Coastal Engineering 53 (2006) 573 – 588

[Link]/locate/coastaleng

Empirical parameterization of setup, swash, and runup

Hilary F. Stockdon a,⁎, Rob A. Holman b , Peter A. Howd a , Asbury H. Sallenger Jr. a
a
Center for Coastal and Watershed Studies, U. S. Geological Survey, St. Petersburg, FL, United States
b
College of Oceanic and Atmospheric Sciences, Oregon State University, Corvallis, OR, United States

Received 27 May 2005; received in revised form 19 December 2005; accepted 21 December 2005
Available online 21 February 2006

Abstract

Using shoreline water-level time series collected during 10 dynamically diverse field experiments, an empirical parameterization for extreme
runup, defined by the 2% exceedence value, has been developed for use on natural beaches over a wide range of conditions. Runup, the height
of discrete water-level maxima, depends on two dynamically different processes; time-averaged wave setup and total swash excursion, each of
which is parameterized separately. Setup at the shoreline was best parameterized using a dimensional form of the more common Iribarren-based
setup expression that includes foreshore beach slope, offshore wave height, and deep-water wavelength. Significant swash can be decomposed
into the incident and infragravity frequency bands. Incident swash is also best parameterized using a dimensional form of the Iribarren-based
expression. Infragravity swash is best modeled dimensionally using offshore wave height and wavelength and shows no statistically significant
linear dependence on either foreshore or surf-zone slope. On infragravity-dominated dissipative beaches, the magnitudes of both setup and
swash, modeling both incident and infragravity frequency components together, are dependent only on offshore wave height and wavelength.
Statistics of predicted runup averaged over all sites indicate a −17 cm bias and an rms error of 38 cm: the mean observed runup elevation for
all experiments was 144 cm. On intermediate and reflective beaches with complex foreshore topography, the use of an alongshore-averaged
beach slope in practical applications of the runup parameterization may result in a relative runup error equal to 51% of the fractional variability
between the measured and the averaged slope.
Published by Elsevier B.V.

Keywords: Wave runup; Swash; Wave setup; Remote sensing

1. Introduction The goal of this work is to devise a simple parameterization

for maximum runup elevation, improving upon an earlier
When ocean waves approach a coast, the majority of wave empirical formula for wave runup by Holman (1986). These
energy is dissipated across the surf zone by wave breaking. elevations, in turn, can be used as input into a storm impact
However, a portion of that energy is converted to potential model (Sallenger, 2000).
energy in the form of runup on the foreshore of the beach In the discussion below, it will be apparent that wave height, H,
(Hunt, 1959). This wave runup is important to coastal deep-water wave length, L0, wave period, T, and beach
planners, nearshore oceanographers, and coastal engineers steepness, β, form a commonly accepted environmental para-
because these motions deliver much of the energy responsible meter set. Of these, deep-water wavelength and period are
for dune and beach erosion (Ruggiero et al., 2001; Sallenger, assumed to be interdependent, linked by the linear dispersion
2000). Thus, understanding the magnitude and longshore relationship,
variability of extreme runup is critical to accurate prediction
of the impacts on protective dunes and adjacent properties. gT 2
L0 ¼ : ð1Þ
2p

⁎ Corresponding author. Tel.: +1 727 803 8747x3074; fax: +1 727 803 2032. The three independent parameters may provide a first-order
E-mail address: hstockdon@[Link] (H.F. Stockdon). description of a beach environment and are often expressed in
0378-3839/$ - see front matter. Published by Elsevier B.V.
doi:10.1016/[Link].2005.12.005
574 H.F. Stockdon et al. / Coastal Engineering 53 (2006) 573–588

terms of a non-dimensional surf similarity parameter, commonly will investigate the degree to which simple parameterizations
called the Iribarren Number (Battjes, 1974), provide useful predictions of extreme runup on natural
beaches.
b
n¼ : ð2Þ
ðH=L0 Þ1=2 2. Background

The Iribarren number can be interpreted as a dynamic beach Runup, R(y, ti), is defined here as the set of discrete water-
steepness, comparing beach slope to the square root of deep- level elevation maxima (Fig. 1a), measured on the foreshore,
water wave steepness. with respect to still water level (that which would occur in the
In early laboratory experiments of monochromatic waves absence of waves). The values of runup depend on the long-
on planar beaches, all quantities were well defined. However, shore location, y, time, t, and the discrete times of maxima, ti.
application of this scaling to natural beaches introduces se- Runup results from two dynamically different processes: (1)
veral complications. For random waves, both wave period and maximum setup,bηN(y), the time-averaged water-level eleva-
wave height become statistical measures, often described by tion at the shoreline, and (2) swash, S(y, ti), the time-varying,
the peak period, Tp, and the root mean square (rms) or signi- vertical fluctuations about the temporal mean (Fig. 1a).
ficant wave height, Hrms or Hs, (defined as 2.8 and 4 times the Setup, the super-elevation of the mean water level, is driven
standard deviation of the time series of sea-surface elevation, by the cross-shore gradient in radiation stress that results from
respectively). Additionally, wave-height measures can be ex- wave breaking (Longuet-Higgins and Stewart, 1963, 1964). The
pressed in deep water (H0), at the break point (Hb), or locally. relationships between setup and environmental conditions, and
Definition of a single beach slope becomes difficult on natural resulting expressions for setup at the shoreline, have been the
beaches with typically concave profiles and is further com- topics of many research studies. Bowen et al. (1968) wrote a
plicated by the common presence of offshore sandbars. Esti- simplified expression for setup by assuming normally incident
mation of runup statistics under these complicated conditions shallow-water waves whose height within the surf zone was
may be handled by appropriate numerical models along with limited to a constant fraction of the local water depth, γ = H / h.
knowledge of boundary conditions (Raubenheimer and Guza, The resulting expression was
1996; Raubenheimer et al., 1995) such as the full incident bgN
wave spectra and bottom bathymetry. However, such an ap- ¼ 0:38g; ð3Þ
Hb
proach is incompatible with broad application to coastal zone
management problems because details of the offshore profile where Hb is the breaking wave height. Measured setup values
and incident spectra are not always available. Therefore, we were found to be greater than that predicted by theory due to the

Fig. 1. Water-level time series (a), extracted from timestack in Fig. 2, indicating individual runup maxima, R, setup at the shoreline, bηN, and swash excursion,
S. Significant swash statistics, S, were calculated from the spectra (b) of the water-level time series. The 2% exceedence value of runup, R2, was calculated from
the cumulative PDF of the discrete measures of R (c).
 H.F. Stockdon et al. / Coastal Engineering 53 (2006) 573–588 575

