditions simulation does not have access to all the
information available to the historical human operators, and does not attempt to represent
the judgment calls that occur throughout the historical record.

Consequently, the Existing Conditions alternative did not intend to reproduce historical events.
Simulations of familiar major floods exhibit differences from observed values. For example, in
Figure 44, the New Year's Day flood of 2006 results in spillway flow in the Existing Conditions
simulation, because the simulated levels were much closer to the rule curve at the start of the
flood event, providing less flood control storage than the actual operators reserved. The superior
performance of the human operators might have been due to judgment or simply the product of
unique circumstances during the weeks prior to the event. The HEC-ResSim model does not
attempt to address either explanation, and simply follows the rules using standard assumptions.
This normalization of the Existing Conditions (Baseline) alternative is necessary to permit
relative comparisons with the results from simulations of modified operations.

Besides the operational differences, comparisons between Existing Conditions alternative results
and historical observations were also difficult because the reservoir inflows for the simulation
were not the historical inflows. The modeling team worked hard to ensure that the basin flows
computed with HEC-HMS resembled the available observed values for all the validation events,
but differences were unavoidable.

The most challenging event for comparing Existing Conditions versus observed results was the
1986 flood. The HEC-HMS model parameters were adjusted as much as reasonable to produce
simulated inflows to Lake Mendocino that resembled historical inflows, but limited precipitation
data available for the area unavoidably resulted in much higher inflows than observed for the
1986-flood event. The reservoir pool (Figure 50) for both the Existing Conditions simulation
and the historical data entered the event on guide curve, and returned to guide curve after the
event finished. The greater area under the simulated outflow curve (Figure 51) than under the
observed releases represents the additional amount of inflow given to the HEC-ResSim model.
The simulated and observed outflows both shut down from 14 through 16 February. The inflow
volumes until that time were relatively close, and at the end of 16 February, the simulated levels
were only two feet higher than observed.

The simulated inflows peaked on 17 February, and the model continued to minimize releases
even as the pool reached the spillway and began uncontrolled flows. Historical inflows
apparently began receding sooner than the simulated values, because the operators were able to
start making Flood Control Schedule 2 releases sometime on 17 February. The historical pool
level remained well below the spillway, while the greater volume and later recession of the
model inflows resulted in four days of spillway flow. In fact, the Existing Conditions 1986 peak
level for Lake Mendocino exceeded the simulated 1964 peak.

After substantial analysis and discussion, the modeling team agreed to move forward with the
Existing Conditions results despite the differences between the 1986 simulated results and
historical observations. The only option to achieve better agreement was to modify the
precipitation data to achieve lower runoff volumes. This option was considered unnecessary for
the relative comparison of alternatives, which applied the same inflows to each water
management strategy.

100
PR-100 Chapter 6 – Reservoir Simulation - Model & Data Development (Deterministic)

Figure 51 Validation of 1986 Event

101
Chapter 6 – Reservoir Simulation - Model & Data Development (Deterministic) PR-100

102
PR-100 Chapter 7 – Hydraulic - Model & Data Development (Deterministic)

Hydraulic
Model and Data Development
(Deterministic)
The FIRO Existing Conditions alternative is based on models that were developed for previous
studies. The individual models were updated and validated against observed data, with
subsequent refinements after preliminary integration into HEC-WAT and the deterministic
simulations. In the course of formulating the FIRO alternatives, further modifications were
made to the HEC-HMS and HEC-ResSim existing condition model alternatives to support
exchange of forecast information. Chapters 5 through 8 provide information on the data and
models (e.g., HEC-HMS, HEC-ResSim, HEC-RAS, HEC-FIA) development for the HEC-WAT
deterministic simulations. Chapter 9 provides information on what other data and model
development was needs for the HEC-WAT FRA simulations.

This chapter details the hydraulic (HEC-RAS) data and model development for the Existing
Conditions alternative and the HEC-WAT deterministic simulations.

Hydraulic Data Development

Channel and floodplain information needed by the HEC-RAS model was extracted from a ten-
meter DEM and aerial imagery data using HEC-GeoRAS. Additional data not extracted from
GIS data was defined in the HEC-RAS model, such as detailed channel data and the dimensions
of bridge openings. The geometric data generated by HEC-GeoRAS was imported into HEC-
RAS. This data included the river system schematic for the Russian River downstream of
Coyote Valley Dam and, Dry Creek downstream of Warm Springs Dam, station-elevation points
for cross-sections, reach lengths between cross-sections, ineffective flow areas in the floodplain,
the location of bridges, station-elevation information for bridge decks, location and station-
elevation information for lateral and inline structures, and elevation-storage information for
storage areas. Additional data were entered into the HEC-RAS model based on information
gathered during field visits as well as standard engineering practices to estimate model
parameters.

Cross-Section Elevation Data

The cross-sections extracted from the DEM data did not contain information below the water
surface. This lack of information affects both the river channel geometry and the computation of
the water surface profiles. The channel flow area is very important to the smaller events;
however, for larger events the channel flow area may not be as important. In order to get an
HEC-RAS unsteady-flow model up and running smoothly, it is imperative to have a reasonable
definition of the main channel in the HEC-RAS model. Additionally, in order to calibrate the
103
Chapter 7 – Hydraulic - Model & Data Development (Deterministic) PR-100

HEC-RAS model to historic events, the main channel must be incorporated into the geometry.
The HEC-RAS Channel Design/Modification tool was used to modify the cross-sections to
include the channel below the water surface - a low-flow channel was cut into each cross-section.
An example is displayed in Figure 52, for a cross-section where the main channel was added to
the cross section. A trapezoidal channel shape (Figure 52) was cut into the original cross-section
developed from the DEM data. Channel depth and width estimates were based on observations
made in the field. Additionally, channel invert elevations were obtained from FEMA (Federal
Emergency Management Agency) studies that were done for both Mendocino and Sonoma
Counties. These invert elevations were used to set the bottom elevations of the trapezoidal
channel cuts. The low-flow channel was added to improve model stability by ensuring that there
was water in the channel during lower flows, like those at the beginning of a simulation. Adding
the low-flow channel will have little effect on peak stages during very high flow events but will
affect low-flow simulation results.

Figure 52 HEC-RAS - Example of Cutting a Low-Flow Channel into a Cross-Section

Manning's n Values
Manning's n values for the main channel were estimated from site visits and were estimated for
the channel as a whole, rather than separate values for the channel bottom and the channel banks.
A large portion of the river corridor has extensive thick brush and trees on the banks of the main
channel. Channel Manning's n values were estimated by weighting the base n value of the
bottom with larger n values for the channel banks. Overbank Manning's n values were estimated
using aerial images to define areas of similar land use, then assigning an n value for that land use
104
PR-100 Chapter 7 – Hydraulic - Model & Data Development (Deterministic)

type. For example, thick forested areas were assigned an n value of 0.12; urban areas with a high
density of buildings were assigned a value of 0.15; fields with rows of nut trees were assigned a
value of 0.06; open fields with high grass was assigned values of 0.05; etc. After all of the cross
sections were assigned initial Manning's n value estimates, further refinement of the Manning's n
values were made during the HEC-RAS model calibration process.

Bridges and Culverts

The Russian River has eighteen bridges below Coyote Valley Dam and one on Dry Creek below
Warm Springs Dam. Bridge locations and bridge deck information was extracted from the
terrain data imported into the HEC-RAS model. Additional information, such as the dimensions
of bridge openings, roadway elevations, bridge railings, bridge abutments, and piers, were
entered into the HEC-RAS model directly. For each of the bridges, detailed bridge plans were
desired, but HEC was unable to obtain plans for the FIRO study. As a result, approximate bridge
dimensions were estimated from site visits to each bridge. A typical site visit for a bridge
consisted of the following: measuring the length of the bridge and the bridge opening;
measuring the height of the bridge deck and any railings; measuring the width of the piers and
pier spacing inside of the bridge opening; measuring the height from either the water surface or
the ground to the elevation of the low chord of the bridge deck inside of the bridge opening;
estimating the depth of the water; and taking photos of the bridge. This information was entered
into the HEC-RAS model as approximate bridge information. Elevations for the top of road and
low chord were estimated using site visit data by taking the height from the top of the water
surface/ground to the low chord of the bridge deck, then adding the elevations from the terrain
model at the bridge location. Each of the bridge modeled on the Russian River along with the
bridge's location by river mile (RM), are detailed in Table 21.

Storage Areas
Storage areas were used to model the available storage in smaller tributaries that would incur
backwater from the Russian River. The lateral structure feature in the HEC-RAS model was
used to connect the main channel to a storage area. Simulated water surface elevations in storage
areas will result in a horizontal water surface. The location and station elevation data of the top
of the lateral structures were extracted by GIS from the DEM data.

Hydraulic Model Development

The original purpose of the hydraulic model was to evaluate the downstream impacts of potential
dam failure scenarios for the Coyote Valley Dam. The HEC-RAS model was used to simulate
Coyote Valley Dam failure scenarios and route the resulting breach hydrographs to the Pacific
Ocean. Because this model was created for large flood flows, the model was expected to perform
well for low-frequency flood events. Steady-flow calibration of the one–dimensional model was
performed using observed high water mark information at gaged locations for four flood events.
Further, two historic flood event simulations were used to verify the model's performance.
Output from the HEC-RAS simulations was used to develop flood inundations maps for different
scenarios.

105
Chapter 7 – Hydraulic - Model & Data Development (Deterministic) PR-100

Table 21 Russian River Bridges by Location used in the HEC-RAS Model

Location by
Bridge Description River Mile
Vichy Springs Rd Bridge. 96.07
Talmage Rd Bridge 94.48
Hopland HWY 175 Bridge 79.72
Hwy 101 Upper Bridge 78.44
Hwy 101 Lower Bridge 68.96
Cloverdale Bridge 64.63
Hwy 128, Geyserville Bridge 54.40
Alexander Valley Road Bridge 48.47
Healdsburg Railroad Bridge 33.80
Healdsburg Ave. Bridge 33.72
Hwy 101 at Healdsburg 33.27
Wohler Rd Bridge 24.825
Hacienda Bridge - River Road 21.31
Plaza Rd Hwy 116 at Guerneville 15.20
Older Guerneville bridge 15.155
Bohemian Hwy Bridge 9.97
Moscow Rd Bridge 5.88
Hwy 1 Bridge 1.75

Boundary Conditions
The HEC-RAS model relied on a normal depth assumption at the outlet to the Pacific Ocean
using a friction slope of 0.0003 feet, with tidal effects not being included. Flows into the most
upstream cross-section on Dry Creek came from the Warm Springs Dam, using observed
outflows for calibration runs or HEC-ResSim releases for the HEC-WAT simulations. Flows
into the most upstream cross-section on the Russian River came from Coyote Valley Dam using
observed outflows for calibration runs. HEC-ResSim releases were then used for the HEC-WAT
simulations.

The other boundary condition hydrographs came from the HEC-HMS model (Figure 53), with
HEC-HMS inputs to the HEC-RAS model being flows from subbasin elements, except for the
inflow to the SantaRosaCr storage area. HEC-HMS combined the contributions from the three
subbasins (Santa Rosa CR10, Santa Rosa CR 20, Santa Rosa – Laguna) and routed them to the
Russian River.

Model Calibration
Observed peak stage and flow data were obtained for four stream gages on the Russian River,
from which four events were used to calibrate the HEC-RAS model - February 1986, January
1995, January 1997, and January 2006. In addition to the gages, high water marks for all four
events were obtained from the Rio Theatre building just upstream of the Bohemian Bridge
crossing in the town of Monte Rio (Figure 54).

A steady flow plan was created to perform the initial calibration of the HEC-RAS model to the
observed data. Peak flows and observed stages were entered into the model at all gage locations
106
PR-100 Chapter 7 – Hydraulic - Model & Data Development (Deterministic)

Figure 53 HEC-RAS Boundary Conditions

Figure 54 Observed High Water Marks on the Theater at Monte Rio

107
Chapter 7 – Hydraulic - Model & Data Development (Deterministic) PR-100

along with the Monte Rio high water marks. The model was run and Manning's n values were
adjusted to improve the fit of the computed water surface elevations to the observed data.

Because the majority of the main channel was estimated from site visits and entered into the
model as a trapezoidal channel, drastic changes to the Manning's n values were not made to get
exact matches of gaged data. Therefore, it is assumed that most of the error between the
computed and observed water surface elevations are due to the lack of detailed channel cross
section data. Results for HEC-RAS steady-flow simulations are presented in Table 22 (these
results are based on final geometry selected after refinement for unsteady-flow simulations).
Profile plots for the calibration events covering the upper Russian River are detailed in Figure
55, and profile plots for the lower Russian River are detailed in Figure 56.

Table 22 Water Surface Elevation Results from Steady-Flow Simulation on the Russian River
1986 1995 1997 2006
Gage Location Obs. Comp. Obs. Comp. Obs. Comp. Obs. Comp.
Hopland 523.5 526.4 522.9 525.3 519.4 523.2 527.3 527.0
Cloverdale 384.3 384.9 384.0 384.5 380.3 380.9 386.9 388.0
Healdsburg 105.6 105.5 106.0 105.8 104.4 104.4 103.0 103.5
Guerneville 71.5 71.6 70.9 70.2 67.9 67.9 67.2 68.7
Monte Rio *
42.9 43.0 41.4 41.5 39.1 39.2 39.9 40.0
*This location is not a gage. Water surface elevations are from high water marks.

Figure 55 Steady-Flow Water Surface Profiles with Observed High Water Mark Data, Upper Half of
Russian River

108
PR-100 Chapter 7 – Hydraulic - Model & Data Development (Deterministic)

Figure 56 Steady-Flow Water Surface Profiles with Observed High Water Mark Data, Lower Half of
Russian River

In addition to the observed data at gages and high water marks, additional information for the
extent of flooding at two points for the 2006 event were obtained. One of the flooding location's
extent (approximately RM 90.5) was based on talking to a property owner whose land was
flooded during the 2006 event. The computed flood boundary (Figure 57) generated from the
HEC-RAS model, along with a pink marker in Figure 57, displays the observed flooding extent

Figure 57 HEC-RAS Computed Flood Extent for the 2006 Event Compared to Observed Marker
109
Chapter 7 – Hydraulic - Model & Data Development (Deterministic) PR-100

from the discussion with the property owner. A second location (Figure 58) was indicated at the
intersection of Eastside Road and Wohler Road (approximately RM 24.3), displayed in Figure 58
as a pink diamond.

Figure 58 HEC-RAS Computed Flood Extent for the 2006 Event Compared to Observed Marker at
Eastside Road and Wohler Road

After the initial HEC-RAS model calibration was performed using a steady flow assumption, the
1997 and 2006 events were also run using an unsteady flow assumption. The flow data for these
runs were developed using the previously described HEC-HMS model. Therefore, the HEC-
RAS model was subject to flow errors based on the results of the HEC-HMS model. Hourly
stage data was not available for these unsteady flow runs (just peak stages). Model comparisons
were limited to observed and computed flows. For the most part, agreement between the
computed and observed values was acceptable. In the lower portion of the basin, the computed
values were lower than the observed, which is partly due to the computed flows from the
hydrologic (HEC-HMS) model being lower than the observed values. This discrepancy is
believed to be due to the poor quality of the precipitation data for the lower half of the
watershed.

Examples of some of the observed versus computed hydrographs for various gages are shown in
Figures 59 through 62. As shown in Figures 59 through 62, the hydrologic (HEC-HMS) model
flow compares well to the observed data for the upper half of the watershed (Hopland and
Cloverdale Gages), but begins to diverge from the observed data near the Healdsburg gage. The
largest errors are shown at the lower end of the HEC-RAS model at the Guerneville gage. The
computed hydrographs do not contain enough volume of water when compared against the
observed hydrographs.

110
PR-100 Chapter 7 – Hydraulic - Model & Data Development (Deterministic)

Figure 59 Observed versus Computed Flow at Hopland Gage, 2006 Event

Figure 60 Observed versus Computed Flow at Cloverdale Gage, 2006 Event

Minor adjustments to the Manning's n values were made from the results of the unsteady flow
runs for the 1997 and 2006 events. Major changes to the Manning's n values were not justified
as most of the error can be attributed to the flow estimates and not the Manning's n values. Final
calibrated Manning's n values for the main channel ranged from a high of 0.055 (in the steep
canyon areas) and a low of 0.03 (at the lower end of the model). Typical values for most of the
111
Chapter 7 – Hydraulic - Model & Data Development (Deterministic) PR-100

Figure 61 Observed versus Computed Flow at Healdsburg Gage, 2006 Event

Figure 62 Observed versus Computed Flow at Guerneville Gage, 2006 Event

channel were in the 0.04 to 0.05 range. Overbank Manning's n values ranged from 0.035 (for
open areas with no vegetation) to 0.15 (in the high-density urban areas). Computed water
surface profiles for the 2006 event, with diamonds shown for observed data at the gages, are
shown in Figures 63 and 64.

112
PR-100 Chapter 7 – Hydraulic - Model & Data Development (Deterministic)

Figure 63 Computed Water Surface Profile with Observed Data, Upper Half of the Russian River,
2006 Event

Figure 64 Computed Water Surface Profile with Observed Data, Lower Half of the Russian River,
2006 Event.

113
Chapter 7 – Hydraulic - Model & Data Development (Deterministic) PR-100

HEC-RAS Outputs Used in HEC-WAT

Deterministic Simulations
The results of the HEC-RAS model informed stage-frequency and flow-frequency analyses, and
provided hydraulic data for computing flood damage to structures and potential life loss. The
HEC-RAS Unsteady Flow Analysis dialog box (Figure 65), displays the settings that were used
for the HEC-RAS, unsteady flow analysis. The Hydrograph Output Interval (Figure 65) was
set to hourly, for consistency with the HEC-HMS model. The Mapping Output Interval
(Figure 65) was set to "1 Hour", allowing the RAS Mapper Tool of HEC-RAS to produce arrival
time grids suitable for subsequent life-loss calculations in HEC-FIA. The Detailed Output
Interval (Figure 65) was set to "1 Day".

Figure 65 HEC-RAS Unsteady Flow Dialog Box – Unsteady Flow Analysis Settings

114
PR-100 Chapter 7 – Hydraulic - Model & Data Development (Deterministic)

Time-Series
The HEC-RAS model specified results for flow and stage hydrographs for thirteen cross-section
locations for the deterministic simulations. These locations represented gages, calibration points,
and confluences, providing input to the HEC-FIA model, and output for processing regarding
study metrics. These settings were established using the HEC-RAS Stage and Flow Output
Locations Selector (Figure 66). The output was stored as HEC-DSS records in the appropriate
output directory within the HEC-WAT study.

Figure 66 HEC-RAS Stage and Flow Output Locations Selector

The HEC–RAS model simulation also automatically wrote hydrograph output at numerous other
cross-sections and storage areas. This information allowed modelers to review the HEC-RAS
model simulation and results to verify model performance or investigate interesting results.

Gridded
The RAS Mapper Tool (HEC-RAS) produced gridded output needed as inputs to the HEC-FIA
model, and for visual analysis and output maps. As shown in Figure 67, the output consisted of
stored maps containing depth, product of depth and velocity, and arrival time.
115
Chapter 7 – Hydraulic - Model & Data Development (Deterministic) PR-100

Figure 67 RAS Mapper Results Maps

The depth and depth times velocity grids were configured to reflect the maximum values for
each cell during the event. The arrival time grid was configured to represent the time (in hours
since the start of the simulation) that was required for each cell to reach a depth of two feet.

116
PR-100 Chapter 8 – Consequence - Model & Data Development (Deterministic)

Consequence
Model and Data Development
(Deterministic)
The FIRO Existing Conditions alternative is based on models that were developed for previous
studies. The individual models were updated and validated against observed data, with
subsequent refinements after preliminary integration into HEC-WAT and the deterministic
simulations. In the course of formulating the FIRO alternatives, further modifications were
made to the HEC-HMS and HEC-ResSim existing condition model alternatives to support
exchange of forecast information. Chapters 5 through 9 provide information on the data and
models (e.g., HEC-HMS, HEC-ResSim, HEC-RAS, HEC-FIA) development for the HEC-WAT
deterministic simulations. Chapter 10 provides information on what other data and model
development was needs for the HEC-WAT FRA simulations.

This chapter details the consequence (HEC-FIA) data and model development for the Existing
Conditions alternative and the HEC-WAT deterministic simulations.

Consequence Model Development

Differences in flood damage estimates provide a more comprehensive measure of relative flood
risk than comparing stage-frequency relationships at key locations. Flood damage differs across
the watershed, and the vulnerability of different areas may change for different scales of events.
HEC-FIA offers a mechanism to estimate system consequences in a consistent and scalable
manner, allowing modelers to get the full picture of flood risk from various water management
alternatives, and have a better understanding of flood risk.

The HEC-FIA model computed flood damage from inundation results for each alternative. The
same HEC-FIA model was used to analyze each flood event. Consequences used in this analysis
were limited to flood damage to structures and contents. With additional data and parameters,
the HEC-FIA model could be extended to compute flood damage to cars and other property,
crops losses and other agricultural damage, as well as indirect losses such as reduced commerce
and employment. The version of HEC-FIA used for deterministic simulations included life loss
estimates, but only as a demonstration. The parameters for flood warning were all assumed
without any real data, and the results were used only as an additional quality control indicator.

One of the goals of the FIRO analysis was to provide useful tools to USACE District offices.
For that purpose, the version of HEC-FIA used for the deterministic compute, includes Impact
Response reports for the gage locations at Hopland, Healdsburg, and Johnson's Beach. These
reports were not relevant to the FIRO analysis.

117
Chapter 8 – Consequence - Model & Data Development (Deterministic) PR-100

HEC-FIA Simulation Modes

HEC-FIA offers four modes of simulation for flood damage:

a) Cross-Section Only – HEC-FIA uses hydrographs from the HEC-RAS output at every
cross-section and storage area to interpolate the maximum inundation at each structure.
HEC-RAS can also compute other hydraulic characteristics, including arrival time of user
defined depth thresholds, and duration above ground. The cross-section only mode is the
most computationally intensive mode in HEC-FIA. The cross-section only mode requires
that the HEC-RAS model writes results to a DSS file for all locations, but does not
require the use of the RAS Mapper Tool for inundation grids.

b) Grids Only – HEC-FIA reads the hydraulic characteristics at each structure from grids
generated by the RAS Mapper Tool. Supported hydraulic characteristics include
maximum inundation, depth times velocity grid, arrival time grids and duration grids.
The Grids Only mode is the least computationally intensive mode in HEC-FIA, and
requires no DSS output from HEC-RAS. This mode does not support HEC-FIA features
that require hydrographs for timing information, such as Impact Response reports or
Warning Issuance rules.

c) Grids and Cross-Sections - Hydraulic characteristics from gridded data takes precedence
over interpolations from cross-sectional data. This mode, pairs the accuracy of gridded
data with the flexibility of hydrograph data for warning issuance and inundation
configurations. The grids and cross-sections mode became operational within HEC-
WAT during the course of the FIRO study. This mode was established as the preferred
approach.

d) CCP Only - HEC-FIA uses hydrographs from HEC-HMS, HEC-ResSim, or HEC-RAS at

key locations used as index points for an entire reach. Each CCP is associated with a
structure that contains aggregated stage-damage information for the entire reach, as in
HEC-FDA. This simulation mode may not be supported in future versions of HEC-FIA.

HEC-FIA Model Development

The HEC-FIA interface (Figure 68) displays the schematic that represents the Russian River
watershed. The individual components of the Russian River HEC-FIA model are listed in the
Study Pane along the left side of the main window (Figure 68). Construction of the HEC-FIA
model can be most easily described by discussing the components of the Study Tree (Figure 68)
from the top to the bottom of the tree.

Map Layers
The first (top) folder of the Study Tree is Map Layers (Figure 68), which allows HEC-FIA
modelers to display many types of GIS layers in the Map Window (right side of the HEC-FIA
interface). The layers may provide visual reference data, or serve as the first step for importing
data as one of the other types of model elements.

118
PR-100 Chapter 8 – Consequence - Model & Data Development (Deterministic)

Figure 68 HEC-FIA Interface

Watershed
The Watershed folder (Figure 68) holds four types of physical data, along with Watershed
Configurations, which define different potential combinations of the physical data. The first
type of physical data, Terrain Grids, points to the location of the terrain grid used whenever
ground elevations are required in HEC-FIA calculations. Ordinarily only one terrain model is
used, and it is the same terrain model referenced by the RAS Mapper Tool when building
inundation grids. The FIRO HEC-FIA model follows this practice, and points to a copy of the
.vrt and associated files used by HEC-RAS Mapper. The DEM was retrieved from the USGS
National Elevation Dataset.
119
Chapter 8 – Consequence - Model & Data Development (Deterministic) PR-100

Stream Alignment is the second type of element, and is where the watershed stream alignment
is defined. There is usually only one stream alignment per HEC-FIA model and for the FIRO
study, the stream alignment was shared by the HEC-HMS, HEC-ResSim, and HEC-WAT
models.

The third element type defines the mapping of hydrograph results from HEC-RAS to locations
defined within the HEC-FIA model (Figure 69). The mappings were generated automatically by
importing shapefiles produced by the RAS Mapper Tool. The fourth type of element allows for
the definition of CCPs, which were not used by the FIRO study.

Figure 69 HEC-FIA Cross-Section Mappings Editor

As visible in Figure 68, the HEC-FIA model contains two "watershed configurations". Figure 70
displays the watershed configuration used for the deterministic simulations, XSforRules, which
provided hydrograph mappings at three locations. The other watershed configuration, SDGO,
used the same definition as XSforRules, except for not specifying an HEC-FIA element for cross-
sections. SDGO was used for the FRA simulations and focused solely on structure damage
computations using the Grids Only compute mode in HEC-FIA and did not require hydrograph
data.

Figure 70 HEC-FIA Watershed Configuration Editor

120
PR-100 Chapter 8 – Consequence - Model & Data Development (Deterministic)

Geographic Data
The Geographic Data folder (Figure 68) references polygon shapefiles that allow modelers to
break out results by different regions, and to characterize life loss parameters specific to different
areas. From the HEC-FIA Impact Area Editor (Figure 71), using the geographic data
information, models can define default parameters for a siren warning system and default
mobilization timeline. The FIRO HEC-FIA model designated separate Impact Areas centered on
Guerneville and Healdsburg, and defined subsequent Warning Issuance Scenarios for use in
stand-alone and deterministic simulations.

Figure 71 HEC-FIA Impact Area Editor

Inundation Data
From the Inundation Data folder (Figure 68), modelers connect defined Watershed
Configurations to a specific set of HEC-RAS results, creating an Inundation Configuration.
From the HEC-FIA New Inundation Configuration wizard (Figure 72), the first dialog requires
a name for the Inundation Configuration, a description (optional), and selection of an HEC-
FIA compute mode. The second dialog of the wizard (Figure 72) requires that the modeler select

121
Chapter 8 – Consequence - Model & Data Development (Deterministic) PR-100

Figure 72 HEC-FIA New Inundation Configuration Wizard

a Watershed Configuration and select what types of gridded data will be required. For
example, in Figure 72, the molder is defining an Inundation Configuration, with Grids and
Cross Sections as the HEC-FIA compute mode; a Watershed Configuration (XSforRules); and,
three grids. The three grids selected represent the maximum depth in each cell during an event
(Inundation Grid), the maximum product of depth and velocity in each cell (Depth*Velocity
Grid), and the amount of time required for the depth in each cell to reach a threshold (Life Loss
Arrival). The FIRO analysis used the typical default depth of two feet for the Life Loss Arrival
grid.

When running HEC-FIA in stand-alone, a further definition is required for an Inundation

Configuration. The modeler will need to define an Event that includes specifying information
based on what HEC-FIA compute mode has been selected for the Inundation Configuration.
For example, in Figure 73, a modeler has created an Event that includes specifications for one of
the HEC-FIA validation simulations using hydrographs from an HEC- DSS file, and grids
produced by the RAS Mapper Tool. Defining Events applies only to standalone HEC-FIA
simulations, since the hydraulic data is managed as part as the HEC-WAT compute. The HEC-
WAT Model Linking Editor uses the HEC-FIA inundation configuration to define the data
required for HEC-FIA to compute an event.
122
PR-100 Chapter 8 – Consequence - Model & Data Development (Deterministic)

Figure 73 Defining an Event for an HEC-FIA Inundation Configuration

Inventory Data
HEC-FIA uses four different types of inventory data: Structures, Critical Infrastructure, Impact
Response, and Agricultural. The most important one for the FIRO study was the Structure
Inventory.

Structure Inventory Data

The Structure Inventories folder (Figure 68) allows HEC-FIA modelers to define different
categories of structures. The Russian River HEC-FIA model used the standard four damage
categories: Residential, Commercial, Public, and Industrial. Each structure was classified
according to a structure occupancy type as belonging to one of the four damage categories.
Specific structures for HEC-FIA are described in specially formatted inventories typically
generated from point shapefiles.

Structure Occupancy Types

Structure occupancy types define default parameters and relationships for each structure in a
structure inventory. The FIRO study used only the standard HAZUS definitions that are built
into HEC-FIA. HEC-FIA allows modelers to override these defaults with specific values for any
given structure, but the FIRO analysis used only the defaults. Such added accuracy could be
useful if certain valuable structures figure prominently in the damage computations.

From the HEC-FIA Structure Occupancy Type Editor (Figure 74) the modeler can define a
structure occupancy type. For example, in Figure 74, a structure occupancy type has been
created by a modeler that contains a depth-damage function for the structure, content, and car.
The most common occupancy type in the Russian River inventory is One-Story Residence with
No Basement (ORSNB). The HEC-FIA software can sample the depth-damage relationships to
represent uncertainty, but this feature was not used in the FIRO analysis, therefore, the depth-
damage relationships were used in a deterministic manner. Note that in Figure 74, the damage
for this occupancy type occurred at depths less than zero, i.e., lower than the first floor elevation.

123
Chapter 8 – Consequence - Model & Data Development (Deterministic) PR-100

Figure 74 Structure Occupancy Type Editor –Depth-Damage Functions Tab

In addition, from the Structure Occupancy Type Editor (Figure 75), the modeler can define
other parameters about a structure occupancy type like the foundation height, structure value,
content to structure value, automobile data, and more. In the example, Figure 75, HEC-FIA
treats the Foundation Height (first floor elevation) as three feet above the ground level. If
HEC-RAS results provided an inundation depth of three feet above ground at such a structure,
the HEC-FIA model would assign damage to the structure amounting to 13.40 percent of the
structure value. Also note (Figure 75) that the value of the contents for a one-story residence
with no basement is fifty percent of the structure value.

Structure Inventory
The structure inventory (RevisedfromSPK) used for consequence analysis in HEC-FIA was
assembled through GIS analysis of 2013 parcel data for Sonoma and Mendocino Counties.
Figure 76 provides an example of the parcel data near Mark West Creek and Highway 101 in
Santa Rosa. The white lines represent the parcel polygons, and the upper table shows some of
the attributes, including assessed value. The blue dots represent the structure inventory point
locations used in HEC-FIA, with point locations assigned as the centroid of each parcel. The
points were cropped to a bounding polygon that HEC-FIA offers based on the extents of the
HEC-RAS model. This effort resulted in 65,522 structures relevant to the HEC-FIA structure
inventory.
124
PR-100 Chapter 8 – Consequence - Model & Data Development (Deterministic)

Figure 75 HEC-FIA Structure Occupancy Type Editor – General Info/Uncertainty Parameters Tab

Figure 76 Parcel Data

125
Chapter 8 – Consequence - Model & Data Development (Deterministic) PR-100

The parcel data was processed to remove unnecessary attributes and add attributes for damage
category and HAZUS occupancy type. The processed parcel data was then joined with
Longitudinal Employer-Household Dynamics (LEHD) data (US Census Bureau, 2010) to
include certain populations. The lower table in Figure 76 displays the parcel data (attribute
fields) that were imported to create the actual HEC-FIA structure inventory. The second field
(Figure 76) contains the unique name for each structure (COM4 29448), which consists of an
occupancy type name with an appended sequence number. The third field (VAL_STRUC)
contains the assessed structure value in United States dollars. For structures where no structure
value was provided, $5,000 was assumed. The fourth and fifth fields list the damage category
(ST_DAMC) and occupancy type (ST_DMG) names, respectively. The sixth field (NUM_STOR)
contains the number of stories associated with a structure. The last four fields (Figure 76)
contain data used for life loss calculations, listing the number of people over 65 and under 65
likely to be in the structure by day or night.

The level of detail in the structure inventory allows HEC-FIA to generate consequence estimates
for very large events across wide areas, yet still support a finely detailed examination of which
structures showed the most impact. Despite the detail, close inspection of the data revealed some
shortcomings.

The centroid assumption worked very well on smaller parcels, but not as well on larger parcels.
Terrain data plays an important role in depth estimates, and structures may show inappropriate
damage unless the ground is very level. Rooftops visible in the satellite view did not always
match up with an inventory point. Occupancy types can be ambiguous, erroneous, and quickly
become out-of-date. For instance, in Figure 76, Structure A was described as "industrial" related
to "food/drug/chemicals" for the structure, but most recently Structure A appears to be a
combination of retail and office purposes. In the example (Figure 76), Structure B (mobile
homes) and Structure C (apartments), likely would begin to incur damage at the southern units
before reaching damaging depths at the centroid location. Structure D (commercial) represented
a wine estate and gardens, with a large two-story building. This study did not attempt to correct
such shortcomings in the structure inventory. Table 23 summarizes the data used in the HEC-
FIA structure inventory.

The total value of the structure inventory exceeded 14 billion dollars, with 71 percent of the
value in residential structures. Commercial structures accounted for about half of the non-
residential value, and industrial about one-third. Figure 77 helps visualize the values for
different categories within the structure inventory. This analysis did not attempt to refine
structure values using other data sources, or index the values to a given base year.

Critical Infrastructure
HEC-FIA allows modelers to identify a separate inventory of point locations of special interest to
flood emergencies, such as police stations, schools, campgrounds, wastewater plants, etc. HEC-
FIA evaluates the specified locations against the hydraulic data for a given event and generates a
report of affected sites. The HEC-FIA model used for the FIRO analysis did not include a
critical infrastructure inventory.

126
PR-100 Chapter 8 – Consequence - Model & Data Development (Deterministic)

Table 23 Summary of Structure Inventory

Occupancy Value of
Type Description Count Structures
AGR1 Agriculture 2,012 $559,373,650.00
COM1 Retail Trade (i.e., Store) 1,381 $1,053,667,812.00
COM2 Wholesale Trade (i.e., Warehouse) 0
COM3 Personal and Repair Services (i.e., Service Station) 502 $192,024,311.00
COM4 Professional/Technical Services (i.e., Offices) 753 $483,794,901.00
COM5 Bank 57 $43,732,392.00
COM6 Hospital 94 $96,066,428.00
COM7 Medical Office/Clinic 103 $66,104,423.00
COM8 Entertainment & Recreation (i.e., Restaurants/Bars) 228 $90,528,800.00
COM9 Theaters 9 $34,390,291.00
COM10 Parking (i.e., Garages) 40 $1,074,069.00
EDU1 Grade Schools 46 $48,932,359.00
EDU2 Colleges/Universities (not including group housing) 0
GOV1 General Services (Offices) 58 $7,956,502.00
GOV2 Emergency Response (i.e., Police/Fire Station/EOC) 0
IND1 Heavy Industrial 27 $36,595,462.00
IND2 Light Industrial 818 $899,205,937.00
IND3 Food/Drug/Chemicals 139 $422,657,848.00
IND4 Metals/Minerals Processing 18 $1,757,334.00
REL1 Religion/Non-Profit (i.e., Church) 152 $87,285,915.00
RES1-1SNB Single Family Dwelling 1 Story no Basement 35,966 $5,044,805,261.00
RES1-2SNB Single Family Dwelling 2 Story no Basement 10,885 $2,221,657,766.00
RES1-3SNB Single Family Dwelling 3 Story no Basement 92 $1,910,683.00
RES1-4SNB Single Family Dwelling 4 Story no Basement 11 $2,152,940.00
RES1-5SNB Single Family Dwelling 5 Story no Basement 11 $1,370,647.00
RES1-6SNB Single Family Dwelling 6 Story no Basement 1 $100,424.00
RES1-7SNB Single Family Dwelling 7 Story no Basement 4 $817,222.00
RES1-10SNB Single Family Dwelling 10 Story no Basement 1 $64,000.00
RES1-20SNB Single Family Dwelling 20 Story no Basement 1 $159,610.00
RES2 Mobile home 882 $135,970,900.00
RES3A Multi Family Dwelling Duplex 9,909 $1,434,169,901.00
RES3B Multi Family Dwelling 3-4 Units 811 $181,955,201.00
RES3C Multi Family Dwelling 5-9 Units 144 $66,943,306.00
RES3D Multi Family Dwelling 10-19 Units 78 $ 67,275,106.00
RES3E Multi Family Dwelling 20-49 Units 51 $94,339,434.00
RES3F Multi Family Dwelling 50+ Units 88 $471,155,980.00
RES4 Temporary Lodging (i.e., Hotel/Motel) 150 $ 208,661,470.00
RES5 Institutional Dormitory (College Housing, Jails) 0
RES6 Nursing Home 0
Total: 65,522 $14,088,658,285.00

127
Chapter 8 – Consequence - Model & Data Development (Deterministic) PR-100

Figure 77 Structure Value Breakdown

Impact Response
HEC-FIA can document the timing where water levels rose or fell across various thresholds,
such as the flood stage impact levels used by NWS River Forecast Centers. This information
lists several key river stages for each gage location, with associated road closures and areas
experiencing flooding. The Impact Response inventory for the FIRO watershed included the
NWS flood stage impacts for the Hopland, Healdsburg, and Johnson's Beach gage locations.
The Impact Response reports were not used in the FIRO analysis, but were included in the
model for subsequent use in CWMS and training applications.

