The Hybrid Single-Particle

Lagrangian Integrated
Trajectory model

1
OVERVIEW – DAY 2
Cleanup

Part 3: Concentration

• Set-up
• Equations
• Case Study
• Apply
• Simple Scenario
• Single Particle
• Multiple Particles
• Puffs
2
CLEANUP
Before we begin, we will clean out our working
directory.
Go to ADVANCED→CLEANUP WORKING
It will then ask you if you want to continue.
Click CONTINUE.
Hysplit will then close and if you go to your hysplit
folder, there will be the new folder you just
created and your working directory will be
empty.
Now, re-open hysplit and hit RESET on the bottom
to change all settings back to default.

3
CONCENTRATION
CONCENTRATION - SETUP
• Very similar inputs compared to
Trajectory.
• Additional menu for POLLUTANT,
DEPOSITION and GRID set up.
•You can specify more than one pollutant
and its deposition properties as well as
multiple grids.
CONCENTRATION - SETUP
GRID OPTIONS DEPOSITION OPTIONS
POLLUTANT OPTIONS

6
CONCENTRATION - EQUATIONS
Concentrations in Hysplit are determined in three different ways:
1. 3-D Particle:
2. Top-Hat:
3. Gaussian:

Notes:
• Different schemes can be used for vertical and horizontal motion at the same time. The motion
of the horizonal determines the method in which the concentration is calculated.
•The incremental concentrations are added to each grid cell or node each advection time step
for all particles or puffs that intersect that point.
•The final average concentration is the incremental sum divided by the number of time steps in
the concentration averaging period.
7
CONCENTRATION – CASE STUDY
We will be using data from the Cross Appalachian Tracer
Experiment (CAPTEX):
• Six 3-hour releases of perfluorocarbon (C7F14) tracers: four in
Dayton, Ohio and two in Sudbury, ON
• Samples collected at 84 sites, 300-800km from the release over
3-6 hour averaging periods for 48-72 hours
• Aircraft also collected data for short time periods (6-10 mins):
the lowest level was 914m MSL and the highest was 2134m MSL
•More information can be found here:
https://journals.ametsoc.org/view/journals/apme/56/8/jamc-
d-16-0345.1.xml

8
CONCENTRATION – SIMPLE SCENARIO
•Go to CONCENTRATION→SETUP RUN.
•Change all the inputs to match the CAPTEX
scenario, as we’ve done before for the
trajectory case.
*Note to add in new meteorology, make
sure to CLEAR the existing files first.
•Then, go to CONCENTRATION→RUN
MODEL.
•Lastly, go to
CONCENTRATION→DISPLAY→CONCENT
RATION→CONTOURS and except the
defaults by clicking EXECUTE DISPLAY. 9
CONCENTRATION – SIMPLE SCENARIO
The concentration pattern represents the
average particle mass in each grid cell
divided by the cell volume.
The plot shows 12-hour averaged
contours of concentrations from 1x10-12
to 1x10-15.
If one particle (mass = 1/2500) were
contained in a grid cell of volume 5x5x1
km (2.5x1010 m3), then the air
concentration would be about 1x10-14
units m-3.
The concentration here though is
averaged so the lower concentrations
here are due to reduced residence times.
10
CONCENTRATION – SINGLE PARTICLE
We will keep the same setup as previously. However,
we will now edit the Grids menu.
Go to CONCENTRATION→SETUP→POLLUTANT,
DEPOSITION AND GRIDS SETUP.
Next, click on SPECIE 1 under POLLUTANT (there will
only be one for this case).
We will be changing the EMISSION RATE and HOUR
OF EMIISION, but not the overall volume. All mass is
now released on one time step though.
Make the changes as shown here.
Hit SAVE.

11
CONCENTRATION – SINGLE PARTICLE
Next we will change the GRID
settings. Click GRID 1.
All we are changing here is the
height of the levels to make sure
we have a large enough domain
to track the single particle.
Hit SAVE to close all menus and
go back to the hysplit home.

12
CONCENTRATION – SINGLE PARTICLE
Lastly, before running the model we need to ensure it
is only including one particle, not its default of 2500.
Go to ADAVANCE→CONCENTRATION→MENU #4
Change 2500 to 1.
Click SAVE to exit everything.
Now RUN THE CONCENTRATION MODEL and
display the results as CONTOURS, accepting all
defaults as we have before.
You should see a narrow plume similar to the
trajectories we’ve modelled.
13
 CONCENTRATION – SINGLE PARTICLE
But what if we didn’t model as high of a
domain?
Go back to the GRIDS menu and change the
height to 500.
RERUN the model and plot the results as
contours.
For this situation, a single particle is insufficient
for a realistic simulation.
•Particles that mixed to the upper-levels of the
boundary layer travel at faster speeds.
•Winds aloft also tend to be more clockwise
than those at the surface.