asymptotic approach of setup to the beach surface (Bowen et al., Thus, coefficients from data gathered at a single experiment (such
1968). Examining data from gently sloping beaches of southern as in Eq. (5)) may vary with sandbar configuration and profile
California, Guza and Thornton (1981) found bηN to be pro- shape, introducing significant amounts of noise into empirical
portional to the significant wave height (10-m depth), Hs, with a relationships. Baldock and Holmes (1999) found in numerical
constant of proportionality equal to 0.17. In contrast, analysis of simulations that incident band swash saturation was related to
data from a beach in Duck, NC (Holman and Sallenger, 1985) bore-driven swash, which also scales with wave period and beach
showed that a direct correlation between setup and wave height slope.
was highly scattered and that non-dimensional setup (normal- The relationship between ξ0 and the distribution of runup
ized by offshore wave height) scaled better with the Iribarren was examined on a wider range of Iribarren space by Nielsen
number, ξ0. The Guza and Thornton (1981) data were found to and Hanslow (1991). They found that the vertical scaling for
be consistent with the Holman and Sallenger (1985) relationship runup distributions was proportional to ξ0 for steep beaches,
at low Iribarren numbers. While the Holman and Sallenger further supporting the original formulation of Hunt (1959),
(1985) relationship between ξ0 and bηN significantly reduced Eq. (4), and the empirical formulation of Holman (1986), Eq. (5).
the scatter in the data, there remained a tidal dependence. The However, for beaches with β b 0.1,they suggest that the di-
relationship failed during the lowest tides when waves signi- mensional vertical scaling of runup distributions may be
ficantly dissipated over an offshore sandbar. Using observations independent of beach slope and proportional to (H0L0)1 / 2
of setup on beaches in Australia, Hanslow and Nielsen (1993) (Nielsen and Hanslow, 1991). The behavior of swash on a
found that the maximum setup on dissipative beaches showed highly dissipative beach (ξ0 b 0.25, β b 0.02) was studied by
no dependence on beach slope and, therefore, could not be Ruessink et al. (1998) who found that the swash signal was
scaled with ξ0. dominated by energy in the infragravity band (frequencies, f,
Swash, S, is generally defined as the time-varying location of b0.05 Hz) and scaled with H0. Ruggiero et al. (2001) also
the intersection between the ocean and the beach. According to studied runup under highly dissipative conditions and found
Miche (1951), monochromatic waves are composed of two parts: that the elevation of R2 scaled best with H0.
a progressive component whose energy is dissipated during wave Several studies have examined the relative roles of infra-
breaking and a standing, reflected component that has its maxi- gravity and incident (f N 0.05 Hz) band swash (SIG and Sinc, res-
mum at the shoreline. Swash represents this standing component pectively) for particular field sites. Guza and Thornton (1982)
whose amplitude cannot exceed some critical value that is showed that infragravity swash height increases linearly with
dependent on both β and T0. Early studies on swash and runup offshore significant wave height, while energy in the incident
were conducted in laboratories in order to determine the impacts band becomes saturated due to dissipation across the surf zone.
of waves on structures. Based on Miche's ideas, Hunt (1959) This linear dependence of SIG on H0 has been confirmed by
stated that there is a critical wave steepness below which wave several other studies (Holman and Bowen, 1984; Howd et al.,
energy will be totally reflected by a beach with a planar slope. 1991; Raubenheimer and Guza, 1996); however, the constants of
Under surging wave conditions, there is little dissipation of wave proportionality were found to vary between sites and with the
energy across the beach slope, and the majority of the energy is wave and beach conditions present during each experiment.
reflected on the steep beach slope (Hunt, 1959). For the more Howd et al. (1991) examined the relationship between infra-
common situation of breaking waves, energy is dissipated across gravity motions at the shoreline and offshore wave height at a
the surf zone. Using a critical value based on wave steepness and number of field sites and found their ratio to be dependent on ξ0:
β (Iribarren and Nogales, 1949), which defines a threshold larger constants of proportionality between infragravity motions
between breaking and non-breaking conditions, Hunt (1959) and wave heights were observed on beaches with larger ξ0 values.
proposed an empirical formulation for vertical wave up-rush, R, Ruessink et al. (1998) noted that the literature presents a wide
that can be re-written in terms of ξ, range of constants of proportionality for the ratio SIG / H0,
particularly between dissipative and reflective beaches, also
R suggesting that the ratio may depend on ξ0.
¼ Kn; ð4Þ
H Based on the previous studies described above, we propose
the following general relationship for the elevation of extreme
where Hunt assumes that H ≈ H0. (2%) runup, R2, for any data run:
Using field data collected from Duck, NC, Holman (1986)
found a clear relationship between 2% exceedence value of runup, S
R2, normalized by Hs (in 18-m depth) and ξ0, R2 ¼ bgN þ ; where ð6Þ
2

R2
¼ 0:83n0 þ 0:2; ð5Þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Hs S¼ ðSinc Þ2 þ ðSIG Þ2 ; and
bgN; Sinc ; SIG ¼ f ðH0 ; T0 ; bf Þ ð7Þ
which includes a linear dependence on the foreshore beach slope,
βf. For experiments on natural beaches, wave breaking over where T0 is the deep-water wave period. The specific goal of
offshore sandbars can significantly attenuate offshore waves. this work is to improve the predictive equation for runup on
576 H.F. Stockdon et al. / Coastal Engineering 53 (2006) 573–588

natural beaches by extending Holman's (1986) original analysis

to data sets from 10 experiments representing a wide variety of
beach and wave conditions and by separately parameterizing the
individual runup processes: setup and swash.
In the next section, we describe our methods including
details of the runup and swash statistics, the environmental
parameters measures at each site, and the specific field expe-
riments. Next, we present the results of our analysis, showing
the parameterization of setup, incident swash, and infragravity
swash, and evaluate the performance of the empirical para-
meterizations. The consequences of a longshore variable-topo-
graphy on the parameterizations are also examined. In the
Discussion section, the use of breaking wave height and surf-
zone slope in the model, as alternatives to H0 and βf, is eval-
uated. Finally, improved parameterizations for setup and swash
under dissipative conditions are presented.

3. Methods

3.1. Runup measurement technique and statistics

All runup data in this study were collected using video

techniques that were developed at the Coastal Imaging Lab at
Oregon State University and previously tested extensively against
in situ runup instruments (Holland et al., 1995; Holman and Guza,
1984). Holman and Sallenger (1985) discussed difficulties with
digitizing faint downwash, particularly on very low-sloping
beaches. While variations in digitization can introduce noise in
swash and setup statistics, Holman and Sallenger showed that
they did so in a way that cancelled when runup peak elevations
were found: slight lows in swash height were balanced by highs in
set-up and visa versa. Other errors associated with video runup
measurements are discussed in detail by Holman and Guza (1984)
and Holland et al. (1995). The video pixel resolution, dependent
on the field of view of the camera, the height of the camera, and
the distance to the observed ground location (Holland et al.,
1997), was typically 5–15 cm in the vertical with corresponding
horizontal resolutions of 20–80 cm.
Cross-shore transects of pixel intensity (example from Duck,
NC, Fig. 2a) were sampled at 1 or 2 Hz over 17- to 120-min Fig. 2. Camera view from Duck, NC (a) and runup timestack (b). The cross-
record lengths, depending on the site. This created timestacks of shore transect in (a) indicates the location of a single runup array. In the 120-s
pixel intensity on which runup and rundown can be seen as a timestack of pixel intensity (b), each horizontal slice is the spatial variability of
intensity at a single time step. The leading edge of swash is digitized through
white edge moving back and forth in the swash zone (Fig. 2b).
time (heavy, white line) and then converted into a time series of water-level
The leading edge of runup was digitized from cross-shore elevations.
timestacks of pixel intensity and then, using published photo-
grammetric relationships (Holland et al., 1997), was converted spectra, PSD(f), of the continuous water-level time series
to time series of water level elevation measured relative to mean (Fig. 1b). The significant swash height, S, was calculated as,
sea level. For each experiment, 17-min records were extracted
qX
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
from the longer time series to minimize the effects of changing
S ¼ 4⁎ PSDðf Þdf ; ð8Þ
tide levels on the location of wave breaking and on the area of
the foreshore over which swash propagates. In order to calculate
wave setup, measured tidal curves were removed from each similar to the calculation of significant wave height (Guza and
time series. All statistics presented are representative of eleva- Thornton, 1980). The significant swash in the incident (dis-
tions measured relative to the still-water level. tinction based on wave frequency, f N 0.05 Hz, and not on
Both continuous and discrete statistics were calculated from direction of propagation) and infragravity (f b 0.05 Hz) bands
raw time series (Fig. 1a). After subtraction of the setup, bηN, the were calculated by summing only over frequencies within the
17-min time average, swash statistics were calculated from the specified limits.
 H.F. Stockdon et al. / Coastal Engineering 53 (2006) 573–588 577