Agricultural Grids
The FIRO analysis did not address agricultural damage, and no inventory of agricultural
information was included in the model configuration.

Warning Issuance Scenario

The FIRO analysis did not directly consider life loss, but the feature was implemented to
demonstrate the capability within HEC-WAT. The loss-of-life calculations also helped identify
irregularities regarding data that was not obvious from just flood damage results, since the
calculations involved different interplay among depth grid, terrain model, and structure
inventory. The life loss calculations were not performed in the FRA simulations.

128
PR-100 Chapter 8 – Consequence - Model & Data Development (Deterministic)

Warning issuance assumptions matter greatly in modeling life loss, and HEC-FIA provides
several options to characterize how flood warnings are generated. This analysis created a simple
rule for both polygons in the previously defined Impact Area data, which were centered on the
gages in Healdsburg and Guerneville. The rules generated warnings for each separate region one
hour after the respective gages reached flood stage. The warning rules for the Healdsburg region
was defined in the HEC- FIA Warning Issuance Scenario Editor (Figure 78) and was only
used in the HEC-WAT deterministic simulations.

Figure 78 HEC-FIA Warning Issuance Editor

Alternatives
Now that the modeler has defined all of the HEC-FIA model elements, an HEC-FIA alternative
(Figure 79) can be created. For example, in Figure 79, in the HEC-FIA Alternative Editor, the
model has created an HEC-FIA alternative, SDLLIR (Structure Damage Life Loss Impact
Response). The modeler has chosen an Impact Area Set, Structure Inventory, Impact

129
Chapter 8 – Consequence - Model & Data Development (Deterministic) PR-100

Figure 79 HEC-FIA Alternative Editor

Response, and Warning Issuance. In addition, to computing structure damage and life loss, and
generate impact response reports, the modeler is using a Grids and Cross-Sections compute
mode (based on hydraulic output for the 1964 event) for the Inundation Configuration.

For each HEC-FIA alternative (Figure 79), the modeler will define the settings for the various
HEC-FIA simulation types. For example, in Figure 79, the SDLLIR alternative has been setup to
simulate life loss computations and is the HEC-FIA alternative used in the HEC-WAT
deterministic simulations.

Time Window
Standalone runs during HEC-FIA model development required definition of simple time
windows corresponding to the duration of the flood event consequences (Figure 80). The start of
the time window must align with the start time of the HEC-RAS simulation used to generate
arrival time grids. Using the same start time for HEC-RAS and HEC-FIA can also help avoid
confusion setting certain Warning Issuance parameters and interpreting Critical Infrastructure
report output. HEC-FIA time window definitions are not used in HEC-WAT, which provides
them to HEC-FIA along with the hydraulic data.

130
PR-100 Chapter 8 – Consequence - Model & Data Development (Deterministic)

Figure 80 HEC-FIA Time Window Dialog Box

Simulations
Standalone runs during HEC-FIA model development required definition of HEC-FIA
simulations (Figure 81) to define a compute. HEC-FIA simulations combine an HEC-FIA
alternative with a time window, and one of the events defined under Inundation Data. In an
HEC-WAT simulation, HEC-WAT is managing inputs and time windows, so HEC-FIA
simulations only applied to standalone uses of HEC-FIA during model development and
validation.

Figure 81 HEC-FIA Simulation Manager

HEC-FIA Results
For each compute, HEC-FIA delivered results through three separate mechanisms:

1) GIS layers and database files for access through HEC-FIA and HEC-WAT, or for
external analysis using spreadsheets or GIS software

131
Chapter 8 – Consequence - Model & Data Development (Deterministic) PR-100

2) interactive displays on the HEC-FIA or HEC-WAT Map Window

3) standard reports

Output Files
The HEC-WAT deterministic simulations for this analysis generated various sets of HEC-FIA
output files for each annual event. Examples of the HEC-FIA output files are displayed in
Figure 82. The SDLLIR.dss file (Figure 82) contained hydrograph time series data that were used
in the Impact Response rules, and the times that each rule was triggered. The name of the DSS
file reflects the name of an HEC-FIA event. The EconResults shapefile listed the flood damage
results at structures within the inundated area (Figure 83).

Figure 82 HEC-FIA Output Files per Deterministic Event

Figure 83 Attributes for GIS Output of Damage at Flooded Structures

HEC-FIA also produced shapefiles with key intermediate and final results for life loss at the
inundated structures (Figure 82). The geo-referenced results per structure offered many
opportunities for analysis of each event, which provided a better understanding of the flood
consequences along the Russian River. The FIRO study examined how structure damage might
(or might not) vary according to alternative water management operations, or the scale of the
event. Sections 8.12 and 11.5.3 of this report provide example of the GIS analyses. HEC-FIA
also writes aggregated results to the HEC-WAT simulation.dss file, totaling consequences for all
the events in an HEC-WAT simulation.

132
PR-100 Chapter 8 – Consequence - Model & Data Development (Deterministic)

Map Output
HEC-FIA map output provided modelers with a comprehensive view of how the data interacts to
generate the results. View controls regarding data layers and display of elements allowed
modelers to configure the map to bring out the salient features of the input or output. For
example, in Figure 84, the structure inventory is displayed as squares that are color-coded
according to the structure value, with the most valuable structures rendered as dark green and the
least valuable as yellow. The maximum depth grid (generated by HEC-RAS) for the 2006 event
was added as a map layer to HEC-FIA, and overlaid on a Google® background map (Figure 84).
The HEC-FIA structure tool was used to select a structure that is located in the depth grid layer,
causing the structure to appear as a bright yellow square with a red border, and enabling the
modeler to open a dialog listing descriptive information about the selected structure. The HEC-
FIA Grid Query Tool allowed the modeler to display the depth in the cell that contains the
structure in a tooltip (Figure 84). The structure selected (Figure 84) experienced a water depth of
1.11 feet, and with a foundation height of three feet, the result is -1.89 feet of depth relative to
the first floor. From the depth-damage function (Figure 74), for occupancy type RES1-1SNB,
the event was sufficient to cause minor damage.

Figure 84 HEC-FIA Map Window – Grid Query Tool

133
Chapter 8 – Consequence - Model & Data Development (Deterministic) PR-100

A more direct approach to evaluating structure damage results within a HEC-FIA map window is
displayed in Figure 85. Display of the structure inventory was removed from the map window,
and the shapefile that contains the HEC-FIA simulation results was added as a new map layer.
The simulation results layer only contains structures where the HEC-RAS depth grid is greater
than zero. The symbology of the HEC-FIA damage points was adjusted to render the least
damaged structures as yellow and the most damaged structures as red (Figure 85). The modeler
using the HEC-FIA Shapefile Query Tool and selecting the same structure displays the output
columns for the structure (Attributes Table, Figure 85), including $563 structure damage and
$271 to contents.

Figure 85 HEC-FIA Map Window - Shapefile Query Tool

Standard Reports
The HEC-FIA software provides several reports that allows the modeler to review pertinent
inputs and results. The reports are detailed in the following sections and are accessible from the
HEC-FIA interface by right clicking on a simulation in the HEC-FIA results pane (Figure 86).

Individual Structure Damage Report

From the HEC-FIA Individual Structure Damage Report (Figure 87), the modeler can validate
the HEC-FIA model, and examining relationships between the hydraulic results and
consequences. Structures can be sorted by depth or damage, and by selecting a structures, the
model can zoom the map window on the structures of interest.
134
PR-100 Chapter 8 – Consequence - Model & Data Development (Deterministic)

Figure 86 HEC-FIA Main Window - Accessing HEC-FIA Reports

Figure 87 HEC-FIA - Individual Structure Damage Report

135
Chapter 8 – Consequence - Model & Data Development (Deterministic) PR-100

Aggregated Consequence Report

The HEC-FIA Aggregated Consequence Report (Figure 88) by default displays consequence
results by damage category. The modeler can also report consequences in terms of occupancy
type or impact area, which allows modelers to aggregate by user-defined geographic boundaries.
The aggregated information was occasionally reviewed to verify the reasonableness of results,
but otherwise was not used for this analysis.

Figure 88 HEC-FIA - Aggregated Consequence Report

Detailed Life Loss Report

The life loss modeling performed in the HEC-FIA standalone analysis did not attempt to produce
accurate results. The modeling used a default flood warning system and warning issuance
parameters for only a portion of the population at risk. However, the results assisted during
validation of the structure inventory by providing another method to identify unreasonable
results. Demonstration of the capability to estimate life loss may also be relevant to future
applications of the HEC-FIA Russian River model.

The Detailed Life Loss Report (Figure 89) indicates 0.8 fatalities, and that most of the risk
occurred in the Chance Zone, due to depth criteria. Differences in the modeling assumptions,
structure inventory, and population input prevented any serious efforts at validation of the life
loss results. However, Bulletin No. 161 (DWR, 1965) reported that the 1964 flood killed one
person in Mendocino County and none in Sonoma County.
136
PR-100 Chapter 8 – Consequence - Model & Data Development (Deterministic)

Figure 89 HEC-FIA - Detailed Life Loss Report

Structure Damage Validation

Accurate determination of flood damage can be difficult because of flawed structure information
and uncertain hydraulic inundation data, as well as uncertain, incomplete, and inconsistent
reporting of actual damage. The validation goal for HEC-FIA in the FIRO study was only to
generate "ball-park" estimates that loosely reflect damage from historical reports.

The process began with a "reality check" type of simulation, using whatever hydraulic data was
available, to show that the HEC-FIA model produced results in a credible order of magnitude.
The model with provisional hydraulic data yielded a flood damage estimate of about half the $31
million dollars of damage estimated for the 1997 event in the Sonoma County Flood Hazard and
Risk Assessment. The team considered this reasonable, in view of the assumptions and
uncertainties involved, and proceeded with refinements to the structure inventory.

Initial Structure Inventory Adjustments

The only hydraulic data available during the early phases of the HEC-FIA model development
were from HEC-RAS calibration runs. Those inputs allowed the HEC-FIA modelers to make a
137
Chapter 8 – Consequence - Model & Data Development (Deterministic) PR-100

first pass at identifying structures with inappropriate damage. For example, in Figure 90, an early
version of the HEC-FIA model is displayed, illustrating a typical structure inventory review at an
area just east of the Highway 116 Bridge in Guerneville, CA. The blue points (Figure 90)
represent structures in the inventory. The satellite background map provides a helpful reference
for ground-truthing for each structure. The HEC-FIA Individual Structure Damage report
(Figure 90) provides a table of each structure located in an inundated area. The HEC-FIA report
displays results for each structure, and the HEC-FIA modeler can sort by decreasing depth
(Depth of Flooding). HEC-FIA also provides modelers with a helpful feature that allows the
modeler from the Map Window to select a structure and from the Individual Structure
Damage report; the structure is highlighted on the report (Figure 90).

Figure 90 Structure Inventory Adjustments

For example, in Figure 90, structures along the Riverlands Road appear to be about 200 feet too
close to the river, resulting in flood depths of up to ten feet for the simulated 1997 event. The
structure points were manually adjusted to the nearest rooftop along the road using HEC-FIA's
Structure Editing Tool, which appears at the left of the map window as an icon with the image
of a red house. The most significant errors in structure locations were screened by sorting the
Individual Structure Damage report by depth and damage, and which allows the HEC-FIA
modeler to zoom in on suspicious results.

138
PR-100 Chapter 8 – Consequence - Model & Data Development (Deterministic)

Non-Damaging Flow Adjustments

The initial screening based on hydraulic output for large events still resulted in thousands of
structures with shallow inundation or minor damage, with no obvious way to tell which results
were appropriate. The next step was to run the HEC-FIA model for an "almost damaging" event,
and make sure that no damage was reported.

The HEC-FIA modelers selected 2007 as the most recent year in the 1950 to 2010 analysis
period where the gage for the Russian River at Guerneville reached "action stage" (NWD
terminology), but not "flood stage". The peak discharge for 2007 was assumed non-damaging
across the watershed. No documentation of flood damage for 2007 was found after checking
reference materials and searching online. The HEC-RAS modeler provided provisional HEC-
RAS results for the minor event of 10 February 2007. The HEC-FIA simulation initially
reported damage of approximately $6,195,620 across 180 structures. Almost all of the damage
occurred in the lower Russian River.

Closer examination of the results by the HEC-RAS and HEC-FIA modeling teams revealed
issues in four locations: Hacienda, Guerneville, Monte Rio, and Duncans Mills. The issues
generally stemmed from inadequacies in the terrain data used for defining cross-sections,
computing inundation depths, and determining first-floor elevations. One issue (Figure 91) was
that structures sustained damage despite hydraulic results staying well within the bank. The
DEM elevation did not adequately reflect the high ground situating two structures within the
Mays Canyon storage area. Inaccuracies in the terrain data also led to inappropriate inundation
extents, which reached far enough to damage (slightly) the structure just northeast of the other
two structures.

Figure 91 Johnson's Beach Structure Adjustments

139
Chapter 8 – Consequence - Model & Data Development (Deterministic) PR-100

HEC-FIA modelers can specify a first-floor elevation for each structure, which overrides the
value determined from the ground elevation. However, for the FIRO analysis this was deemed
too cumbersome for the number of adjustments required at each structure. The region features
many expensive structures, with various degrees of floodproofing. HEC-FIA modelers found
that simply dragging the structures to the west side of the road was an expedient and reasonable
way to avoid spurious damage. Figure 92 displays an example of such inventory adjustments.

Figure 92 Structure Relocations East of Guerneville across from Rio Nido

After reviewing the most troublesome locations, the HEC-RAS and HEC-FIA modelers
determined that no changes to the hydraulic model or terrain data were warranted. The HEC-
RAS model was not designed for great accuracy at non-damaging flows. Acquiring more
detailed terrain data and regenerating the models was beyond the scope of this study. The 74
structures along the lower Russian River were manually relocated in the structure inventory to
avoid damage for the 2007 event. This action left 106 structures elsewhere along the river
incurring $688,081.50 for the presumably non-damaging 2007 event. Almost all of these
structures showed minor flooding, according to a variety of different circumstances.

A typical situation (Figure 93), where structures experienced very shallow inundation, was most
likely due to limitations of the terrain. The structures in the inventory were of different
occupancy types with different assumed foundation heights. The locations and number of
structures in the inventory corresponded poorly to rooftops visible from the reference satellite
data. With so much uncertainty in the data and such little damage, the solution again consisted
of simply dragging the points over to the nearest road or high ground.
140
PR-100 Chapter 8 – Consequence - Model & Data Development (Deterministic)

Figure 93 Structure Damage in Agricultural Area near Cloverdale

Historical Flood Damage Flow Validation

After completing revisions to the structure inventory and as more hydraulic results became
available, the damage results from HEC-FIA were compared to several recent events. Table 24
shows a summary of consequences from recent Russian River floods, as provided in the Sonoma
County Flood Hazard and Risk Assessment. The information provided in Table 24 is for the
whole county, although most of the damage occurred along the Russian River. The HEC-FIA
totals included damage in Mendocino County, but it only accounts for a small portion of the
total. The reported HEC-FIA damage only considered damage to structures and contents. The
damage estimates in Table 24 show that the HEC-FIA model calculated reasonable flood damage
for use in relative comparisons.

Comparisons to older events are difficult due to differences in the structure inventory, and changes to
the river system and floodplains, such as drainage improvements in Laguna De Santa Rosa or
construction of the Warm Springs Dam. The 1964 event applied to the contemporary structure
inventory resulted in $65 million dollars of damage, although Bulletin No. 161 (DWR, 1965)
from DWR reported only $6.5 million dollars of damage at the time of the flood. A 1987 article
by the Los Angeles Times reported flood damages of $37.5 million along the Russian River from
the 1986 flood, while the HEC-FIA model computed $45.7 million.

141
Chapter 8 – Consequence - Model & Data Development (Deterministic) PR-100

Table 24 Flood Damage Validation Events

Loss HEC-FIA
Date Estimates Consequences Damage
Over 50 roads closed. 15,000 residents without power.
Total displaced persons exceeded 2,000, of which 456
January 8-31, 1995 $21 million
flood victims were evacuated by air. 13 medical cases
$35
were treated and 2 flood-related fatalities occurred.
million
Over 100 roads closed. 45,000 residents without power.
March 7-15, 1995 $13.3 million At least 3,000 residents displaced. Up to 30 containers of
possible toxic materials identified in the flood zone.
Up to 200 roads were closed and/or damaged, some due
to major slides. 12,000 residents without power. Over
December 30, 1996- $11
$31 million 1,200 victims evacuated their residences and 2 storm-
January 4, 1997 million
related deaths occurred. Sewage and treatment plants
overflowed.
Over 100 roads closed due to flooding and landslides
Approximately 50,000 county residents without power.
December 30, 2005 $104 million 2106 properties inundated, 67 declared uninhabitable. $37
– January 3, 2006 Unknown number of self-evacuations. Laguna million
Wastewater Treatment Plant flooded with partially-
treated sewage spill into the Laguna de Santa Rosa

Analysis of HEC-FIA Results

Flooded structures were joined with the nearest cross-section using GIS analysis, allowing
damage to be accumulated per river mile, as shown in Figure 94 for the 2006 event. Damage
along Dry Creek is included on Russian River Mile 32.00. The analysis shows that little damage
occurred above Healdsburg. The slope (i.e., damage/mile) appears uniform from Monte Rio
through Guerneville, but uneven from Guerneville through Healdsburg. The jumps primarily
result from structures in storage areas added at the same river mile. Flow changes in the
hydraulic model from lateral inflow hydrographs may have also contributed to abrupt changes.

Figure 94 Russian River Damage by River Mile for the 2006 Event
142
PR-100 Chapter 9 – HEC-WAT Model Development (Deterministic)

HEC-WAT
Model Development
(Deterministic)
The individual model alternatives for the FIRO study have been updated and validated separately
in "standalone" mode, but were closely coordinated using HEC-WAT as described in this
chapter. This chapter also details the creation of the HEC-WAT FIRO study; importing the
finalized individual model alternatives into the study; creating tools that setup the HEC-WAT to
run the deterministic simulations; and then computation of study results.

Planning and Execution

Modelers held a "kick-off" meeting to discuss what types of desired results are required from
HEC-WAT, and identify the input requirements of the individual models. Much of the
information took further investigation and numerous follow-up discussions in sub-groups.

The meetings identified the data needs of each constituent model, and the locations where results
from one model could be used as input to another model. The HEC-HMS, HEC-ResSim, and
HEC-RAS modelers used models from prior studies to develop provisional models for the FIRO
study suitable for integration into HEC-WAT. Most of the coordination consisted of establishing
a naming convention and aligning the schematic elements of the models.

HEC-WAT provides flexibility in a study workflow by allowing modelers to work either in the
shared framework or in stand-alone mode, depending on the requirements of particular tasks.
The Russian River modeling teams found it most convenient to generally work separately on
developing the stand-alone models, but occasionally import the model alternatives into HEC-
WAT to share updates and confirm suitability of schematic changes. The streamlined process
for importing and linking constituent model alternatives in HEC-WAT made it very practical to
re-assemble them on an ad-hoc basis. This flexibility proved very helpful as the modelers
continued development and learned more about the system and study requirements.

An initial deterministic HEC-WAT model was assembled as soon as provisional HEC-HMS and
HEC-ResSim models became available. These two models determined the quantities and
movement of water within the Russian River watershed, and identified most of the locations
where models needed to exchange results in the Russian River system. This initial HEC-WAT
study was not run to produce results, but served to finalize shared locations and naming
conventions, and identify further compatibility issues between the individual models.
Subsequent HEC-WAT studies were assembled that also included provisional HEC-RAS and
HEC-FIA models.

143
Chapter 9 – HEC-WAT Model Development (Deterministic) PR-100

The provisional HEC-WAT studies created during development of the individual models were
configured to run in the deterministic mode. Deterministic simulations allowed modelers to
leverage observed data and calibration events to verify that the individual models performed as
expected when integrated within HEC-WAT. Additionally, the deterministic simulations
provided results that addressed the success of different reservoir operating strategies in filling
Lake Mendocino to the desired summer pool (the FRA simulations only addressed annual flood
risk). After importing the individual model alternatives, they were assembled into HEC-WAT
simulations, and run to produce the final deterministic results.

Creating the FIRO HEC-WAT Study

Creating a new HEC-WAT study required a name, a disk location, and a coordinate system
(Figure 95), and the choice of whether to create a default alternative. Note that the FIRO study
name included the word "final". The choice distinguished the HEC-WAT study from all the
earlier provisional versions created to coordinate parameters and names among the component
models, and identify remaining data requirements. This study used the standard coordinate
system adopted by the USACE Modeling, Mapping and Consequences (MMC) Production
Center, USA Contiguous Albers Equal Area Conic. This study also used the default alternative
name of Existing Conditions.

Figure 95 Creating the HEC-WAT Study

HEC-WAT supports a wide variety of plug-ins, which may be enabled or disabled as needed for
each project. From the HEC-WAT main window, from the Files menu, click Study Details, the
Study Details dialog box will open (Figure 96). Click the Plugins tab, from the table enable or
disable the appropriate plug-ins for the FIRO study. Figure 96 displays the enabled plug-ins for
the FIRO study: Hydrologic Sampler, HEC-HMS, HEC-ResSim, HEC-RAS, HEC-FIA, the
Time Window Modifier, and the Time Window Interval Generator.

144
PR-100 Chapter 9 – HEC-WAT Model Development (Deterministic)

Figure 96 HEC-WAT – Study Details Dialog Box – FIRO Study Plug-Ins

Model Imports
The final versions of each standalone model, including calibration/validation data and output,
were archived for future reference. A cleaned-up copy of each model, containing only the
desired model alternative and necessary supporting data, were imported into the HEC-WAT
FIRO study. Importing the minimum necessary model configurations helped ensure the fastest
execution, tidiest results, and easiest troubleshooting in the HEC-WAT environment.

HEC-ResSim
The HEC-ResSim model was developed with the understanding that it would be the first model
imported into the HEC-WAT study, and provided the basic schematic elements. Coordination
among the modeling team ensured that the HEC-ResSim model contained a comprehensive
stream alignment, and a superset of the important junctions in the watershed. Using HEC-
ResSim as the initial model to import was also recommended by the HEC-WAT software
development team as a good standard practice.
145
Chapter 9 – HEC-WAT Model Development (Deterministic) PR-100

The ResSim import in HEC-WAT allows for either importing the entire HEC-ResSim watershed
or selected HEC-ResSim alternatives. For the FIRO study, the choice was made to import the
entire HEC-ResSim watershed. Figure 97 shows the HEC-WAT interface immediately after the
import. The map window displays the stream alignment, reservoirs and common computation
points that were imported from the HEC-ResSim model.

Figure 97 HEC-WAT Study after Importing HEC-ResSim Model

The Study Pane (Figure 97, upper left) display the Study Tree for the HEC-WAT FIRO study.
From the Models folder, HEC-ResSim is selected, from the Content Pane (Figure 97, lower
left); all HEC-ResSim model alternatives that were imported are displayed. Note that there were
sixteen HEC-ResSim model alternatives imported. As previously described, these resulted from
an iterative process of building and running provisional models. The list of HEC-ResSim
alternatives included the base HEC-WAT alternatives described in Section 4.9. The four HEC-
ResSim alternatives ending with "_F" represent the HEC-ResSim model alternatives used in the
FRA simulations. The remaining HEC-ResSim alternatives, which end with "S", were used for
troubleshooting within HEC-ResSim and were not used in the final FIRO analysis.

146
PR-100 Chapter 9 – HEC-WAT Model Development (Deterministic)

HEC-HMS Model Alternatives

The HEC-HMS import occurred next, which resulted in two HEC-HMS model alternatives
(runs) being imported in the HEC-WAT FIRO study. One of the HEC-HMS model alternative
was designed for deterministic simulations ((RUN)Basin Average 60 year), and the other HEC-
HMS model alternative if for the FRA simulations.

From the Study Tree (Figure 98), from the Models folder, click on HMS, the HEC-HMS model
alternatives will display in the Content Pane (Figure 98). Right-click on an HMS model
alternative name, from the shortcut menu (Figure 98) modelers can review the basin model,
meteorological model, and simulation run, for the selected HEC-HMS model alternative.

Figure 98 HEC-HMS Model Alternative

Alternately, from the HEC-WAT main window, from the Programs Toolbar, the modeler can
execute the HEC-HMS software by clicking . This selection provides a convenient way to run
the full HEC-HMS program in stand-alone mode, which allowed modelers to review the
imported data, and ensure use of the proper compute settings for the HEC-HMS model
alternatives. All the precipitation and evapotranspiration data needed for the deterministic
simulation ((RUN)Basin Average 60 year) was also imported. From the HEC-WAT Content
Pane, by right clicking on an HEC-HMS model alternative and from the shortcut menu (Figure
98), the modeler is able to execute an HEC-HMS simulation run. This action allows the modeler
to verify that all information for the selected HEC-HMS model alternative imported correctly.
The HEC-HMS model for the deterministic simulations was the only one executed immediately
after import, since it had no external data dependencies.

After verifying a successful import, the "source" elements in the HEC-HMS basin model that
represented the outflows from Lake Mendocino and Lake Sonoma were removed. These
elements allowed HEC-HMS to include and route observed reservoir releases during stand-alone
simulation, but were not needed in the HEC-WAT FIRO study, since the HEC-ResSim model
alternative will be providing the releases.
147
Chapter 9 – HEC-WAT Model Development (Deterministic) PR-100

HEC-RAS and HEC-FIA Model Alternatives

The HEC-RAS import occurred next, with HEC-RAS model alternatives (plans) that were
named POR and FRA being imported into the HEC-WAT FIRO study. From the Study Tree,
from the Models folder, click on RAS, the HEC-RAS model alternatives will display in the
Content Pane. Right-click on a RAS model alternative name, from the shortcut menu modelers
can review the geometry data for the selected HEC-RAS model alternative. Alternately, from
the HEC-WAT main window, from the Programs Toolbar, the modeler can execute the HEC-
RAS software by clicking . This selection allows modelers to access HEC-RAS functions,
such as the RAS Mapper Tool for verifying the spatial reference projection file, the terrain
association, and results maps.

The HEC-FIA import occurred next, with the HEC-FIA model alternatives named SDLLIR and
SDDepthGridOnly being imported into the HEC-WAT FIRO study. From the Study Tree, from
the Models folder, click on FIA, the HEC-FIA model alternatives will display in the Content
Pane. Right-click on an FIA model alternative name, from the shortcut menu modelers can
review the imported HEC-FIA model alternative parameters. For the SDLLIR model alternative,
which is used for an HEC-WAT deterministic simulation, parameters reviewed were impact
areas, cross-sections, structure inventory, watershed configuration, impact response, and warning
issuance. Alternately, from the HEC-WAT main window, from the Programs Toolbar, the
modeler can execute the HEC-FIA software by clicking . This action allowed the modelers to
access the terrain grid definition.

Other Model Entries

There are several HEC-WAT options available, that will aid in the computational speed of HEC-
WAT simulations, especially for FRA simulations. For the HEC-WAT FIRO study, model
alternatives for both the Time Window Modifier and the Time Window Interval Generator were
created.

Time Window Interval Generator

The Time Window Interval Generator (TWIG) divides continuous computes into annual
events and generates one-year time window intervals by default. To create a TWIG alternative,
from the Study Pane, from the Models folder, right-click on Time-Window Interval
Generator, from the editor provide a name for the TWIG alternative, and be sure Use
Simulation Start Time is selected.

Time Window Modifier

The Time Window Modifier allowed modelers to establish separate time windows for each
event according to user-defined inputs and criteria. To create a time window modifier
alternative, from the Study Pane, from the Models folder, right-click on TimeWindow

148
PR-100 Chapter 9 – HEC-WAT Model Development (Deterministic)

Modifier, from the shortcut menu, click New. From the Create New Time Window Modifier
dialog box (Figure 99), enter a name, click OK. The Time Window Modifier Alternative
Editor will open (Figure 100); the modeler can now define the time window modifier
alternative.

Figure 99 HEC-WAT - Create New Time Window Modifier Alternative Dialog Box

Figure 100 HEC-WAT - Time Window Modifier Alternative Editor for Deterministic Simulations

The Time Window Modifier alternative displayed in Figure 100 determines the time of the peak
flow for a given input hydrograph, and establishes a time window. The Time Window Modifier
alternative displayed in Figure 100 was used for the HEC-WAT deterministic simulations. This

149
Chapter 9 – HEC-WAT Model Development (Deterministic) PR-100

alternative began the simulation time window five days prior to the peak and extended the
simulation time window ten days after the peak flow. The compute window was longer on the
recession side of the storm to ensure that the modeling fully captured any system interactions
that could affect flood peaks downstream. Analysis of the deterministic simulation results
determined this was unnecessary, so a separate Time Window Modifier alternative was created for
the HEC-WAT FRA simulations, which limited the time window to five days on either side of the
peak flow.

Create a Program Order

An HEC-WAT program order specifies a sequence of software that will be used in an HEC-
WAT simulation. There is an HEC-WAT default program order, but for the HEC-WAT FIRO
study, two program orders were created. From the HEC-WAT main window, from the WAT
Tools toolbar, click . The Program Order dialog box will open (Figure 101); from this dialog
box, the modeler will configure the program orders. For example, displayed in Figure 101, is the
program order used for all of the deterministic simulations in the FIRO analysis. HEC-HMS
uses its embedded precipitation and evapotranspiration data to generate continuous streamflows
for the entire simulation time window. HEC-ResSim then uses that information to calculate
continuous reservoir releases for the entire simulation time window.

Figure 101 HEC-WAT - Program Order Dialog Box

After HEC-ResSim runs, the TWIG model alternative was used to divide the analysis period into
separate annual events, and the Time Window Modifier model alternative was used to establish
the starting and ending dates for each annual maximum flood event. Subsequent software in the
program order can be run for the entire simulation time window, or just for each event. The
subsequent models in this analysis were HEC-RAS and HEC-FIA, which were only run for the
maximum event in each year.
150
PR-100 Chapter 9 – HEC-WAT Model Development (Deterministic)

Create Analysis Periods

Analysis periods define the start and end of each HEC-WAT simulation, and were defined in the
Analysis Period Editor (Figure 102). For example, the analysis period displayed in Figure 102,
POR_Full, was used in deterministic simulations that covered the full period of available

Figure 102 HEC-WAT - Analysis Period Editor

historical data. For the HEC-WAT FIRO study, three analysis periods were defined. Results for
the deterministic simulations were generated using a water year that spanned 1950 through 2010
(Figure 102). A similar analysis period (POR_Short) covering only the water years 1961
through 1965 was used for troubleshooting, and quick checks before launching a compute for the
full period. The third analysis period (FRA) spanned arbitrary dates for the FRA simulations, as
described in Section 10.6.1.

Creating Simulations
An HEC-WAT simulation is the combination of an HEC-WAT alternative and an analysis
period. By default, HEC-WAT creates a simulation name based on the alternative and analysis
period, but the modeler can create a simulation name. The modeler will select whether a
simulation, is deterministic or an FRA simulation; select a program order; and select the
individual model alternatives for the HEC-WAT simulation. All of this setup is done from the
HEC-WAT Simulation Editor (Figure 103), where the modeler can also adjust the time window
for each model alternative. Once an HEC-WAT simulation has completed, the modeler will
have access to model alternatives and HEC-WAT results.

For example, Figure 103 displays an HEC-WAT deterministic simulation, Base_POR_Full that
is based on the Existing Conditions HEC-WAT alternative. The Base_POR_Full simulation uses
the model alternatives designed for an HEC-WAT deterministic simulation. The HEC-HMS
((RUN(Basin Average 60 yr) and the HEC-ResSim (Baseline) model alternatives compute for the
full time window defined in the analysis period (POR_Full). The HEC-RAS (POR) and HEC-
FIA (SDLLIR) model alternatives compute for only the time window established by the Time
Window Modifier (Annual Peaks_POR) model alternative. The HEC-WAT Simulation Editor
(Figure 103) provides a great deal of flexibility for the modeler.

151
Chapter 9 – HEC-WAT Model Development (Deterministic) PR-100

Figure 103 HEC-WAT - Simulation Editor

Model Linking
After defining a simulation, the inputs and outputs for each model alternative in the program
sequence are coordinated using the HEC-WAT Model Linking Editor (Figure 104). HEC-
HMS is the first model in the program sequence in an HEC-WAT deterministic simulation. For

Figure 104 HEC-WAT – Model Linking Editor

152
PR-100 Chapter 9 – HEC-WAT Model Development (Deterministic)

example, in Figure 104, the Model Linking Editor is displaying the linking for a selected
simulation (Base_POR_Full) and for a specific model alternative (HMS-(RUN)Basin Average 60
yr). This example happens to be for an HEC-HMS model alternative, where for a deterministic
simulation there are no linkages. By default, HEC-HMS will link to all boundary conditions
already established within the HEC-HMS model alternative.

Now for the HEC-ResSim model alternative in the selected simulation (Figure 105), linkages are
a mixture of DSS data and subbasins flows computed by the HEC-HMS model alternative.
HEC-ResSim inputs from DSS files provided data spanning the 1950 to 2010 analysis period.
Lookback storages (i.e., initial reservoir state information) were taken from results of
preliminary simulations, since neither reservoir existed in 1950. Data for Lake Pillsbury, the
Hydrologic Index, and Potter Valley Project were both historical and simulated. The other input
pathnames contained repeating patterns of diversions or losses along the Russian River.

Figure 105 HEC-WAT – Model Linking Editor – HEC-ResSim Model Alternative Linkages

From the Model Linking Editor (Figure 106), the team linked the Healdsburg flow from the
HEC-ResSim model alternative as the hydrograph used by the Time Window Modifier model
alternative. Healdsburg was selected because of its central location along the Russian River.
The HEC-ResSim output hydrograph was selected to account for potential timing differences in
the peak stage, because of different Lake Mendocino releases in the HEC-WAT alternatives.
However, the resulting event time windows rarely differed among the HEC-WAT alternatives.
153
Chapter 9 – HEC-WAT Model Development (Deterministic) PR-100

Figure 106 HEC-WAT – Model Linking Editor - Time Window Modifier Alternative Linkage

Linking for the HEC-RAS model alternative (Figure 107) is mainly to the HEC-HMS model
alternative flow results. The two reservoir outflow hydrographs from the HEC-ResSim model
alternative are also linked to two locations in the HEC-RAS linkages.

Figure 107 HEC-WAT – Model Linking Editor - HEC-RAS Model Alternative Linkages

154
PR-100 Chapter 9 – HEC-WAT Model Development (Deterministic)

Linking for the HEC-FIA model alternative is displayed in Figure 108. All of the inputs to HEC-
FIA came from the HEC-RAS model alternative, and remained the same across all deterministic
simulations.