Runup statistics, R, were defined as the elevation of Table 1

individual water-level maxima above the still-water level Average environmental conditions for each experiment
P P P P
(Fig. 1a), merging contributions from setup and swash (Eq. (6)). Site Date Hs ðmÞ T0 ðsÞ bf n0 Fr N (Nsingle)
The 2% exceedence value for runup, R2, was calculated from Duck, NC 5–25 Oct 1.7 12.0 0.11 1.4 ± 0.5 149 (36)
the cumulative probability density function of runup elevations (Duck82) 1982
(Fig. 1c). This statistic, which represents the value of R that will Scripps Beach, CA 26–29 Jun 0.7 10.0 0.04 0.6 ± 0.1 42 (41)
1989
be exceeded 2% of the time, is often used in engineering
Duck, NC 6–19 Oct 1.5 8.9 0.10 0.9 ± 0.4 1829 (138)
applications (Holman, 1986) and may be important for use in (Delilah) 1990
scaling the impacts of severe storms on beaches. San Onofre, CA 16–20 Oct 0.8 14.9 0.10 2.2 ± 0.3 59 (59)
1993
3.2. Environmental parameters Gleneden, OR 26–28 Feb 2.1 12.4 0.08 0.9 ± 0.2 52 (42)
1994
Terschelling, NL 2–22 Apr 2.4 8.0 0.02 0.1 ± 0.04 41 (6)
To make comparisons between sites where wave heights 1994
were measured in water depths varying between 7 and 20 m, an Terschelling, NL 1–21 Oct 1.4 8.1 0.01 0.1 ± 0.05 27 (8)
effective deep-water significant wave height, H0, was calculat- 1994
ed. To estimate H0 during each water-level time series, mea- Duck, NC 3–21 Oct 1.5 12.1 0.08 1.1 ± 0.3 975 (52)
(Duck94) 1994
surements of significant wave height from local buoys and
Agate Beach, OR 11–17 Feb 2.5 13.2 0.02 0.2 ± 0.05 126 (14)
instrument arrays, Hs, were reverse shoaled to deep water using 1996
linear wave theory assuming a shore-normal approach. While Duck, NC 3–30 Oct 1.3 9.5 0.10 1.2 ± 0.7 491 (95)
this procedure neglects local generation, friction, white-cap- (SandyDuck) 1997
ping, refraction, and diffraction, it allows for inter-comparisons ξ0 ± σ — mean Iribarren number ± the standard deviation.
of deep-water equivalent wave heights between different field N — number of individual estimates of runup statistics, includes multiple
sites. Breaking wave height, Hb, was calculated by shoaling Hs longshore lines.
Nsingle — number of individual estimates of runup statistics used in bulk
over local bathymetry using Thornton and Guza's (1983) wave
parameterization.
transformation model, with the limiting ratio γ = Hrms / h equal to
0.42 (Thornton and Guza, 1982) and the bore dissipation coef-
ficient, B, equal to 1.0. While B was not tuned due to lack of during these experiments represent a full range of ξ0 with
appropriate field observations at each site, model error has been Terschelling and Agate Beach representing the dissipative end
shown to be less than 10% when B is within 25% of its optimal (ξ0 b 0.3) and San Onofre representing the most reflective
conditions (n0 ¼ 2:2). Mean beach profiles, averaged over the
P
value (Thornton and Guza, 1983). Hb was defined as the wave
height occurring at the onset of wave breaking, identified as the duration of each experiment, illustrate the differences in beach
seaward-most location where dissipation exceeded 20 W/m2. slope and offshore bathymetry (Fig. 3).
Alternatively, the cross-shore location of breaking and Hb may The bulk of the data (91%) was collected at the U.S. Army
be defined by the percent of waves breaking. Corps of Engineers Field Research Facility (FRF) in Duck,
For each water-level time series, βf was defined as the ave- NC, and represents intermediate to reflective conditions (as
rage slope over a region ± 2σ around bηN, where σ is the defined by Wright and Short (1983), 0.3 b ξ0 b 4.0). The typical
standard deviation of the continuous water level record, η(t). profile here is characterized by an offshore sandbar and a mean
Surf-zone slope, βsz, is also considered here and was defined for foreshore slope of 0.1. Hs and T0 were recorded hourly at a
each timestack as the slope between the shoreline (the cross- Waverider buoy located in ∼18 m of water. Tide level was
shore position of bηN), xsl, and the cross-shore location of wave measured every 6 min at a NOAA tide gauge mounted at the
breaking, xb. On barred profiles, the magnitude of βsz may end of the FRF pier. Bathymetry was sampled roughly daily
change significantly over the course of a day as the break point during the Duck experiments in 1990, 1994, and 1997 and
shifts from the sandbar at low tide toward the shoreline at high every few days during the 1982 experiment using the Coastal
tide. These cases will be of special interest when trying to Research Amphibious Buggy (Birkemeier and Mason, 1984).
isolate the relative importance of foreshore and surf-zone slope. Additional high-resolution foreshore topographic surveys were
usually collected daily during all experiments. During Duck82
3.3. Field experiments (1982), runup was digitized from 35-min time-lapse photog-
raphy along selected longshore locations (Holman, 1986)
Swash and runup data from 10 field experiments were following methods outlined in Holman and Guza (1984),
compiled for this study: Duck, North Carolina, USA (1982, resulting in 149 estimates of runup and swash statistics. The
1990, 1994, 1997); Scripps Beach, California, USA (1989); San runup array during the Delilah experiment (1990) consisted of
Onofre, California, USA (1993); Terschelling, the Netherlands 26 longshore lines with a longshore spacing, dy, of 10 m,
(1994, 1994); Gleneden, Oregon, USA (1994); and Agate sampled at 2 Hz (Holland and Holman, 1993). During the
Beach, Oregon, USA (1996). Table 1 provides details of expe- Duck94 experiment (1994), 120-min runup timestacks were
riment dates, mean wave and beach conditions (calculated over collected at up to 48 equally spaced (dy = 5 m) cross-shore
the durations of each experiment), as well as the number of lines (Holland and Holman, 1996). The runup array from
runup measurements collected. The beach and wave conditions SandyDuck (1997) consisted of five cross-shore transects from
578 H.F. Stockdon et al. / Coastal Engineering 53 (2006) 573–588

obtained from an experiment at Gleneden, OR, in late February

1994. A linear offshore profile (β = 0.025) was extrapolated
from a survey that extended from the beach across the surf zone.
Incident wave conditions were measured at a CDIP buoy
located approximately 190 km to the southeast in 11 m of water.
Tides measured at Yaquina Bay, OR, located 35 km to the south,
were corrected so that the values were representative of the tides
at Gleneden. 120-min video runup timestacks were collected at
2 Hz multiple times a day along a single cross-shore transect.
Data from two beaches dominated by dissipative conditions
(ξ0 b 0.3)are also included in the data set, providing a small (5.3%
of the total data) but important extension into highly dissipative
domains. Two data sets were collected from Terschelling, the
Netherlands, in April and October 1994 by Ruessink et al. (1998).
Terschelling features multiple offshore sandbars with offshore
slopes generally 0.005 or less, as measured from a single offshore
bathymetry profile. Daily beach surveys were conducted and an
alongshore-averaged foreshore slope was calculated for the ana-
lysis of runup statistics at seven cross-shore transects (Ruessink et
al., 1998). Hs, as measured by an offshore buoy located approx-
imately 5 km offshore in 15 m water depth, ranged from 0.5 m to
almost 4 m. 45-min runup timestacks were collected at 2 Hz
along nine cross-shore lines once a day. Because the only beach
slope available for this site was an alongshore-averaged fore-
shore slope, the runup and swash statistics computed at each
cross-shore line were alongshore averaged before being com-
Fig. 3. Time-averaged profile (topography and bathymetry) from each pared in bulk to the other sites. (Runup statistics were fairly
experiment illustrating differences in foreshore slope and offshore profiles: longshore uniform during each data run. Over the duration of the
SandyDuck (a), Duck94 (b), Delilah (c), Duck82 (d), San Onofre (e), Gleneden experiment, the mean and standard deviation, μ ± σ, of long-
(f), Scripps Beach (g), Agate Beach (h), and Terschelling (i). The profiles are
offset 2.5 m in the vertical. Mean sea level is indicated by the horizontal dotted
shore-variable swash and setup were 69 ± 8 and 16 ± 5 cm,
lines. The vertical dashed line marks the cross-shore location of MSL for each respectively.) The second dissipative site is Agate Beach, OR, a
profile (x = 0). The average profile was calculated over the time period of the multiple-barred, low-sloping beach where runup data were col-
experiment at a single cross-shore transect. lected in February 1996 as a part of a larger experiment on the
nearshore dynamics of high-energy beaches (Ruggiero et al.,
which approximately 500 17-min swash timestacks were 2001). Foreshore slopes were measured daily using GPS topo-
collected. graphic surveys. Offshore survey data were not available for this
More limited data exist from three other intermediate/re- experiment; however, a typical offshore slope was measured
flective sites. The Uswash experiment was conducted June 26–29, from surveys collected in 1998 to be ∼0.01. Wave height and
1989 at Scripps Beach, CA, a fine-grained sandy beach period were obtained from a CDIP buoy located approximately
(Holland et al., 1995). Offshore bathymetry was not measured 170 km to the southeast in 64 m water depth. Measured tides
on a daily basis and was approximated offshore of the 1-m depth were taken from a NOAA gauge located at Yaquina Bay, OR,
contour using an offshore slope of 0.01 (Holland et al., 1995). approximately 8 km from Agate Beach. 120-min runup time-
Wave height and period were measured every 6 h in 7 m water stacks were collected at 1 Hz at five cross-shore transects.
depth, a few hundred meters offshore, as a part of the Coastal
Data Information Program (CDIP) (Seymour et al., 1985). Six 4. Results
8-Hz data runs were collected over the course of the experiment
(Holland et al., 1995). A more reflective data set was collected For the initial analysis of bulk runup statistics, the complications
during a October 1993 experiment in San Onofre, CA. The of longshore variability were avoided by selecting a single cross-
offshore slope was measured to be 0.013 from a cross-shore shore transect from each experiment. Limiting the data to one
profile, surveyed early in the experiment (October 10, 1993), runup line per experiment also helped to minimize the bias towards
that extended approximately 100 m offshore. Wave data were intermediate conditions since the most extensive longshore co-
obtained from a CDIP buoy located offshore of Oceanside, CA, verage occurred during the Duck experiments. This resulted in 491
in approximately 10 m of water, a few kilometers southeast of independent measures of runup and swash, distributed as follows:
the field site. Tide measurements were obtained from a NOAA Duck82 (36), Uswash (41), Delilah (138), San Onofre (59),
gauge located in La Jolla, CA. 120-min timestacks of runup Terschelling (14), Gleneden (42), Duck94 (52), Agate Beach (14),
were collected at 2 Hz at a single cross-shore transect over a and SandyDuck (95). The statistics of longshore variability from
5-day period. The final data set on an intermediate beach was multiple-transect runs are examined later in the paper.
 H.F. Stockdon et al. / Coastal Engineering 53 (2006) 573–588 579

The assumption that the 2% exceedence level for runup been common in the literature. In a regression between non-
approximately equals bηN + S / 2 (Eq. (6)) was first tested. S is dimensional R2 (normalized by H0) and ξ0, it is found that small
calculated as four times the square root of the swash variance errors in data from small wave cases will map to large errors in
(4σ) and, for a Gaussian process, would encompass 95.4% of ξ0 and can introduce large errors in regression statistics. Thus, in
the values. The remaining 4.6% defines the extreme high and order to avoid undue influence of these errors and to better
low values and is split evenly between the tails of the distri- weigh storm events, the dimensional parameterization is pre-
bution. Accordingly, the actual value of the extreme low value ferable to its non-dimensional counterpart. Regressions through
of runup (the 98% exceedence value) would be defined as the the origin are used, rather than allowing least squares intercepts,
mean (setup) minus half of the spread (2σ, or S / 2), and the in order to avoid non-physical results in setup and swash mo-
extreme high value of runup (the 2% exceedence value) would dels. For example, a significant intercept in a relationship bet-
be defined as the mean (setup) plus half of the spread. The ween wave height normalized swash and ξ0 would results in a
squared-correlation between the measured values of bηN + S / 2 value of swash when H0 = 0. In most models presented here, the
and R2 is 0.94, which is significant at the 99% confidence level computed intercept was not significantly different than zero,
(Fig. 4). The slope of the regression is 1.1, which reflects the supporting the selection of regression through the origin me-
slightly non-Gaussian nature of natural swash. To test whether thodology. To evaluate each empirical model, the squared-cor-
the observed swash was Gaussian, the skewness and kurtosis of relation, ρ2, referred to hereafter simply as correlation, and the
each swash distribution were calculated for all experiments, 95% significance level, ρsig2
, are presented as a measure of the
except Duck82, and compared to values defining Gaussian linear relationship between the two parameters. The goodness of
distributions (skewness = 0, kurtosis = 3). The mean kurtosis for fit of the empirical model is measured using an rms error, rmse.
all data was 2.9 (σ = 0.80). The mean skewness for all swash
distributions was 0.19 (σ = 0.38) indicating that natural swash is 4.1. Setup
slightly skewed. To account for this skewness, the slope of the
regression in Fig. 4 is included in Eq. (6), resulting in a more No single constant of proportionality between bηN and H0 was
complete definition of runup, found (ρ2 = 0.30, rmse = 25.3 cm; Fig. 5a). The setup parameter-
  ization was improved by inclusion of L0 (T0) and βf (Fig. 5b),
S
R2 ¼ 1:1d bgN þ : ð9Þ modeling setup using a dimensional, Iribarren-like form (ξ0H0),
2
bgN ¼ 0:35bf ðH0 L0 Þ1=2 : ð10Þ
The empirical models described in the following sections
will be presented in dimensional space and based on regressions The squared-correlation of the dimensional model is ρ2 = 0.48
2
forced through the origin. Synthetic tests based on samples with (ρsig = 0.01) and rmse= 21.3 cm. A summary of the regression
known statistical characteristics indicate a danger in carrying coefficients, squared-correlations, and rms error for all suggested
out the least-squares analysis in non-dimensional space, as has parameterizations is presented in Table 2.

Fig. 4. The sum of setup and half of the swash excursion plotted against the 2% runup peak elevation. The dashed line is a 1 : 1 line. The heavy solid line is the best fit to
the data (m = 1.1, b = 0.10, ρ2 = 0.94).
580 H.F. Stockdon et al. / Coastal Engineering 53 (2006) 573–588

Fig. 5. (a) Setup vs. wave height (ρ2 = 0.30) and (b) setup vs. βf(H0L0)1 / 2 (ρ2 = 0.48). The βf(H0L0)1 / 2 dimensional parameterization also has the lowest rms error,
21.3 cm.

Setup data were divided into low-, mid-, and high-tide over the βf-dependent parameterization (Eq. (10)) that was
groupings with divisions defined as 1/3 (low to mid) and 2/3 derived from the data set as a whole.
(mid to high) of the tidal range observed at each site. As was
observed in the Duck82 data by Holman and Sallenger 4.2. Incident and infragravity swash
(1985), the goodness of fit of the parameterization for bηN
varied during different stages of the tide (Fig. 6). When low- On dynamically different beaches, energy in the incident and
tide values of bηN were examined separately from those infragravity frequency bands will contribute varying amounts to
occurring during mid- and high tide, the correlation between total swash (Guza and Thornton, 1982). Swash heights within
bηN and βf(H0L0)1 / 2 fell to ρ2 = 0.29 (Fig. 6a). For the mid- these two bands are forced by different processes and therefore
and high-tide values, the correlation remained high, ρ2 = 0.52 require separate parameterizations. Incident swash is best para-
(Fig. 6b). It was suggested by Holman and Sallenger (1985) meterized by a dimensional version of an Iribarren-like rela-
that offshore morphology and surf-zone slope play a larger tionship (Fig. 7a),
role in bηN at low tide, with breaking patterns over barred
topographies producing more complex patterns in radiation Sinc ¼ 0:75bðH0 L0 Þ1=2 : ð11Þ
stress gradients (Raubenheimer et al., 2001). However, the
use of surf-zone slope in any form of a setup parameterization The correlation for the model is ρ2 = 0.44 (ρsig 2
= 0.01) and
for the low-tide data did not improve the model performance rmse = 46.9 cm, lower than that for the non-dimensional version.
When the same dimensional model is used to describe the
infragravity band (Fig. 7b), the correlation remains high, ρ2 =
Table 2
Regression parameters for components of runup model 0.56 (rmse = 34.2 cm). However, when local foreshore slope is
removed from the expression (Fig. 7c), the correlation im-
Quantity Model Slope, m Intercept, ρ2 (ρ2sig) rmse
modeled input ba (cm)
proves to ρ2 = 0.65 (ρsig
2
= 0.01) and the rms error decreases to
25.7 cm. The magnitude of infragravity swash is, therefore,
All βf(H0L0)1/ 2 0.35 (± 0.01) 0 0.48 (0.01) 21.3
linearly independent of βf and best parameterized as
bηN
sites Sinc βf(H0L0)1/ 2 0.75 (± 0.03) 0 0.44 (0.01) 46.9
SIG (H0L0)1 / 2 0.06 (± 0.002) 0 0.65 (0.01) 25.7
SIG ¼ 0:06ðH0 L0 Þ1=2 : ð12Þ
ξ0 b 0.3 bηN (H0L0)1 / 2 0.016 (± 0.003) 0 0.68 (0.22) 11.9
S (H0L0)1 / 2 0.046 (± 0.004) 0 0.78 (0.22) 15.7 The absence of beach slope in this model may be
a
Regressions forced through origin to avoid non-physical intercepts. somewhat counter-intuitive given that the presence of
 H.F. Stockdon et al. / Coastal Engineering 53 (2006) 573–588 581

Fig. 6. Setup vs. βf(H0L0)1 / 2 during low tide (a) and mid- and high tide (b). The correlation of the model during low tide (ρ2 = 0.29) is significantly lower than that for
mid- and high-tide values (ρ2 = 0.52).