Figure 108 HEC-WAT – Model Linking Editor - HEC-FIA Model Alternative Linkages

Running an HEC-WAT Deterministic Simulation

Now that the deterministic simulations have been setup, the model is ready to compute HEC-
WAT simulations. The HEC-WAT provides several places within its interface to compute a
simulation, whatever the choice; the Compute Progress dialog box will open (Figure 109). The
Compute Progress dialog box displays progress information and warning messages from the
individual model alternatives of the selected simulation, as well as HEC-WAT information.

Figure 109 HEC-WAT - Compute Progress Dialog Box

155
Chapter 9 – HEC-WAT Model Development (Deterministic) PR-100

The sixty-year HEC-HMS run required about 45 minutes on a laptop with four 2.70 GHz CPU's.
The subsequent HEC-ResSim compute lasted 75 minutes. For each event, the time window
modifier took eight seconds, HEC-RAS ran for 2.5 to three minutes, and HEC-FIA required ten
to fifteen seconds. The total compute time for the HEC-WAT deterministic simulation
Base_POR_Full was five hours.

Viewing Deterministic Results in HEC-WAT

The HEC-WAT interface was used for interactive viewing of results using the standard output
features. Figure 110 displays a convenient check on the Lake Mendocino reservoir modeling
results after running the HEC-WAT simulation Base_POR_Short. From an HEC-WAT
Simulation Map Window, right-click on the reservoir element that represents Lake Mendocino.
From the shortcut menu (Figure 110), the modeler can chose to select HEC-ResSim report and
plots. For example, in Figure 110, the model has chosen to view the HEC-ResSim default plot
for Lake Mendocino.

Figure 110 HEC-WAT - HEC-ResSim Standard Plot

156
PR-100 Chapter 9 – HEC-WAT Model Development (Deterministic)

The HEC-WAT simulation map window also provides shortcut menus that allow modelers to
choose from elements that are common across the various model alternatives in an HEC-WAT
simulation. From an HEC-WAT Simulation Map Window (Figure 111), right-click on the
Hopland Gage location. This location serves as a junction in both HEC-HMS and HEC-ResSim,
so the shortcut menu that displays contains a Junction item, that provides a menu for HEC-HMS
and HEC-ResSim. The shortcut menu also lists several other available elements. HEC-WAT
determines what elements are available in the area and list them in the shortcut menu. Therefore,
from the Hopland Gage Location a modeler could review results from all of the model
alternatives for the selected HEC-WAT simulation.

Figure 111 HEC-WAT – Simulation Map Window - Elements

The modeler can also access various model alternative results from the Results menu of the
HEC-WAT main window (Figure 112). Since the FIRO study workflow involved concurrent
development by several team members, the summary results and interactive map display were
useful for "spot checks" to verify expected model functioning.

Figure 112 HEC-WAT – Results Menu

157
Chapter 9 – HEC-WAT Model Development (Deterministic) PR-100

HEC-WAT Study Directory Structure

As an HEC-WAT study is built, HEC-WAT organizes the directories and files that it creates in
the study directory (Figure 113). In this example, the study directory is /RussianRiver. As data
is entered, other files and subdirectories are created.

Figure 113 HEC-WAT Study Folder Hierarchy for FIRO Study

HEC-WAT modeling results are stored in the runs folder of an HEC-WAT study. A modeler
can access the folder directly from the file structure of the computer or from the HEC-WAT
interface. From the Study Tree, from the Simulations folder, right-click on a simulation, from
the shortcut menu click Browse Run Folder (Figure 114), a File Explorer window (similar to
Figure 113) will open to the runs of the selected HEC-WAT simulation.

Figure 114 HEC-WAT – Simulation Browse Run Folder

158
PR-100 Chapter 9 – HEC-WAT Model Development (Deterministic)

Under the runs folder (Figure 113), the first level of sub-folders organizes results according to
the HEC-WAT alternative. The results for each HEC-WAT alternative are organized into
separate folders for each HEC-WAT analysis period. The combination of an alternative and
analysis period in HEC-WAT defined a simulation, and the corresponding unique set of results.

The structure for a deterministic simulation is displayed in Figure 113; the structure for an FRA
simulation is different. For a deterministic simulation, the structure for results (Figure 113)
consists primarily of a DSS file (simulation.dss) with the same name as the HEC-WAT
simulation, and sub-folders for each event of the simulation. The separate event folders were
created by the TWIG alternative. Each folder contained HEC-RAS and HEC-FIA model results
for the annual maximum flood within each year of the total simulation time window.

DSS Results
The HEC-WAT simulation.dss file serves as the hub of data exchange among the models and
storage of simulation results. The file is accessible from the computer file system, or from the
HEC-WAT main window, from the Results menu; click on Simulation DSS File, which opens a
DSS viewer window.

The simulation.dss file consolidates the output produced by each model alternative across the
analysis period, as specified by the output parameters for each model. This file serves as the
primary source of data for post-processing and analysis of results for a simulation. The version
(F-Part) of the database pathnames is concatenated from an HEC-WAT alternative and analysis
period name, the name of the model, and the name of the model alternative. For example, the
model results in the simulation.dss file (Figure 113) were identified by the following version
information (F-Part):

EXISTING C:POR_SHORT:HMS-(RUN)BASIN AVERAGE 60/

EXISTING C:POR_SHORT:RESSIM-BASELINE/
EXISTING C:POR_SHORT:RAS-POR/
EXISTING C:POR_SHORT:FIA-SDLLIR/

This pathname design allowed convenient handling of data for comparative analysis. For
example, levels for Lake Mendocino for each alternative in the five-year deterministic simulation
were stored as:

//LAKE MENDOCINO-POOL/ELEV/01SEP1960 - 01SEP1965/6Hour/EXISTING C:POR_FULL:RESSIM-BASELINE/

//LAKE MENDOCINO-POOL/ELEV/01SEP1960 - 01SEP1965/6Hour/WO_CVD:POR_FULL:RESSIM-WO_CVD/
//LAKE MENDOCINO-POOL/ELEV/01SEP1960 - 01SEP1965/6Hour/ENCROACH:POR_FULL:RESSIM-ENCROACH/
//LAKE MENDOCINO-POOL/ELEV/01SEP1960 - 01SEP1965/6Hour/COMBINED:POR_FULL:RESSIM-COMBINED/
//LAKE MENDOCINO-POOL/ELEV/01SEP1960 - 01SEP1965/6Hour/ENCROACHWI:POR_FULL:RESSIM-ENCRWIF/

HEC-DSSVue allowed modelers to open five of the HEC-WAT DSS files, then quickly sort and
select the records of interest. The study workflow made frequent use of this ability to quickly
plot and compare results from different alternatives. The design also facilitated other analyses of
the results. For instance, the reservoir levels listed above from the deterministic simulations
were processed through the Cyclic Analysis Math Function of HEC-DSSVue and exported to
Microsoft Excel® for the determination of sampling parameters of initial reservoir levels for the
FRA simulation, as described in Chapter 10 (Section 10.3.2).
159
Chapter 9 – HEC-WAT Model Development (Deterministic) PR-100

Event Folders
Each event simulated with HEC-RAS and HEC-FIA in the deterministic simulations resulted in
sub-folders for the calculations and model-specific output. The HEC-FIA event folders
contained the shapefile results described in Chapter 8 (Section 8.7). In the case of HEC-RAS,
the event folder contained the entire model dataset, which could be opened in HEC-RAS and re-
computed for detailed examination. The gridded hydraulic results from HEC-RAS were also
stored under the event folder for use in analysis and reports.

HEC-WAT Deterministic Simulation HEC-FIA

Analysis Tool
At the time of this study, HEC-WAT did not aggregate HEC-FIA results for deterministic
simulations. However, a utility software was created to process the HEC-FIA events folders
across HEC-WAT alternatives, and provided the results shown in Figure 115. The utility
software displayed the HEC-WAT alternative names and analysis period names in a "tree"
format resembling the directory structure. The tool allowed analysts to select multiple structures
for a simulation and display summary statistics. As displayed in Figure 115, for each structure,
the number of times when each structure was damaged; cumulative damage; maximum amount
of damage; and, which event caused the maximum damage, is provided.

Figure 115 HEC-WAT – HEC-FIA Analysis Tool – Simulation Total Map Display

The utility tool rendered results very quickly, and provided flexible options for loading different
background maps or other shapefiles. Besides aggregating damage from deterministic
simulations, the tool proved useful for fast data analysis without running ArcGIS® or HEC-FIA.

160
PR-100 Chapter 9 – HEC-WAT Model Development (Deterministic)

For example, Figure 116 displays structures in southern Healdsburg damaged by the simulated
1964 flood, rendered with a large red house for the most damage and a small yellow one for the
least.

Figure 116 HEC-WAT – HEC-FIA Analysis Tool - Single Event Map Display

Creating Alternative Simulations

The HEC-WAT Alternative and Simulation Manager allows modelers to create a new
simulation based on an existing one. Defining the PoR_Full analysis period leads to an empty
cell in the HEC-WAT Alternative and Simulation Manager, highlighted in blue in Figure 117.
Right-click on the empty cell, from the shortcut menu, click Create Simulation From. From
the Select Simulation dialog box, select the simulation based on the simulation for the Existing
Conditions alternative and the PoR_Short analysis period. The resulting new simulation is
identical except for the analysis period, including the linking and other options. This process
minimized the selections required to define new simulations, helping avoid mistakes.

Figure 117 HEC-WAT – Alternative and Simulation Manager

161
Chapter 9 – HEC-WAT Model Development (Deterministic) PR-100

Creating new HEC-WAT alternatives, establishes additional rows in the HEC-WAT Alternative
and Simulation Manager, and provides the capability to create new simulations based on
simulations that have already been created. This process makes the simulation identical except
for the HEC-ResSim model alternative. To get the model linking correct in these new
simulations, the modeler used a feature called "dynamic linking" (Figure 118).

Figure 118 HEC-WAT - Model Linking Editor - Dynamic Linking.

The Dynamic Linkage Editor (Figure 119), allows the modeler to replace the two HEC-RAS
inputs from the Baseline HEC-ResSim model alternative with outflows from the HEC-ResSim
model alternative corresponding to the HEC-WAT simulation.

Figure 119 HEC-WAT - Dynamic Linkage Editor

All the HEC-WAT simulations for the FIRO alternatives were created by copying HEC-WAT
simulations based on the Existing Conditions alternative and changing the HEC-ResSim model
alternative and linking. This workflow minimized configuration errors, and helped ensure that
the alternatives being compared held "all else equal".

162
PR-100 Chapter 9 – HEC-WAT Model Development (Deterministic)

Running Multiple Simulations with the Compute

Manager
HEC-WAT provides a Compute Manager to run a batch of unattended deterministic
simulations (Figure 120). For example, in Figure 120, the settings selected would run all five of
the deterministic simulations for the full analysis period.

Figure 120 HEC-WAT - Compute Manager

163
Chapter 9 – HEC-WAT Model Development (Deterministic) PR-100

164
PR-100 Chapter 10 – Flood Risk Analysis Simulation

Flood Risk Analysis Simulation

The FRA simulations extended the FIRO study to consider storms outside the events experienced
in the period of record. The probabilistic analysis generated synthetic storms and modeled the
consequences, allowing a more robust understanding of the frequency of relevant watershed
outcomes. This analysis captured the natural variability of flooding in the watershed by
sampling parameters for the precipitation frequency curve, hyetograph shape, and date of the
flood event, and initial conditions regarding soil moisture and reservoir storage. For the sake of
simplicity, other parameters within the individual models, such as Manning's n values or depth-
damage relationships, for the FIRO analysis, were not considered for the Monte Carlo analysis.

For computational efficiency during the Monte Carlo simulation, the models used for the FRA
simulations were modified to include only the calculations and results required for flood event
analysis. The FRA simulations also skipped latter parts of the program sequence for events with
very little potential for causing damage.

The hydrologic sampling generated a dataset of 5,000 annual maximum precipitation events.
HEC-WAT provided the date and time-window of each generated event to the other models in
the HEC-WAT program sequence, aligning the timelines of each simulation. The event date was
also important for the two models that were performing parameter sampling in the FIRO study,
informing initial soil moisture conditions in the HEC-HMS model and initial reservoir levels in
the HEC-ResSim model.

The HEC-HMS model computed the streamflows resulting from the synthetic hyteographs
produced by the hydrologic sampling, and provided the streamflows as input to the HEC-ResSim
and HEC-RAS models. The HEC-ResSim model generated reservoir outflows that were input to
the HEC-RAS model. The HEC-RAS model computed downstream flows, stages and depth
grids, which were input to the HEC-FIA model and used to estimate damage to structures and
their contents.

To allow reproducible results for well-aligned comparison of alternatives, the HEC-WAT

hydrologic sampling algorithm produced the same series of synthetic storm events for each
HEC-WAT FRA simulation. The HEC-HMS model then sampled the same series of starting soil
moisture storages for the events. This capability was accomplished using an HEC-WAT feature
that allows modelers to maintain the same pseudo-random seeds for each FRA simulation.
Consequently, the same series of hydrologic events were provided to each HEC-WAT
simulation, in other words, the only difference among the simulations was the water management
alternative. The consistent pseudo-random seeds also allowed the HEC-ResSim model to sample
the starting reservoir level at the same cumulative probability for each FRA event, but since the
distributions differed for each water management alternative, the resulting initial storage values
differed as well.

HEC-WAT FRA simulations were created, with all new model alternatives that were built
specifically for FRA simulations. Specifically, the HEC-ResSim model alternative is defined for
165
Chapter 10 – Flood Risk Analysis Simulation PR-100

modeling forecast-informed operations. Advantages or disadvantages of the reservoir operation

alternatives are quantified as differences with respect to EAD or 1% AEP flow at Guerneville.

Meteorologic Data Development

In HEC-WAT, the hydrologic sampling process provides system-wide flood events for an HEC-
WAT FRA simulation. For precipitation sampling, the hydrologic sampling process requires a
distribution of event probability date, a basin-average precipitation frequency curve, and
hyetograph shape sets, which includes hyetographs for all subbasins in the HEC-HMS model of
the FIRO watershed. These hydrologic sampling inputs were developed from historical
precipitation and runoff data.

For the FIRO study, an eight-day basin average precipitation frequency curve was used for
randomly generating events. An eight-day period was selected based on a critical duration
analysis of the ten largest historical flood events. A new precipitation frequency curve was
generated once per realization to account for knowledge uncertainty when defining the
precipitation frequency curve. HEC-WAT sampling parameters are described in Section 5.1.3.
For the FIRO study, each realization had 500 flood events with ten realizations computed per
HEC-WAT FRA simulation.

For each event the hydrologic sampling process, first samples an event date from the distribution
describing event date (the flood season). Next, an eight-day precipitation magnitude from the
basin-average precipitation frequency curve is sampled from a subbasin hyetograph shape from
user specified precipitation shape sets. A shape set consisted of one hyetograph for each subbasin
in the HEC-HMS model, which as a set, captured a temporal and spatial rainfall pattern. For
example, precipitation from the 1964 event was used to create a single eight-day precipitation
shape set containing precipitation hyetographs for each subbasin. The hydrologic sampling
process sampled event precipitation depth with a sampled hyetograph shape set, creating scaled
hyetographs for each subbasin. (The shape set is scaled in order for the basin average to equal the
sampled event precipitation depth.) HEC-WAT supplied the scaled hyetographs from the
hydrologic sampling process to the HEC-HMS model, and then the hyetographs were applied to
the subbasin elements within the HEC-HMS basin model.

Critical Duration
The critical duration analysis informed event duration for subsequent precipitation frequency and
precipitation shape set analyses. The critical duration was selected as the event duration
producing the greatest flood impact.

Ten major historical events in the Russian River basin were evaluated to inform the critical
duration estimate. Plots of incremental precipitation and flow were reviewed to assess the
temporal extent of historical storms. Graphical inspection distinguished the start and end time of
events. Flows at Hopland, Healdsburg, and Guerneville (Hacienda) were examined to verify
similar runoff responses at different locations on the river. Time values were based on
observations for Guerneville, since it showed the longest response time due to its downstream
position, and best represented the flood damage center.

166
PR-100 Chapter 10 – Flood Risk Analysis Simulation

Table 25 provides a summary of major event duration. Figures 121 through 130 illustrate the
precipitation and discharge for each event, with the shaded period indicating the event duration.
Historical analysis yielded an average duration for major events of 8.1 days, so subsequent
precipitation frequency and precipitation shape set analyses assumed an eight-day event duration.

Table 25 Historical Event Critical Duration Analysis

Start End Duration
Event Time Time (days)
Dec 1955 15-Dec-1955 12:00 23-Dec-1955 12:00 8.0
Dec 1964 18-Dec-1964 12:00 24-Dec-1964 12:00 6.0
Jan 1978 12-Jan-1978 00:00 19-Jan-1978 12:00 7.5
Jan 1983 22-Jan-1983 00:00 30-Jan-1983 00:00 8.0
Feb 1986 12-Feb-1986 00:00 21-Feb-1986 00:00 9.0
Jan 1995 03-Jan-1995 00:00 15-Jan-1995 00:00 12.0
Mar 1995 08-Mar-1995 00:00 15-Mar-1995 00:00 7.0
Dec 1996 26-Dec-1996 00:00 03-Jan-1997 00:00 8.0
Feb 1998 31-Jan-1998 00:00 08-Feb-1998 00:00 8.0
Dec 2005 26-Dec-2005 00:00 02-Jan-2006 00:00 7.0

Figure 121 December 1955 Event

167
Chapter 10 – Flood Risk Analysis Simulation PR-100

Figure 122 December 1964 Event

Figure 123 January 1978 Event

168
PR-100 Chapter 10 – Flood Risk Analysis Simulation

Figure 124 January 1983 Event

Figure 125 February 1986 Event

169
Chapter 10 – Flood Risk Analysis Simulation PR-100

Figure 126 January 1995 Event

Figure 127 March 1995 Event

170
PR-100 Chapter 10 – Flood Risk Analysis Simulation

Figure 128 January 1997 Event

Figure 129 February 1998 Event

171
Chapter 10 – Flood Risk Analysis Simulation PR-100

Figure 130 December 2005 Event

Flood Season Distribution

The flood season analysis determined the likelihood that a major event occurs in a given month.
An hourly time series of an eight-day precipitation accumulation yielded a population of major
events that informed the seasonal distribution. The HEC-DSSVue Find Peaks Tool filtered the
hourly record of eight-day precipitation accumulation finding 32 major events, as shown in
Figure 131.

Events were binned by month and a histogram of event occurrence was created. Figure 132
shows the flood season probability distribution that was entered in the HEC-WAT Hydrologic
Sampling Editor (Figure 132), on the Event Sampling tab. An empirical distribution was
selected and the cumulative probability values with corresponding date were entered. This
distribution was later sampled to generate the date of each flood event produced by the FRA
simulation.

Precipitation Frequency
The precipitation frequency relationship provides the precipitation volume corresponding to a
sampled exceedance probability. For the FIRO study, the precipitation frequency analysis
analyzed Russian River basin average eight-day precipitation accumulations. An area average of
incremental precipitation from each of the Russian River subbasins yielded an hourly record of
basin average precipitation for the study period, and for water years 1951 through 2010. The

172
PR-100 Chapter 10 – Flood Risk Analysis Simulation

Figure 131 Peaks Found using the HEC-DSSVue Find Peaks Tool

Figure 132 HEC-WAT - Hydrologic Sampling Editor - Event Sampling Tab

173
Chapter 10 – Flood Risk Analysis Simulation PR-100

hourly record of basin average incremental precipitation was accumulated over eight days and
the annual maximum eight-day precipitation depths were extracted. A GEV distribution was
estimated from the annual maximum series. Table 26 provides the fitted location, scale, and
shape parameters of the estimated GEV distribution.

Table 26 GEV Parameters for Precipitation Frequency Curve

Parameter Value
Location 7.527
Scale 2.714
Shape 0.057

Figure 133 displays the precipitation frequency curve entered in the HEC-WAT Hydrologic
Sampling Editor (Figure 133) on the Frequency Curve Tab, from the Distribution Type List
(Figure 133), GEV was selected and the distribution parameters were defined. A sixty-year
equivalent record length was specified for defining the uncertainty around the estimated
frequency curve. The equivalent years of record is used when re-sampling a precipitation
frequency curve for each realization within a simulation. In the case of sixty years, sixty-points
will be randomly sampled from the user entered frequency curve and then a new GEV
distribution will be fit to the sixty data points. This analysis chose to include ten realizations for
each FRA simulation; therefore, ten different precipitation frequency curves were re-sampled
from the user defined precipitation frequency curve.

Figure 133 HEC-WAT - Hydrologic Sampling Editor - Frequency Curve Tab

174
PR-100 Chapter 10 – Flood Risk Analysis Simulation

Hyetograph Locations
The basin average precipitation for an event was disaggregated to individual subbasins, resulting
in separate precipitation magnitudes for each location. From the HEC-WAT Hydrologic
Sampling Editor (Figure 134), on the Hyetograph Location Tab, the modeler is designating
the locations for which precipitation hyetographs must be produced from the randomly sampled
basin average precipitation. Specifying the area for each location allows the hydrologic
sampling process to ensure that the weighted average of the locations remains consistent with the
basin average precipitation.

Figure 134 HEC-WAT - Hydrologic Sampling Editor - Hyetograph Location Tab

175
Chapter 10 – Flood Risk Analysis Simulation PR-100

Precipitation Shape Sets

Precipitation shape sets were defined to provide sets of hyetograph shapes for sampled storms
that captured a temporal and spatial pattern for each event. The individual subbasin hyetographs
that comprise a shape set are scaled to reproduce the randomly sampled basin average
precipitation depth. Figure 135 displays sample hyetographs from the shape set for the
December 1955 event at the WF Russian and Russian 10 subbasins. The WF Russian subbasin is
the most northern catchment, while the Russian 10 subbasin represents the lowest reach from the
ocean to the city of Guerneville. The timing of hyetograph peaks during the 1995 event was
similar across the watershed, but the lower portions of the basin received heavier precipitation.

Figure 135 Shape Set Hyetographs for December 1955 Event at Russian 10 and WF Russian Subbasins

Historical precipitation from ten major events were used to define the hyetograph shape sets.
Table 25 summarized the historical events. The shape sets assume an eight-day event duration.
Table 27 displays the eight-day period selected for each historical event defining precipitation
hyetographs for shape sets. In HEC-DSSVue, using the Math Tool, the modeler extracted the
hyetograph shape for each of the nineteen subbasins in the HEC-HMS model for each of the
historical events. After sampling an event date, the hydrologic sampling process determines the
simulation start and end times corresponding to the temporal extent of the precipitation shape set.
Preceding and trailing buffers were added to the hyetograph shapes to extend the time window
used by the FIRO model alternatives. Three days of no precipitation preceded the hyetograph
providing a warm-up period prior to the event. Ten days of no precipitation followed the
hyetograph allowing time for excess precipitation to route through the reservoir and downstream.

176
PR-100 Chapter 10 – Flood Risk Analysis Simulation

Table 27 Shape Set Definition

Start End Duration
Event Time Time (days)
Dec 1955 16-Dec-1955 00:00 24-Dec-1955 00:00 8
Dec 1964 18-Dec-1964 19:00 26-Dec-1964 19:00 8
Jan 1978 12-Jan-1978 00:00 20-Jan-1978 00:00 8
Jan 1983 22-Jan-1983 00:00 30-Jan-1983 00:00 8
Feb 1986 12-Feb-1986 00:00 20-Feb-1986 00:00 8
Jan 1995 06-Jan-1995 11:00 14-Jan-1995 11:00 8
Mar 1995 07-Mar-1995 12:00 15-Mar-1995 12:00 8
Dec 1996 26-Dec-1996 00:00 03-Jan-1997 00:00 8
Feb 1998 31-Jan-1998 00:00 08-Feb-1998 00:00 8
Dec 2005 25-Dec-2005 23:00 02-Jan-2006 23:00 8

A shape set for the December 1955 event that was provided to the HEC-WAT Hydrologic
Sampling Editor, which was defined on the Shape Sets tab, is displayed in Figure 136. The
DSS records containing each hyetograph for the shape set are identified in the table (Figure 136).
Note: that the FIRO study assumed that each historical shape set could be appropriately scaled to
any basin average magnitude, so the exceedance probability filter controls were set for the shape
sets allowing their selection for any size event. The FIRO study also made no assumptions
regarding the relative likelihood of shape sets, and assigned them all equal weights.

Figure 136 HEC-WAT - Hydrologic Sampling Editor - Shape Sets Tab

177
Chapter 10 – Flood Risk Analysis Simulation PR-100

Hydrologic Modeling Uncertainty Analysis

A modified version of the calibrated HEC-HMS model was prepared for the HEC-WAT FIRO
study. The HEC-HMS model used for the deterministic simulations could also have been used
for the FRA simulations, but the modelers chose to create a separate version customized for
efficient execution of the specific flood risk output required for the FIRO study. All event
simulations and corresponding basin models, meteorologic models and control specifications
were removed from the HEC-HMS model, eliminating unnecessary computations, reduced input
data requirements, and minimized disk input/output (I/O). Simplifying the HEC-HMS model
also reduced the chance for modeler mistakes, and made linking the HEC-HMS model within the
HEC-WAT FIRO study efficient.

For the modified FRA HEC-HMS model, an HEC-HMS simulation type of Uncertainty
Analysis was added. The HEC-HMS uncertainty analysis (Figure 137) for the FRA simulations
included a modified basin model (FRA), a modified meteorologic model (Basin Average FRA),
and revised Modified Puls storage discharge relationships. The HEC-HMS Start and End Dates
(Figure 137) and Start and End Times (Figure 137) are somewhat arbitrary for an uncertainty
analysis simulation within HEC-HMS. The dates and times should line up with the meteorologic
data referenced within the meteorologic model. The Time Interval (Figure 137) is the time step
within the simulation and Total Samples (Figure 137) is the number of times HEC-HMS will
sample model parameters (including initial conditions) within the uncertainty analysis. When
HEC-HMS is part of an HEC-WAT FRA simulation, Total Samples should be set to one since
HEC-HMS uncertainty analysis will sample model parameters once per event in an HEC-WAT
FRA simulation.

Figure 137 HEC-HMS Uncertainty Analysis

Basin and Meteorologic Model Changes

The HEC-HMS uncertainty analysis and corresponding basin and meteorologic models were
prepared as follows:

• Meteorologic Model. A new HEC-HMS meteorologic model, Basin_Average_FRA was

created for the HEC-WAT FRA simulation, from the HEC-HMS deterministic
178
PR-100 Chapter 10 – Flood Risk Analysis Simulation

meteorologic model Basin Average 60 yr. The evapotranspiration method was set to
None. The FRA simulation was based on events lasting only fifteen days, so modeling
evapotranspiration was unnecessary.

• Basin Model. A new HEC-HMS basin model, FRA was created for the HEC-WAT FRA
simulation, from the HEC-HMS deterministic basin model Continuous. Observed flow
was set to None for all locations with observed flow because there was no observed flow
for the hypothetical events.

• Storage Discharge. Modified Puls storage-discharge relationships were extrapolated for

reaches below Cloverdale and on Dry Creek to handle extreme events sampled during the
HEC-WAT hydrologic sampling process. Storage discharge relationships were updated in
the Routing.dss file (located in the data subdirectory of the hms folder), by manually
maintaining the slope of the storage-discharge curve at the upper end.

Parameter Sampling for Initial Soil Moisture

As explained in Section 5.2.1, subbasins in the HEC-HMS deterministic model used the Deficit
and Constant loss method, except for subbasin EF Russian 20, which used the SMA loss method.
Frequency analysis of daily values extracted from the sixty-year deterministic simulation
provided empirical distributions of soil moisture for each month. Among distributions available
in HEC-HMS that can be specified as seasonal, the Beta distribution best represented soil
moisture uncertainty. The Beta distribution parameters, alpha and beta, were estimated to
approximate the Empirical distributions from the deterministic simulation (Figure 138).

Figure 138 Calibrated Beta Distributions for Computed Soil Moisture for WF Russian Subbasin

179
Chapter 10 – Flood Risk Analysis Simulation PR-100

Beta distribution parameters for the relevant soil moisture methods were entered in the HEC-
HMS model for each month. Since subbasin EF Russian 20 used the SMA loss method, monthly
distributions were initial soil content. Subbasin WF Russian used the Deficit and Constant loss
method, so in the HEC-HMS uncertainty analysis model, monthly distributions for subbasin WF
Russian were set for initial deficit in the HEC-HMS model (Figure 139). The remaining
subbasins in the HEC-HMS model were linearly related to the initial deficit sampled for the WF
Russian subbasin. The HEC-HMS software provides an option that linearly relates uncertainty
parameters by applying a linear regression relationship. A linear regression with error term was
defined for the other remaining Deficit and Constant loss subbasins in order to compute similar
initial loss values (Figure 140). The slope of the regression was determined by dividing the
maximum deficit for a given subbasin by the maximum deficit of subbasin WF Russian.

Figure 139 HEC-HMS - Uncertainty Parameter Specified for Subbasin WF Russian

Figure 140 HEC-HMS - Uncertainty Parameter for Subbasin Russian 70 - Specified as a Linear
Regression of Subbasin WF Russian

180
PR-100 Chapter 10 – Flood Risk Analysis Simulation

The maximum deficit for each subbasin was determined during the period of record calibration.
If the regression slope is greater than one, the initial deficit will be greater than the value
sampled for subbasin WF Russian. For example, if the regression slope is less than one, the
initial deficit will be less than the value sampled for subbasin WF Russian. The varying slope
values preserve the spatial pattern of deficit found during model calibration. The error term of
the regression was de-emphasized because calibration results showed a strong linear relationship
across subbasins.

Selected HEC-HMS Outputs

By default, the HEC-HMS software does not save output from an HEC-HMS uncertainty
analysis. Therefore, the user must choose locations and output time-series to be saved, from the
HEC-HMS Uncertainty Analysis Editor (Figure 141). For the FIRO study, it was necessary to
choose all locations where results from HEC-HMS could be linked to the HEC-ResSim or HEC-
RAS models.

Figure 141 HEC-HMS - Uncertainty Analysis Editor

Reservoir Modeling
The event-based modeling approach implemented for the FRA simulations focused on flood-risk.
Seasonal reservoir modeling considerations, such as refill success or the effect of summer
depletions, were represented in the starting storage of the reservoir. The HEC-ResSim model
alternatives used for the FRA simulations was based on the network developed for the
deterministic simulation, but modified for efficient modeling of flood operations. All elements
and rules not related to flood operations were removed, and outputs were minimized.

The HEC-ResSim network for the FRA simulation contained a separate operation set for each of
the four HEC-ResSim model alternatives used in the FRA simulations. The HEC-ResSim
network for the FRA simulations also included Monte Carlo parameters for each HEC-ResSim

181
Chapter 10 – Flood Risk Analysis Simulation PR-100

model alternative to sample initial reservoir storage. The network also contained logic for
randomly perturbing the precipitation forecast information used by the EncroachWIF model
alternative.

Network Changes for Efficient Flood Operation

Modeling
For the FRA simulations, the HEC-ResSim diversion elements were removed from the HEC-
ResSim network, since the winter withdrawals are insignificant in relation to river flows during
flood operations. The SCWA Diversion Dam was also deleted from the network. The operation
sets for each water management alternative remained the same, except that the minimum flow
rules for Coyote Valley Dam were removed, as well as rules and state variables related to the
hydrologic index. The SWRCB Decision 1610 rules were also removed from the Warm Springs
Dam. The Potter Valley Project was assumed not to be operating during flood events, so the
inflow diversion values were set to zero.

To reduce computational activity and the size of the saved HEC-ResSim results for the FRA
simulations, the output settings for the HEC-ResSim model alternatives were set to save only the
results necessary to the study metrics and other models. For example, Figure 142 displays the
output settings regarding HEC-ResSim junction elements (ResSim Alternative Editor, DSS
Output tab, Junctions tab).

Figure 142 HEC-ResSim Output Selection for FRA Simulations

182
PR-100 Chapter 10 – Flood Risk Analysis Simulation

Sampling of Starting Reservoir Levels

The distribution of starting reservoir levels is an example of natural variability in the watershed
transformed according to a given water management alternative. For FRA simulations, the
lookback elevation of Lake Mendocino was varied for each event by randomly sampling from a
set of seasonal empirical distributions of pool elevation developed for each alternative. For
simplification, the starting level of Lake Sonoma was fixed at 451.5 feet NGVD (top of
conservation). Varying the initial pool was unnecessary since it was very unlikely to affect
releases from Warm Springs Dam, and the assumptions are held equal for each Lake Mendocino
water management alternative.

The seasons for the Lake Mendocino empirical distributions were determined by deriving a daily
time-series from the six-hour pool levels resulting from the 1950 through 2010 results, and
performing a seasonal frequency analysis. The modelers examined a cyclic analysis that was
computed with HEC-DSSVue, which summarized variability in pool levels by date to identify
periods where the pool levels displayed consistency regarding median value, standard deviation,
and the spread between minimum and maximum. Figure 143 displays the cyclic analysis for the
Existing Conditions alternative. The blue line (Figure 143) indicates the mean Lake Mendocino
level for each day of the year, while the magenta line (Figure 143) indicates the median. The
other lines (Figure 143) display the highest, the 95th percentile, the 5th percentile, and the lowest
levels for each day of the year. For example, in Figure 143 empirical distributions were
calculated for each month of the flood season, except December and March. Levels in these
transitional months showed greater change in variability across the month, and were
characterized in half-month seasons.

Figure 143 HEC-DSSVue Cyclic Analysis of the Existing Conditions Alternative's Deterministic Pool
Levels
183
Chapter 10 – Flood Risk Analysis Simulation PR-100

The empirical probability distributions for each season were estimated in Microsoft Excel®
based on daily values from the sixty-year deterministic results. Figure 144 displays the general
shape of the distributions for each season using histograms. The Existing Conditions
alternative's pool levels in November and the first half of December show an unexplained
bimodal distribution, but almost entirely fall below the winter pool of 737.5 feet NGVD,
including a few extremely low levels. The reservoir (Coyote Valley Dam) elevations begin
tending toward the winter pool level in the second half of December, and are usually very close
to the pool level during January, February, and the first half of March, when the range becomes
narrowest. The higher elevations of the distributions spread during the spring as the reservoir
attempts to fill to the summer pool of 761.8 feet NGVD.

Figure 144 Empirical Distributions for Deterministic Existing Conditions Alternative Coyote Valley
Dam Levels
184
PR-100 Chapter 10 – Flood Risk Analysis Simulation

A similar frequency analysis was performed for the period-of-record Coyote Valley Dam level
results from the Encroach and Combined water management alternatives. For example, Figure
145 displays the distribution graphics for the seasons chosen for the Combined water
management alternative. As would be expected from the different guide curve, the levels tend to
be higher than under the Existing Conditions alternative in November, but similar in January and
the first half of February. The Combined water management alternative levels are higher than
those under the Existing Conditions alternative for the second half of February and later,
allowing the FRA simulations to model the impact of the reduction in late winter flood control
storage under the Combined water management alternative.

Figure 145 Empirical Distributions for Deterministic Combined Alternative Coyote Valley Dam Levels

The starting storage distributions for the Encroach water management alternative were applied to
the EncroachWIF alternative, since the deterministic results for the two alternatives matched so
closely.

The random sampling of initial reservoir level was configured in the HEC-ResSim Alternative
Editor, from the Run Control tab, from the Alternative Type box; the modeler selected Monte
Carlo for the four FRA HEC-ResSim model alternatives (Baseline_F, Encroach_F,
Combined_F, EncroachWIF_F). The Lake Mendocino lookback elevation was set to a constant
185
Chapter 10 – Flood Risk Analysis Simulation PR-100

value of 761.8 feet NGVD in the Lookback tab (HEC-ResSim Alternative Editor) for all FRA
alternatives (the value was replaced during sampling). The Monte Carlo tab of the HEC-ResSim
Alternative Editor provides a wizard (Figure 146) for creating a random variable to replace the
constant lookback elevation with a sampled value (with the distribution described in Figure 145,
according to the natural variability of the data), as shown in Steps 1 and 2 in Figure 146. Steps 3
and 4 of the Random Variable Wizard (Figure 146) allowed the modeler to define the
variability according to user-defined seasons, and enter the exceedance probabilities into a table
(Step 4, Figure 146).