significant infragravity band energy is prescribed by the β- experiment and for all data as a measure of the bias of the
dependent ξ 0 . However, the inclusion of β f in the estimator. The rms error of the differences, Δrms, was used
parameterization of the actual magnitudes of SIG has not to measure the scatter or noise of the estimate. Mean values
for observed setup, swash, and runup (bgN, S¯ , and R̄
P
been supported by our observations. The non-dimensional
form of Eq. (12) shows that the efficiency of infragravity respectively) are presented in order to evaluate the
energy generation is inversely dependent on deep-water wave significance (or relative magnitudes) of the mean difference
steepness, H0 / L0, where low values are indicative of swell errors. Summaries of the error statistics are presented in
conditions (for a given value of H0). Infragravity band Table 3.
motions have previously been shown to correlate best with The mean error between setup estimates and observations,
averaged over all experiments, DbgN , is − 3 cm, while the rms
P
energy from swells rather than that from seas (Elgar et al.,
error, ΔbηNrms, is 21 cm (bgN ¼ 49 cm). There is little bias in
P
1992).
setup estimates on intermediate and reflective beaches
(DbgN ¼ 2 cm; bgN ¼ 57 cm); however, on dissipative bea-
P P
4.3. Evaluation of swash and runup parameterization
ches (ξ0 b 0.3), the bulk parameterization underestimates setup
elevation by 16 cm (bgN ¼ 27 cm). Accordingly, the largest
P
Estimated setup, bηNe, swash, Se, and runup peaks, R2e, were
calculated from measured offshore wave height, wave period, and underestimates of setup occur for large wave events (H0 N 1.5 m,
DbgN ¼ −7 cm) and on the most gentle beach slopes (βf ≤ 0.02,
P
foreshore beach slope using the empirical parameterizations for
P
setup, incident swash, and infragravity swash (Eqs. (10)–(12)) in DbgN ¼ −17 cm).
The mean difference in swash, DS , averaged over all data, is
P
Eq. (9). Performance of the parameterizations was measured in
terms of dimensional differences, Δ, from observations, 34 cm and the rms difference, ΔSrms, is 46 cm (S¯ = 149 cm).
The bulk parameterization for runup contains an − 18 cm bias
DbgNðtÞ ¼ bgNe ðtÞ−bgNðtÞ; ð13Þ (R̄ = 144 cm). The rms difference error between runup esti-
mates and observations, (ΔR)rms, is 38 cm. The bias in the
DSðtÞ ¼ Se ðtÞ−SðtÞ; and ð14Þ
runup measurements is two times larger on intermediate and
reflective beaches (DR ¼ −18 cm) than on dissipative sites
P
DRðtÞ ¼ R2e ðtÞ−R2 ðtÞ ð15Þ
(DR ¼ −9 cm); however the observed runup magnitudes are
P

where bηN(t), S(t), and R2(t) are the observed values at each also larger on intermediate and reflective sites (R̄ = 148 cm)
time, t. The mean difference errors, Δ̄, were calculated for each than dissipative ones (R̄ = 84 cm).
582 H.F. Stockdon et al. / Coastal Engineering 53 (2006) 573–588

Fig. 7. (a) Incident (ρ2 = 0.44, rmse = 46.9 cm) and (b) infragravity swash (ρ2 = 0.56, rmse = 34.2 cm) parameterized in a dimensional form of the traditional Iribarren-
based equation. (c) The parameterization of infragravity swash improves when beach slope is removed from the model (ρ2 = 0.65, rmse = 25.7 cm).

4.4. Longshore variability since SIG showed little or no linear dependence on βf, spatial
variations in the infragravity band swash are expected to be less
Runup and incident swash have been shown to be dependent than those in the incident band. To explore the dependencies on
on foreshore beach slope. Accordingly, beaches with significant beach slope, the longshore variability of swash and runup was
longshore-variable slopes should exhibit similar longshore examined using longshore runup arrays from the Delilah (26
variability in swash excursions and runup elevation. However, lines; total array length, Y, =250 m), Duck94 (35 lines; Y = 165 m),
and SandyDuck (6 lines; Y = 75 m) experiments. Spatial var-
Table 3 iability observed at each site was characterized by calculating
Accuracy of setup, swash, and runup parameterizations, given in (cm) the longshore standard deviation of the total swash excursions,
σS(t), as well as that of the incident, σSinc(t), and infragravity,
P P P
Experiment DbgN ΔbηNrms DS ΔSrms DR ΔRrms
Duck82 3 34 48 58 9 37 σSIG(t), components. Additionally, spatial squared-correlations,
Scripps Beach −5 9 18 19 2 10 ρs2, were calculated between S(y), Sinc(y), SIG(y) and βf (y) in
Delilah −6 13 42 57 − 32 41
San Onofre 11 16 17 22 −4 17 Table 4
Gleneden −2 9 20 27 − 17 27 Longshore variability of swash and correlations to foreshore beach slope
Terschelling a −6 14 25 31 −1 13
Experiment σS σSinc σSIG %ρ2s (S, βf) %ρ2s (Sinc, βf) %ρ2s (SIG, βf)
Duck94 − 33 40 54 67 − 62 69
(m) (m) (m) significant significant significant
Agate Beach − 25 28 34 39 − 16 27
(%ρs N 0) (%ρs N 0) (%ρs N 0)
SandyDuck 11 20 31 37 −2 34
Average error, all data −3 21 34 46 − 18 38 Delilah 0.24 0.23 0.15 45.8 (97.0) 50.0 (100) 13.9 (80.0)
Duck94 0.20 0.24 0.19 31.0 (77.8) 55.2 (100) 38.0 (0)
bgN ¼ 49 cm
P
S̄ = 149 cm R̄ = 144 cm SandyDuck 0.18 0.18 0.14 55.6 (100) 58.6 (100) 14.8 (100)
Mean observed values of setup, swash, and runup are presented in the last row in The first three columns indicate the spatial standard deviation of swash. The last
order to examine mean error magnitudes relative to observed values. three columns list the percentage of spatial correlations significant at the 95%
a
The results of the two Terschelling field campaigns are combined in these confidence level. The value in parenthesis indicates the percentage of significant
statistics. correlations that are positive.
 H.F. Stockdon et al. / Coastal Engineering 53 (2006) 573–588 583