Figure 146 HEC-ResSim Random Variable Wizard

The FRA simulation results of each alternative were reviewed to verify that the sampling
produced populations of initial Coyote Valley Dam levels that matched the input distributions.

Additionally, spot checks verified proper handling of the pseudo-random seeds for a given event.
For example, Event 37 from Lifecycle 79 started on 15 January, and sampled an exceedance
probability of 0.80. The starting Lake Mendocino levels for the same event under the Existing
Conditions, Encroach, and Combined alternatives were determined by interpolating the 0.80
exceedance probability from the respective distributions containing that date. The Monte Carlo
sampling feature of HEC-ResSim allowed the FIRO analysis to reflect the likelihood of much
higher antecedent levels for Lake Mendocino at the start of FRA events under the FIRO
alternatives than for the Existing Conditions alternative.

186
PR-100 Chapter 10 – Flood Risk Analysis Simulation

Hydraulic Modeling
The HEC-RAS model alternative (i.e., HEC-RAS plan) used for the FRA simulations was
identical to the one used for the deterministic simulation, except for reduced output. The name
of the HEC-RAS model alternative is FRA. In the HEC-RAS Unsteady Flow Analysis Editor
(Figure 65), the hydrograph output interval remained set to hourly, for consistency with the
HEC-HMS model alternative. However, the mapping output interval was set to Max Profile,
since only maximum depth output was required. The HEC-RAS detailed output interval was set
to 1 Month, which avoided writing unneeded output for the fifteen-day FRA simulations. The
HEC-RAS Stage and Flow Output Locations Selector (Figure 66) included only cross-sections
for the gage locations used for comparison metrics. The RAS Mapper Tool produced only
maximum depth grid.

Consequence Modeling
The HEC-FIA model alternative, SDDepthGridOnly, used for the FRA simulations only
calculated flood damage for structures and contents for each event. The life loss and impact
response calculations were removed, and the damage calculations proceeded as a grids-only
compute using just the depth grid computed by HEC-RAS.

HEC-WAT FRA Modeling

The FRA simulations for the FIRO study were configured to simulate 5,000 events; this number
of events (Section 4.7.1) provides adequate information for estimating the 1% ACE flow/stage
estimates and EAD for each 500-year realization. For the 1% ACE flow/stage estimates and
EAD results, uncertainty estimates were not done during the FIRO analysis because of the
limited number of realizations that were used during the analysis.

FRA Analysis Period

Before creating FRA simulations, the time window (analysis period) was defined. Figure 147
displays the analysis period used for the FRA simulations: 01Jan3001 - 31Dec3050. The use of
calendar years instead of water years aligned with the input distribution formats for the
hydrologic sampling, HEC-HMS, and HEC-ResSim model alternatives. The choice of years far
into the future reinforced the synthetic nature of the FRA simulations, and helped avoid
confusion with historical events or the deterministic simulation results.

The FIRO study adopted a time window of fifty years, which is commonly used in USACE
planning studies, even though lifecycle planning was not part of the study. Consequently, to
produce 5,000 events, the FRA simulations each involved 100 lifecycles, contained within ten
realizations.

187
Chapter 10 – Flood Risk Analysis Simulation PR-100

Figure 147 HEC-WAT - Analysis Period for FRA Simulations

Defining FRA Simulations

The FRA simulations were identical except for the HEC-ResSim model alternative selected.
Simulation parameters were initially defined for the HEC-WAT Base FRA simulation (Figure
148). Using the HEC-WAT Alternative and Simulation Manager, the study team was able to
create other simulations from the Base FRA simulation, and substitute the appropriate HEC-
ResSim model alternative for each new FRA simulations (Section 4.10).

Figure 148 Creating a New FRA Simulation – Base FRA

For the FIRO study, the team selected ten realizations (Figure 148), with the hydrologic
sampling process capturing knowledge uncertainty by sampling a new precipitation frequency
curve for each realization; each realization contained 500 annual floods.
188
PR-100 Chapter 10 – Flood Risk Analysis Simulation

The FIRO FRA simulations used a different Program Order from the deterministic simulations
(Figure 148). Since the FIRO study analyzed single annual events, the Time-Window Interval
Generator was not relevant for the FRA simulations. In addition, the FRA simulations used a
different Time Window Modifier alternative. The deterministic results established that
modeling fifteen-days with HEC-RAS was unnecessary to capture the downstream flood peak.
To reduce compute time, the Time Window Modifier alternative for the FRA simulations
covered only ten days.

The relative comparisons performed for the FIRO study required simulations for each FIRO
alternative to model the same FRA flood events. The HEC-WAT design supports this
requirement using "pseudo-random" sampling, which ensures that models share identical initial
seeds and sampling parameters will produce identical Monte Carlo sampling results. The HEC-
WAT Simulation Editor (Figure 149) for an FRA simulation displays the seed values used for
all the FRA simulations. When creating the FRA simulations from the Base FRA simulation, the
seed values from the Base FRA simulation were copied to the new FRA simulation.

Figure 149 HEC-WAT – Simulation Editor - Initial Seeds for FRA Simulations

189
Chapter 10 – Flood Risk Analysis Simulation PR-100

The Output tab of the HEC-WAT Simulation Editor (Figure 149) allows modelers to specify
filename filters for files to preserve, such as "*.dbf". By default for FRA simulations, HEC-
WAT runs the models for individual events in a temporary workspace, and optionally deletes all
the model files after successful computes. For final, or production, simulations, the option to
delete all model files, and only save results, should be selected to avoid overwhelming computer
storage.

FRA Linking
The HEC-WAT Model Linking Editor (Figure 150) setup for the FRA simulations generally
corresponds to the model linking described for the deterministic simulations. The main
difference was linking the hydrologic sampling output precipitation to the corresponding HEC-
HMS model inputs, as shown in Figure 150.

Figure 150 HEC-WAT – Model Linking Editor - FRA Simulation - HEC-HMS Linking

Other differences between the model linking for the deterministic and FRA simulations occurred
because versions of the HEC-ResSim and HEC-FIA model alternatives required fewer inputs for
the FRA simulations.
190
PR-100 Chapter 10 – Flood Risk Analysis Simulation

Skip Rules
Not all of the synthetic events in the FRA simulations were large enough to cause flood damage.
In order to reduce compute time, an HEC-WAT Model Skip Rule alternative was defined for the
FRA simulations. The Model Skip Rule alternative was setup so that HEC-RAS and HEC-FIA
did not run for non-damaging flows in the Russian River. For example, in Figure 151, the
HEC-WAT Model Skip Rules Editor displays a simple conditional rule in Row 1 of the table.

Figure 151 – HEC-WAT – Model Skip Rules Editor

The condition is true for an event in the Base_FRA simulation if the peak flow at the Guerneville
gage (computed by the HEC-ResSim model alternative) remained below 30,000 cfs. The actions
specified for the true condition were to skip the HEC-RAS and HEC-FIA model runs. The same
rule was created for the FRA simulations for the Encroach, Combined, and EncroachWIF
alternatives.

The non-damaging flow of 30,000 cfs was inferred from the deterministic simulation results.
The lowest non-zero damage amount in the Existing Conditions results was $195,000 for 1973,
and damage was incurred at two structures in Guerneville. Other results and anecdotal historical
information indicated that Guerneville was the most easily damaged location along the river.
The "barely damaging" HEC-ResSim flow corresponding to the 1973 event was 32,000 cfs. The
flows computed for Guerneville by the HEC-HMS model were confirmed as similar (Figure
152). The HEC-ResSim flow was selected for use in the rule because the flow included releases
from Coyote Valley Dam.
191
Chapter 10 – Flood Risk Analysis Simulation PR-100

Figure 152 Determination of Non-Damaging Flow – Guerneville Gage

Output Variable Editor

HEC-WAT aggregates results from the FRA simulations according to selections defined in the
Output Variable Editor (Figure 153). The Output Variable Editor allows modelers to specify
which model results HEC-WAT should save (simulation.dss file) and whether to generate
frequency curves for each output using the maximum value for each event. HEC-WAT
computes a separate frequency curve for each realization, and in addition, computes the full
frequency curve from all realizations for each output, and the 5- and 95-percent confidence limits
(although the 90% confidence interval is not useful with only ten realizations being simulated).

The outputs selected for the FIRO study included all of the results used for the comparison
metrics: Lake Mendocino levels from the reservoir model; flows and stages for key Russian
River locations from the hydraulic model; and, flood damage from the consequence model.
Results were also reviewed for basin average precipitation from the hydrologic sampling
process, and Lake Mendocino inflow computed by the hydrologic model (HEC-HMS).

The other outputs selected from HEC-HMS and HEC-ResSim relate to sampling of initial soil
moisture and starting reservoir storage, for use in validating the FRA simulation results.

192
PR-100 Chapter 10 – Flood Risk Analysis Simulation

Figure 153 HEC-WAT - Output Variable Editor

Running the FRA Simulations

HEC-WAT was designed to accommodate a wide range of study requirements and complex
models and computer environments, and modelers should expect to refine the models and
simulation parameters over the course of trial FRA simulations. Initiating an FRA simulation
brought up the dialog shown in Figure 154, which allows control over conduct of the simulations
and the output produced.

The FRA simulations were configured to stop in case of an error, ensuring that the results were
complete and successful. HEC-WAT allows modelers to run an entire FRA simulation, or only
specified lifecycles or realizations. Both of these options were used in the course of the FIRO
study. The ability to resume FRA simulations at specified lifecycles proved especially useful for
occasions when the simulations were interrupted due to factors external to HEC-WAT, such as
forced re-boots or resource constraints on a machine.

The workflow of the FIRO study involved numerous provisional FRA simulations of the
Existing Conditions alternative, leading to a subsequent final simulation. The provisional
simulations of the Existing Conditions alternative identified numerous issues during the course of
simulating 5,000 events. Sometimes limitations of the models calibrated using deterministic
data were exposed by the wider variety of hydrologic scenarios in the FRA simulation. For
example, the scale of synthetic events exceeded the initial definitions for the Mod-Puls storage-
discharge paired data curves in the models. Other provisional computes allowed modelers to
leverage improved understanding of the system as they analyzed provisional results, refined
choices, and corrected oversights.

From the HEC-WAT Run FRA Simulation dialog box (Figure 154) the model has two options
available on what to do with FRA simulation results. The Save Lifecycle DSS Files option
(Figure 154) is where only hydrographs and other results from the individual models for each
193
Chapter 10 – Flood Risk Analysis Simulation PR-100

Figure 154 HEC-WAT - Run FRA Simulation Dialog Box

event are saved. The Save Model Folders option (Figure 154) is where everything from an FRA
simulation is saved for each event. By default, the Save Model Folders option is not selected to
conserve computer storage. During the FIRO study, this option was occasionally used to provide
full model results per event when issues arose. The information was helpful for troubleshooting
and identifying model improvements required for some synthetic events. In other cases, the
Save Model Folders option was used to provide detailed information about the flood risk
impacts illustrated by certain events.

Another FRA simulation option available to the modeler is the ability to stop and restart the
software applications (plug-ins) that are part of the program sequence of a simulation. From the
HEC-WAT Run FRA Simulation dialog box (Figure 154), from the Restart Plugins box, the
modeler can restart the plug-ins for HEC-HMS, HEC-ResSim, and HEC-RAS between events.
For the final FRA simulation for the FIRO study, all plug-ins were selected in order to minimize
the amount of virtual memory used by the individual software applications during an FRA
simulation. Restarting the plug-ins (or software applications) is a way to force the release of any
memory referenced by the software applications.

194
PR-100 Chapter 10 – Flood Risk Analysis Simulation

From the Run FRA Simulation dialog box (Figure 154), an additional window will display a
summary of the simulation definitions and parameters that are part of the simulation, and the
ability to start the simulation. For the final FIRO FRA simulations, four different computers were
required to run the FRA simulations in parallel. Each of the FRA simulations ran for more than
three days, with each realization averaging about eight hours of simulation time. Results for
each FRA simulation occupied about 16GB of disk space. The HEC-WAT Distributed Compute
feature of HEC-WAT provides the study team the ability to automatically spread the processing
load across multiple machines, but was not available for the FIRO study.

FRA Simulation Results

An FRA simulation can generate a great deal of results, which can be accessed directly from the
underlying folder structure within the FIRO study directory, or from the HEC-WAT interface
through plots and reports. Results include individual model outputs for specific events (provided
the Save Model Folders option was selected) and statistical summaries according to lifecycle,
realization, or total population.

Results from the HEC-WAT Interface

From the HEC-WAT interface, the modeler must select which results will be viewed by selecting
the desired realization, lifecycle, and event. Once this selection has been made, modelers can
view results by right clicking on individual model elements in a Simulation Map Window. In
addition, from the HEC-WAT interface, from the Results menu, the modeler can access model
output for the selection.

From the HEC-WAT Output Variable Editor, the aggregated results that were identified by the
modeler are available. The results included in these reports can be filtered to represent
individual lifecycles, all lifecycles in a realization, or all realizations (i.e., the full population of
events). The Histogram Viewer (Figure 155) provides plots of the relative frequencies of
occurrence for the ranges of result values. For example, Figure 155 displays the distribution of
damage results for all lifecycles in selected realization. The total frequency adds up to 0.70
because 151 of the 500 events were skipped.

The Output Variable Viewer (Figure 156) plots a point for the maximum value from each
event. For example, displayed in Figure 156, are the maximum stages at Johnson's Beach for all
realizations simulated under the Existing Conditions alternative. The events on the X-axis
(Figure 156) correspond to the fifty-year span of the analysis period used for the FRA
simulations. The 3,403 points plotted (Figure 156), display the maximum stage for each event
from the 100 lifecycles (1,597 events were skipped). The Output Variable Viewer was helpful
for identifying the largest events, and placing them in context with the other events.

The Frequency Viewer (Figure 157) plots the exceedance frequency for the output variables on
a probability scale. For example, Figure 157 displays the basin average precipitation sampled
from frequency curves for each realization, along with an "expected probability" curve that
combines results from all events into one frequency curve. The Frequency Viewer (Figure 157)
also plots curves for the 5- and 95-percent confidence limits, but recall that the ten realizations
used in the FIRO analysis are insufficient to accurately compute those limits.

195
Chapter 10 – Flood Risk Analysis Simulation PR-100

Figure 155 HEC-WAT - Histogram Viewer

Figure 156 HEC-WAT – Output Variable Viewer

196
PR-100 Chapter 10 – Flood Risk Analysis Simulation

Figure 157 HEC-WAT – Frequency Viewer

Results Stored in DSS Files

For the FIRO study detailed analysis and reporting of results was primarily performed with
HEC-DSSVue and Microsoft Excel®, using the data stored in the runs folder of the HEC-WAT
study. The folder structure (Figure 158) for an FRA simulation differs from the structure for a
deterministic simulation (Figure 114). For example, in Figure 158, under the runs folder, under
the FRA folder are the sub-folders for each realization, and the simulation.dss file with the same
name as the simulation.

The Base_FRA.dss (Figure 158) file contains the values for the aggregated results specified in
the Output Variable Editor (Figure 153). The maximum value for each event is stored with a
pathname that is constructed from the HEC-WAT simulation name, the software application and
associated model alternative name, the type of parameter stored, location identifiers, and the
HEC-WAT alternative name and analysis period name. For example, in the table below, the
highest Lake Mendocino Lake levels for the FRA simulation of the Existing Conditions
alternative were stored in similar pathnames:

/Base_FRA/ResSim-Baseline_F/Event-Elev/Output Variable/Lake Mendocino-Pool Elev 6HOUR/C:000000|Existing C:FRA/

/Base_FRA/ResSim-Baseline_F/Event-Elev/Output Variable/Lake Mendocino-Pool Elev 6HOUR/C:000001|Existing C:FRA/
…
/Base_FRA/ResSim-Baseline_F/Event-Elev/Output Variable/Lake Mendocino-Pool Elev 6HOUR/C:000009|Existing C:FRA/
197
Chapter 10 – Flood Risk Analysis Simulation PR-100

Figure 158 FRA Results Folder Structure – Simulation Level

Note that the F-Part defines each pathname as part of a collection. Each pathname contained the
results for one of the ten realizations, and each realization corresponded to a collection. Figure
159 displays a tabulation of one of the pathnames. Each column lists the peak for fifty events of
each lifecycle in the realization. The FRA results used for the comparison metrics were mainly
analyzed in Microsoft Excel®. The HEC-WAT output design made it very convenient to
tabulate all realization pathnames for an output variable, select all, and then paste the 5,000
values into the spreadsheet in one operation.

Figure 159 Example Output Variable DSS records

198
PR-100 Chapter 10 – Flood Risk Analysis Simulation

The Base_FRA.dss file also contains frequency curves computed for the output variables (if
indicated on the Output Variable Editor) and the random seed values provided to each model
alternative for each realization, lifecycle, and event.

Each realization folder contains sub-folders with results for the lifecycles within that realization.
For example, in Figure 160, the results for each lifecycle include at least a log file, and a
lifecycle.dss file. The log file can be useful for troubleshooting, since it contains the run-time
messages from the models, execution times, and status information during the HEC-WAT
simulation. The lifecycle.dss file contains time-series results for each event, such as a
hyetograph for a subbasin, or flow hydrographs at a river gage location. In addition, Figure 160
illustrates the case where the Save Model Folders option for the HEC-WAT simulation was not
selected. With this option not selected, sub-folders for each event of the lifecycle were created
and contained model alternative data.

Figure 160 Example Lifecycle Results Folder

199
Chapter 10 – Flood Risk Analysis Simulation PR-100

200
PR-100 Chapter 11 – Existing Conditions Alternative Results

Existing Conditions Alternative Results

The Existing Conditions alternative results served several purposes for the FIRO study:

1) The Existing Conditions alternative deterministic results provided the basis for model
calibration or validation, and established the credibility of the modeling in the context of
familiar events.

2) The hydrologic model results for the Existing Conditions alternative served as a common
input to each water management alternative.

3) The Existing Conditions alternative deterministic results for soil moisture and reservoir
levels were used to determine sampling distribution parameters for these antecedent
conditions during the FRA simulations.

4) The Existing Conditions alternative deterministic results provided the baseline data for
comparing relative storage gains due to water management alternatives.

5) The Existing Conditions alternative FRA results provided the baseline data for relative
comparisons for flow and stage frequency results at key locations.

6) The Existing Conditions alternative FRA results provided the baseline data for
comparisons of EAD throughout the Russian River system.

The Existing Conditions deterministic results also served as the "proving ground" for the
modeling team to fully identify and produce the data and parameters required for the study.
Planning and coordination helped clarify the requirements for each team member, but actually
using the different model results together sometimes identified issues that had not been fully
anticipated. The HEC-WAT framework made it easy for modelers to "check-in" with each other
to verify that the individual pieces fit together and produced acceptable results. Almost all of the
model and data questions arose in the course of computing the Existing Conditions (Baseline)
alternative. After determining that the Existing Conditions alternative for both the deterministic
and FRA results was satisfactory, running the other HEC-WAT alternatives was relatively easy.

Meteorologic Modeling Observations

The meteorological inputs used for calibration of the hydrologic model are discussed in Section
10.1. These data, along with observations regarding the precipitation characteristics for the
deterministic modeling analyses, were used to develop the inputs for the hydrologic sampling
process, as described in Section 5.1. Table 28 shows an example of the hydrologic sampling
output for an event. In the example (Table 28), Event 38 from Lifecycle 28 resulted in the third
highest precipitation amount sampled for the analysis. The 0.07 percent exceedance probability

201
Chapter 11 – Existing Conditions Alternative Results PR-100

Table 28 Hydrologic Sampler Results for Event 38 from Lifecycle 28

Water Year: 38
Start: 1 January 3038 00:00
End: 1 January 3039 00:00
Number of Events: 1
Start End Expected
Date & Time Date & Time Probability Shape Set Type Location Magnitude
1/13/3038 2/3/3038 7.00E-04 dec_1955 Hyeto Austin Cr 51.00
5:00:00 AM 5:00:00 AM Hyeto Big Sulphur Cr 32.39
Hyeto Dry Creek 10 38.58
Hyeto Dry Creek 20 37.87
Hyeto Dry Creek 30 31.51
Hyeto EF Russian 10 22.89
Hyeto EF Russian 20 21.13
Hyeto Green Valley 25.74
Hyeto Laguna 21.02
Hyeto Russian 10 38.67
Hyeto Russian 20 31.31
Hyeto Russian 30 29.93
Hyeto Russian 40 28.73
Hyeto Russian 50 26.11
Hyeto Russian 60 24.22
Hyeto Russian 70 24.58
Hyeto Santa Rosa Cr 10 26.47
Hyeto Santa Rosa Cr 20 27.56
Hyeto WF Russian 22.15
Hyeto BASIN AVERAGE 29.23

corresponded to 29.23 inches as the basin average eight-day total for the event. The hyetograph
shape set sampled was for the December 1955 historical event, and 13 January was sampled as
the start date of the FRA simulation. The hyetograph locations defined for the 1955 historical
event reflected the heaviest precipitation in the subbasins along the Lower Russian River and
Dry Creek, with the locations near Lake Mendocino receiving relatively lighter precipitation
totals. The magnitude (Table 28) is the total precipitation in each subbasin and is defined in the
hyetograph shape set. The aggregate results from the hydrologic sampling process regarding
event date, total precipitation, and shape set selection were analyzed to "back-check" the results
for consistency with the inputs and validate the adequacy of the sampling.

A histogram of the ten shape sets sampled during the FRA simulation are displayed in Figure
161. The height of each bar represents the incidence of events using each shape set as a percent
of the total events sampled. For comparison, Figure 161 displays the defined incidence (gold
line), the 50% confidence interval (dashed green lines), and the 90% confidence interval (dashed
red lines). There is no hard metric to judge suitability of any particular set of results. Either too
little or too much variation in the incidence of selection requires a closer look at the sampling
parameters and output. For the FIRO study with ten equally weighted shape sets and 5,000
samples, only one or two of the shape set selections would typically fall outside the 90%
confidence interval. None of the shape sets in the FIRO analysis sampled outside the 90%
202
PR-100 Chapter 11 – Existing Conditions Alternative Results

Figure 161 Frequency of Shape Set Selection

confidence interval, indicating that the shape sets used in this analysis were sampled a bit more
evenly than expected. Alternately, normal variation would suggest about half of the shape sets
should fall outside the 50% confidence interval. The results displayed in Figure 161 conform to
this expectation, with five of the ten shape sets sampled within the 50% confidence interval, four
shape sets displaying incidence slightly greater than the 50% confidence interval and one
occurring below the confidence interval. Overall, the shape set selection fell within reasonable
ranges.

Figure 162 compares the event dates sampled during the FRA simulation with the empirical
distribution input to the hydrologic sampling process defining the flood season, and the observed
values used to generate the definition. The gold circles (Figure 162) represent FRA events; the
blue line (Figure 162) represents the input distribution during the storm season, indicating
consistency between the input and output. Note that the modelers chose to allow events to occur
throughout the entire month of March, even though the latest observed storm occurred on 16
March.

The precipitation magnitudes sampled during the FRA simulation with the GEV distribution
input to the hydrologic sampling process are displayed in Figure 163, along with the observed
values that were used to generate the precipitation magnitude. The gold circles (Figure 163)
represent FRA events; the blue line (Figure 163) represents the input distribution, indicating
consistency between the input and output. The close alignment of the input GEV precipitation
frequency curve and the 5,000 sampled values demonstrates that the limited number of
realizations used for the FIRO analysis did not produce a biased sample. As expected in a
random sample, the consistency waivers for the largest events, and the seven largest events
sampled are substantially higher than the input GEV curve. Such variation in the rarest events is
typical of random sampling, where a sample of 5,000 events can potentially include a one-in-
one-million event, or an absence of 500-year and larger events. To determine whether the
outcomes reasonably reflect the input curve, modelers must review both the re-sampling of
203
Chapter 11 – Existing Conditions Alternative Results PR-100

Figure 162 Frequency of Event Date Selection

Figure 163 Frequency of Eight-Day Total Precipitation

precipitation frequency curves for each realization, which reflects the uncertainty in the input
precipitation frequency distribution, and sampling of individual events from each frequency
curve. The confidence limits (Figure 163) were calculated using HEC-SSP for the GEV curve

204
PR-100 Chapter 11 – Existing Conditions Alternative Results

input to the hydrologic sampling process for this analysis. The confidence interval displays the
knowledge uncertainty in the input frequency curve, defined from the gage record length of
sixty-years. The precipitation amounts sampled for the FIRO study all fall within the 90%
confidence interval, indicating that the FRA simulations reasonably represent the uncertainty
expected.

The visual comparison of the sampling results for the realizations shown in Figure 157 provides
a preliminary indication of the how well the FRA analysis represented uncertainty regarding the
size of events. A more detailed illustration is provided in Figure 164, where the ten re-sampled
GEV precipitation frequency curves for the realizations are represented by solid plotting lines,
each differing from the input curve to capture knowledge uncertainty. The 500 events sampled
from each precipitation curve are plotted as points using the same color. Note that the three
largest events occurred under different realizations of the precipitation frequency curve,
indicating that the sampling was not disproportionately affected by a high frequency curve in one
realization. The largest basin average precipitation of 32.98 inches occurred in Event 39 from
Lifecycle 13. The number of frequency curves above or below the input frequency curve varied
slightly across the range of exceedance frequencies, but did not result in a bias to the mean
results (as seen by the alignment of sampled values to input curve in Figure 163). Figure 164
also displays the 90% confidence intervals computed for the input GEV precipitation frequency
curve. The array of frequency curves spans the 90% confidence interval and reasonably captures
the uncertainty expected at the 1% exceedance probability used for the flood risk metrics for the
relative comparisons in this analysis. The suitability of these sampled precipitation frequency
curves for the realizations is a fortunate outcome for the FIRO study. Remember, in a typical
USACE planning study, ten realizations are insufficient to define the uncertainty distribution at
the 1% exceedance probability.

Figure 164 Realization Frequency Comparisons

205
Chapter 11 – Existing Conditions Alternative Results PR-100

Hydrologic Modeling Observations

Outflows from subbasins (Figure 19) were the primary results from HEC-HMS. These subbasin
flows were used for simulating reservoir operations and for hydraulic computation of river
stages. The Existing Conditions HEC-HMS model and resulting subbasins flows were applied to
each HEC-WAT alternative.

The deterministic HEC-HMS model was developed and calibrated as described in Section 5.2.
The continuous hourly simulation produced results for water years 1950 through 2010. Figure
165 displays a typical portion of HEC-HMS simulation results, covering water year 2005, for the
subbasin above the Calpella gage, which used the SMA method of modeling losses. The upper
plot (Figure 165) shows basin average precipitation, the middle plot shows modeled versus
observed runoff, and, the lower plot shows soil moisture storage. The storms circled in red
(Figure 165) during October helped moisten the soil, but produced very little runoff. Light rains
in November and December similarly produced little flow in the river, but combined with
climate factors to maintain the soil moisture.

Figure 165 HEC-HMS Deterministic Simulation Example Results

The peak flow for the water year occurred in December, after a sudden heavy precipitation event.
The event was brief and isolated, providing a limited increase of inflow volume to Lake
206
PR-100 Chapter 11 – Existing Conditions Alternative Results

Mendocino. However, the event served to saturate the soil, which dried only slightly during the
following weeks of clear weather. Wet weather set in from the last week of December through
the first week of April. None of the storms were remarkably strong, but kept the soil nearly
saturated during those months and maintained elevated inflows to the reservoir.

Clear weather and changing climate factors during April allowed the soil to partially dry before a
series of storms in May. The last storm of the sequence produced the second-highest peak flow
of the year, but the wet weather ceased without producing a sustained increase in river flow. The
soil dried quickly afterwards, and the June storms circled in red (Figure 165) produced no
significant runoff.

The continuous subbasin flows were necessary inputs for the reservoir model to simulate
reservoir conservation operations in the deterministic simulations. The flow allowed analysis of
successive events during the entirety of the flood seasons, and modeling of summer depletions at
Lake Mendocino. The continuous subbasin flows were necessary for modeling the success of
the alternatives in re-filling the reservoir. The sixty years of deterministic reservoir modeling
using the continuous subbasin flows also provided the basis for the sampling parameters used to
determine starting reservoir storage for FRA events, as explained in Section 10.3.2. Finally, the
sixty years of continuous hydrologic modeling results from the deterministic analysis provided
the basis for the sampling parameters used to determine the antecedent soil moisture for each
FRA event, as described in Section 10.2.2.

Validation was performed on the FRA results regarding the sampling for the SMA loss method
used in the EF Russian 20 subbasin, and the Deficit and Constant loss method used in the WF
Russian subbasin. The review compared the sampled results for each month to the
corresponding input distribution parameters. Figure 166 illustrates the consistency between the
SMA sampled for each February event (green circles) and the input Beta distribution defined for
that month (thick green line). The black dots (Figure 166) represent the daily values used to
define the beta distribution. Note the absence of values around fifty percent of soil moisture
capacity; the driest 56 points occurred in 1977 and 1991 and were significantly lower than the
rest of the data.

Figure 166 Validation of Sampling for SMA Loss Method - EF Russian 20 Subbasin – February Events

207
Chapter 11 – Existing Conditions Alternative Results PR-100

A similar comparison validated the soil moisture sampling for the Deficit and Constant loss
method used in the WF Russian subbasin. Figure 167 illustrates the comparison for sampling of
December events. Figure 168 displays the Median plotting positions for the sixty annual
maximum inflows to Lake Mendocino for the deterministic simulations values as green circles.
In Figure 168, the plotting positions for the 5,000 FRA values define the output frequency curve,
and are plotted as a solid green line. FRA results with exceedance probabilities greater than
twenty percent were omitted, since the FRA modeling was designed to represent large events
with the potential to cause flood damage, and not the events that produced the lower part of the
frequency curve. The dashed green lines (Figure 168) show the spread of the

Figure 167 Validation of Sampling for Deficit and Constant Loss Method - WF Russian Subbasin –
December Events

frequency curves among the ten realizations of the FRA results, by plotting the maximum and
minimum values for a given frequency. The dashed green lines (Figure 168) do not represent an
actual uncertainty-based confidence interval in the modeling effort, which could only be defined
with many more realizations of the input precipitation curve, and might require including
additional model parameters in the uncertainty analysis. The top ranked event of the 500 values
was assigned a frequency of 0.14 percent per the Median plotting position formula. The largest
FRA flows that plot outside the frequency range of the realization frequency curves are displayed
in Figure 168 as points to emphasize the greater uncertainty for such rare floods.

Comparing the frequency of annual peak reservoir inflows provided an expedient check on the
consistency between the deterministic and FRA results. HEC-WAT makes comparing peaks
values quick and convenient, but if modelers determined a need for further investigation, they
could easily perform an analysis incorporating volumes to provide a more appropriate metric for
a study of reservoir flood operations. Figure 168 confirms the consistency in results between the
deterministic and FRA simulations. The flows for the largest deterministic events plot below the
FRA frequency curve, reflecting the uncertainty associated with the limited sample represented
in the deterministic simulation. The largest peak inflow among the FRA events occurred in

208
PR-100 Chapter 11 – Existing Conditions Alternative Results

Figure 168 Annual Peak Flow Frequency Curve – Coyote Valley Dam Inflow – Existing Conditions

Event 39 from Lifecycle 13, corresponding to the largest precipitation event (Figure 163). The
rankings of the peak inflows for events differed from the ranking of the basin average
precipitation total, due to spatial and temporal variations in the shape set hyetographs, and
sampled initial soil moisture.

Reservoir System Modeling Observations

The deterministic HEC-ResSim model was developed and validated as described in Section 4.3.
The continuous six-hour simulation produced results for water years 1951 through 2010.
Deterministic results from the HEC-ResSim model alternative for the Existing Conditions
provided the reservoir pool levels used as the baseline metric of potential additional storage
benefits of forecast-informed reservoir operations, as described in Chapter 16 (Section 16.1).
FRA results from the HEC-ResSim model alternative for the Existing Conditions provided the
reservoir pool frequencies used as the baseline metric of potential hydraulic loading risks on the
dam and spillway, as described in Chapter 16 (Section 16.4).

Existing Conditions Deterministic Lake

Mendocino Levels
The Existing Conditions results represent a normalized baseline for relative comparisons
regarding alternative water management strategies. The continuous simulation dataset allowed
209
Chapter 11 – Existing Conditions Alternative Results PR-100

statistical comparisons among alternatives, as well as comparisons for selected events. As

explained in Section 6.9, the Existing Conditions deterministic results did not reproduce
observed data, due to operations for the Potter Valley Project, Coyote Valley Dam, and Warm
Springs Dam that changed over time, and the challenge of detailed hydrologic simulations based
on limited historical data.

However, the Existing Conditions deterministic results were developed to provide a reasonable
representation of both historical water supply conditions and short-term flood events. The
simulated Coyote Valley Dam levels for the analysis period are displayed in Figures 169 through
172 along with observed pool levels, in order to illustrate situations where the Existing
Conditions dataset resembled or differed from the observed record.

Figure 169 Simulated and Observed Lake Mendocino Levels from1950 - 1964

The green line in Figures 169 through 172 displays the Existing Conditions level for Lake
Mendocino from the deterministic simulation. The dashed purple line (Figures 169 through 172)
indicates the contemporary guide curve used throughout the analysis period. The black line
(Figures 169 through 172) indicates the historical lake level, with a small triangle marking the
peak for the water year. This section notes some general differences and details specific
differences throughout the record. The most significant differences are:

- The Coyote Valley Dam spillway flowed only once in the historical record, while the
simulation allowed reservoir storage to reach that high three times.

210
PR-100 Chapter 11 – Existing Conditions Alternative Results

Figure 170 Simulated and Observed Lake Mendocino Levels from 1965 – 1979

Figure 171 Simulated and Observed Lake Mendocino Levels from 1980 - 1994
211
Chapter 11 – Existing Conditions Alternative Results PR-100

Figure 172 Simulated and Observed Lake Mendocino Levels from 1995-2010

- The Existing Conditions reservoir levels were depleted much lower than the historical
levels in several years, commonly falling to around 710 feet NGVD. Such low levels
only became common during the most recent years of the observed record. For most of
the historical record, Lake Mendocino received significant dry-season diversions from
the Potter Valley Project and faced fewer demands on its storage than currently exist.

In Figure 169, the spikiness of the green line during the winters of 1950 to 1951, and 1955 to
1956 reflect numerous flood operations, but all within Flood Control Schedule 2 or lower. Three
of the years between 1950 and 1964 provided sufficient spring inflow to fill Coyote Valley Dam
past 750 feet. There were several dry years but no extreme droughts during the first fifteen years
of the analysis period. The driest summer preceded the flood event in December 1964, in which
the reservoir level exceeded the spillway for the simulated and observed datasets. The early
observed Coyote Valley Dam levels match poorly with the simulated Existing Conditions levels,
mainly because of different regulation practices that reflect reliable supply and lesser demands
on the system.

The Lake Mendocino levels simulated for years from 1965 through 1975 (Figure 170) reflected
relatively moderate hydrologic conditions. Each of these winters featured normal flood control
operations, which briefly reached Flood Control Schedule 3 (Appendix A) in two of the years.
The most striking element of Figure 170 is the drought of 1976 through 1977. The simulated
levels fell much lower than the observed reservoir levels, due to higher demands and lower
Potter Valley Project flows assumed for the baseline. The Existing Conditions dataset generally
shows greater depletions than the historical record for the same reasons. The simulated flood
peak for the 1978 calibration event resembles the historical "drought-buster" event.
212
PR-100 Chapter 11 – Existing Conditions Alternative Results

Figure 171 includes 1982 and 1983, which are two of the twelve years where the simulation
maintained higher summer levels than the observed record. Wet springs in 1982 and 1983
allowed relative success toward filling Lake Mendocino to the summer guide level, and avoided
excessive summer depletions in the simulation results, although historical operations followed a
lower guide curve during the summer. The peak simulated lake level for the 1983 validation
event reasonably matched the observed value, although the timing differed due to historical
operational decisions. The springs of 1984 and 1985 illustrated the most common differences
between the simulated and observed levels, where the simulated reservoir levels failed to fill as
high in the spring due to lower Potter Valley Project flows, and were depleted more because of
greater downstream demands. The simulated level for the 1986 validation event substantially
exceeded the observed peak, as explained in Section 6.9. The simulated reservoir levels from
1986 and 1994 generally resembled the observed values during the late winter and early spring,
but were otherwise lower due to greater depletion.