order to determine the amount of longshore variability in swash in runup elevation, δR2, was defined as (R2avg − R2e(y)) / R2e(y),
that may be attributed to a longshore-variable foreshore slope. where R2avg is the estimate of runup calculated from a single
Table 4 presents the details of longshore variability (the longshore-averaged beach slope and R2e(y) is the runup
standard deviation of S, Sinc, and SIG) and the correlation calculated using βf(y) measured at each longshore line. The
analysis (the percentage of significant correlations and of error in a runup estimate calculated from a single longshore-
positive significant correlations) for each of the three experi- averaged beach slope was found to equal 51% of δβ. For
example, a 20% difference between bf and βf(y) would result in
P
ments considered.
At Delilah and SandyDuck, the total swash excursion was a 10.2% error in estimated runup. Maximum alongshore
most longshore variable on days when the beach topography variability of βf was observed during the Delilah experiment
was highly three dimensional, either in the form of a regular when megacusps were present (longshore spacing of ∼200 m,
cusp field, a large megacusp embayment, or welded swash bars δβ = − 0.25 to 0.75). Here, runup predictions based on a
(σS = 40–50 cm). The values of ρs2(S, βf) showed a positive longshore-averaged beach slope may be underestimated by
12% on foreshore locations that are more steeply sloped than bf
P
correlation between total swash and beach slope, indicating that
increases in S corresponded to increases in βf, consistent with and overestimated by as much as 38% on more gently sloped
expectations. Correlations were significant at the 95% confi- regions. On a well-developed cusp field, longshore variability in
dence level 46% (Delilah) and 56% (SandyDuck) of the time foreshore slope (δβ = − 0.3 to 0.5) may result in runup values
(Table 4). Correlation values were higher and more often sig- 15% lower than the true value on steep cusp horns and
nificant for rhythmic, large-scale spatial variations in slope (i.e., overestimates of 26% within the more gently sloped cusp
megacusps) than for irregular, short-scale slope variations. The embayments. Again, swash circulation within a cusp field may
time-averaged value of σSinc was higher than σSIG, indicating be affecting the observed runup amplitudes, partly explaining
that most of the observed longshore variability in total swash why a change in beach slope leads to a 51% change in runup,
was contained within the incident band. Additionally, Sinc was even though the parameterizations show a linear relationship
significantly and positively spatially correlated to βf for 50% between the two.
and 59% of the cases at Delilah and SandyDuck, respectively
(Table 4). Less variability was observed in SIG and correlations 5. Discussion
with βf were significant only 15% of the time (Table 4),
supporting the results that the magnitude of infragravity swash The proper choice of wave height in the parameterization of
has little or no linear dependence on foreshore beach slope. total swash, S, was explored as a part of this study. The most
During Duck94, when a well-developed cusp field was pre- rigorous comparison between Hs (a locally measured significant
sent, there was less of a relationship between the magnitude of wave height), H0 (deep-water equivalent wave height), and Hb
the longshore variability observed in total swash and that ob- (breaking wave height) could be completed for the Duck
served in beach slope. Spatial correlations between S and βf were experiments because accurate, daily bathymetry data were avail-
significant only 31% of the time. Correlations between Sinc and able for shoaling the waves and determining an approximate
βf were significant and positive 55% of the time. When the cusp break point. At this site, the dimensional Iribarren-based para-
field was particularly well-developed, longshore variability meterization for S calculated using Hb (ρ2 = 0.46, rmse = 57.9 cm)
within the infragravity band increased. Significant correlations showed no improvement over the similar swash parameterization
between SIG and βf were more than twice as common during calculated using H0 (ρ2 = 0.46, rmse = 56.4 cm). On the dissipative
Duck94 than during the other experiments (Table 4). Surpris- beaches, where daily offshore bathymetry data were not available,
ingly, significant values of ρs2(SIG, βf) were always negative, an approximation of Hb was made by shoaling Hs over the single
indicating that Sinc and SIG were out of phase within the cusp offshore profile available for each site. On these beaches, Hb was
field. This relationship between SIG, Sinc, and beach slope may found to improve the performance of the regression: the
be related to longshore-variable dynamics due to swash correlation was higher, ρ2 = 0.80 (ρsig2
= 0.22) than that for the
circulation within a cusp field, as opposed to a simple cross- swash parameterization based on H0 (ρ2 = 0.67). However, be-
shore flow which has been assumed for the empirical parame- cause daily profiles were not measured at these sites and the
terizations presented. An exploration of the longshore variability values of Hb are approximate, the exact relationship between Hb
of setup, swash, and runup and the complex dynamics of inci- and S cannot be clearly defined here.
dent and infragravity swash, particularly on highly three-di- While Hb was not shown to offer a significant improvement
mensional topography, will be presented in detail in a over H0 (at Duck), the question of which measure of wave height
subsequent paper. is most appropriate for practical use in runup parameterizations
The dependence of total swash on βf will have implications on remains a relevant one. Because of refraction, frictional
the practical use of the bulk parameterization of runup on bea- dissipation across the shelf (Herbers et al., 2000), and white-
ches with complex foreshore topography. In order to evaluate capping, the wave height measured at a buoy located in deep
how much error can be expected in estimates of R2 if a water may be significantly higher than that which actually
longshore-averaged slope, bf , is used instead of a more accurate
P
reaches the nearshore, so runup predictions using deep-water
measure of slope at each longshore line, βf(y), a relative slope buoy measurements may be anomalously high. Analysis at
difference was compared to a relative runup error. Relative slope Duck, where wave height measurements are available in deep
difference, δβ, was defined as ðbf −bf ðyÞÞ=bf ðyÞ. Relative error
P
water, 18 m water depth, and 8 m water depth, shows a somewhat
584 H.F. Stockdon et al. / Coastal Engineering 53 (2006) 573–588

stronger relationship between runup and waves measured at 8 m. conditions where breaking occurs on the bar at low tide (gentle
While H0 was used in this analysis as an equal measure of wave βsz) and near the shoreline at high tide (steep βsz). Data runs
height between different sites, in practical applications, it may be where these conditions were met were isolated from the larger
preferable to use local wave measurement, reverse shoaled to an data set. This unique situation allows for a changing surf-zone
equivalent deep-water value. slope while the input wave conditions remain relatively
Some researchers have suggested that the slope of the surf constant. On most topographies observed, the magnitude of
zone (βsz) might be more directly related to swash height than runup and swash did not change dramatically over the change in
βf; therefore, its use in the empirical parameterization would tide or related changes in xb. Both R2 and S correlated well with
likely improve runup estimates (Holman and Sallenger, 1985; βf over the entire tidal cycle. (See Section for 4.4 for detailed
Nielsen and Hanslow, 1991). To test this idea, βsz was correlations between S and βf.) These results again indicate that
substituted in the dimensional, Iribarren-based parameteriza- βsz offers no significant improvement over βf for incident-band-
tion for total S. When the entire data set was used, the dominated sites. (Note: Variability in the magnitude of Sinc was
correlation of the model (ρ2 = 0.03, ρsig 2
= 0.01) decreased observed over a few tidal cycles at Duck94 when beach cusps
significantly from the similar model calculated using βf were present on the foreshore, perhaps related to the complex
(ρ2 = 0.68, ρsig
2
= 0.01). The data from Delilah, Duck94, and interactions between incident and infragravity band swash on
SandyDuck were examined separately because the detailed, this rhythmic topography.)
daily bathymetric surveys provided more accurate estimates While the goal of this work was to present a parameterization
of βsz. Here, the correlation for the model for S using βsz was for extreme runup that is useful and accurate on a broad spec-
not significantly different than zero showing that, on these trum of beaches, it is important to address the errors that occur
beaches, the foreshore slope has more of an influence on during dissipative conditions (ξ0 b 0.3) when R2 is estimated
swash processes than the surf-zone slope. When dissipative using the bulk parameterization (Eqs. (10)–(12)). Under these
beaches were isolated from the entire data set and examined extreme end-member conditions, increased dissipation likely
separately, the use of estimated βsz showed similar predictive becomes a significant term in momentum balances and the
capabilities (ρ2 = 0.71, rmse = 20.1 cm) as calculations based on parameterization from steeper conditions no longer works.
βf (ρ2 = 0.67, rmse = 21.3 cm). However, the correlation of the When the dissipative beaches are isolated from the whole data
model for swash on dissipative beaches is highest when set, the correlation of the parameterization of setup improves
slope is completely removed from the parameterization (ρ2 = 0.67). Interestingly, when βf is removed from the para-
(Sα(H0L0)1 / 2, ρ2 = 0.78). In order to explore the effects of meterization of setup for dissipative conditions, there is no
βsz on runup more directly, the variations in S were studied as decrease in the correlation of the model. This suggests that the
the tide rose and fell over a barred profile at Duck. Since βsz inclusion of βf in this parameterization is not necessary, sup-
is defined between the shoreline and the break point (xb), a porting earlier work which found that shoreline setup on dis-
significant change in βsz is observed under certain wave sipative beaches was not linearly dependent on beach slope

Fig. 8. Parameterization of (a) setup (ρ2 = 0.68, rmse = 11.9 cm) and (b) swash (ρ2 = 0.78, rmse = 15.7 cm) during dissipative conditions only (ξ0 b 0.3) using (H0L0)1 / 2.
For both models, the correlations are equal or higher than when the bulk parameterizations are used for the dissipative conditions subset.
 H.F. Stockdon et al. / Coastal Engineering 53 (2006) 573–588 585