Figure 172 compares the Lake Mendocino levels for the Existing Conditions dataset versus
historical values for 1995 through 2010. The simulated inflow for the springtime conditions
during these years were generally either very dry or very wet. The spring and summer Existing
Conditions reservoir levels departed significantly from the observed values in some cases.
Spring supplies in 1995, 1999, 2003, 2005, and 2006 would have allowed water managers to
come much closer to the current summer pool than occurred under historical operations. Lower
supplies and higher demands in the simulation resulted in less ability to fill the reservoir in the
simulations from 2000-2002. The simulated results most closely match the observed record for
2007 and later, because the Existing Conditions operations and diversion assumptions reflect the
historical conditions during that period. The exception is in 2009, when higher simulated
inflows allowed the reservoir to fill higher than the observed elevation.

The Existing Conditions reservoir level for the January 1995 flood event was 3.5 feet higher than
the observed peak due to greater inflow for the simulated event, rather than a difference in
operation. The March 1995 and 1997 events closely matched the historical levels. The 1998
Existing Conditions levels were ten feet less than observed, due to lower simulated inflows. The
third spillway activation in the Existing Conditions results occurred during the 2006 event. The
simulated reservoir levels exceeded the historical peak by 7.5 feet, primarily due to the lower
observed level than simulated level before the storm. Figure 44 displays a detailed comparison
of the observed and Existing Conditions levels for the 2006 event.

Existing Conditions Lake Mendocino Elevation

Frequency Results
As with other sampled parameters, the modelers verified that the starting level of Lake
Mendocino for the FRA events represented the intended distributions for each season, as
discussed in Section 10.3.2. As an example, Figure 173 confirms the consistency between the
starting storage values for each January event (purple circles) and the corresponding input
empirical distribution (thick purple line). The black dots (Figure 173) represent the daily values
used to define the input distribution. Note the apparent discontinuity for the 28 lowest points,
which occurred in 1977. The probabilities assigned in the distribution based on the limited data
set may underestimate the rarity of that year.
213
Chapter 11 – Existing Conditions Alternative Results PR-100

Figure 173 Validation of Sampling for Coyote Valley Dam Initial Pool Level – January Events –
Existing Conditions Alternative

Figure 174 compares the frequency curve of the FRA reservoir pool level results against the
plotting positions of results from the deterministic Existing Conditions simulations. As with
Figure 168, the green circles represent the sixty annual maximums for the deterministic
simulations, while the solid line shows the frequency curve defined by the 5,000 FRA values.
The largest events are shown as points to emphasize the greater uncertainty for such rare floods.
The dashed lines indicate the maximum and minimum elevations from the frequency curves for
the results of each realization. As noted in the discussion of Figure 168, these lines simply
reflect the spread of results among the ten realizations, and do not represent an actual
uncertainty-based confidence interval.

In Figure 174, the maximum annual pool levels for the five percent and smaller events are
heavily influenced by reservoir regulation, as demonstrated by the flat segment of the frequency
curve at 761.8 feet NGVD, corresponding to the guide curve top of conservation. The
exceedance frequency curve (Figure 174) crosses the 1% ACE mark on the horizontal axis at
770.55 feet NGVD (773.42 feet NAVD). This "100-year" level at Lake Mendocino for the
Existing Conditions alternative serves as one of the key baseline metrics for relative comparisons
of downstream flood risk in the other FIRO alternatives. The spillway was overtopped in 261 of
the 5,000 events (5.2% ACE). The total duration of spillway flow across all 261 events was
17,460 hours. None of the FRA events overtopped the Coyote Valley Dam.

The largest flood of the FRA simulation was in Event 39 from Lifecycle 13, corresponding to the
largest precipitation magnitude and peak inflow to Lake Mendocino. Rankings of the peak
reservoir level for other FRA events differed from rankings according to precipitation magnitude
214
PR-100 Chapter 11 – Existing Conditions Alternative Results

Figure 174 Annual Peak Pool Elevation Frequency Curve – Lake Mendocino – Existing Conditions

or peak inflow, due to variations in inflow volume, reservoir operations, and sampled initial pool
levels. The flood event simulated for Event 39 from Lifecycle 13 started in late February, as
shown in Figure 175. The eight-day basin average total precipitation totaled 33.0 inches. For
reference, the spillway design flood was based on a three-day total of 27.80 inches or
precipitation. The event used the December 2005 shape set, which featured three large storms
with similar hyetographs throughout the subbasin locations.

Sampling resulted in an initial reservoir elevation about one foot below the guide curve. The
heavy rain on 7 through 8 March pushed the reservoir pool almost four feet over the spillway.
The uncontrolled outflows resulting from this initial storm reached 3,900 cfs, but did not increase
the peak flow downstream. The second storm arrived late on 10 March, and caused the pool to
rise almost fifteen feet above the spillway crest – about 4.5 feet from the top of the dam.
Simulated flow over the spillway reached 37,900 cfs and largely coincided with the peak flow
downstream.

Lake Mendocino offered limited protection to downstream locations during this second part of
the event, as evidenced by the 10,000 cfs attenuation between the inflow and outflow curves
(Figure 175). Note that the routing and consequence computations in this analysis assumed that
the dam, spillway, and other components of the Coyote Valley Dam always performed as
intended, regardless of the loading applied to them. This assumption may be very uncertain at
the scale of Event 39 from Lifecycle 13, but applies equally across the alternative water
management strategies compared in this analysis.
215
Chapter 11 – Existing Conditions Alternative Results PR-100

Figure 175 Coyote Valley Dam Results for Largest FRA Event

Significance of Initial Pool Levels

The dry season depletions modeled under the deterministic simulations sometimes affected the
initial reservoir elevations for flood events in the following water year. Figure 176 illustrates
how the peak reservoir level for a year was sometimes related to the success in filling the
reservoir during the previous spring.

The storm in early January 1995 occurred while the reservoir was still heavily depleted,
providing a large amount of incidental flood protection. The event pushed the reservoir well into
the storage zone for Flood Control Schedule 3. For contrast, the much smaller storm in late
January 1997 was also able to bring the reservoir pool into Flood Control Schedule 3, because
prior inflows raised the pool to guide curve ahead of the storm.
216
PR-100 Chapter 11 – Existing Conditions Alternative Results

Figure 176 Lake Mendocino Existing Conditions Deterministic Results 1995 - 1998

Note that the wet spring of 1995 allowed for a high degree of success in reaching the target
summer pool level for Lake Mendocino. The relatively high reservoir levels carried through the
summer, until autumn guide curve operations brought the lake to the winter pool level. If the
January 1995 storm had occurred during the winter of 1995 - 1996, the event would have easily
caused flow over the spillway.

Substantial amounts of incidental storage reduced the impact of several early season storms in
the Existing Conditions results. Most notably, the events of January 1954, December 1955,
December 1964, December 1978, and January 1995 would have caused much higher reservoir
levels if the lake had not been so depleted at the beginning of the storms. Lake Mendocino
reached 766.91 feet NGVD for the 1964 event of the Existing Conditions alternative, compared
to 768.92 feet NGVD for the 1986 flood and 768.36 feet NGVD for the 2006 flood.

Hydraulic Modeling Observations

The deterministic results from the HEC-RAS model alternative for the Existing Conditions
alternative provided the hydraulic inputs for development of the HEC-FIA consequence model,
which computed the AAD of the Existing Conditions alternative. FRA results from the HEC-
RAS model alternative for the Existing Conditions alternative provided the downstream flow and
stage frequencies used as the baseline metric for assessing downstream flood risk, as described in
Chapter 16 (Section 16.4). The FRA results from HEC-RAS provided the hydraulic inputs to the
consequence model for computing EAD for the baseline alternative.

Travel Times and Downstream Residence

The Coyote Valley Dam regulation manual lists routing times for the crest of flood waves to
travel downstream (Table 29). The higher the flow, the faster the peak travels downstream. The
travel times of flood peaks in the results for the Existing Conditions alternative closely matched
the values in the Coyote Valley Dam regulation manual (Figure 177).
217
Chapter 11 – Existing Conditions Alternative Results PR-100

Table 29 Coyote Valley Dam Reach Length and Flood Wave Travel Time (reference: Table 14, CVD
Regulation Manual)
Travel Time, Hours
Discharge, cfs
Length
Miles 400 1,000 2,000 4,000 6,000 8,000 10,000 20,000 40,000
Forks of the Russian River
to Hopland 14 11.0 9.0 7.5 6.5 6.0 6.0 5.5 5.0 4.5
Hopland to Cloverdale 16 12.5 9.0 7.0 5.5 5.0 4.5 4.0 3.0 2.5
Cloverdale to Healdsburg 28 18.5 13.0 10.5 9.5 8.5 8.0 7.5 6.5 6.0
Healdsburg to Guerneville 16 43.0 31.0 26.0 21.0 19.0 18.0 16.5 14.0 13.0
*These travel times are approximate

The routing of Coyote Valley Dam outflow changes carried significance beyond the travel time
for peaks. The West Fork Russian River and Hopland maximum flow rules in the Coyote Valley
Dam regulation manual were designed to prevent Lake Mendocino releases from adding to flood
peaks downstream. Flood operations aim to minimize outflows from the East Fork Russian
River in order to avoid contributing to peak flows at the confluence with the West Fork Russian
River or at Hopland. However, small portions of earlier releases can remain in the Russian River
when downstream tributaries drive the peak flows in the lower Russian River.

Figure 177 illustrates the downstream flow impacts of waiting an extra six hours to reduce the
outflows from Coyote Valley Dam during the 2006 event. The four green hydrographs (Figure
177) display the results for Existing Conditions alternative for Coyote Valley Dam releases from
the HEC-ResSim model, and the flows modeled with HEC-RAS at Hopland, Healdsburg, and
Johnson's Beach in Guerneville. The first storm on 28 December pushed the reservoir into flood
operations, with "backfilling" releases reaching 3,150 cfs on 29 December. The operators
ramped down the outflows on the 30 December as the main storm arrived.

Figure 177 Downstream Impact of Six-Hour Delay to Stopping Coyote Valley Dam Release

218
PR-100 Chapter 11 – Existing Conditions Alternative Results

The purple hydrographs in Figure 177 represents the results of delaying the ramp-down by six
hours. The HEC-RAS model was re-run, with the Coyote Valley Dam outflows from HEC-
ResSim for 30 December delayed by six hours, resulting in the purple hydrographs (Figure 177)
at the three downstream gages. These new results differed only slightly from the original values.
The separations between the hydrographs shows that Hopland exceeded 8,000 cfs sooner than in
the original simulation and Healdsburg rose slightly faster on the 31 December. The impact at
Guerneville was very small, as routing effects spread the difference across a wider time span.
The inset graph (Figure 177) shows that the diminishing tail of additional flow increased the
peak at Johnson's Beach by less than 100 cfs (0.1 percent).

Figure 178 displays the impact of waiting a full day to ramp down the releases from Lake
Mendocino. In this case, the delayed reduction of outflow causes greater impact to the
downstream locations. The releases of 30 December still reside in the channel when the peak
flows occur on the 31 December and 1 January, and result in about one-percent more flow in the
peak at Johnson's Beach.

Figure 178 Downstream Impact of One Day Delay to Stopping Coyote Valley Dam Release

These comparisons based on the 2006 event illustrate how the West Fork Russian River and
Hopland maximum flow rules effectively prevent Coyote Valley Dam flood operations from
increasing downstream flood damage. The miniscule differences shown in Figure 177 seemed
representative of operations in general, despite differences in runoff hydrographs among
simulated events, and hydraulic complexities of the routing. Modelers were unable to find a case
among the deterministic results where these rules allowed significant Coyote Valley Dam
releases to remain resident in the lower Russian River during peak flow. FIRO-based
219
Chapter 11 – Existing Conditions Alternative Results PR-100

alternatives offer the opportunity to act sooner when confronted with potential flood events,
which will further reduce the chance of impacts from Coyote Valley Dam operations on
downsteam locations. However, Figure 178 demonstrates that modifications to the release
constraint at Coyote Valley Dam can have downstream impacts that some might consider
significant. FIRO rules should be constructed to avoid situations where incorrect forecasts could
delay flood operations. Note also that Coyote Valley Dam releases, could affect downstream
flood peaks from changes such as extending the ramp-down period, or relaxing the 8,000 cfs
limitation. The amount of impact and frequency of occurrence could be quantified through
modeling.

Flow and Stage Frequencies

Figures 179 through 184 display the flow and stage frequencies curves for the Existing
Conditions at Hopland, Healdsburg, and Guerneville. In Figures 179 through 184, the frequency
curve is defined by the plotting positions of the 5,000 FRA peak values and are plotted as a solid
green line. Values with exceedance probabilities greater than twenty percent were omitted, since
the FRA modeling was designed to represent large events with the potential to cause flood
damage. The dashed green lines (Figures 179 through 184) display the spread of the frequency
curves among the ten realizations, by plotting the maximum and minimum values for a given
frequency. The dashed green lines (Figures 179 through 184) do not represent an actual

Figure 179 Annual Peak Flow Frequency Curve – Hopland – Existing Conditions Alternative

220
PR-100 Chapter 11 – Existing Conditions Alternative Results

Figure 180 Annual Peak Water Surface Elevation Frequency Curve – Hopland – Existing Conditions
Alternative

uncertainty-based confidence interval in the modeling effort, which could only be defined with
many more realizations of the input precipitation curve, and might require including additional
model parameters in the uncertainty analysis. The top ranked event of the 500 values was
assigned a frequency of 0.14 percent per the Median plotting position formula. The largest FRA
flows that plot outside the frequency range of the realization frequency curves are shown
(Figures 179 through 184) as points to emphasize the greater uncertainty for such rare floods.

The frequency curve from the FRA results should be compared to the frequency curve using the
historic observations. The use of "expected probability of exceedance" reflects the fact that the
probability of exceeding a flow is itself a random variable due to sample error, which causes
parameter uncertainty. The expected probability curve reflects the expected long-run proportion
of exceedances of a threshold flow when accounting for both natural variability and sample error
caused by short record length (i.e., small sample size).

The flow-frequency results (for each gage location in Figures 179 through 184) are comprised of
annual maximum flows that were extracted from the HEC-RAS model alternative within the
FRA simulation along with an expected probability curve computed during the FRA simulations.
Figures 179 through 184 display the Median plotting positions for the sixty annual maximum
flows (1951 – 2010) from the deterministic simulations as green circles (Figures 179 through
184), and the observed annual maximum flows as hollow black diamonds (Figures 179 through
221
Chapter 11 – Existing Conditions Alternative Results PR-100

Figure 181 Annual Peak Flow Frequency Curve – Healdsburg – Existing Conditions Alternative

184). A Log Pearson Type III distribution was fit to the annual maximum series using Bulletin
17B (Interagency Advisory Committee on Water Data: IACWD, 1982) techniques and the
resulting parameters were used to estimate an expected probability curve using a Monte Carlo
algorithm to account for short record lengths. For the most part, the flow frequency curves
computed from the models match the flow frequency curve from observed data.

The comparison of modeled and observed flows contains heterogeneous data results from with-,
without-project, and the historical operations differ from Existing Conditions. Differences in the
modeled and observed stages are due to a lack of detailed terrain data used to define the channel
cross sections, lack of bathymetric information to define the channel geometry, and the HEC-
RAS model was calibrated to a limited number of flood events. A ten-meter terrain model was
used to create the channel geometry. The ten-meter terrain model did not include the channel
geometry below the low flow water surface. The low flow channel was manually cut into the
channel geometry using results from flood insurance models and a site visit to the Russian River.
The HEC-RAS model was originally developed for a dam breach simulation and only calibrated
and validated to large flood events. Differences in the modeled and observed stages could be
reduced by using a more detailed terrain model that included detailed bathymetric data and
calibrating and validating the model to a range of flood events. Other factors that will contribute
to differences include uncertainties in the flow values from the hydrologic simulations and
differences in the gage location relative to the HEC-RAS cross section used to gather the
modeled results.
222
PR-100 Chapter 11 – Existing Conditions Alternative Results

Figure 182 Annual Peak Water Surface Elevation Frequency Curve – Healdsburg – Existing Conditions
Alternative

The modelers decided that the Existing Conditions alternative flow and stage results were
acceptable to use as the baseline for relative comparisons of downstream impacts of alternative
water management plans. Refinements to the hydraulic models, which may provide closer
agreement between the simulated and observed flow and stage frequency results, could be
considered for subsequent analyses.

Hopland
Figure 179 compares the observed flow-frequency at Hopland with hydraulic model results from
the deterministic and FRA simulations. The deterministic results closely matched the frequency
curve based on the observed record, reflecting both the careful calibration of the hydrology
model and agreement of results between the hydrologic models. The two largest deterministic
events plotted lower than the FRA frequency curve, reflecting the uncertainty of plotting
positions based on the short period-of-record. Consistent with the results at Lake Mendocino,
the largest flow of 143,000 cfs occurred during Event 39 from Lifecycle 13. The exceedance
frequency curve crosses the 1% ACE mark on the horizontal axis at 55,800 cfs (Figure 179).
This "100-year" flow at Hopland for the Existing Conditions alternative serves as one of the key
baseline metrics for relative comparisons of downstream flood risk under the other alternatives.

223
Chapter 11 – Existing Conditions Alternative Results PR-100

Figure 183 Annual Peak Flow Frequency Curve – Guerneville (Hacienda) – Existing Conditions
Alternative

Figure 180 compares the Hopland stage-frequency results for the deterministic and FRA
simulations with observed stage-frequency values. The observed stages plot lower than the
simulated results, especially for smaller events. The deterministic stage-frequency results plot
lower than the FRA stage-frequency results. The exceedance frequency curve crosses the 1%
ACE mark on the horizontal axis at 531.77 feet NAVD. This "100-year" stage at Hopland for
the Existing Conditions alternative serves as one of the key baseline metrics for relative
comparisons of downstream flood risk under the other alternatives.

Healdsburg
Figure 181 compares the observed flow-frequency at Healdsburg with hydraulic model results
from the deterministic and FRA simulations. The deterministic results closely matched the
frequency curve based on the observed record, reflecting the careful calibration of the hydrology
model. Although, as at Hopland, the flows for the two largest deterministic simulation events
plot below the FRA frequency curve. The exceedance frequency curve crosses the 1% ACE
mark on the horizontal axis at 96,366 cfs. This "100-year" flow at Healdsburg for the Existing
Conditions alternative serves as one of the key baseline metrics for relative comparisons of
downstream flood risk under the other alternatives.

224
PR-100 Chapter 11 – Existing Conditions Alternative Results

Figure 184 Annual Peak Water Surface Elevation Frequency Curve – Guerneville (Hacienda) – Existing
Conditions Alternative

Figure 182 compares the observed stage-frequency at Healdsburg with hydraulic model results
from the deterministic and FRA simulations. The observed stage-frequency values plot four to
five feet higher than the deterministic simulation results. Consistent with the flow-frequency
results, stages for the two largest deterministic simulation events plot below the FRA frequency
curve. The exceedance frequency curve crosses the 1% ACE mark on the horizontal axis at
105.59 feet NAVD (Figure 182). This "100-year" stage at Healdsburg for the Existing
Conditions alternative serves as one of the key baseline metrics for relative comparisons of
downstream flood risk under the other alternatives.

Guerneville (Hacienda)
Figure 183 compares the observed flow-frequency at Guerneville (Hacienda) with results from
the deterministic and FRA simulations. The deterministic results plot lower than the flow
frequency curve based on the observed record, reflecting differences in routing between the
HEC-HMS and the HEC-RAS results, possibly due to additional storage modeled in HEC-RAS
model below Healdsburg (the hydraulic model could be improved with additional effort). The
exceedance frequency curve crosses the 1% ACE mark on the horizontal axis at 98,700 cfs
(Figure 183). This "100-year" flow at Guerneville (Hacienda) for the Existing Conditions
alternative serves as one of the key baseline metrics for relative comparisons of downstream
flood risk under the other alternatives.
225
Chapter 11 – Existing Conditions Alternative Results PR-100

Figure 184 compares the observed stage-frequency at Guerneville (Hacienda) with results from
the deterministic and FRA simulations. Corresponding to the difference in flows-frequency, the
simulated stage-frequency results plots lower than the observed values. The exceedance
frequency curve crosses the 1% ACE mark on the horizontal axis at 69.96 feet NAVD (Figure
184). This "100-year" stage at Guerneville (Hacienda) for the Existing Conditions alternative
serves as one of the key baseline metrics for relative comparisons of downstream flood risk
under the other alternatives.

Consequence Modeling Observations

The deterministic HEC-FIA model was developed and validated as described in Section 8.1. The
HEC-FIA model produced average annual damage for structures and contents at $6.1 million
dollars. Simulation results showed ten lives lost across the sixty-year analysis period. The flood
damage computed for individual events in the deterministic simulations are displayed in Figure
185. FRA results from the HEC-FIA model alternative for the Existing Conditions alternative
provided EAD values, which were used as the baseline metric for comparing flood risk, as
described in Section 11.3. The EAD for the Existing Conditions alternative was $7.6 million
dollars.

Figure 185 Flood Damage from the Deterministic Simulations

Annual Average Damage versus EAD

The difference between the AAD damage determined by deterministic and the EAD results from
the FRA simulations in HEC-WAT illustrates the shortcomings of analyses based only on the
limited sample presented by the historical hydrologic record. While a deterministic analysis may
226
PR-100 Chapter 11 – Existing Conditions Alternative Results

intuitively seem sufficient to demonstrate the viability of a FIRO based plan, it should be noted
that such a deterministic analysis represents one very limited sequence of events. A different
sequence of events over the same analysis period may lead to different results. The FRA
simulations modeled 5,000 synthetic events, which were based on a flood frequency analysis of
historical precipitation data. In addition, this analysis also provided a more robust assessment of
flood frequency in terms of risk and uncertainty.

The FIRO study generated the 5,000 simulated events in ten realizations, with each realization
containing ten lifecycles of fifty events each. Averaging the damage across the fifty events of
each lifecycle provides a rough analogy to the average damage calculated across the sixty events
used in the deterministic simulation. Figure 186 displays the ranked lifecycle damage, which
ranged from $1.0 million for Lifecycle 61 to $22.9 million for Lifecycle 13. The $6.1 million
AAD from the deterministic simulation covered the years 1950 through 2010, and ranks at the 35
percent percentile of the lifecycle damage results for the FIRO study.

Figure 186 Average Damage Results from Deterministic and Lifecycle Simulations for the Existing
Conditions Alternative

Figure 187 displays shows the significance of computing damage over a wide range of potential
events. The plotted points (Figure 187) represent the averages for each lifecycle in order of
simulation. The yellow line (Figure 187) traces the running mean, describing the EAD if the
analysis had been limited with the inclusion of each lifecycle. The vertical lines (Figure 187)
indicate the ten realizations that sampled different precipitation frequency curves. By chance,
the $9.1 million average computed from the first lifecycle was relatively close to the $7.6 million
average across all 100 lifecycles. However, the next lifecycle contained events with
substantially greater damage that averaged $17.1 million. The running mean jumped from $9.1
million after the first lifecycle to $13.1 million after including the second lifecycle (i.e., first 100
total events). Average damage for the next ten lifecycles range from $5.2 to $12.7 million,
diluting the influence of Lifecycle 2 on the running mean. The running EAD increased abruptly
with the inclusion of Lifecycle 13, in which Event 39 caused more than $682.8 million in flood
damage. The remaining lifecycles included high damage values in Lifecycles 75, 76, 77, and 79.
227
Chapter 11 – Existing Conditions Alternative Results PR-100

Figure 187 Variation of EAD across FRA Lifecycles

Damage values from Realization 8 increased the running average by about three percent;
however, this was balanced out when thirteen of the final fourteen lifecycles computed less than
$7 million dollars in damage.

In the particular sequence of events for this analysis, the running mean stabilized toward a result
of $7.6 million across all 100 lifecycles (5,000 events). However, this analysis still reflects only
a sample of potential events, and a different set of 100 lifecycles may provide a different answer.
Note that after Lifecycle 13 the running average EAD based on 650 events was more than fifty
percent greater than the average based on 5,000 events, reflecting how a very large event in a
small population can distort the results. Also, note that the running average continued to decline
slightly as the relatively low damages in the final lifecycle results were included. Future studies
should consider a convergence analysis to determine EAD. Nevertheless, the EAD calculated in
this analysis reflects a fuller range of potential flood events than the historical record, and
provides a much more robust estimate than possible with deterministic analysis.

Mendocino versus Sonoma County

As described in Section 8.12, most of the flood damage occurred in the lower Russian River. The
upper Russian River has fewer structures at risk. Figures 188 and 189 display the depth grids and
inundated structures for each County defined in the Existing Conditions alternative for the December
1964 flood. Figure 188 displays the inundation extents within Mendocino County for the simulated
1964 flood. The darker purple (Figure 188) indicates greater depth. The blue squares (Figure 188)
228
PR-100 Chapter 11 – Existing Conditions Alternative Results

Figure 188 Existing Conditions Alternative Inundation Extents for 1964 Event in Mendocino County

229
Chapter 11 – Existing Conditions Alternative Results PR-100

display the 52 structures damaged by flooding. Note that the structures are mainly located on the
periphery of inundated areas around Ukiah and Hopland. Only eleven of the damaged structures
were flooded deeper than one foot above grade, and the damage in Mendocino County amounted to
only $1,453,000 of the $64,940,000 total for the watershed.

Figure 189 displays the locations of the 821 structures in Sonoma County with flood damage in 1964
event under the Existing Conditions alternative. Most of the $63,487,000 of damage occurred along
the lower Russian River near Monte Rio and Guerneville, and around Healdsburg. Note that dozens
of structures along Laguna De Santa Rosa incurred damage in areas of shallow inundation.

Figure 189 Existing Conditions Alternative Inundation Extents for 1964 Event in Sonoma County
230
PR-100 Chapter 11 – Existing Conditions Alternative Results

Damage Locations for Extreme Events

As described in Section 8.12, about half of the flood damage under the Existing Conditions
alternative of the 2006 event occurred between Monte Rio and Guerneville, with significant
additional increments at Laguna de Santa Rosa and Healdsburg. Figure 190 displays the results of
the same analysis applied to a range of other events. Since the scale of damages varied so widely
among the events, the Y-axis was normalized to the cumulative percent of total damage.

Figure 190 Proportion of Russian River Structure Damage by River Mile across Events

Although the 1964 event caused almost twice the damage of the 2006 event, the distribution
along the Russian River remained very similar. The pattern differed for the largest synthetic
events. In general the bigger the event, the larger the proportion attributed to Laguna de Santa
Rosa, reflecting the denser population in the southeast portion of the watershed. Healdsburg and
the upper Russian River incurred greater proportions of the total damage for the very large
events, while the damage along the Russian River below Forestville diminished in proportion.

The rarity of events large enough to inundate the area around Windsor-Santa Rosa-Sebastopol
limited their influence on the EAD. Analysis of FRA results for lesser events validated the
choice of Guerneville as the location offering the most consistent relationship between flow and
event damage. However, the relative significance of various Russian River reaches to the total
flood damage results could be useful for prioritizing improvements to the hydraulic and
consequence models.
231
Chapter 11 – Existing Conditions Alternative Results PR-100

Verification of Time Window Modifier

Parameters
As noted in Chapter 10 (Section 10.6.2), review of the Existing Conditions alternative results
showed a ten-day simulation was long enough to compute the peak levels and flows along the
Russian River with the hydraulic model. Review of selected FRA events confirmed that ten days
was enough time to model the peaks of large synthetic events that caused heavy spillway flow.
Wider time windows would be needed to analyze the performance of alternatives using forecasts
longer than five days.

The selection of the peak at Healdsburg for centering the time window was also confirmed to
provide sufficient time to model the peaks throughout the Russian River system. The only
exception noted in the deterministic analysis occurred in 1995. As displayed in Figure 191, the
two 1995 events (January and March) caused very similar downstream peaks, with the March
event slightly higher. This double peaked event caused the hydraulic modeling and consequence
computation for 1995 to be based on the March event, although the January event was more
damaging in the historical record. Some consideration was given to Guerneville as a better
choice for identifying the peak for the year, but it also showed the March peak flow as higher.
The modeling team decided it was acceptable to use the March 1995 event in the Existing
Conditions alternative as the basis for the relative comparisons of this study. Running the
hydraulic model for the March event also served to add more focus on the critical reservoir-
filling period.

Figure 191 Existing Conditions Alternative 1995 Flows

232
PR-100 Chapter 11 – Existing Conditions Alternative Results

Verification of Skip Rule Parameters

The modelers analyzed the Existing Conditions alternative results to verify the assumptions used
for defining the skip rule, described in Chapter 10 (Section 10.6.4). The results displayed in
Figure 192 confirm that flows at Guerneville provided the most consistent relationship to the
total damage across the range of FRA events, with the least spread across the plotted values. For
example, in Figure 192, events with a peak flow of 30,000 cfs at Hopland result in damage
ranging from $2 million to $116 million dollars. Events with a peak flow of 100,000 cfs at
Guerneville resulted in damage ranging only from $96 million to $112 million.

Figure 192 Relationship between Event Damage and Peak Flow at Selected Locations

The non-damaging flow threshold of 30,000 cfs at Guerneville was effectively established during
development of the consequence model according to historical data, as described in Section 8.10.
Figure 193 concentrates on the least-damaging events in the plot, demonstrating a "back-check"
that Guerneville peak flows less than 31,900 cfs resulted in no damage during the FRA
simulations.

Figure 193 Least-Damage Flows at Selected Locations

233
Chapter 11 – Existing Conditions Alternative Results PR-100

Integrated Modeling Results

The HEC-WAT simulations (deterministic or FRA) produced detailed results for each element of
each model, allowing modelers to bring a variety of analysis tools to bear on the results, and
examine the relationships among the data. As an example, Figure 194 illustrates use of a plot
template from HEC-DSSVue to display a composite of key model results for the deterministic
simulation of the Existing Conditions alternative. The upper plot (Figure 194) displays HEC-
HMS results of the basin average precipitation with Lake Mendocino inflow, and the middle plot
(Figure 194) displays the Lake Mendocino level from the HEC-ResSim model alternative. The
lower plot (Figure 194) displays the HEC-RAS computed stages at Healdsburg, along with the
event damage computed by the HEC-FIA model alternative.

Figure 194 Selected Existing Conditions Alternative Hydrographs from Deterministic Simulation

From the HEC-DSSVue plot (Figure 195), the modeler was able to zoom to any time period for
detailed examination. For example, in Figure 195, a closer look reveals that Lake Mendocino
attained the full summer pool due to some late season rain during the Spring of 2005, and
required guide curve releases during October in order to draw down to the winter pool. The pool
was at guide curve when a series of storms began in late December, and was still in the flood
pool when the largest storm arrived in early January.

234
PR-100 Chapter 11 – Existing Conditions Alternative Results

Figure 195 Selected Existing Conditions Alternative Hydrographs for Storm Event 2005 - 2006

The river rose more than five feet above flood stage at Healdsburg, and the event caused the
fourth highest damage in the deterministic simulation. Soon after Healdsburg began to recede,
the reservoir level rose above the spillway. Using HEC-DSSVue to review results represents a
frequent approach employed by modelers, who often arrange similar plots for different time
periods or different alternatives for comparison to explain model results.

Analysis of individual model parameters and results in HEC-DSSVue, Microsoft Excel®, and
ESRI ArcGIS®, allows modelers to derive relationships such as the damage per river mile, or
seasonal distributions of reservoir levels for a given water management alternative. The statistics
and performance metrics presented in this report were all mined from model results. The metrics
were saved within the simulation.dss file and consisted of annual maximums from the time-series
output and damage information from the HEC-FIA model alternative.

235
Chapter 11 – Existing Conditions Alternative Results PR-100

236
PR-100 Chapter 12 – Without CVD (WO_CVD) Alternative

Without CVD (WO_CVD) Alternative

Coyote Valley Dam offers limited flood protection to the lower Russian River basin because of
the dam's location (upper Russian River) and the dam only controls a small fraction of the
drainage area. Historical data shows that contributions from tributaries below the Coyote Valley
Dam drive the flood peaks on the lower Russian River basin. Questions arose during initial
discussions on the FIRO study, regarding how much additional flood risk forecast-based
reservoir operations at the Coyote Valley Dam could possibly contribute. The WO_CVD
alternative, models the watershed without the dam, and provided context for the flood risk
damage in the FIRO study. The WO_CVD alternative was not under consideration as a FIRO
alternative and was not modeled in an FRA simulation.

Formulation of the WO_CVD Alternative

The WO_CVD alternative originated from the Existing Conditions alternative, linking Lake
Mendocino inflows as input to the HEC-RAS model instead of the reservoir outflow. FIRO
partners found the preliminary results from the WO_CVD alternative interesting and suggested
the alternative be included in the FIRO analysis as useful reference information. The modeling
team refined the approach by creating a new network in the HEC-ResSim model, which replaced
Lake Mendocino with a routing reach. This approach provided better results because Warm
Springs Dam operates to avoid flows greater than 35,000 cfs at Guerneville, and made different
releases in the absence of the Coyote Valley Dam.

Results
Figure 196 represents typical results, comparing selected river flows with and WO_CVD
alternative results for a minor event in March 1995. The green line (Figure 196) displays the
Existing Conditions alternative flows immediately below Coyote Valley Dam, and the stages at
Hopland, and at Guerneville (Hacienda). The red lines (Figure 196) display the results for these
locations without the Coyote Valley Dam. The high flows coming from the East Fork Russian
River without the dam largely coincide with high flows from the West Fork Russian River and
other tributaries, resulting in a higher peak stage at Hopland. The peak stages from the
WO_CVD alternative were higher throughout the Russian River, although the difference
diminished the further down the Russian River.

The results from the WO_CVD alternative show that Coyote Valley Dam reduced the hydrograph
peaks across all the events within the HEC-WAT deterministic analysis. Review of the detailed
results showed that adding to the peak flow at Hopland always resulted in higher stages at
downstream locations. This observation justified a simplifying assumption during plan
formulation of the FIRO alternatives, which used a trial-and-error approach to identify effective
forecast-based rules and establish optimum parameters based on results from the deterministic

237
Chapter 12 – Without CVD (WO_CVD) Alternative PR-100

Figure 196 Influence of Coyote Valley Dam for the March 1995 Flood Event

modeling. Since routing flows to Hopland within HEC-ResSim was adequate to indicate
whether more flood damage was caused along the river, the modelers could omit running HEC-
RAS and HEC-FIA during the inner iterations of the formulation process to screen ideas and
roughly estimate parameters.

When consequence modeling was included, the flood damage computed for individual events in
the deterministic simulations are displayed in Figure 185. The $10.0 million annual average
damage computed for the WO_CVD alternative was much higher than the $6.1 million computed
for the Existing Conditions alternative.

The difference in flood depth for the December 1964 event between the Existing Conditions
alternative and WO_CVD alternative is displayed in Figure 197. The difference was two feet or
less for most of the inundated area, but the agricultural land along the Russian River west of
Windsor suffered up to four feet of additional flooding without the Coyote Valley Dam.

For future analysis, the WO_CVD alternative might be useful for calculating USACE project
benefit.

238
PR-100 Chapter 12 – Without CVD (WO_CVD) Alternative

Figure 197 Difference in Depths for the 1964 Flood Event

239
Chapter 12 – Without CVD (WO_CVD) Alternative PR-100

240
PR-100 Chapter 13 – Simple FIRO - Encroach Alternative

Simple FIRO – Encroach Alternative

The Encroach alternative demonstrated one of the simplest FIRO-based approaches for Lake
Mendocino. The alternative added a single rule to the Coyote Valley Dam operations defined for
the Existing Conditions alternative, allowing the reservoir to store water in the flood pool until a
large storm appeared in the (perfect) forecast.