(Hanslow and Nielsen, 1993). The suggested parameterization incident band on dissipative beaches is also revealed by looking
for setup on dissipative sites, bηNd (Fig. 8a), is at the ratio of Sinc to SIG, υ, plotted against ξ0 (Fig. 10). Cutoff
values between dissipative, intermediate, and reflective beaches
bgNd ¼ 0:016ðH0 L0 Þ1=2 ð16Þ
are defined using an Iribarren equivalent of the Wright and Short
(ρ2 = 0.68, ρsig
2
= 0.22, rmse = 11.9 cm). On dissipative beaches, (1983) surf-similarity values. Dissipative beaches (ξ0 b 0.3) are
frictional dissipation of large waves over extremely wide, low- dominated by infragravity energy (υ b 1) for 90% of the data. The
sloping surf zones begins to play a larger role in shoreline energy on intermediate beaches is closely split between incident-
processes. For a given H0 (and T0), the values of bηN on a and infragravity-dominated conditions (47% incident energy).
dissipative beach will be lower than on a reflective or intermediate On the reflective beaches (ξ0 N 1.25), the swash is dominated by
beach. (This is seen in Fig. 5a where the values of bηN for incident energy (υ N 1) for 90% of the data. This shows that
Terschelling and Agate Beach fall below the below the data shoreline motions on reflective beaches are dominated by energy
cluster and the best-fit line.) Earlier work on setup has shown that in the incident band while shoreline motions on dissipative
bηN decreases for lower values of γ (Bowen et al., 1968) and that beaches are dominated by energy in the infragravity band, again
γ values are lower on dissipative sites (Bowen et al., 1968) and, supporting what has been observed and explained using data
particularly, on lower sloping beaches (Sallenger and Holman, collected from single sites (Guza and Thornton, 1982; Holman
1985). and Sallenger, 1985).
The behavior of swash under extremely dissipative condi- Because total swash on dissipative beaches, Sd, is composed
tions is also different than during reflective and intermediate mostly of energy within the infragravity band, it is best para-
conditions. On low-sloping, high-energy beaches, energy in the meterized using a form similar to the bulk model describing
incident band is saturated and increases in H0 contribute only to swash in the infragravity band (Eq. (12)). Modeling both the
increases in the infragravity band (Guza and Thornton, 1982; incident and infragravity bands together,
Holman and Sallenger, 1985; Ruessink et al., 1998). This was
clearly observed in the swash data when dissipative beaches Sd ¼ 0:046ðH0 L0 Þ1=2 ð17Þ
(ξ0 b 0.3) were separated from the intermediate and reflective
beaches, and Sinc and SIG for the two subsets of data were plotted (Fig. 8b). The correlation for this dissipative-specific parame-
against (H0L0)1 / 2 (Fig. 9). On intermediate and reflective terization (ρ2 = 0.78, ρsig
2
= 0.22) is significantly higher than that
beaches, swash in both the incident (Fig. 9c) and infragravity for the parameterization of S on dissipative beaches which
(Fig. 9d) bands increases as H0 (and T0) increase. On the includes βf (ρ2 = 0.67). Additionally, the rms error is reduced
dissipative beaches, the magnitude of the infragravity swash from 21.3 to 15.7 cm when βf is removed from the expression.
grows with increasing H0 (Fig. 9b) while the incident band Using a small subset of data from infragravity-dominated Agate
swash is completely saturated (Fig. 9a). The saturation of the Beach, Ruggiero et al. (2004) found a slope dependence in

Fig. 9. Incident and infragravity swash plotted against (H0L0)1 / 2 for dissipative (a, b) and intermediate/reflective (c, d) beaches. On dissipative beaches, the incident
band (a) is saturated while the magnitude of the infragravity band (b) continues to grow with increasing (H0L0)1 / 2. On intermediate and reflective beaches, both
frequency bands respond to increases in (H0L0)1 / 2.
586 H.F. Stockdon et al. / Coastal Engineering 53 (2006) 573–588

Fig. 10. Ratio of incident to infragravity swash variance (ν) plotted against the Iribarren number. The vertical dashed lines mark the cutoff values between dissipative
(ξ0 b 0.3),intermediate, and reflective beaches (ξ0 N 1.25). Values above the horizontal line at log(ν) = 1 are incident dominated while those below the line are
infragravity dominated.

spatially variable swash. However, when all swash data, inclu- were also significantly reduced. The bias decreased from 29 to
ding the subset used by Ruggiero et al., are examined together, 13 cm while the ΔSrms decreased from 34 to 16 cm. The
no significant linear slope dependence exists. It has been evaluation of the estimators across all sites, using the dissi-
suggested by other researchers (Ruessink et al., 1998; Ruggiero pative-specific parameterizations when ξ0 b 0.3, is also pre-
et al., 2001) that swash on highly dissipative beaches should be sented in Table 5.
scaled using wave height alone; however, the correlation for this The final, general expression for runup on all beaches,
model is only 0.37 and the rms error is larger (29.5 cm). The based on the entire data set and substituting Eqs. (10)–(12)
model using (H0L0)1 / 2 has a significantly higher correlation into Eq. (9), is
and lower rms error, suggesting that the inclusion of wave !
period allows for improved predictive capabilities. Based on 1=2 ½H0 L0 ð0:563b2f þ 0:004Þ1=2
R2 ¼ 1:1 0:35bf ðH0 L0 Þ þ
these dissipative-specific parameterizations and substituting 2
into Eq. (9), runup on sites where ξ0 b 0.3 may be calculated
ð19Þ
as
and may be used over the full range of beach conditions.
Given the dissipative-specific formulations, it may seem lo-
R2 ¼ 0:043ðH0 L0 Þ1=2 : ð18Þ
gical for intermediate- and reflective-specific parameteriza-
tions as well. However, when conditions where ξ0 N 0.3 are
The improved performance of this model is given in Table 5. considered separately, the coefficients of the setup and
The setup bias for dissipative sites was reduced from −16 to 3 incident swash parameterization change less than 0.5% and
cm. The bias and noise for swash for individual dissipative sites the coefficient of the infragravity parameterization changes
∼2.8%. On reflective beaches (ξ0 N 1.25), where swash is
Table 5 dominated by incident energy, the complete expression for
Accuracy of setup, swash and runup parameterizations for dissipative sites (cm)
P P P
runup 2% exceedence elevations (Eq. (19)), can be simplified
Experiment DbgN ΔbηNrms DS ΔSrms DR ΔRrms by assuming that the infragravity contribution (the 0.004
a
Terschelling 8 14 10 12 3 15 term) to total runup is negligible. Here, incident swash and
Agate Beach −1 9 16 17 −9 23 setup have the same parametric dependencies and can be
Average error, ξ0 b 0.3 3 12 13 16 −5 21 combined,
Average error, all sites −2 21 33 46 − 17 38

R2 ¼ 0:73bf ðH0 L0 Þ1=2 : ð20Þ

P
bgN (cm) S̄ (cm) R̄ (cm)
While this simplified form is more convenient for practical
ξ0 b 0.3 27 85 84
All sites 49 149 144 applications, the rms error under reflective conditions is 47 cm,
larger than that for the full expression (Eq. (19), rmse = 32 cm).
Mean observed values of setup, swash, and runup are presented in the last two
rows in order to examine mean error magnitudes relative to observed values. Therefore, our final recommendation is the broad use of the full
a
The results of the two Terschelling field campaigns are combined in these expression (Eq. (19)), with an exception (Eq. (18)) for ex-
statistics. tremely dissipative conditions.
 H.F. Stockdon et al. / Coastal Engineering 53 (2006) 573–588 587

6. Conclusions underestimate the elevation of runup peaks. The rms difference

between estimated and measured runup was 38 cm.
The elevation of extreme runup peaks, given by the 2% The longshore variability of runup was examined during the
exceedence value, R2, is dependent on the sum of two dyna- Delilah, Duck94, and SandyDuck experiments where runup
mically different processes, the time-mean setup, bηN, and data were collected over extensive longshore arrays. On days
swash, defined in terms of the significant swash height, S, and when foreshore slope was longshore variable, runup, in par-
computed as four times the square root of the swash variance. ticular incident band swash, was also spatially variable.
Extreme runup is defined as the sum of setup and half of the Differences between longshore observed runup and runup
total swash excursion (Eqs. (6) and (9)). Empirical formulations predictions made using a single longshore-averaged foreshore
for each of the components have been developed using carefully slope may be as much as 38% when the foreshore topography is
defined water-level, wave, and topography statistics from 10 highly three-dimensional (for example, within a megacusp
field experiments spanning a wide range of environmental field). Longshore variability in foreshore slope may result in a
conditions. This data set represents a major expansion on the relative runup error equal to 51% of the fractional variability
range of conditions for which empirical relationships have been between the measured and averaged slope.
tested.
Dimensional setup is best parameterized (ρ2 = 0.48, rmse = Acknowledgments
21.3 cm) using foreshore beach slope, estimated over the region of
significant swash activity (bηN ± S / 2), and offshore wave height The authors thank Todd Holland, Peter Ruggiero, and
and wavelength (Eq. (10)), the dimensionally equivalent form of Gerben Ruessink for sharing their swash time series and for
an Iribarren number dependency. The significant swash excursion providing expert knowledge on the details of their experi-
can be decomposed into incident (fq0 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
N 0.05 Hz) and infragravity ments. We also thank John Stanley, Cindy Paden, Joe Haxel,
(f0 b 0.05 Hz) frequency bands, S ¼ ðSinc Þ2 þ ðSIG Þ2 , each of which Logan Mitchell, David Sedivy, Nick Parazoo, and Dan Clark
is be modeled separately. Dimensional incident swash scales with for their assistance with runup digitization. Wave data for the
foreshore beach slope, offshore wave height, and offshore wave- different experiments were provided by the U.S. Army Corps
length (Eq. (11), ρ2 = 0.44, rmse= 46.9 cm). Dimensional infra- of Engineers' Field Research Facility, the NOAA National
gravity swash also scaled well withβ(H0L0)1 / 2; however, when Buoy Data Center, the Coastal Data Information Program, and
foreshore slope was removed from the equation, the correlation of Britt Raubenheimer. We thank Bob Guza for his thoughtful
the model improved (Eq. (12), ρ2 = 0.65, rmse = 25.7 cm). Addi- review of the manuscript. This research was funded by the
tionally, the use of the surf-zone slope (defined as the average USGS National Assessment Program.
slope from the break point to the mean swash location) in the
parameterization offered no significant improvements, even on References
days when wave breaking was occurring on the sandbar.
The above relationships for setup and swash show large biases Baldock, T.E., Holmes, P., 1999. Simulation and prediction of swash oscillations
under the extreme dissipative conditions of two of the field sites, on a steep beach. Coastal Engineering 36 (3), 219–242.
Battjes, J.A., 1974. Surf Similarity, Proceedings of the 14th Conference of
perhaps reflecting the increasing role of bottom friction on very Coastal Engineering. ASCE, pp. 466–480.
wide surf zones in the dynamic balances. For Iribarren numbers Birkemeier, W.A., Mason, C., 1984. The CRAB: a unique nearshore surveying
less than 0.3, setup was best parameterized using only offshore vehicle. Journal of Surveying Engineering 110, 1–7.
wave conditions (Eq. (16), ρ2 = 0.68, rmse = 11.9 cm). Similarly, Bowen, A.J., Inman, D.L., Simmons, V.P., 1968. Wave ‘set-down’ and ‘set-up’.
Journal of Geophysical Research 73 (8), 2569–2577.
the total swash, merging both frequency bands, was best parame-
Elgar, S., Herbers, T.H.C., Okihiro, M., Oltman-Shay, J., Guza, R.T., 1992.
terized using only offshore wave height and wavelength (Eq. (17), Observations of infragravity waves. Journal of Geophysical Research 97
ρ2 = 0.78, rmse = 15.7 cm). (C10), 15573–15577.
Substituting the suggested forms of setup and swash, the Guza, R.T., Thornton, E.B., 1980. Local and shoaled comparisons of sea surface
final parameterization for the 2% exceedence value of runup elevations, pressures and velocities. Journal of Geophysical Research 85
peaks on all natural beaches is (C3), 1524–1530.
Guza, R.T., Thornton, E.B., 1981. Wave set-up on a natural beach. Journal of
! Geophysical Research 86 (C5), 4133–4137.
1=2 ½H0 L0 ð0:563b2f þ 0:004Þ1=2
R2 ¼ 1:1 0:35bf ðH0 L0 Þ þ : Guza, R.T., Thornton, E.B., 1982. Swash oscillations on a natural beach. Journal
2 of Geophysical Research 87 (C1), 483–491.
Hanslow, D., Nielsen, P., 1993. Shoreline set-up on natural beaches. Journal of
Under extremely dissipative conditions, estimates of R2 may be Coastal Research SI (15), 1–10.
Herbers, T.H.C., Hendrickson, E.J., O'Reily, W.C., 2000. Propagation of swell
improved using the dissipative-specific parameterization
across a wide continental shelf. Journal of Geophysical Research 105 (C8),
19729–19737.
R2 ¼ 0:043ðH0 L0 Þ1=2 for n0 [Link] Holland, K.T., Holman, R.A., 1993. Statistical distribution of swash maxima on
natural beaches. Journal of Geophysical Research 98 (C6), 10271–10278.
The performance of the runup parameterizations was tested Holland, K.T., Holman, R.A., 1996. Field observations of beach cusps and
swash motions. Marine Geology 134, 77–93.
at each site using data collected along a single transect. The Holland, K.T., Raubenheimer, B., Guza, R.T., Holman, R.A., 1995. Runup
mean difference between the estimated and measured runup was kinematics on a natural beach. Journal of Geophysical Research 100 (C3),
− 17 cm, indicating that the parameterization tends to slightly 4985–4993.
588 H.F. Stockdon et al. / Coastal Engineering 53 (2006) 573–588

Holland, K.T., Holman, R.A., Lippmann, T.C., Stanley, J., Plant, N., 1997. Raubenheimer, B., Guza, R.T., Elgar, S., 2001. Field observations of wave-
Practical use of video imagery in nearshore oceanographic field studies. driven setdown and setup. Journal of Geophysical Research 106 (C3),
IEEE Journal of Oceanic Engineering 22 (1), 81–92. 4629–4638.
Holman, R.A., 1986. Extreme value statistics for wave run-up on a natural Ruessink, B.G., Kleinhaus, M.G., van den Beukel, P.G.L., 1998. Observations
beach. Coastal Engineering 9, 527–544. of swash under highly dissipative conditions. Journal of Geophysical
Holman, R.A., Bowen, A.J., 1984. Longshore structure of infragravity wave Research 103 (C2), 3111–3118.
motions. Journal of Geophysical Research 89 (C4), 6446–6452. Ruggiero, P., Komar, P.D., McDougal, W.G., Marra, J.J., Beach, R.A., 2001.
Holman, R.A., Guza, R.T., 1984. Measuring run-up on a natural beach. Coastal Wave runup, extreme water levels and the erosion of properties backing
Engineering 8, 129–140. beaches. Journal of Coastal Research 17 (2), 407–419.
Holman, R.A., Sallenger Jr., A.H., 1985. Setup and swash on a natural beach. Ruggiero, P., Holman, R.A., Beach, R.A., 2004. Wave run-up on a high-energy
Journal of Geophysical Research 90 (C1), 945–953. dissipative beach. Journal of Geophysical Research 109 (C6).
Howd, P.A., Oltman-Shay, J., Holman, R.A., 1991. Wave variance partitioning Sallenger, A.H., 2000. Storm impact scale for barrier islands. Journal of Coastal
in the trough of a barred beach. Journal of Geophysical Research 96 (C7), Research 16 (3), 890–895.
12781–12795. Sallenger Jr., A.H., Holman, R.A., 1985. Wave-energy saturation on a natural
Hunt, I.A., 1959. Design of seawalls and breakwaters. Journal of Waterways and beach of variable slope. Journal of Geophysical Research 90 (C6),
Harbours Division, ASCE 85 (WW3), 123–152. 11939–11945.
Iribarren, C.R., Nogales, C., 1949. Protection Des Ports, XVIIth International Seymour, R.J., Sessions, M.H., Castel, D., 1985. Automated remote recording
Naval Congress, Lisbon, pp. 31–80. and analysis of coastal data. Journal of Waterway, Port, Coastal, and Ocean
Longuet-Higgins, M.S., Stewart, R.W., 1963. A note on wave set-up. Journal of Engineering 111 (2), 388–400.
Marine Research 21, 4. Thornton, E.B., Guza, R.T., 1982. Energy saturation and phase speeds measured
Longuet-Higgins, M.S., Stewart, R.W., 1964. Radiation stresses in water waves; on a natural beach. Journal of Geophysical Research 87 (C12), 9499–9508.
a physical discussion, with applications. Deep-Sea Research 11, 529–562. Thornton, E.B., Guza, R.T., 1983. Transformation of wave height distribution.
Miche, R., 1951. Le pouvoir réfléchissant des ouvrages maritimes exposés à Journal of Geophysical Research 88 (C10), 5925–5938.
l'action de la houle. Annales des Ponts et Chaussées 121, 285–319. Wright, L.D., Short, A.D., 1983. Morphodynamics of beaches and surf zones in
Nielsen, P., Hanslow, D.J., 1991. Wave runup distributions on natural beaches. Australia. In: Komar, P.D. (Ed.), CRC Handbook of Coastal Processes and
Journal of Coastal Research 7 (4), 1139–1152. Erosion. CRC Press, Boca Raton, pp. 35–64.
Raubenheimer, B., Guza, R.T., 1996. Observations and predictions of run-up.
Journal of Geophysical Research 101 (C10), 25575–25587.
Raubenheimer, B., Guza, R.T., Elgar, S., Kobayashi, N., 1995. Swash on a
gently sloping beach. Journal of Geophysical Research 100 (C5),
8751–8760.