Formulation of Encroach Alternative

The Encroach alternative initially intended to help address questions discussed among the FIRO
team about whether forecast information could have improved operations in 2013. Figure 198
displays the observed level of Lake Mendocino and the guide curve. After a storm in late
December 2012, USACE followed the regulation plan and released water out of the lake's flood
storage when downstream river conditions allowed. No further storms of significance occurred
that winter, and USACE was unable to fill Lake Mendocino in the spring. Unfortunately, this
was also the beginning of a severe drought in the area, which lasted four years. USACE was
criticized for "short-sighted" operations and "outdated" policies that wasted water and
compounded the problems of the subsequent drought. Some of the most reliable forecasts in
California are those predicting an absence of rain, and various commenters suggested USACE
could have been more relaxed about clearing the flood storage after that final December storm in
2012. The Encroach alternative represented one possible approach to implement the suggestion.

Figure 198 Lake Mendocino Levels (2013)

241
Chapter 13 – Simple FIRO - Encroach Alternative PR-100

The process of formulating the new rule used in the Encroach alternative consisted of numerous
simulations of the deterministic HEC-ResSim model with different variations on the new rule.
The Encroach alternative results were compared to the results from the Existing Conditions
alternative to judge performance of the new rule according to four criteria:

- effectiveness for filling Lake Mendocino to the 761.8 feet NGVD summer target, which
represented the benefit to all conservation uses

- not increasing peak flow at Hopland, which represented a proxy metric for not increasing
downstream flood damage

- not increasing incidence of spillway flow, in order to avoid additional dam safety
concerns

- not increasing the number of flows at Hopland above 8,000 cfs, for maintaining the same
degree of compliance with the limitation in the existing water control manual.

The new rule used in the Encroach alternative required establishing the definition of a large
storm, and setting constraints on storage accumulation. A final refinement to the Encroach
alternative restricted its applicability to late winter and spring.

The new rule in the Encroach alternative, took the form of a simple variable guide curve, with
the target pool level toggling between floor and ceiling storage values according to the presence
or absence of a large storm in the forecast. The Encroach alternative operating set within the
HEC-ResSim model (Figure 199) does not include a specific rule in the Flood Control zone
related to encroachment, since the alternative simply modifies the active guide curve during a
simulation. The rule was implemented using a combination of HEC-ResSim state variables;
("dummy" rules were added to the Emergency Zone), which ensured that HEC-ResSim
evaluated the associated state variables at each reservoir simulation time-step.

Figure 199 HEC-ResSim - Reservoir Rules for the Encroach Alternative

242
PR-100 Chapter 13 – Simple FIRO - Encroach Alternative

The ceiling guide curve served as the active guide curve in the absence of a large storm in the
forecast, resulting in conservation operations while the pool was below 761.8 feet NGVD.
Reservoir storage accumulated up to the ceiling guide curve as inflows allowed. The floor
storage was set to the existing guide curve (737.5 feet NGVD), for most of the winter, then rising
in March and April. When the forecast included a large storm, the guide curve switched to the
floor value, which in this case amounted to reverting to the Existing Conditions alternative
operations. During the absence of a large storm, storage was accumulated, which encroached
into the flood pool, and was handled according to the standard flood operation rules. The
reservoir modelers experimented with different ceiling limits at different times during the winter,
but the additional complexity did not result in better performance against the four metrics. The
experimentation included various limitations on the rate of storage accumulation, but these were
also found ineffective.

The modelers defined a "large storm" as one totaling three inches or more of precipitation over
the catchment above Coyote Valley Dam during the next five days. Precipitation was chosen as
the most familiar forecast product, and characterizing the precipitation as a storm total best
suited the analysis of reservoir storage impacts. The observed precipitation was used as the
"perfect" forecast information. Forecast durations of two-, three-, and five-days were evaluated
with different thresholds for total precipitation. Longer forecast durations were not considered,
in keeping with the intention of a simple rule based on well-known forecast information with a
relatively high level of public trust. The five-day forecast period provided the best performance,
as expected since it supplied the most information with the same certainty as the shorter
forecasts. The modelers experimented with different precipitation thresholds, and determined
that three inches over five-days captured the most storage over the course of the sixty-year
deterministic simulation period, while still meeting the other three criteria.

Over the course of the simulations, the modelers noted that the largest historical storms occurred
in December and January, and that encroachments during early winter were most likely to cause
additional flood risk. In addition, storage accumulations from early in the winter often proved
unhelpful for filling toward the summer pool goal, since subsequent storms reverted operations
according to the Existing Conditions alternative guide curve. Maximizing the storage captured
from the final storm of the wet season seemed to be the most important action. Following further
experimentation, the new rule was allowed to go into effect on 15 January. This date helped
avoid encroachments during the period of highest flood risk, but accumulated storage when
storms were more than likely to be the final storm of the season. The modelers further
experimented with modified thresholds in May in order to avoid unnecessary flood operations
very late in the season, but the results did not justify complicating the plan.

The Encroach alternative formulated for this analysis represents a simple modification to current
water management practices, with very aggressive operations using perfect forecast information,
as part of a "bookend" analysis to determine the potential storage gain of FIRO operations. The
Encroach alternative was not intended as a proposal for actual project implementation.
However, the formulation of the Encroach alternative provided insights, which could be useful if
similar precipitation-based alternatives are considered when developing future modifications to
the Lake Mendocino water management practices.

243
Chapter 13 – Simple FIRO - Encroach Alternative PR-100

Typical Operations
Typical operations under the Encroach alternative occurring in the deterministic simulation
during winter of 2007 are displayed in Figure 200. The top plot window (Figure 200) contains
the hyetograph in light blue, along with the flag indicating the presence of a large forecasted
storm in red. The next plot window (Figure 200) shows the level of Lake Mendocino. The level
under the Encroach alternative is solid red (Figure 200), while the Existing Conditions
alternative level is shown in solid green (Figure 200). The dashed red line (Figure 200) indicates
the variable guide curve used under the Encroach alternative, while the dashed green line (Figure
200) shows the Existing Conditions alternative guide curve. The third plot window (Figure 200)
display the Coyote Valley Dam releases under the same alternatives, along with reservoir inflow
in black (Figure 200). The bottom plot window (Figure 200) shows the flows computed by the
HEC-ResSim model at Hopland under the two alternatives.

Figure 200 Typical Operations under the Encroach Alternative

244
PR-100 Chapter 13 – Simple FIRO - Encroach Alternative

Relatively dry conditions prevailed in early January 2007, and the variable guide curve switched
from 737.5 to 761.8 feet (NGVD) on 15 January, as the Encroach alternative rule activated.
Inflows barely exceeded the minimum conservation flows during the next few weeks, so very
little storage accumulated. As a large storm approached in early February, the precipitation flag
changed to "true", and the variable guide curve reverted to 737.5 feet NGVD. When the
outflows under the Encroach alternative reached 2,000 cfs, at this point the outflows were
ramped down to minimum as the reservoir level retreated to the normal guide curve (Symbol A,
Figure 200).

The storm produced reservoir inflows reaching 6,000 cfs, and resulted in Hopland flows of
almost 13,000 cfs. Operations under both the Existing Conditions and Encroach alternatives
minimized releases during the peak flows. However, the Existing Conditions alternative began
"backfilling" as the river receded on 10 February (Symbol B, Figure 200). The forecast
precipitation flag had cleared by that time, returning the variable guide curve to 761.8 feet, so the
Encroach alternative maintained minimum releases and allowed the pool to rise.

Later in February, another storm approached and triggered the forecast precipitation flag
(Symbol C, Figure 200). The variable guide curve reverted and flood operations commenced
with the pool encroached into Flood Control Schedule 2. Releases under the Encroach
alternative went to 4,000 cfs, then down to 2,000 cfs as the pool fell within Flood Control
Schedule 1. Under, the Encroach alternative, the flood pool was unable to empty before the
second storm arrived, and was still ramping down releases when runoff arrived. Consequently,
the Encroach alternative increased the peak flow at Hopland by 2,000 cfs over the Existing
Conditions (Symbol D, Figure 200) alternative. This instance of higher flow at Hopland had no
impact on flood damage, since flows in 2007 were insufficient to cause damage. In addition, the
Encroach alternative made no impact on the flow frequency curves used for comparison in this
analysis, because the highest peak of the year had occurred two weeks earlier. The instance did
not affect the metrics for compliance with the Hopland rule, since the Existing Conditions
alternative flow was above 8,000 cfs for the same length of time.

The second storm was smaller than the first, and the forecast precipitation flag cleared before the
runoff peaked. The simulation of the Encroach alternative reduced outflows to minimum and
stored the remaining inflows. The Existing Conditions alternative simulation increased releases
according to Flood Control Schedule 1, and lowered the lake until reaching the 737.5 feet winter
pool level on 28 February. On the next day, the Existing Conditions alternative guide curve
began rising toward the summer pool. Residual inflows were sufficient to keep the Existing
Conditions alternative simulation on guide curve for the first week of March, but the lake ended
up only reaching 742 feet NGVD. The Existing Conditions alternative filled the reservoir to 758
feet NGVD.

In general, the Encroach alternative did not result in significantly busier dam operations (i.e.,
more outflow changes) than under the Existing Conditions alternative. The number of additional
flood operations in advance of storms were generally balanced by fewer backfilling operations
after the storms passed.

245
Chapter 13 – Simple FIRO - Encroach Alternative PR-100

Results
The Encroach alternative performed significantly better at refilling the reservoir than the
Existing Conditions alternative, as described in the study comparison metrics presented in
Chapter 16. Although the flood damage was equivalent between the Encroach alternative and
Existing Conditions alternative during the formulation phase based on the deterministic analysis,
the FRA analysis resulted in greater flood damage under the Encroach alternative.

The highest level for Lake Mendocino simulated under the Encroach alternative for the
deterministic simulation occurred for the December 1964 event. Section 16.5.2 explains how the
Encroach alternative resulted in higher damage than the Existing Conditions alternative because
the reservoir entered the early-season storm in a less-depleted condition; a major finding from
the analysis. Any alternative that results in higher reservoir levels at the end of the fall compared
to the Existing Conditions alternative will have relatively less available storage in the reservoir
for early season storm events.

Section 16.5.3 describes a case from the FRA simulations where the Encroach alternative
resulted in higher damage from a late-season event due to encroachment in the traditional flood
pool. The worst case in the deterministic simulation results of additional flood damage due to
encroachment occurred in simulations of the 1986 event, where damage for the Encroach
alternative were one-hundredth of a percent higher than under the Existing Conditions
alternative.

Comparison of Coyote Valley Dam Levels

Figure 201 compares the Lake Mendocino levels with the baseline results for the first half of the
deterministic simulation analysis period. The Encroach alternative succeeded far more than the
Existing Conditions alternative at filling the reservoir to the summer pool level. The higher lake
levels carried throughout most of the year. The two sets of results matched only under guide
curve operations during early portions of unusually wet winters, or very rarely, during long
periods of flood operations.

The peak reservoir level under the Encroach alternative of 770.98 feet NGVD exceeded the
Existing Conditions alternative reservoir level for the December 1964 flood event by 4.2 feet.
The difference was because the Encroach alternative entered the storm with the pool less
depleted than the Existing Conditions alternative. Section 16.5.2 compares the simulations for
the December 1964 flood event. The additional spillway flow slightly increased downstream
flood peaks, resulting in $64,000 more damage under the Encroach alternative. The difference
was not considered significant for an event causing $65 million in flood damage.

The reservoir levels under the Encroach alternative performed better than the Existing
Conditions alternative through much of the drought of 1977, but reached the same degree of
depletion in August. Note that the reservoir nearly reached the spillway in 1974 under the
Encroach alternative. Figure 202 displays the pool levels for that winter. The Encroach and
Existing Conditions alternative resulted in identical operations during the early winter of 1973 -
1974. A storm in the middle of January led to flood operations under the Existing Conditions
alternative, while the Encroach alternative filled the reservoir to the summer pool level. Wet
246
PR-100 Chapter 13 – Simple FIRO - Encroach Alternative

Figure 201 Coyote Valley Dam Levels under the Encroach Alternative (1950 – 1980)

Figure 202 Coyote Valley Dam Levels under the Encroach Alternative (1974)

conditions in the drainage area above the lake kept inflows high enough to continue guide curve
operations until a small storm in late February. The precipitation was insufficient to revert the
variable guide curve, but raised the lake to within a few inches of the spillway. Heavier storms
at the start and the end of March did trigger the precipitation threshold and caused flood
247
Chapter 13 – Simple FIRO - Encroach Alternative PR-100

operations to clear most of the encroached storage. On 30 March, the level simulated under the
Encroach alternative was less than a foot higher than the level under the Existing Conditions
alternative, rendering the different operations during the preceding winter practically moot.
Operations after 30 March under the Encroach alternative stored the inflow from this last storm
of the season, and carried the pool into the dry season almost nine feet higher.

Figure 203 compares the Lake Mendocino levels with the baseline results for the second half of
the deterministic simulation analysis period. In 1983, 1995, 1999, 2003, and 2010 storms during
wet conditions that were too small to trigger the precipitation threshold pushed the lake up close
to the spillway, similar to the previously described 1974 operations. A storm in May 2005
caused the pool, under the Encroach alternative, to reach the spillway crest for one time step.
Preventing spillway flow in these years during the trial simulations led to the selection of the
three-inch precipitation threshold for use with the "ceiling curve" of 761.8 feet NGVD.

Figure 203 Coyote Valley Dam Levels under the Encroach Alternative (1980 – 2010)

The 1986 event resulted in the second-highest level for Lake Mendocino under the Encroach
alternative deterministic simulation results. The peak reservoir level was 0.54 feet higher than
for the Existing Conditions alternative, due to encroachment in the flood pool when the event
occurred, as shown in Figure 204. The encroachment resulted in slightly earlier and higher
amounts of spillway flow, which slightly increased the peak at Hopland, but not enough to make
a significant difference downstream. The 1986 flood damage computed under the Encroach
alternative exceeded the Existing Conditions alternative total of $46 million by only $4,400.

The largest difference between the peak reservoir levels for the Encroach alternative and the
Existing Conditions alternative in the deterministic simulation results occurred in January 1995
(Figure 205). The peak reservoir level of 767.6 feet NGVD was more than five feet higher than
simulated for the Existing Conditions alternative. As with the 1964 event, the different results
248
PR-100 Chapter 13 – Simple FIRO - Encroach Alternative

Figure 204 Encroach Alternative Operations for the 1986 Event

Figure 205 Coyote Valley Dam Levels for Encroach Alternative (1995)
249
Chapter 13 – Simple FIRO - Encroach Alternative PR-100

were due to incidental additional early season flood control storage for the Existing Conditions
alternative, as described in Section 11.3.3. The January spillway flow under the Encroach
alternative occurred too late to increase downstream flood peaks. Both simulations computed the
same $35 million damage result, based on the March 1995 event. Note that the Encroach
alternative switched to the upper guide curve just after the January event and stored the water
until releasing most of the water prior to the March event.

Compliance with Hopland Rule

The modelers compared frequency distributions of the deterministic results regarding Hopland
flows for the Encroach alternative and Existing Conditions alternative, and found the differences
negligible at that level of aggregation. In order to analyze differences on a release-by-release
basis, modelers used analyses (Figure 206) to evaluate compliance with the rule requiring water
managers to avoid Hopland flows greater than 8,000 cfs.

Figure 206 Comparison of Hopland Flows over 8,000 cfs for Encroach Alternative

The upper plot (Figure 206) displays the six-hour Hopland flows computed by HEC-ResSim for
each simulation, with flows below 8,000 cfs being replaced with the value 8,000. The Existing
Conditions hydrograph plotted first, so that the Existing Conditions flow was higher wherever
green is visible above the red (Figure 206). The lower plot (Figure 206) shows the results of
subtracting the Existing Conditions curve from the Encroach curve. The positive values indicate
occasions of additional Hopland flows exceeding 8,000 cfs due to the Encroach alternative.
Negative values (Figure 206) represent times that the Encroach alternative avoided flows of
8,000 cfs or more at Hopland. Figure 206 demonstrates that the results show very little
difference in net compliance with the rule.
250
PR-100 Chapter 13 – Simple FIRO - Encroach Alternative

Downstream Flood Damage

In general, the damage computed under the Encroach alternative was the same as the Existing
Conditions alternative, and occasionally slightly higher. The Encroach alternative cumulative
total of $363.2 million over the sixty-year deterministic simulation was $117,000 more than
under the Existing Conditions alternative. The $64,000 difference from the December 1964
event accounted for more than half the total. The other major differences were $28,000 for the
storm in February 1968, and $18,000 from the storm in February 1958. In both these situations,
the highly encroached pool caused brief releases of 4,000 cfs under Flood Control Schedule 2 in
response to the precipitation forecast. The releases increased the peaks downstream by very
small amounts, in the manner described in Section 11.4.1. Damage for the 1986 simulation
using the Encroach alternative exceeded the Existing Conditions damage for that event by
$4,000, and four others totaled $2,000 above the Existing Conditions damage. The Encroach
alternative slightly reduced damage in two deterministic simulation events, when "backfilling"
by the Existing Conditions simulation increased downstream peaks by small amounts.

The FRA simulation resulted in more occasions than the deterministic simulation where the
Encroach alternative caused higher damage than the Existing Conditions alternative. Most of the
damage differences between the FRA simulations of the Encroach and Existing Conditions
alternatives were due to higher flood peaks on the river resulting from higher and earlier spillway
flow during very large events either early or late in the flood season. The higher spillway flow
under the Encroach alternative occurred because of the alternative's tendency to have less flood
storage available at the beginning of events. Figure 207 illustrates the difference for Event 5
from Lifecycle 30, which was one of the events where the biggest difference in damage between
the two alternatives occurred. The event ranked as the 21st most damaging for the Existing
Conditions alternatives at $133.3 million, but the eleventh most damaging for the Encroach
alternative at $167.8 million.

Starting lake levels for the 18 January event were interpolated at the 6.7 percent exceedance
frequency from the input distributions defined for each alternative. For the Existing Conditions
alternative, the empirical distribution for the month of January resulted in an initial lake level of
743.48 feet NGVD, in other words, about six feet above the guide curve. The empirical
distribution defined for the Encroach alternative in late January resulted in a much higher
starting pool of 758.88 feet NGVD. Event 5 from Lifecycle 30 used the January 1995 shape set
and featured 21.95 inches of basin-average precipitation, which ranked ninth-highest for the
FIRO study. The Existing Conditions alternative was almost able to absorb the initial main
storm before spillway flow added to peaks downstream. The Encroach alternative began
uncontrolled spillway flow a day sooner, which substantially increased peaks downstream and
resulted in $33.5 million dollars more damage than the Existing Conditions alternative.

Figure 208 displays a frequency plot of the difference in flood damage for each FRA event
between the Existing Conditions and the Encroach alternative. The graph (Figure 208) uses
Weibull plotting positions as a display technique in order to show the extreme values in the tails
of the distribution. Since the ranking was based on damage difference instead of damage itself,
the X-axis does not represent annual exceedance frequency of event size. However, the largest
differences in damage totals occurred for large floods with certain characteristics, where small
increments of inundation depth represent large increments of damage.
251
Chapter 13 – Simple FIRO - Encroach Alternative PR-100

Figure 207 Encroach Alternative for Event 5 from Lifecycle 30

Figure 208 Differences between Damage for Existing Conditions and Encroach Alternatives for each
FRA Event
252
PR-100 Chapter 13 – Simple FIRO - Encroach Alternative

There was no difference in flood damage for 3,673 of the 5,000 events in an FRA simulation.
The difference in basin-wide damage exceeded $10,000 for 670 of the events. Positive
differences reflect events where the Encroach alternative resulted in lower damage than the
Existing Conditions alternative. The FRA simulation resulted in 81 events where the Encroach
alternative caused at least a million dollars more damage than the Existing Conditions
alternative. The largest differences were characteristic of events occurring during late January
through March, as explained previously for the February 1986 event. Some of the events where
the Encroach alternative resulted in greater damage than the Existing Conditions alternative also
resulted from higher pools entering early season storms, in the manner described previously for
the 1964 and 1995 events.

The deterministic results did not show any significant damage reductions under the Encroach
alternative, but the FRA simulation resulted in two events with approximately $3 million dollars
less damage computed for the Encroach alternative than the Existing Conditions alternative, as
shown by the left-most points in Figure 208. Event 31 of Lifecycle 31 resulted in $61.8 million
damage for the Encroach alternative, ranking as 144th most damaging event. Event 8 of
Lifecycle 52 resulted in $46.5 million and ranked 225th most damaging event. Both events
occurred in the first days of January, with 14.8 inches of basin average precipitation causing
spillway flow early enough to increase downstream peaks. The difference in outcomes was due
to the difference in the empirical distributions used for initial lake levels in late December. The
starting pool for the events sampled at the upper ends of the pool level distributions (1- and 2-%
exceedance), where the Existing Conditions alternative empirical distribution showed a higher
spread than the Encroach alternative. Consequently, the Encroach alternative lake levels started
about five feet lower than the Existing Conditions alternative. A few other simulations showed
slightly less damage under the Encroach alternative, these simulations reflected small differences
on downstream peaks. The Encroach alternative stored water after an initial storm, rather than
backfilling as the main storm approached.

The cases of higher damage under the Encroach alternative did not significantly affect the EAD.
The higher damage values were too rare among the 5,000 events, and the differences in damage
too small compared to the total damage amounts. The $7.75 million EAD across 5,000 events
was $140,000 higher than the Existing Conditions alternative. The difference was not considered
significant for this analysis, but might be significant in other studies. The plan formulation stage
of this study was based solely on deterministic results as a matter of convenience. If the analysis
required a very strict interpretation of maximizing success at re-filling the reservoir without
increasing flood risk, then FRA simulations would be necessary during formulation of the
alternative.

253
Chapter 13 – Simple FIRO - Encroach Alternative PR-100

254
PR-100 Chapter 14 - Complex FIRO - Combined Alternative

Complex FIRO – Combined Alternative

Where the Encroach alternative added one FIRO rule to the Existing Conditions operations, the
Combined alternative demonstrated a more comprehensive makeover of operations using
multiple rules based on different forecast information. The Combined alternative used forecasted
hydrologic flows instead of precipitation. As with the Encroach alternative, the Combined
alternative used observed values as "perfect" forecast information for deterministic rules.

Formulation of Combined Alternative

The Combined alternative modified the guide curve and substituted different rules, but did not
add more operational rules than used in the Encroach alternative. The Combined alternative
operations substituted an advance release rule based on reservoir inflow forecasts for the
precipitation rule, and replaced the Existing Conditions alternative maximum flow constraints
with a limit based on forecasted unregulated flows at Hopland. The Combined alternative
represents a more complex formulation approach because the guide curve was re-shaped for best
performance in conjunction with its FIRO rules.

The process of formulating the Combined alternative consisted of numerous simulations of the
deterministic HEC-ResSim model. The initial phase consisted of separate evaluations of
individual measures. The most promising ideas were carried forward to another phase of
simulations for refinement in consideration of trade-offs among them. As with the Encroach
alternative, the results were compared to the Existing Conditions alternative deterministic dataset
to judge performance of the trial according to four criteria:

- effectiveness for filling Lake Mendocino to the 761.8 feet NGVD summer target, which
represented the benefit to all conservation uses

- not increasing peak flow at Hopland, which represented a proxy metric for not increasing
downstream flood damage

- not increasing incidence of spillway flow, in order to avoid additional dam safety
concerns

- not increasing the number of flows at Hopland above 8,000 cfs, for maintaining the same
degree of compliance with the limitation in the existing water control manual.

Figure 209 displays the operation set used for the Combined alternative, with arrows pointing to
the rules differing from the Existing Conditions alternative. Three rules using hydrologic
forecast information were considered as individual measures. Reservoir inflow, soil moisture,
and downstream local flows were chosen as familiar types of information for use in reservoir

255
Chapter 14 - Complex FIRO - Combined Alternative PR-100

Figure 209 HEC-ResSim - Reservoir Rules for the Combined Alternative

operating rules, and could be readily available as short-term forecasts using CWMS. Comparable
information is also generated as part of routine forecasts performed by the CNRFC, and could be
provided to USACE District offices that perform real-time operations. The forecast information
in the Combined alternative analysis were observed values produced by the HEC-HMS model of
the Existing Conditions alternative. The HEC-HMS model computed flows that were the same
across all HEC-WAT alternatives. The forecast duration was limited to five days for consistency
with the Encroach alternative.

The forecasted reservoir inflow was used with an advance release rule (Pre_Release, Figure
209), following a simple deterministic implementation that borrows from the recent update to the
operations for Folsom Dam (California). The rule compares the inflow volume surplus from the
guide curve for one-, two-, three-, and five-day forecasts. The rule calculates a minimum release
based on the worst case among available forecasts in order to achieve guide curve at the end of
the forecast period. The rule re-computed at each reservoir simulation time-step. The result of
the advance release calculation was implemented as a minimum flow rule in both the Flood
Control and Conservation zones of the Combined HEC-ResSim operations set. The outflows
computed by the advance release rule were constrained by the other maximum, minimum, and
rate of change rules.

256
PR-100 Chapter 14 - Complex FIRO - Combined Alternative

The forecasted flows at Hopland, excluding releases from Coyote Valley Dam, were used with a
rule for maximum flow at Hopland (Max_Rel_Hopland). The Max_Rel_Hopland rule calculated
the allowable current release to come as close to 8,000 cfs as possible, including unregulated
flows and prior releases. The rule was also added to both the Flood Control and Conservation
zones, in place of the RR_Ukiah_flow block of rules used in the HEC-ResSim Baseline operation
set described in Section 6.6. The purpose of the Max_Rel_Hopland rule was to maximize
efficiency for advance releases and backfilling after peaks. The rule was included as a
demonstration of a potential use of a type of forecast information, but made a negligible
improvement in performance of the Combined alternative against the four criteria. The main
reason was that the RR_Ukiah_flow block of rules, defined in the Existing Conditions alternative,
was very efficient in its compliance with the Hopland 8,000 cfs limit, despite a lack of forecast
information, leaving little room for improvement. The Max_Rel_Hopland rule also used a very
simplified linear routing for releases, which occasionally caused it be to less accurate than the
Existing Conditions alternative rule.

Forecasted soil moisture information was considered in a separate rule during the initial
screening trials of individual measures. The rule made minimum releases of 100-, 200-, or 500-
cfs based on HEC-HMS results regarding the future wetness of the basin above the Coyote
Valley Dam. The rule was explored as a hedging strategy to provide more reservoir space during
conditions with the most vulnerability to large storms. This rule was abandoned during the
alternative formulation screening trials because the rule required long periods of elevated
releases in order to substantially affect the flood control storage, and diminished reservoir re-fill
success. The decision in the screening trials for the Combined alternative to ignore rules based
on forecasted soil moisture does not pertain to consideration of other rules based on current soil
moisture. Water managers traditionally factor the antecedent wetness in the watershed in their
decisions, and some regulation manuals contains formal rules based on such information. The
state of current soil moisture conditions makes intuitive sense as a strategy to hedge against
uncertainty in forecasts. Future analyses of proposed water management operations at Coyote
Valley Dam should revisit the use of current basin wetness information in conjunction with
different types of forecast information. For the PVA, the modeling team chose not to include
operations based on forecasted soil moisture due to ineffective performance in screening trials.
The team did not explore operations based on current soil moisture in the PVA for reasons of
expedience and simplicity.

Formulating the Combined alternative involved trade-offs among the storage zone definitions
and the implementations of the various rules regarding the four performance criteria. The
alternative retained a fixed guide curve approach, with the following modifications considered as
individual measures (Figure 23):

1) Earlier start of fill. Allowing the pool to rise sooner than 1 March, was effective for
capturing more storage, but increased flood risk from late-season storms. 1 February
was selected in the end.

2) Allow faster fill. A steeper slope for the "summer shoulder" of the guide curve was
effective for capturing more storage, but increased flood risk from late-season storms.
Allowing the pool to fill as soon as 10 March was selected in the end.

257
Chapter 14 - Complex FIRO - Combined Alternative PR-100

3) Enforce slower drawdown. A milder slope for the "winter shoulder" was attractive for
flexibility in releases for fishery management, but could increase flood risk from very
early storms. Reaching the winter pool on 30 November was selected in the end.

4) Raise winter pool level. Reducing the flood pool for the whole winter was effective for
capturing more storage, but increased flood risk. Winter pool levels ranged from 742.5
feet NGVD to 761.8 feet NGVD (i.e., year-round summer pool). Raising the winter pool
level during the screening trials was abandoned, and is not displayed in Figure 25.
Future analyses might revisit the idea, especially exploring the trade-offs for a higher
guide curve later in the flood season.

5) Raise summer pool level. Filling the reservoir higher was effective for capturing more
storage, but increased flood risk. This idea was abandoned as unrealistic based on
feedback from the USACE District office, which strongly wished to reserve the three
feet of freeboard below the spillway.

The other new rule (Dummy_Stor_Zone) displayed in Figure 209 served only as a modeling
technique required during initialization of the reservoir model.

Typical Operations
Figures 210 and 211 illustrate typical flood control operations under the Combined alternative
using blue lines. The plots (Figures 210 and 211) show the Existing Conditions alternative
results in green for comparison. The winter of 1982 through1983 was the wettest of the sixty-
year deterministic simulation analysis period. Although the timeframe lacked a large event, the
flood season provides an example of regulation plan performance through frequent flood
operations.

Figure 210 Coyote Valley Dam Levels under the Combined Alternative (1983)

258
PR-100 Chapter 14 - Complex FIRO - Combined Alternative

Figure 211 Coyote Valley Dam Flood Operations under Combined Alternative (1983)

Coyote Valley Dam reservoir levels under the Combined alternative finished water year 1982
about five feet higher than the Existing Conditions alternative, but both released storage during
autumn to draw the pool down to the winter guide curve level of 737.5 feet NGVD. The Existing
Conditions alternative reduced the storage to the winter pool level by releasing 800 cfs during
the final week of October 1982, as seen around Symbol A (Figure 210). The more gradual
decline under the revised guide curve for the Combined alternative resulted in minimum releases
until mid-November, when outflows of 500 cfs began forcing the pool level down. The later
"winter shoulder" revision to the Combined alternative guide curve generally avoided sudden
bursts of outflow in October, which were observed in the Existing Conditions alternative.
During a meeting of the FIRO group in April 2017, fishery experts expressed concern that such
October outflows could send false signals to migrating fish. Autumn operations did not
historically adhere to guide curve as strictly as the simulation of the Existing Conditions
alternative, so the October bursts may be more simulation artifacts than real problems to avoid.
However, the FIRO group expressed a preference for carrying more storage into November,
because it provides flexibility for management of fisheries.

Three small storms occurred later in November (Symbol B, Figure 210), and the advance release
rule caused outflows of up to 1,500 cfs. These forecast-based releases dipped the pool below its
guide curve ahead of each storm, and then allowed the pool to re-fill following the receding limb
of the forecast inflow hydrograph. The advance release rule raised outflows gradually ahead of

259
Chapter 14 - Complex FIRO - Combined Alternative PR-100

the 1 December storm, in contrast to the Existing Conditions alternative, which abruptly
increased outflows to back-fill after the peak downstream and clear the flood pool.

The advance release rule was unable to clear enough storage from the Conservation pool in
advance of the pair of storms in the second half of December. The rule computed desired
outflows ranging up to 6,500 cfs, but was limited to 2,000 cfs by the MaxReleaseWCM-FC (i.e.,
Flood Control Schedule 1). Consequently, the pool rose above guide curve, although less than
under the Existing Conditions alternative.

Figure 211 illustrates similar operations ahead of the largest storm of the winter in late January
1983. The Pre_Release rule caused outflows from 200 to 600 cfs ahead of the small storm
during 14 through 18 January. Reservoir inflows are displayed as the dashed black line (Figure
211) plotted along with the outflows. Just as the small storm passed and the reservoir level under
the Combined alternative returned to guide curve, a series of larger storms appeared in the
forecast.

The advance releases were again limited to 2,000 cfs (Symbol C, Figure 211), and then ramped
down to the 25 cfs minimum as the second of the three storms triggered the Hopland flow
criterion (Symbol D, Figure 211). The greater accuracy of the Max_Rel_Hopland rule allowed
the Combined alternative to ramp down one time step later than the Existing Conditions
alternative while still avoiding 8,000 cfs at Hopland. The Max_Rel_Hopland rule
underperformed on 24 January, when it prevented back-filling after the peak by a time step
longer than necessary.

The Combined and Existing Conditions alternatives resulted in the same outflows during the
largest storm of January 1983. The downstream flood peaks and $14 million simulated flood
damage were the same for both alternatives. The reservoir reached Flood Control Schedule 2 in
both results, although the pool under the Combined simulation peaked about four feet lower.
The Max_Rel_Hopland rule helped empty the Flood zone faster by allowing reservoir outflows
to continue at 4,000 cfs throughout 29 January (Symbol E, Figure 211), while the Existing
Conditions alternative reduced outflows temporarily due to briefly rising flows on the West Fork
of the Russian River. The Combined alternative maintained the 2,000 cfs releases under Flood
Control Schedule 1 until the reservoir level reached guide curve on 3 February, when the
outflows ramped down according to the decreasing rate of change rule (Symbol F, Figure 211).

The Combined alternative stored the inflows from several small storms in February according to
the earlier fill and faster allowable rise to guide curve (Symbol G, Figure 210). The advance
release rule controlled the outflows, ranging up to 1,800 cfs, until 25 February. The Existing
Conditions alternative toggled in and out of the 2,000 cfs release from Flood Control Schedule 1
during the same weeks in February. Heavier storms during the last week of February and first
week of March triggered the Hopland criteria under both simulations, with the
Max_Rel_Hopland rule more effective at maintaining just under 8,000 cfs through the period.

The advance release rule again governed outflows for the Combined alternative simulation
through several other small storms in March and April (Symbol H, Figure 210). Releases ranged
from 900 to 1,200 cfs, aside from one very brief reduction caused by the Hopland rule. As a
note, regarding future alternative formulation, current information about soil moisture in March
and April could have been used as criterion to avoid the advance releases as unnecessary. The
260
PR-100 Chapter 14 - Complex FIRO - Combined Alternative

pool attained the summer level of 761.8 feet NGVD about two months earlier than the target date
for the Existing Conditions alternative guide curve. The lake level eventually peaked at 757.5
feet NGVD in June under the Existing Conditions simulation.

The simulation results do not support any conclusions about the net number of outflow changes
due to the advance release rule. At first glance, these simulations suggest that the Combined
alternative may result in more dam operations (i.e., outflow changes) than under the Existing
Conditions alternative. The Combined alternative resulted in numerous small outflow
adjustments in advance of storms according to the forecasted inflow, which did not occur under
the Existing Conditions alternative. However, the Combined alternative tended toward fewer
backfilling flow changes than the Existing Conditions alternative, with more periods of constant
2,000 cfs releases after storms according to the Flood Control Schedule 1. The 2,000 cfs flow
rule is a simplified assumption for modeling situations subject to significant operator discretion,
and the model results may not adequately represent the real-life net impact on the number of
outflow changes.

Results
The study comparison metrics for the Combined alternative are presented in Chapter 15. Section
15.5.1 describes how differences in the outflows between the Combined alternative and the
Existing Conditions alternative can result in very small differences in subsequent peaks in the
Russian River downstream.

The highest level for Lake Mendocino simulated under the Combined alternative for the
deterministic simulation occurred for the December 1964 event. As noted elsewhere in this
analysis, simply filling the pool successfully to the summer guide curve level can increase the
flood risk from large events early in the following winter. Section 15.5.2 explains how the
Combined alternative resulted in lower damages than the Existing Conditions alternative because
of the advance release rule, despite entering the early-season storm with a higher reservoir level.

Section 15.5.4 describes a case from the FRA simulations where the Combined alternative
resulted in higher damage from a late-season event due to guide curve changes for earlier and
faster fill-up. The deterministic simulation results did not include any cases of flood damage
resulting from early fill under the Combined alternative, although the event of March 1995 came
close.

Comparison of Coyote Valley Dam Levels

Figure 212 compares the Lake Mendocino levels from the Combined alternative with the results
from the Existing Conditions alternative for the first half of the deterministic simulation analysis
period. The Combined alternative succeeded far more than the Existing Conditions alternative at
filling the reservoir to the summer pool level. As with other alternatives, the higher lake levels at
the end of flood season carried throughout most of the year.

The Combined alternative did not the fill the pool as reliably as the Encroach alternative,
because the Combined alternative did not start storing inflow until 1 February, and the guide
curve constrained the amount retained until 1 March. As displayed in Figure 213, the Combined
261
Chapter 14 - Complex FIRO - Combined Alternative PR-100

Figure 212 Coyote Valley Dam Levels under the Combined Alternative (1950 – 1980)

Figure 213 Distribution of Lake Mendocino Storage on 10 May

262
PR-100 Chapter 14 - Complex FIRO - Combined Alternative

alternative matched the Encroach alternative performance in filling the reservoir to the summer
pool in about half the years of the deterministic simulations. The Combined alternative was
effective at avoiding spillway flow or adding to downstream flood peaks during December and
January. The Combined alternative simulation of the December 1964 event contained the
reservoir below the spillway, as discussed in Section 15.5.2.

Like the Encroach alternative (Figure 202), the Combined alternative almost resulted in spillway
flow in the 1974 simulation (Figure 214). The level of Lake Mendocino reached 764.5 feet
NGVD. However, the high pool levels under the Combined alternative occurred due to storms
on 31 March through 1 April while the reservoir was full. The Combined alternative releases did
not increase downstream peaks, and the late-season storms caused no flooding.

Figure 214 Coyote Valley Dam Levels under the Combined Alternative (1974)

Figure 215 compares the Lake Mendocino levels with the baseline results for the second half of
the deterministic simulation analysis period. These years showed the same improvements over
the baseline results as the earlier period, with more success at filling the reservoir to the summer
pool, and ending with less depletion throughout the dry season. The three occasions of spillway
flow under the Combined alternative occurred in 1986, 1995, and 2006.

Figure 216 explains the simulation of the 1986 event, which resulted in the highest Coyote
Valley Dam elevation of the deterministic simulation under the Combined alternative. The
reservoir was beginning to fill in early February when the storm appeared in the inflow forecast.
The Coyote Valley Dam outflows increased according to the advance release rule as the event
approached, until the 2,000 cfs maximum for Flood Control Schedule 1 controlled releases. The
advance releases reduced the pool to 737.51 feet NGVD, so that the reservoir started the storm
coincidentally at practically the same level as the Existing Conditions simulation. The outflows
ramped down to the minimum 25 cfs during the flood, resulting in the same releases, pool levels,
downstream peaks, and damages as the baseline results. The Combined alternative stopped
backfilling sooner than the Existing Conditions alternative and stored the receding inflows
according to its higher guide curve.
263
Chapter 14 - Complex FIRO - Combined Alternative PR-100

Figure 215 Coyote Valley Dam Levels under the Combined Alternative (1980 – 2010)

Figure 216 Combined Alternative Operations for the 1986 Event

264
PR-100 Chapter 14 - Complex FIRO - Combined Alternative

Figure 217 illustrates the main trade-off regarding flood risk by comparing the performance of
the plan for the January and March events. The Existing Conditions simulation entered January
with Lake Mendocino highly depleted and well below guide curve when the large storm
occurred. The incidental flood space allowed the reservoir to absorb the event without spillway
flow. The lake level under the Combined alternative was at guide curve prior to the January
storm. The advance release rule allowed the reservoir to clear enough storage, which allows the
reservoir to contain the January flood below the spillway. Note that the January events consisted
of two peaks, with the higher second peak reduced under the Combined alternative. The two
peaks occurred because flows at Hopland receded briefly between the successive storms and the
Combined alternative backfilled more aggressively than the Existing Conditions simulation.

Figure 217 Coyote Valley Dam Levels under the Combined Alternative (1995)

The earlier fill of the guide curve used by the Combined alternative resulted in spillway flow for
most of 15 March, with a peak reservoir level of 766.67 feet NGVD. The March event also
featured a pair of successive storms. The Combined alternative stored all the inflow from the
first storm in order to raise the lake close to its guide curve (i.e., the summer pool level of 761.6
feet NGVD). In addition, the advance release rule was slow to react to the second storm because
the pool remained below guide curve for a short time after the first storm. By the time the
advance release rule attempted to clear storage before the second storm, the rule was constrained
by the flow limit at Hopland. The Combined alternative resulted in no additional flood damage
because the spillway flow occurred after the peak passed downstream. In fact, the lower releases
of the Combined alternative in the time between the March storms very slightly reduced the
downstream peaks for the second storm, and resulted in $3,400 less flood damage.

Compliance with Hopland Rule

The modelers compared frequency distributions of the Hopland flows for the Combined and
Existing Conditions deterministic simulations, and found the differences too few and small to
265
Chapter 14 - Complex FIRO - Combined Alternative PR-100

analyze in terms of frequency. Instead, the modelers used the data displayed in Figure 218 to
evaluate compliance with the rule requiring water managers to avoid Hopland flows greater than
8,000 cfs.

The upper plot (Figure 218) displays the six-hour Hopland flows computed by the HEC-ResSim
model for each simulation, with flows below 8,000 cfs, which were replaced with the value
8,000. The Existing Conditions hydrograph plotted first, so that the baseline flow was higher
wherever green is visible above the blue (Figure 218). The lower plot (Figure 218) shows the
results of subtracting the Existing Conditions curve from the Combined curve. The positive
values indicate occasions of additional Hopland flows exceeding 8,000 cfs due to the Combined
alternative. Negative values represent additional times that the Combined alternative avoided
flows of 8,000 cfs or more at Hopland. Figure 218 demonstrates that the Combined alternative
improved net compliance with the rule by a small amount.

Figure 218 Comparison of Hopland Flows over 8,000 cfs for the Combined Alternative

Downstream Flood Damage

The advance release rule did not add to downstream peaks computed with the hydraulic model,
since Flood Control Schedule 1 limited releases to 2,000 cfs. The HEC-RAS model was
configured to have at least 2,000 cfs flowing into the East Fork Russian River, and was not
sensitive to outflow changes below that threshold.

The Combined alternative's cumulative damage total of $363.3 million over the sixty-year
deterministic simulation was $270,000 more than under the Existing Conditions alternative. The
difference was primarily due to the 2006 event, for the Combined alternative damage exceeded
266
PR-100 Chapter 14 - Complex FIRO - Combined Alternative

the Existing Conditions alternative by $299,000, as described in Section 15.5.1. The

deterministic simulation for 1970 resulted in $25,000 less flood damage for the Combined
alternative than the Existing Conditions alternative. The peak stages downstream occurred on 25
January, after the fifth of six successive storms beginning on 9 January. The Combined
simulation made a brief advance release prior to the first storm, reducing the reservoir level a
few feet lower than the simulation of the Existing Conditions simulation. Subsequent operations
were very similar between the two simulations, with releases limited by Flood Control Schedule
1 or downstream criteria. The Existing Conditions alternative lake level rose briefly into Flood
Control Schedule 2 until 22 January, while the pool under the Combined alternative remained
within Flood Control Schedule 1. Consequently, the Combined alternative maintained slightly
lower releases, and reduced the flow by a small amount from the downstream peaks.

The FRA simulation resulted in more occasions of where the Combined alternative caused higher
damage than the Existing Conditions alternative in the deterministic simulation. Most of the
damage differences between the Combined and Existing Conditions FRA simulations resulted
due to higher flood peaks on the Russian River resulting from higher and earlier spillway flow
during very large events early or late in the flood season. For the March flood, Event 49 from
Lifecycle 51 resulted in the greatest damage difference between the two alternatives. With $29.1
million, more damage for the Combined alternative than the Existing Conditions alternative. The
event ranked as the seventeenth most damaging event for the Existing Conditions alternative at
$142.1 million, but only the ninth most damaging event for the Combined alternative at $171.2
million. The difference was due to high initial pool levels (3.2 percent exceedance). The
corresponding reservoir level for the early March distribution of the Combined alternative was
761.80 feet NGVD, versus 744.65 feet NGVD for the Existing Conditions alternative.

Figure 219 displays a frequency plot of the difference in flood damage for each FRA event
between the Existing Conditions alternative and the Combined alternative. The plot (Figure 219)
uses Median plotting positions as a display technique in order to show the extreme values in the

Figure 219 Differences between Damage for Existing Conditions and Combined Alternatives for each
FRA Event

267
Chapter 14 - Complex FIRO - Combined Alternative PR-100

tails of the distribution. Since the ranking was based on damage difference instead of damage
itself, the X-axis does not represent annual exceedance frequency of event size. However, the
largest differences in damage totals only occurred in some of the largest floods, where small
increments of inundation depth represented large increments of damage.

There was no difference in flood damage for 3,775 of the 5,000 events. The difference in basin-
wide damage exceeded $10,000 for 428 of the events. For 324 events, the Combined alternative
resulted in at least $10,000 more damage than the Existing Conditions alternative, with 41 of
those events resulting in at least $1,000,000 more damage. In 104 events, the Combined
alternative resulted in at least $10,000 less damage than the Existing Conditions alternative, with
nine of those events resulting in at least $1,000,000 less damage.

The left-most point (Figure 219) represents Event 8 of Lifecycle 52, in which the Combined
alternative resulted in $39.9 million dollars of damage, while the Existing Conditions alternative
resulted in $49.5 million dollars of damage. Figure 220 displays the difference in damage due to
a lower initial level of Lake Mendocino for the Combined alternative, resulting in lesser spillway
flow beginning later in the event. Event 8 of Lifecycle 52 provides the most extreme example,
but each of the events in which the Combined alternative resulted in significantly less damage
than the Existing Conditions alternative occurred in late December or in January with starting
pool levels sampled at the high end of the empirical distributions shown in Figures 144 and 145.
The Existing Conditions alternative distributions for those months show a higher spread at the
upper end than the Combined alternative distributions. The difference in the distributions of
initial reservoir elevation occurs because lake levels for the Combined alternative deterministic
simulation did not rise as high as for the Existing Conditions alternative, since the advance
release rule cleared storage before storms rather than afterward.

Figure 220 Combined versus Existing Conditions Alternative Pool Levels for Event 8 from Lifecycle 52

268
PR-100 Chapter 14 - Complex FIRO - Combined Alternative

Due to the perfect forecast assumption of this analysis, the results do not explicitly address the
potential for additional downstream damage by advance release based on a missed forecast,
where a storm arrives sooner than expected. However, the impact would be limited by the Flood
Control Schedule 1 limit of 2,000 cfs, and by routing effects described in Sections 11.4.1 and
15.5.1.

269
Chapter 14 - Complex FIRO - Combined Alternative PR-100

270
PR-100 Chapter 15 – Encroach Alternative with Imperfect Forecast

Encroach Alternative with Imperfect

Forecast
The goal of the Encroach with Imperfect Forecast (EncroachWIF) alternative was to
demonstrate a mechanism usable within a FIRO-based study to compute reservoir operations
with an HEC-ResSim model based on imperfect forecast information, and compare deterministic
and FRA results against an alternative based on perfect information.

Formulation of EncroachWIF Alternative

The EncroachWIF alternative resembled the Encroach alternative, except the simulation
recognized that forecasts have errors, and was based upon imperfect forecast information.
Rather than basing the release decision on the actual future precipitation, the EncroachWIF
alternative relies on forecast precipitation that was randomly perturbed from the observed values.
The Encroach alternative was chosen as the basis of the comparison because of the simplicity of
the forecast data and the water management rule that uses the alternative. For the purposes of
comparison, the EncroachWIF alternative rather naively assumes the forecast is perfect and uses
the same reservoir operations as formulated for the Encroach alternative. A water management
alternative formulated with awareness that the forecast information is imperfect would typically
include mechanisms such as hedging to mitigate the effects of uncertainty.

The perturbation of the forecasts was accomplished by randomizing the five-day total
precipitation at each time step to create a reasonable forecast series. This process or analysis
represented a very simple description of skill, defined as an average forecast error that was some
percentage (e.g., twenty percent) of the forecast magnitude, and described by a Normal
distribution. Using the percentage rather than a constant depth allowed error to increase with
precipitation depth (i.e., skill decreased with forecast magnitude).

Analyzing the impact of uncertainty or forecast skill on the effectiveness of potential water
management operations can be a complex task. This analysis requires consideration of the types
of data being forecasted, the nature of the forecast process, the formulation of reservoir
operations that use the forecast information (perhaps with an awareness of its uncertainty), and
the specific implementation of the reservoir operations, as well as the interplay among these
factors. The approach employed in this study represents one initial attempt at a simple
sensitivity analysis, intended to gain insights regarding the Lake Mendocino operations, and
inform research by FIRO partners regarding forecasting improvements.

All of the logic for the EncroachWIF alternative was implemented within the HEC-ResSim
models. HEC-WAT treated the different HEC-ResSim model alternatives just like another
model alternative with a different water management alternative.

271
Chapter 15 – Encroach Alternative with Imperfect Forecast PR-100

Perturbing Method - Deterministic Simulation

Perturbation of the precipitation information was accomplished by randomizing the five-day total
precipitation at each time step to create a reasonable forecast series, which was generated by
logic added to a state variable script in an HEC-ResSim model alternative. The method of
perturbing the precipitation amounts to creating random forecasts that required the generation of
a series of Uniform [0, 1] pseudo random values, in other words, numbers ranging between 0.0
and 1.0. The random values used to perturb precipitation are generated differently for
deterministic and FRA simulations. For the deterministic simulations, a defined time series of
pre-generated random values is used. Given the initial Uniform [0, 1] random value for a time
step, the state variable script generates subsequent pseudo random values and uses them to
produce perturbed precipitation forecasts from the observed precipitation using an
Autoregressive Lag 1 (AR(1)) process with a Normal error.

For a given event, the initial precipitation forecast is randomized from a Normal distribution with
a mean of the actual depth, and a standard deviation defined as a user-defined percentage of the
actual depth. The forecast in each subsequent time step is generated as an AR(1) value having a
user-defined serial correlation with the previous step. For this analysis, the serial correlation was
a constant value of 0.8. Both of the user-defined values reflect an attempt to capture forecast
skill, and the success of the resulting operation is affected by the values specified.

The HEC-ResSim model's state variable ultimately provides a five-day forecast series as input to
the operations defined for the Encroach alternative. However, to allow for a more flexible
description of forecast error as increasing with lead-time, the forecast error statistics (standard
deviation as percentage of actual) are defined separately for Day 1, Day 2, Day 3, Day 4, and
Day 5. For this analysis, the percent error used for Days 1 through 5 were constant at 0.05, 0.08,
0.13, 0.21, and 0.35. The HEC-ResSim model's state variable script finds the amount of
precipitation in the next 1-, 2-, 3-, 4-, and 5-days to isolate each day, and then computes the
standard deviation for each. The script combines the error contributions from each day, and then
using the generated random U[0, 1] value for each five-day forecast in the series, computes the
AR(1) result based on that five-day sum and the specified serial correlation. This process
generates the perturbed forecast series. As with the Encroach alternative, if the forecasted
precipitation was less than the threshold value of three inches, then the pool was allowed to
encroach into flood zone.

Perturbing Method - FRA

The pre-defined set of random variables for each time step during the deterministic simulations
was unnecessary for the FRA simulations. The perturbing of the precipitation state variable for
the FRA simulations generated random values for each time step of an event using an HEC-
ResSim model Monte Carlo random variable as a seed.

Typical Operations
Simulating the same water management alternative using perfect and imperfect forecast
information resulted in situations where the operations performed both better and worse because
of the forecast error. Figure 221 illustrates the operation of the EncroachWIF alternative in
272
PR-100 Chapter 15 – Encroach Alternative with Imperfect Forecast

Figure 221 March 1987 Simulation of the EncroachWIF Alternative

contrast to the Encroach alternative using the perfect forecast. The upper plot (Figure 221)
displays the precipitation values with time. The solid blue bars (Figure 221) represent the actual
incremental precipitation on six-hour intervals. The blue line in the precipitation plot (upper
plot, Figure 221) shows the actual precipitation total for the next five days. The gold line (Figure
221) shows the imperfectly forecasted five-day total. The areas striped in gold (Figure 221)
indicate when the precipitation flag was set "true" for the EncroachWIF alternative. The red line
(Figure 221) shows the state of precipitation flag for the Encroach alternative using the perfect
forecast.

The lower plot (Figure 221) displays the Lake Mendocino level under the imperfect forecast as
the gold line, with the lake level under the perfect forecast in red. The variable guide curves for
each simulation are displayed as red and gold dashed lines (Figure 221). The dashed green line
(Figure 221) displays the Existing Conditions guide curve.

Reservoir levels carried over from the prior year differed slightly, but the simulations remained
in close agreement through the winter of 1986 through 1987. The few storms that occurred were
insufficient to raise the pool to a level where the Encroach alternative applied, until a storm
during the second week of February. By 12 February, the storm was almost over, the
precipitation flag turned "false", and the variable guide curve for both simulations switched to
273
Chapter 15 – Encroach Alternative with Imperfect Forecast PR-100

761.8 feet NGVD. Both simulations stored the residual inflows from the February storm, and
absorbed the inflow from small storms that occurred on 3 and 5 March. Storms during the
second week of March were heavy enough to cause the perfect forecast simulation to revert to
the lower guide curve for a few days, and release all of the encroached storage. The imperfect
forecast underestimated the storms, and continued to absorb inflow until it reached the higher
guide curve on 18 March. There was little additional precipitation that spring, and the Encroach
alternative using the perfect forecast fell ten feet short of the summer pool level.

Figure 222 displays the operations for the 1986 flood. The EncroachWIF alternative initially
underestimated the event, and the simulation reverted to the lower guide curve twelve hours later
than the simulation under the Encroach alternative. The delay rendered the EncroachWIF
alternative less successful in clearing out the encroached storage, and the lake peaked 0.3 feet
higher than with Encroach alternative. The greater spillway head made the downstream peaks
slightly higher than under the Encroach alternative, resulting in an additional $69,600 to the
$45.7 million simulated event damage.

Figure 222 Simulation of 1986 Event under EncroachWIF Alternative

The EncroachWIF alternative initially overestimated the storm in early March by just enough to
cross the three-inch precipitation threshold, which caused the guide curve to revert thirty hours
earlier than under the Encroach alternative. That was the final storm of the season, and
subsequent inflows proved insufficient for the reservoir to recover the lost volume. The
Encroach alternative successfully re-filled the reservoir to the full summer pool, while the
EncroachWIF alternative fell short by 1.3 feet.

274
PR-100 Chapter 15 – Encroach Alternative with Imperfect Forecast

Results
The study comparison metrics for the EncroachWIF alternative are presented in Chapter 16.
Performance with the EncroachWIF alternative information was nearly the same as with the
Encroach alternative, particularly regarding flood risk metrics.

The highest level of Lake Mendocino simulated under the EncroachWIF alternative was 771.13
feet NGVD during the December 1964 event (Figure 223). The peak level was 0.15 feet higher
than the simulation with Encroach alternative, because of a difference in reservoir level due to
slightly better carryover storage from earlier in the year. The greater spillway head made the
downstream peaks slightly higher than under the Encroach alternative, resulting in an additional
$72,700 to the $64.9 million simulated event damage.

Figure 223 Simulation of 1964 Event under EncroachWIF Alternative

Note that a few days ahead of the storm, the EncroachWIF alternative expected up to 25 inches
of precipitation, while the actual five-day total reached about seventeen inches. The forecast
error had no effect on the simulated reservoir operations, because the reservoir elevations under
both alternatives were below the guide curve when the forecasted precipitation crossed the three-
inch threshold.

The additional damage for the 1964 and 1986 events were the only occasions of higher damage
in the deterministic simulation using the EncroachWIF alternative. Results for events in 1958
and 1968 using the EncroachWIF alternative actually resulted in $10,800 and $27,900 less
damage than the simulations using the Encroach alternative. The EncroachWIF alternatives
275
Chapter 15 – Encroach Alternative with Imperfect Forecast PR-100

reverted the guide curve sooner than the perfect forecast, giving the EncroachWIF alternative a
head start of twelve hours when reducing releases from Flood Control Schedule 2 to Schedule 1.
This option resulted in slightly lower peaks downstream during the subsequent main storm.

The FRA results for the simulations using the Encroach and EncroachWIF alternatives were
even more similar than the deterministic simulation results, since the events started with exactly
the same sampled initial lake level. Differences were due solely to threshold effects regarding
the reservoir release rules, or routing effects of small differences in the timing of Coyote Valley
Dam releases, as described in Section 16.5.1. Damages under the EncroachWIF alternative were
higher in 53 of the 5,000 events, and lower in 51 of the events. Figure 224 illustrates the
differences for each event using the same plotting technique as for the previous alternatives, but
note the very small range of the vertical axis. The largest difference occurred in Event 28 from
Lifecycle 3, which ranked 1,791st of the 5,000 events by total damage. The imperfect forecast
information resulted in the EncroachWIF alternative reacting to a forecasted event twelve hours
later than the Encroach alternative. The Encroach alternative was able to reduce the level of
Lake Mendocino two feet lower than the EncroachWIF alternative, and make Flood Control
Schedule 1 releases prior to the event while the EncroachWIF alternative made Flood Control
Schedule 2 releases. The higher outflow made the downstream peaks slightly higher under then
EncroachWIF alternative. The resulting damage of $1.35 million was $66,000 higher than that
of the Encroach alternative.

Figure 224 Difference between Damage for EncroachWIF and Encroach Alternatives for each FRA
Event

The events shown in Figures 222 and 223 illustrate the main reasons that the water management
operations of the Encroach alternative were insensitive to forecast error regarding performance
metrics for flood damage and spillway flows.

1) The variable guide curve only applies after 15 January, so the forecast information played
no role in most of the biggest simulated events.

276
PR-100 Chapter 15 – Encroach Alternative with Imperfect Forecast

2) Even if the variable guide curve had been in effect during the large over-forecast in 1964,
it would have made no difference to operations. Both perfect and imperfect forecasts
crossed the three-inch threshold at the same time. The operations key off a simple binary
test, and would be similar if 100 inches were forecasted.

3) Even when the forecast timing was off, such as the twelve-hour delay in the simulation of
1986 flood, the impact on storage was typically minor. The amount of encroached
storage that flood control operations could have cleared during that half-day regarding the
1986 event made no significant difference to the dam's ability to protect downstream
communities.

The performance in filling the pool each spring differed from year to year in the results between
the Encroach alternative and the EncroachWIF alternative. As seen in Figure 225, the net result
was a wash regarding performance on the metric for attaining summer pool with perfect and
imperfect forecasts. The forecast error for when the five-day precipitation dropped below three
inches, sometimes proved significant for the final storm of the year, since the final storm affected
when the simulation switched from flood operations to storing water. The accidental success
shown in Figure 221 of filling the reservoir more effectively under the imperfect forecast
simulation for 1987 also occurred in 1953, 1971 and 2004. The forecast error worked both ways,
as shown in Figure 222, and the perfect forecast simulation filled the reservoir more successfully
than the imperfect forecast in 1954, 1960, 1986, 2002, and 2009.

Figure 225 Distributions of Lake Mendocino September 30th Storage

277
Chapter 15 – Encroach Alternative with Imperfect Forecast PR-100

The lack of sensitivity to forecast error observed in this analysis should not be interpreted as
proof that forecast skill matters little to forecast-informed reservoir operations. Instead, the
analysis reflects the complexity of properly simulating the uncertainty associated with forecast
skill. Which includes the impact on the specific reservoir operations under evaluation and the
system being modeled. For the three reasons described above, forecast error resulted in rare and
minor differences to these specific operations based on a low precipitation threshold. The
consequences of those differences were also muted by the distance between Lake Mendocino
and the downstream damage centers. The operations of the Encroach alternative were so simple
that the only forecast uncertainty of consequence to flood operations was the timing of when the
five-day total crossed the three-inch precipitation threshold. The modeling technique used in this
analysis did not perturb the timing of the precipitation, and as such rendered such timing to be
relatively more certain for large events than small ones.

Closer investigation of the results, suggests that refinements to the techniques for representing
forecast uncertainty should be considered in future analyses. The method used here to perturb
observed precipitation may not adequately represent big "misses", in other words, forecasts that
fail to anticipate a significant storm or predict one that does not happen, because the storm
follows an unexpected path or lingers over a sensitive location longer than predicted.
Additionally, choosing twenty percent of the forecast magnitude to represent average forecast
error may have been overly optimistic. This value was chosen without detailed analysis for
expedience in this preliminary study. This assumption, if used in any future studies that employ
similar methods for simulating uncertainty in precipitation forecasts, should be considered
carefully.

278
PR-100 Chapter 16 – Results of Modeling

Results of Modeling
The HEC-WAT results were analyzed to quantify the potential storage that can be made
available under different reservoir operation scenarios for Lake Mendocino and to determine if
the FIRO-based alternatives increased flood risk. The primary analysis consisted of relative
comparisons between the FIRO-based alternatives and the baseline (Existing Conditions
alternative), according to performance metrics described in Section 4.5. Section 16.5.1 provides
additional analysis of specific events and explains other findings and insights regarding flood
risk on the Russian River.

Conservation Storage
The analysis designed FIRO-based alternatives that capture the most storage without a
significant increase in flood damage. Deterministic simulations of sixty years of reservoir filling
and depletion under each reservoir operation alternative were performed to determine the
potential storage gain under a perfect forecast scenario.

The schedule of the existing rule curve calls for Lake Mendocino to reach its full summer pool of
761.8 feet NGVD by 10 May of each year. In practice, the reservoir has seldom attained this
level. One measure of the success of a FIRO-based alternative is the increased likelihood of
refilling to the full summer pool level. Figure 213 compares the frequency of storages simulated
on 10 May for the reservoir operation scenarios, using results from the deterministic simulation.
The FIRO-based alternatives modeled in this analysis typically capture about 20,000 to 29,000
acre-feet more than the Existing Conditions alternative across the full range of dry and wet years.
Using the median as a simple metric, the FIRO-based alternatives all show a gain of 29,000 acre-
feet over the Existing Conditions alternative. This 29 kaf gain represents a reasonable estimate
of the additional storage possible without significantly increasing flood risk, since these
alternatives used aggressive operations leveraging perfect knowledge of future conditions.

Figure 225 displays a similar ranking of storages under each alternative at the end of the water year.
The alternatives that started a season with higher lake levels also finished with higher levels,
although summer depletions compressed the difference in levels among the water management
alternatives. The additional storage during autumn provided under the FIRO-based alternatives
should be helpful for fishery support.

Compliance with Hopland Flow Rule

The water control manual for Coyote Valley Dam calls for releases to avoid causing flows
greater than 8,000 cfs at the Hopland gage. Unregulated flows typically exceed this threshold
several times per year and cause no direct flood damage. For the sake of compliance with
reservoir operating criteria, the forecast-informed alternatives developed in the FIRO study were
formulated to avoid increasing the incidence of flows above 8,000 cfs at the Hopland gage. The
metrics were computed based solely on results from the deterministic simulation, since the
279
Chapter 16 – Results of Modeling PR-100

annual events generated in the FRA results did not address the frequency of multiple incidences
of such flows during a given year. Table 30 and Figure 226 summarize results from the
deterministic simulations, demonstrating that the FIRO alternatives can improve observance of
this regulation rule.

Table 30 Hopland Rule Compliance Comparison

Hours Percent of Time that
with Flow Greater Flows Exceeded
Alternative than 8,000 cfs 8,000 cfs
Existing Conditions 3,828 .729%
Encroach (perfect forecast) 3,552 .676%
Combined (perfect forecast) 3,438 .655%
EncroachWIF (imperfect forecast) 3,522 .671%

Figure 226 Comparison of Hopland Flows above 8,000 cfs from Deterministic Simulations

Flood Damage
EAD for the Russian River basin serves as the bottom line for the relative comparisons of flood
risk, since EAD integrates the direct economic consequences of the alternatives across the
watershed and throughout the range of hydrologic events. Other metrics such as changes in
flood frequency at key locations are important to identify possible transfers of risk between
locations, and trade-offs among different forms of flood risk.
280
PR-100 Chapter 16 – Results of Modeling

Table 31 shows damage results from the HEC-WAT simulations for the deterministic and FRA
simulations. The alternatives featuring FIRO-based plans were formulated during the initial
modeling phase with the goal of capturing as much water as possible without increasing flood
risk. The final simulation demonstrates that none of the alternatives with FIRO-based plans
increased the simulated flood damages by a significant amount.

Table 31 Flood Damage Comparison for Deterministic and FRA Simulations

Annual Average Expected Annual
Damage (60 years) Damage (5000 events)
Alternative millions of dollars millions of dollars
Existing Conditions $ 6.1 $ 7.6
Without Coyote Valley Dam $ 10.0 -
Encroach (perfect forecast) $ 6.1 $ 7.8
Combined (perfect forecast) $ 6.1 $ 7.7
EncroachWIF (imperfect forecast) $ 6.1 $ 7.8

Figure 227 displays the damage computed by the HEC-WAT deterministic simulations. Note
that despite the relatively small catchment above Coyote Valley Dam and the dam's distance
from primary damage centers, Lake Mendocino substantially reduces flood damage in the
Russian River watershed. The flood operations keep the damage in check by temporarily storing
water that would otherwise add to the flood peaks downstream.

Figure 227 Flood Damage from the Deterministic Simulations

Consequences for events shown in Figure 227, and the Annual Average Damages reported in
Table 31, provide helpful historical context for understanding the analysis results, but offer a
limited perspective on actual flood risk. As described in Section 11.5.1, the EAD from the FRA
simulations provide a more robust representation of flood risk.
281
Chapter 16 – Results of Modeling PR-100

Elevation-Frequency at Lake Mendocino

Flood risk considerations for Lake Mendocino primarily consisted of the potential for additional
loading on the dam and spillway (uncontrolled flow). This analysis evaluated FIRO-based plans
that captured the most of the additional reservoir storage without significant increases in flood
damage under a perfect forecast scenario. The plans were formulated with a tolerance for minor
increases in peak lake levels and spillway activation, which were generally assumed not to
represent a significant dam safety risk. The analysis does not consider flood damage to
structures around Lake Mendocino or the upper East Fork of the Russian River.
The FIRO alternatives all resulted in higher stage-frequency relationships for the Lake
Mendocino pool, as shown in Figure 228. Note that the Encroach alternative was not sensitive
to errors in the forecast, so the red and orange lines (Figure 228) reflect the Encroach alternative
without and with forecast error, which almost completely coincide.

Figure 228 Comparison of Lake Mendocino Annual Peak Frequency

The reasons for the differences among the other curves primarily depended on the timing of a
given event. Early in the flood season, the Existing Conditions alternative pool levels were often
substantially below the rule curve, providing incidental extra flood storage. This extra storage
was reflected in summer depletions and the frequent inability to fill the reservoir during the
previous spring. The greater ability of the FIRO alternatives to fill the reservoir resulted in
levels closer to the rule curve for early storms of the following flood season.

The FIRO-based plans also resulted in higher lake levels than the Existing Conditions alternative
for events late in the flood season. The encroachment or early fill operations, which were
utilized in the FIRO alternatives to fill the pool more reliably, resulted in higher lake levels when
heavy storms occurred in February or March.
282
PR-100 Chapter 16 – Results of Modeling

Table 32 summarizes key flood risk metrics for Lake Mendocino under each alternative. The
FIRO alternatives increased the 100-year lake level by less than a foot, and increased the chance
of spillway flow in any given year by about two percent. However, the additional pool elevation
and flow and duration over the spillway did not translate into additional flood risk in the
consequence areas below the dam, since it occurred too late to affect downstream peaks and the
amount of flow going over the spillway was relatively small. Additional analysis would be
needed to evaluate whether there were any dam safety concerns with increasing the hydraulic
loading on the dam and increasing the occurrence of spillway flow.

Table 32 Comparison of Lake Mendocino Frequencies at Key Elevations

Lake Mendocino
Event with 1%
Annual Chance of
Exceedance
WSEL WSEL Annual Exceedance Probability
NAVD NGVD Storage of Spillway Flow
Alternative (feet) (feet) (ac-ft) (NGVD 764.8 feet/NAVD 767.7 feet)
Existing Conditions 773.42 770.55 127,200 0.052
Encroach (perfect forecast) 775.06 772.19 130,400 0.088
Combined (perfect forecast) 774.35 771.48 129,000 0.108
EncroachWIF (imperfect forecast) 775.06 772.19 130,400 0.109

Spillway Activation
Table 33 displays additional summary statistics from the deterministic simulations. Substantial
spillway flow occurs for simulations of the 1964, 1986, 1995 (March), and 2006 events for the
FIRO alternatives. In addition, a storm in May 2005 caused the Encroach alternative to briefly
activate the spillway. The spillway crest elevation is 764.8 feet NGVD. The FRA maximums all
occurred for Event 39 from Lifecycle 13.

Table 33 Comparison of Spillway Flow Durations for Deterministic Simulations

Deterministic Simulations FRA Simulations
(60 years) (5000 events)
Total Total
Hours Max Max Max Hours Max Max Max
Spillway Outflow level level Spillway Outflow level level
Alternative Flow (cfs) (NAVD ft) (NGVD ft) Flow (cfs) (NAVD ft) (NGVD ft)
Existing Conditions 204 6,000 771.79 768.92 17,460 37,900 782.46 779.59
Encroach 324 6,400 773.85 770.98 40,944 38,200 782.57 779.70
Combined 180 6,010 771.74 768.87 29,784 38,000 782.52 779.65
EncroachWIF 324 6,420 774.00 771.13 41,088 38,200 782.57 779.70

The ranking of the events with the highest lake levels was similar but not identical to the ranking
for the longest durations of spillway flow. Events with very similar peak reservoir levels could
have significantly different durations above the spillway crest, depending on the amount of flood
storage available before an event and the shape of the inflow hydrographs. The alternatives
using the encroachment approach reached higher pool levels and experienced substantially
longer spillway flows.

283
Chapter 16 – Results of Modeling PR-100

Flood Frequency at Downstream Locations

Table 34 displays the key flood frequency statistics from the FRA simulation results at the
primary gages along the Russian River. The FIRO-based alternatives result in small increases in
the one-percent flow and stage that were not considered significant for this analysis. These
differences were due to advance releases or spillway flow adding to flood peaks downstream.
The alternatives did not increase the probability of reaching flood stage in any given year. No
flood stage is defined for the Russian River near Guerneville (Hacienda) gage, so Table 34
displays the frequency of exceeding flood stage at the nearby gage at Johnson's Beach.

Table 34 Comparison of Key Frequencies at the Hopland, Healdsburg and Guerneville Gages
Russian River near Hopland Gage - USGS 11462500/NWS HOPC1/RM 84.78
Event with 1% Annual Chance
of Exceedance Annual Exceedance
WSEL Gage Probability of Flood Stage
Discharge NAVD height (Gage 21.00/
Alternative (cfs) (feet) (feet) NAVD 521.46 feet)
Existing Conditions 55,800 531.77 31.31 0.610
Encroach (perfect forecast) 56,900 531.93 31.47 0.611
Combined (perfect forecast) 56,300 531.82 31.36 0.611
EncroachWIF (imperfect forecast) 56,900 531.93 31.47 0.611

Russian River near Healdsburg Gage - USGS 11464000/NWS HEAC1/RM 35.42

Event with 1% Annual Chance
of Exceedance Annual Exceedance
WSEL Gage Probability of Flood Stage
Discharge NAVD Height (Gage 23.00/
Alternative cfs (feet) (feet) NAVD 102.87 feet)
Existing Conditions 96,400 105.59 25.73 0.033
Encroach (perfect forecast) 97,700 105.77 25.91 0.034
Combined (perfect forecast) 96,800 105.63 25.77 0.033
EncroachWIF (imperfect forecast) 97,700 105.77 25.91 0.034

Russian River near Guerneville (Hacienda) Gage - USGS 11467000/NWS RIOC1/RM 21.29
Russian River at Johnson's Beach at Guerneville Gage - USGS 11467002/NWS GUEC1/RM 15.15
Event with 1% Annual Chance
of Exceedance Annual Exceedance
WSEL Gage Probability of Flood Stage
Discharge NAVD Height (Gage 32.00/
Alternative (cfs) (feet) (feet) NAVD 43.68 feet)
Existing Conditions 98,700 69.96 46.94 0.539
Encroach (perfect forecast) 99,700 70.04 47.02 0.540
Combined (perfect forecast) 99,400 70.03 47.01 0.539
EncroachWIF (imperfect forecast) 99,700 70.04 47.02 0.540

Results at other locations below Coyote Valley Dam are consistent with the results at the gage
locations, demonstrating that the forecast-informed operations did not transfer significant flood
risk between communities along the Russian River. Flow-frequency results for the alternatives
are plotted together for these gage locations in the Figures 229 through 231. The figures
demonstrate the similarity of results across the full range of frequencies.

284
PR-100 Chapter 16 – Results of Modeling

Cases for Further Analysis

Examination of results identified three principal mechanisms that caused differences in the flood
damage calculations between the Existing Conditions and FIRO based alternatives:

1) Previous releases still resident in channel can increase the peak during the next event.

2) Storage depleted below the guide curve provides incidental storage to help avoid spillway
flows during early season storms.

Figure 229 Comparison of Hopland Flow Frequency Curves

Higher reservoir levels at the start of events result in more spillway flow. The higher reservoir
levels affect the flood risk regarding late season events for FIRO-based alternatives because they
typically encroach into the flood storage.

These mechanisms operated on too small of a scale, or occurred too rarely, to affect the total
flood risk calculation for the alternatives evaluated in this study. However, the results do show
increases in flood risk for individual events, either from the deterministic or FRA simulations.
The variability in damage is increased under FIRO based alternatives, but the average damage is
similar to the Existing Conditions alternative. The following sections explain each case for use
in designing further analyses.

285
Chapter 16 – Results of Modeling PR-100

Figure 230 Comparison of Healdsburg Flow Frequency Curves

Figure 231 Comparison of Guerneville (Hacienda) Flow Frequency Curves

286
PR-100 Chapter 16 – Results of Modeling

January 2006 - Minor Consequences from

Different Back-Filling Releases
Water managers exercise discretion in managing use of the reservoir flood storage, balancing the
need to make room in the pool for the next storm against the concern of making releases that
contribute to flooding downstream. The Coyote Valley Dam flood operations were designed to
minimize flood stages downstream by limiting outflows when the Russian River begins to rise.
The 1959 water control manual stipulates: "In addition, care must be exercised that releases
from the reservoir do not unnecessarily increase the effective reduction of subsequent flood
peaks due to residual flow in damage reaches in the downstream portion of the river." The
maximum flow rules related to the West Fork Russian River and Hopland gage effectively
prevent large releases during the rising limb of the river hydrograph. In order to determine when
and how much outflow to release as the river recedes downstream ("back-filling" the river
flows), water managers use judgment based on experience and a variety of information,
including forecasts.

The simulations performed in this analysis generated no situations where back-filling releases
added significantly to flood damage because of residual flows in downstream reaches. The lack
of impact could seem encouraging for the development of forecast-informed operations. The
development of forecast-informed operations has the potential to require frequent difficult
decisions higher in the flood pool. A close examination of the 2006 flood event simulation for
the Combined alternative demonstrates, the scale of consequences observed in the results. The
outcome of the 2006 flood event, resulted in the largest damage difference observed between the
Existing Conditions alternative and any of the alternatives in the deterministic analysis. The
shaded area in Figure 232 represents the extra volume released from the Coyote Valley Dam
after the initial storm on 28 December, but before the main storm started on 30 December.
Outflows computed by the advance release rule of the Combined alternative were constrained by
the 4,000 cfs limit of Flood Control Schedule 2, since the lake level was above 746 feet NGVD.
The Existing Conditions simulation made smaller releases, according to the operation logic. The
main storm finished 31 December, just as the spillway began to flow in the simulation for the
Existing Conditions alternative. The spillway began flowing a day later in the simulation for the
Combined alternative, due to earlier advance releases. The spillway flows, were late and did not
affect the downstream river peaks.

However, the additional volume from releases made on 29 December, failed to pass entirely
throughout the entire Russian River system prior to the peak flow in the lower Russian River.
The additional volume under the Combined alternative slightly increased the flood peak of the
2006 event. The difference becomes most noticeable below the confluence with Laguna de
Santa Rosa (modeled in HEC-RAS as the storage area labeled Santa Rosa Creek). The upper
portion of Figure 233 compares the hydrographs from the alternatives at that location, with the
lower portion exaggerating the difference between them, which amounts to an extra 0.04 feet on
the peak stage.

The 0.04 feet difference contributed $299,000 to the $37,192,000 in damage modeled for the
2006 event for the Combined alternative. Most of the damage differential occurs at just a few
industrial structures, where the depth-damage relationship for structure content shows sensitivity.

287
Chapter 16 – Results of Modeling PR-100

Figure 232 Outflows for the Combined Alternative – 2006 Flood Event

Figure 233 Higher Stage Due to Prior Outflows Still Resident Downstream

288
PR-100 Chapter 16 – Results of Modeling

The difference was not significant for the purposes of this study. However, the results
demonstrate that previous releases can result in additional flood damage, even if small. This
observation could be important if future analysis require absolutely no increase in flood risk. In
addition, impacts may vary somewhat among different events due to differences in scale of the
flooding. The travel time for releases to get downstream, included the particular timing of
tributary hydrographs for each event, and the hydraulic effect of different amounts of storage in
the lower river. Subsequent analysis that modifies the limitations based on Hopland and West
Fork criteria should evaluate downstream impacts. Concerns regarding differences in
downstream flood peaks due to releases from Coyote Valley Dam could make the case for longer
lead times to evacuate water stored in Lake Mendocino. The downstream flood peaks under the
Encroach alternative for the 2006 event, were the same for the Existing Conditions alternative,
since the FIRO rule in the Encroach alternative takes effect after 15 January.

December 1964 – FIRO Alternatives can both

Improve and Hinder Flood Operations
The HEC-WAT deterministic results illustrate how FIRO alternatives can show relatively similar
risk of flood damage, but potentially increase or decrease dam safety risk. Figure 234 compares
simulations, Existing Conditions and two FIRO alternatives, which are for the largest event in the
project history. The Existing Conditions alternative (green line, Figure 234) was less successful
in filling Lake Mendocino during the spring of 1964 than the two FIRO alternatives, and entered
December lower than the other alternatives. This more depleted condition provided incidental
flood storage when the 1964 event occurred. The reservoir was almost able to absorb the event,
and spilled for a short period a couple of days after the peak inflow.

Figure 234 Comparison of Plans for 1964 Event

The FIRO alternative using a simple "Encroach" approach (red line) was most successful at
filling the pool during spring 1964, and entered December in the least depleted condition.
Lacking the incidental flood storage enjoyed by the Existing Conditions alternative, the
289
Chapter 16 – Results of Modeling PR-100

Encroach alternative began spillway flow just after the reservoir inflow peaked on 23 December.
This flow over the spillway slightly elevated the peak flow downstream at Hopland. The spill
under the Encroach simulation was delayed enough to make no practical difference in the peak
stage at Guerneville, although it slightly slowed the decline of the receding hydrograph. Lake
Mendocino reached 771.13 feet NGVD in the simulation, and spilled almost 6,000 cfs for three
days.

The FIRO alternatives using the more complex "Combined" approach (blue line, Figure 234)
entered December less depleted than the Existing Conditions alternatives, giving it less incidental
flood storage. However, the Combined alternative's advance release rule lowered the pool in
advance of the storm enough to absorb the event without spillway flow. The advance release
flow caused the river to rise more quickly as it was routed downstream, but made no practical
difference in the peak flow at Hopland or the stage at Guerneville. This scenario illustrates a
potential reduction in dam safety risk under one type of FIRO operation.

Incidental storage occurred during the early flood season for the Existing Conditions alternative,
and accounted for most of the events where greater flood damage happened during one of the
FIRO alternatives. This risk could be mitigated when formulating future FIRO alternatives by
drawing Lake Mendocino down below 737.5 feet NGVD in the fall, to levels more consistent
with the Existing Conditions alternative.

Event 3 from Lifecycle 69 – Spillway Flow Adds

to Downstream Peaks
The FRA simulations enabled modeling of many more possible events than the ones available in
the historical record. Using a wider range of flood events, provides a better representation of the
probability associated with flood events, as explained comparing AAD (Figure 187) estimated
from the observed record to EAD calculated using the FRA simulation. Examination of
individual FRA events may also reveal flood risks in the system that might not be obvious from
modeling flood events limited to the historical record. The FRA simulation exhibited a few
instances where encroachment into the flood pool at Lake Mendocino resulted in substantially
higher damages downstream (Figure 235). The event illustrates how the FIRO alternatives can
result in spillway flow that makes downstream flooding worse than under the Existing
Conditions alternative. Figure 235 also illustrates how the incidental flood storage built into the
baseline alternative affects the relative comparison of flood risk.

Event 3 from Lifecycle 69 applied a basin-average storm total of 20.6 inches to the 1964 shape
set, corresponding to an exceedance frequency of 0.49 percent. This represents a substantially
heavier rain than the 13.6-inch basin average of the historical 1964 event, and ranked
seventeenth out of the 5,000 FRA events based on precipitation magnitude. The spatial
distribution of the 1964 shape set applied the heaviest rainfall in the northern portion of the
basin, with totals for hyetograph locations above Hopland ranging from 25.5 to 27.5 inches.
The temporal distribution featured a pair of small initial storms that saturated the ground and
started the river rising. Over the next three days, a much larger pair of storms led to extensive
flooding.

290
PR-100 Chapter 16 – Results of Modeling

Figure 235 Event 3 from Lifecycle 69 Results for Combined Alternative

The event date was sampled as 14 November, and starting reservoir levels were sampled as a bit
above average for that time of year (32 percent exceedance). The Existing Conditions alternative
resulted in an initial pool of 729.2 feet NGVD, which is 8.3 feet lower than the guide curve of
737.5 feet NGVD for that date. The initial pool level under the Combined alternative was 743.0
feet NGVD, which was right on the guide curve. This difference reflects both the relaxed
autumn drawdown formulated in the Combined alternative, and the generally higher autumn
levels under FIRO alternatives due to greater success in re-filling the reservoir each spring.

The Combined alternative had little opportunity to make advance releases before the main storms
arrived. The event resulted in spillway flow under both alternatives, but the Combined
alternative resulted in a much earlier and higher spillway flow. The spillway flow occurred early
enough to coincide with flood peaks in the lower river. The peak flow at Hopland was 12,900
cfs higher for the Combined alternative, and the peak stage at Johnson's Beach was 0.52 feet
higher. The $197.6 million total damage for Event 3 from Lifecycle 69 under the Combined
alternative was $22.1 million greater than the Existing Conditions alternative.

The same event under the Encroach alternative produced very similar results, with a total
damage of $193.4 million. The damage under the Encroach alternative was less than the
Combined alternative damage because of a lower initial pool level of 737.5 feet NGVD,
291
Chapter 16 – Results of Modeling PR-100

reflecting the guide curve for the Encroach alternative. The Existing Conditions and the
Encroach alternatives share the same guide curve during November, and the $17.9 million
difference between the alternatives represents the impact of flood control storage (incidental)
arising from the difficulty in re-filling the reservoir under the Existing Conditions alternative.

Event 25 from Lifecycle 64 – Late Season

Vulnerability
The February flood simulated as Event 25 from Lifecycle 64 resulted in $133.3 million of flood
damage for the Existing Conditions alternative, ranking as the twenty-first most damaging event.
However, the damage was $9.8 million greater for the Combined alternative (Rank 16) and $34.5
million greater for the Encroach alternative (Rank 11). The differences arise from the late
season flood risk trade-offs designed into each alternative in order to re-fill Lake Mendocino
with greater reliability.

The event sampled 18.41 inches of basin average precipitation, ranking 48th out of the 5,000
events in the analysis. The precipitation was disaggregated spatially and temporally according to
the shape set that represented the December 2005 event. Figure 236 displays the event featured a
large storm followed by an even larger one three days later. The precipitation in the subbasins
ranged from 16.0 to 21.7 inches, with the higher amounts in the west. The rainfall peaks
occurred at approximately the same times across the watershed. Sampling of initial soil moisture
resulted in all subbasins starting close to saturated.

Sampling for the initial reservoir levels yielded relatively high starting values, corresponding to
4.8 percent exceedance in the distributions for each alternative. The Combined alternative
started with a lake level of 745.81 feet NGVD, which was above that alternative's rising guide
curve during the first days of the simulation. The Existing Conditions alternative started at
741.42 feet NGVD, which was about four feet above the guide curve. The Combined and
Existing Conditions alternatives both started the event operating the reservoir according to Flood
Control Schedule 1. For the Encroach alternative, the reservoir started at 761.80 feet NGVD,
which corresponds to the maximum allowable encroachment defined in the water management
plan. The Encroach alternative increased outflows on 6 February in response to the precipitation
forecast, but stopped releases the next day due to the downstream flow constraints with the
reservoir still highly encroached into the flood pool. The first storm pushed the lake level above
the spillway for the Encroach alternative, while the Combined and Existing Conditions
alternatives only reached Flood Control Schedule 2.

The reservoir was still above the spillway for the Encroach alternative when the main storm
arrived. The lake reached 774.14 feet NGVD and outflows exceeded 21,000 cfs, with the peak
reservoir releases adding significantly to the downstream flood peaks. The lake level also rose
above the spillway for the Combined and Existing Conditions alternatives, with peak elevations
and outflows of 773.57 feet NGVD/17,400 cfs and 772.03 feet NGVD/11,000 cfs respectively.
The spillway flows occurred later under the Combined and Existing Conditions alternatives,
resulting in less impact on downstream flood peaks than for the Encroach alternative. Peak stage
at Johnson's Beach for the Encroach alternative was 1.24 feet higher than the Existing
Conditions alternative. The corresponding difference for the Combined alternative was 0.43 feet.

292
PR-100 Chapter 16 – Results of Modeling

Figure 236 Comparison of Simulations for Event 25 from Lifecycle 64

The flood control operations of either FIRO alternative prior to and during the event made little
difference to the results. The consequences were driven by the positioning of the pool prior to
event. The Encroach alternative as formulated in this study typically resulted in greater damage
than the Combined alternative due to flood events in February through April, because the
Combined alternative captured storage earlier and more aggressively. This strategy provided
greater success than the Combined alternative regarding reservoir re-fill during years with little
precipitation in late winter, as evidenced by the left side of Figure 224. Capturing this additional
increment of reliability for re-fill results in a greater chance of "getting caught" with diminished
available flood storage. The slightly higher flood risk metrics of the Encroach alternative
quantify the trade-off.

293
Chapter 16 – Results of Modeling PR-100

The deterministic simulations in this study demonstrated that in February and March the lake
level under the Existing Conditions alternative consistently reached 761.8 feet NGVD in May.
The FRA simulations demonstrated that during February and March the trade-off seems most
acute regarding success of attaining the summer target level versus vulnerability to uncontrolled
spillway flow.

Other Consequences
This analysis was limited to flood damage to structures and contents because the modelers had
access to appropriate input data. HEC-FIA could readily compute flood damage for automobiles
and any other direct damage that can be associated with occupancy types or individual structures.
The hydraulic results from HEC-RAS could be used to estimate damage for roadways, bridges,
and other infrastructure. As described in Section 8.1, this analysis included a very simplified life
loss computation for the deterministic simulations, intended for model validation instead of
representing an accurate estimate. Parameters for flood warning systems and other relevant
inputs could be implemented in HEC-FIA to better assess whether proposed water management
changes posed any risk to life safety.

The analysis also neglected agricultural damage due to flooding. HEC-FIA offers methods for
computing agricultural flood damage, but the parameters, assumptions, and inputs are complex,
and beyond the scope of this analysis. However, agricultural flood damage could be a major
type of consequence along the Russian River, and potentially sensitive to differences in the
seasonality and duration of flooding. Figure 237 illustrates the inundated area around
Geyserville resulting from the December 1964 flood under the Existing Conditions alternative.
The depths were insufficient to damage many of the structures within the inundated area, but
damage to vineyards, orchards, and fields could be disruptive to the regional economy.

This analysis also did not consider indirect economic consequences of flooding, such as
diminished commerce while a bridge was out of service or changes in employment while
businesses are closed and cleaning up. HEC-FIA provides access to ECAM to generate rough
estimates of such shocks to the economy in terms of reductions in labor and capital. This
analysis directly considered environmental consequences only in terms of compliance with
requirements regarding ramping rates and minimum flows. The study did not explore
environmental consequences for different timing of Coyote Valley Dam releases, and left
environmental benefits of FIRO to be inferred from increased storage for the dry season. As
with agricultural impacts, the environmental consequences of flooding depend on more than just
the severity of inundation used to compute structure damage. Seasonality, duration, inter-annual
variability, and other considerations may need consideration. Environmental consequences of
alternative water management plans may require investigation of changes to the river flow
regime, such as variability of flows during spawning season or impacts to sediment
characteristics.

294
PR-100 Chapter 16 – Results of Modeling

Figure 237 Agricultural Inundation for 1964 Event – Existing Conditions Alternative

295
Chapter 16 – Results of Modeling PR-100

296
PR-100 Chapter 17 – Conclusions & Recommendations

Conclusions and Recommendations

The HEC-WAT demonstrated its ability as a tool to fully evaluate stored floodwater using a
systems-based approach, and analyze flood risk impacts including spillway flow activation, for
the full range of feasible flood events. This "bookend" analysis that considered both perfect and
imperfect forecasts shows that forecast-informed reservoir operations deserve serious
consideration for Coyote Valley Dam, and suggests a modeling framework suitable for a USACE
study process to evaluate future proposed alternative operations.

Summary of Observations
The process of formulating the example forecast-informed reservoir operations in this report, and
review of the results, provided insight regarding the system performance of Coyote Valley Dam
across a range of potential flood scenarios. Several observations were made regarding
mechanisms for how Coyote Valley Dam operations can affect flood risk on the Russian River.

1) Lake Mendocino significantly reduces flooding on the Russian River. Coyote Valley
Dam does not prevent flooding at Guerneville and other downstream locations, but it
substantially reduces the peak stages along the river as long as Lake Mendocino remains
able to absorb and delay flow from the East Fork of the Russian River.

2) Spillway flow can increase downstream flood peaks for large events. If flow over the
Coyote Valley Dam spillway starts early enough, the flow can add to flood peaks
downstream. The most significant differences in flood risk observed in this analysis
relate to flow over the uncontrolled spillway.

3) The current reservoir operations for Lake Mendocino include a significant amount
of incidental flood storage. Using Existing Conditions as a baseline for relative
comparison of flood risk among alternative water management plans comes with a
complication in the case of Lake Mendocino. Contemporary water management practices
typically deplete the lake far below the winter flood control pool, so that the reservoir
enjoys a substantial amount of "incidental" storage to absorb early season floods. The
proposed FIRO water management plans succeeded in filling the pool in spring much
higher than the Existing Conditions alternative. However, the higher spring levels also
lead to higher reservoir levels in the fall, diminishing the incidental flood storage, and
resulting in an apparent increase in flood risk compared to the Existing Conditions
alternative. Some might consider the Existing Conditions alternative in this study to be
flawed because incidental flood storage should not be considered as part of the current
official water management plan. Others might insist that the incidental storage exists in
reality, and not accounting for incidental storage represents a tangible increase in flood
risk regarding early season floods.

297
Chapter 17 – Conclusions & Recommendations PR-100

4) There is a trade-off between spillway flow and reservoir re-fill success. Preserving
enough storage space to absorb the next event without spilling is critical to managing
flood risk, while storing inflow from the final storms of the flood season is critical to
attaining a full pool for the dry season. As in actual operations, the FIRO alternatives
analyzed in this report did not know whether a certain event would be the final one of the
season, and employed different strategies to shift the emphasis between these two
objectives during February and March.

5) Reservoir releases prior to a major flood can contribute to flood damages at

downstream locations. A portion of water released from Coyote Valley Dam prior to a
storm may continue to be present when the event affects downstream locations, resulting
in higher flood peaks and damage. The impacts are minor because the releases are small,
due to the limited capacity of controlled releases through the gate, and the operating rules
constraining the releases.

6) There are dam safety implications due to more reservoir storage. The FIRO-based
alternatives formulated for this analysis were effective at capturing more water supply
storage without significantly increasing the EAD. However, the FIRO-based alternatives
also increased the hydraulic loading on Coyote Valley Dam, which caused more frequent
spillway flow, and generated higher spillway flow for longer durations. Subsequent
analyses should incorporate input from the dam safety community, preferably as part of
plan formulation criteria.

7) Monte Carlo only reveals certain events with potential significant damage in the water
management alternatives results, even if an alternative only makes a little difference to
the overall EAD. The specific causes of damage differences should be identified and
used in a re-evaluation of the water management alternatives.

The study results demonstrated that forecast-informed reservoir operations offer tremendous
potential for increasing the storage at Lake Mendocino, which meant more storage available for
water supply uses (29,000 acre-feet) while limiting increased flood risk. The alternative
regulation plans were formulated using a deterministic analysis spanning sixty-years to capture
the additional stored water without significantly increasing flood risk. Then the alternatives were
simulated using the FRA compute option of HEC-WAT to calculate more thorough flood risk
results.

Complexities cut both ways regarding the timing of hydrologic, reservoir regulation, and
hydraulic interactions on the downstream consequences. For example, sometimes releases due to
a heavily encroached pool added slightly to downstream flood peaks, while other times the
forecast information allowed a FIRO alternative to reduce downstream peaks by ramping down
outflows sooner than the Existing Conditions alternative.

FIRO operations at Lake Mendocino might not necessarily show sensitivity to forecast error,
depending on the type of reservoir operation, and the method of simulating forecast error. The
differences between the forecast and observed precipitation tested in this analysis failed to affect
the performance of the plan. However, this complex topic requires further careful analysis in
order to draw conclusions.

298
PR-100 Chapter 17 – Conclusions & Recommendations

Further details about the reservoir model development and formulation process regarding the
alternative water management plans are available from the Hydrologic Engineering Center
(point-of-contact: Mr. Matthew McPherson) upon request.

Suggestions for Future Analyses

The most logical next step for future analysis is to look beyond the "perfect forecast" assumption
used in this study and the fairly simple "imperfect forecasts". Development of realistic FIRO
alternatives should reflect the degree of skill in the forecast, and operations designed with less
faith in the forecast, might capture less storage for water supply uses. This study used a
relatively low three-inch precipitation threshold for attempting to get back to the guide curve.
The low threshold inherently incorporates variability in modeling and forecast uncertainty.
HEC-WAT allows plan formulators to understand performance improvements and trade-offs
across the system while developing alternative water management operations based on uncertain
forecast information. Some additional research may be required regarding characterizations of
forecast skill under both the deterministic and FRA simulation options. Modeling in this
analysis assumed uncompromised performance of the dam and appurtenances under all loading
scenarios. This assumption should be verified with the dam safety community in further
analyses.

Additional research regarding the seasonality of flood risk in the Russian River would be
informative in the development of FIRO alternatives. In particular, better characterizations of
the frequency and severity of events in February, March, and April would be helpful to analyze
the risk trade-off between failure to fill the reservoir and worsening downstream damage by
uncontrolled spillway flow. Software improvements to the hydrologic sampling function of
HEC-WAT, may allow for a more robust analysis of spatial and temporal variability. Longer
forecast periods should be considered, and could be especially useful for constructing rules to
mitigate risk in highly encroached situations. Similarly, the use of antecedent soil moisture
conditions may also be useful for the same purpose.

This analysis leveraged existing models and data for use in developing the PVA, however, the
existing models and data were not developed for this specific application. More detailed terrain
and bathymetry information could support more accurate hydraulic and consequence modeling,
especially regarding minor floods. The decreasing rate-of-change constraints (DROC)
sometimes result in a few inches of lower reservoir pool after the final storm of the year. When
the DROC slows the ramp-down of backfilling flows between events, it can also result in slightly
higher downstream peaks during the second storm. Further analyses might suggest different
forms of the constraint that are effective at avoiding fish stranding without affecting reservoir
performance. For instance, modelers could consider basing the rule on change in river stage at a
downstream location rather than on reservoir releases.

299
Chapter 17 – Conclusions & Recommendations PR-100

300
PR-100 Chapter 18 - References

References

References
DWR,1965. Bulletin No. 161 Flood! December 1964-Janurary 1965. Department of Water
Resources, State of California, Sacramento, California. January 1965.
https://archive.org/details/x5flooddecember196161calirich

FAO, 1998. Crop Evapotranpiration: Guidelines for computing crop water requirements.
Irrigation and Drainage Paper No. 56. Food and Agriculture Orgnatization (FAO) of the United
Nations. Allen, R. G., Pereira, L. S., Raes, D., & Smith, M. 1998.

HEC, 2000. HEC-HMS Technical Reference Manual. CPD-74B. United States Army Corps of
Engineers Hydrologic Engineering Center, Davis, California, USA. March 2000.

HEC, 2009. Hydrologic and Hydraulic Modeling of Coyote Valley Dam for With and Without
Failure of the Dam. PR-72 , United States Army Corps of Engineers Hydrologic Engineering
Center, Davis, California, USA. 2009.

HEC, 2012. Determination of a Hydrologic Index for the Russian River Watershed using HEC-
ResSim. PR-85, United States Army Corps of Engineers Hydrologic Engineering Center, Davis,
California. 2012.

IACWD, 1982. Guidelines for Determining Flood Flow Frequency, Bulletin 17B. U.S.
Geological Survey, Interagency Advisory Committee on Water Data (IACWD), Reston, VA.
1982.

Moriasi, 2007. Model evaluation guidelines for systematic quantification of accuracy in

watershed simulations. Moriasi, D. N., J. G. Arnold, M. W. Van Liew, R. L. Bingner, R. D.
Harmel, and T. L. Veith (2007). Trans. ASABE 50(3), 885-900.

NMFS, 2008. Biological Opinion for Water Supply, Flood Control Operations, and Channel
Maintenance conducted by USACE, SCWA, and MCRRFCWCID in the Russian River
Watershed. Santa Rosa, California. National Marine Fisheries Service (NMFS), 2008.

NOAA, 2018. Atmospheric River Portal. NOAA, Earth System Research Laboratory (ESRL),
Physical Sciences Division. http://www.esrl.noaa.gov/psd/atmrivers/

NOAA, 2018a. Russian River - Hopland (HOPC1). NOAA, California Nevada River Forecast
Center. http://www.cnrfc.noaa.gov/graphicalRVF.php?id=HOPC1

301
Chapter 18 - References PR-100

NOAA, 2018b. Russian River - Healdsburg (HEAC1). NOAA, California Nevada River
Forecast Center. http://www.cnrfc.noaa.gov/graphicalRVF.php?id=HEAC1

NRCS, 1986. Urban Hydrology for Small Watersheds. TR-55, U.S. Department of Agriculture,
Natural Resources Conservation Service, Conservation Engineering Division. June 1986.
https://www.nrcs.usda.gov/Internet/FSE_DOCUMENTS/stelprdb1044171.pdf

OSU, 2012. United States Average Annual Precipitation, 1981-2010 (800m; ASCII GRID).
Oregon State University, PRISM Climate Group, 10 July 2012. http://prism.oregonstate.edu

SCWA, 2015. Lake Mendocino Water Supply Reliability Evaluation Report, Term 17. Sonoma
County Water Agency, Santa Rosa, California. 30 April 2015.

US Census Bureau, 2010. Longitudinal Employer-Household Dynamics (LEHD) data, U.S.

Census Bureau, Center for Economic Studies, Washington, DC. https://lehd.ces.census.gov/data

USACE, 1954. Design Memorandum No. 2, Hydrology and Hydraulic Analysis, Russian River
Reservoir (Coyote Valley), California. United States Army Corps of Engineers, San Francisco
District. 1954.

USACE, 1959. Reservoir Regulation Manual for Coyote Valley Dam. United States Army
Corps of Engineers, San Francisco District. 1959.

USACE, 2003. Coyote Valley Dam and Lake Mendocino, Russian River, Californina, Water
Control Manual; Appendix I of the Master Water Control Manual for the Russian River. United
State Army Corps of Engineers, Sacramento District. Revised September 2003.

USACE, 2006. ER 1105-2-100, Planning – Planning Guidance Notebook, Department of the

Army, Washington, DC 20314-1000. April 2000 (revised January 2006).
http://www.publications.usace.army.mil/Portals/76/Publications/EngineerRegulations/ER_1105-
2-100.pdf

USACE, 2017. ER 1105-2-101, Planning – Risk analysis for Flood Damage Reduction Studies,
Department of the Army, Washington, DC 20314-1000. July 2017.
http://www.publications.usace.army.mil/Portals/76/Publications/EngineerRegulations/ER_1105-
2-101.pdf

USDA, 2015. SSURGO (Soil Survey Geographic) Database". U.S. Department of Agriculture,
Natural Resources Conservation Service. 3 November 2015.

USGS, 2005. Enhanced Historical Land-Use and Land-Cover Data Sets of the U.S. Geological
Survey. U.S. Department of the Interior, U.S. Geological Survey, Data Series 240. Price, C.V.,
Nakagaki, N., Hitt, K.J., & Clawges, R.M. 2 January 2005. http://water.usgs.gov/GIS/dsdl/ds240/

USGS, 2013. USGS National Elevation Dataset (NED). U.S. Department of the Interior, U.S.
Geological Survey. https://catalog.data.gov/dataset/usgs-national-elevation-dataset-ned
302
PR-100 Chapter 18 - References

USGS, 2014. Area- and Depth-Weighted Averages of Selected SSURGO Variables for the
Conterminous United States and District of Columbia. U.S. Department of the Interior, U.S.
Geological Survey, NSDI Note. Wiezorek, Michael E. 10 July 2014.
https://water.usgs.gov/GIS/metadata/usgswrd/XML/ds866_ssurgo_variables.xml

Weather Bureau, 1961). Rainfall Frequency atlas of the United States. Technical Paper Number
40. U.S. Department of Commerce, Weather Bureau, Washington D.C. May 1961

Background Information
ENTRIX, 2004. Russian River Biological Assessment. ENTRIX Inc., Walnut Creek, Califronia.
Sonoma County Water Agency, Santa Rosa, California. 2004

FEMA, 2008. Flood Insurance Study, Sonoma County and Incorporated Areas. Federal
Emergency Management Agency, 2 December 2008.

FERC, 2004. 106 FERC 61,065 - Order Amending License. U.S. Federal Regulatory
Commission, 28 January 2004.

FIRO, 2015. A Comprehensive Plan to Evaluate the Viability of Forecast Informed Reservoir
Operations (FIRO) for Lake Mendocino. FIRO Steering Committee, 2015.

Florsheim, 1995. Geomorphic and hydrologic conditions in the Russian River, California.
Historic trends and existing conditions. Florsheim, J., & Goodwin, P., Philip Williams and
Associantes Ltd., San Francisco, CA. California State Coastal Conservancy, Mendocino County
Water Agency, Circuit Rider Productions, Inc, revised 1995.

Ford, 2016. Review of Sonoma County Water Agency's MATLAB reservoir operation model for
Lake Mendocino and Coyote Valley Dam. David Ford Consulting Engineers, Sacramento,
Califronia. August 2016

Gasith, 1999. Streams in Mediterranean Climate Regions: Abiotic Influences and Biotic
Responses to Predictable Seasonal Events. Gasith, A., & Resh, V. H., Annual Review of
Ecology and Systematics. Vol. 30:51-81. November 1999.

Grandi, 2016. Mel Grandi, personal communication. City of Ukiah, Califronia. 28 June 2016.

Hamill, 2013. NOAA's Second-Generation Global Medium-Range Ensemble Reforecast Dataset.

Hamill, T. M., Bates, T. J., Whitaker, J. S., Murray, D. R., Fiorino, M., Galarneau, Jr, T. J.,
Lapentia, W., American Meteorological Society. Vol. 94, 1553-1565.1 October 2013.
http://dx.doi.org/10.1175/BAMS-D-12-00014.1.

Menne, 2012. Global Historical Climatology Network - Daily (GHCN-Daily) Version 3.22.
Menne, M.J., I. Durre, B. Korzeniewski, S. McNeal, K. Thomas, X. Yin, S. Anthony, R. Ray,
R.S. Vose, B.E.Gleason, and T.G. Houston. NOAA, National Climatic Data Center.
doi:10.7289/V5D21VHZ. 2012.
303
Chapter 18 - References PR-100

NCDC, 2001. Integrated Surface Global Hourly Data. NOAA, National Climatic Data Center.
https://catalog.data.gov/dataset/integrated-surface-global-hourly-data-f6180

NCDC, 2018. U.S. Hourly Precipitation Data (1951-2010). NOAA, National Climatic Data
Center. https://gis.ncdc.noaa.gov/map/viewer/

NMFS, 2016. Letter to U.S. Army Corps of Engineers, Re: Coyote Valley Dam Ramping Rates.
National Oceanic and Atmospheric Administration, National Marine Fisheries Service. 14 April
2016.

Opperman, 2005. Influence of Land Use on Fine Sediment in Salmonid Spawning Gravels within
the Russian River Basin, California. Opperman, J. J., Lohse, K. A., Brooks, C., Kelly, N. M., &
Merenlander, A. M., Canadian Journal of Fisheries and Aquatic Sciences. Vol. 62:2740-2751. 4
July 2005.

PG&E, 2006. Flow Regulation and Verification Plan, Addressing License Article 52 (A) as they
relate to RPA Measures 3 and 5, Potter Valley Hydroelectric Project. FERC Project No. 77.
Pacific Gas and Electric Company, California. 2006.

SCWA, 2016. Fish Habitat Flows and Water Rights Project Draft Environmental Impact
Report. Sonoma County Water Agency, Santa Rosa, California. 19 August 2016.

Sonoma County, 2017. Approved 2016 Sonoma County Hazard Mitigation Plan Update.
County of Sonoma, Santa Rosa, California. 4 May 2017. http://www.sonoma-
county.org/prmd/docs/hazard-mitigation-update/

SWRCB, 2013. In the Matter of Permits 12947A, 12949, 12950, and 16596 (Applications
12919A, 15736, 15737, 19351), Sonoma County Water Agency Order Approving Temporary
Urgency Changes. State of California , California Environmantal Protection Agency, State
Water Resources Control Board, Sacramento, CA. 25 April 2013.

USACE, 1982. Northern California Streams Investigations: Russian River Basin Study. U. S.
Army Corps of Engineers, San Franciso District, San Franciso, California. March 1982.

USACE, 1984. Warm Springs Dam and Lake Sonoma, Dry Creek, California, Water Control
Manual. U.S. Army Corps of Engineers, Sacramento District, Sacramento, California. 1984.

USACE, 2010a. HEC-ResSim, Reservoir System Simulation, (Version 3.1, RC3, Build 42).
U.S. Army Corps of Engineers, Institute for Water Resources, Hydrologic Engineering Center,
Davis, California. July 2010.

USACE, 2010b. HEC-HMS, Hydrologic Modeling System, Version 3.5. U.S. Army Corps of
Engineers, Institute for Water Resources, Hydrologic Engineering Center, Davis, California.
August 2010.

304
PR-100 Chapter 18 - References

USACE, 2010c. Corps Water Management System, Version 2.0. U.S. Army Corps of
Engineers, Institute for Water Resources, Hydrologic Engineering Center, Davis, California.
October 2010.

USACE, 2016. Technical Working Group Meeting. FIRO Prelminary Viability Assessment
Technical Working Group. Bond, Marchia, U.S. Army Corps of Engineers, Sacramento District,
Sacramento, California. 1 February 2016.

USGS, 1971. Turbidity and Suspended Sediment Transport in the Russian River Basin,
Califronia. Ritter, J. H., & Brown, W. H., USGS, Water Resourcdes Division, Menlo Park,
California. Open-File Report 72-316. Prepared in Cooperation with U.S. Army Corps of
Engineers. 1 October 1971.

Whitin, 2016. E-mail Communication from B. Whitin, CNRFC Hydrologist. NOAA - National
Weater Service, California Nevada River Forecast Center, Sacramento, California. 7 November
2016.

305
Chapter 18 - References PR-100

306
PR-100 Appendix A – Water Control Diagram

Appendix A
Water Control Diagram

A-1
Appendix A – Water Control Diagram PR-100

A-2