14
CONCENTRATION – MULTIPLE PARTICLES
We will now model multiple particles.
First change the HEIGHT back to 5000.
Then we will change the AVERAGING time to every 3 hours.
Next, go to: ADVANCED→CONCENTRATION→MENU #4
and change the number of particles from 1 to 100.

15
CONCENTRATION – MULTIPLE PARTICLES
RUN the model.
Go to the
CONCENTRATION→DISPLAY→
CONTOURS and make sure to
click the option for FRAMES
before you click EXECUTE
DISPLAY.
You should have four images
load onto your screen (for every
3 hour period over 12 hours).
We still do not get a good
plume from this, lets try 1000
and 10,000 particles as well.
16
 CONCENTRATION – MULTIPLE PARTICLES
At 10,000 particles we have a much better plume shape. However, it is still much coarser than our test
case. Although this case only had 2500 particles it was averaged over 12 hours instead of the 3 hours we
used here. The model needs to sample a sufficient number of particles over the averaging interval and
concentration grid size to ensure robust results. The ratio of these two should be such that the air
concentration is not sensitive to the particle release rate (determined by trial/error).

17
CONCENTRATION - PUFFS
Instead of following thousands of individual particles, we can follow a mean
trajectory and model the growth of the horizontal (and vertical) distribution. This is
known as a hybrid approach. This can done by two methods in hysplit:
1. Top-Hat: concentration is zero outside and a constant average value inside.
2. Gaussian: follows a normal distribution over a range of 3σ.

**All vertical dispersion is modelled as Top-Hat (or particle). However, horizontal

18
 TOP HAT VS GAUSSIAN
Incremental concentrations for Top-Hat and Gaussian puff are
computed at the center point of each grid cell. Only puffs that
intersect this point will be summed. In contrast, a 3D particle can
be summed anywhere. Volume is a combination of the puff
radius and cell height.

And the standard deviation (σu) is used to calculate horizontal

*Top-Hat used more often because less complex computing and just as accurate
as Gaussian downwind of source. Gaussian only more accurate right at source. 19
 PUFF SPLITTING
What happens when a puff expands to cover several meteorological grid points?
• Top-Hat puff splits horizontally into four puffs, each with 25% of the mass, when
1.54σh > Lh. These new puffs are located at P(x±0.5σh, y±0.5σh).
• A large Gaussian puff splits into five smaller puffs when 3.0σh >Lh. The center puff
gets 60% of the mass while the outside 4 puffs get 10% each. All of which are
located similar to the Top-Hat, but with a plume in the middle.
*σh is the puff horizontal standard deviation.
*Lh is the meteorlogical grid spacing.
• In order to keep array dimensional space, the model also has several rules for
merging and removing puffs as a function of standard deviation.

20
CONCENTRATION – TOP HAT
Go to:
ADVANCED→CONCENTRATION→MENU
#3
We will select the TOP-HAT-HORIZONTAL
PARTICLE-VERTICAL.
Next go to MENU #4 and change the particle
release number back to 1.
RUN the model and display the output in
FRAMES as we did before.

21
CONCENTRATION - GAUSSIAN
Go to:
ADVANCED→CONCENTRATION→MENU #3
We will select the GAUSSIAN-HORIZONTAL
PARTICLE-VERTICAL.
RUN the model and display the output in
FRAMES as we did before.
Here the circular structure is only conserved for 6
hours before it splits.
The area covered after 12 hours though is still
much less than the multiple particle case.
If the center point of either type of puff passes
over a sampling location, the integrated
concentration at that point will be the same.
22
 CONCENTRATION - HYBRID
One advantage of hysplit is its ability to model horizontal dispersion with puffs, but
the vertical mixing is treated as individual particles. Otherwise growing vertical puffs
would have to split very frequently to simulate the boundary layer wind shear.
Modelling the change in dispersion with time can greatly speed up computation time
compared to tracking individual particle trajectories.
Hysplit’s default though is to compute 3D particle trajectory, so it is up to the user to
use other approaches.
We can see this by changing the particle count to 100 through:
ADANVANCED→CONCENTRATION→MENU #4
*Keep plume as Gaussian-Horizontal with Particle-Vertical

23
CONCENTRATION - HYBRID
Comparing the hybrid approach (right) and the 3D particle (left), we see that this
approach is satisfactory. However, depending on your application of hysplit it may
not suffice. ***Look at the concentration ranges

24
 CONCENTRATION – PUFFS SPLITTING
First, change the particle count back to 1 and running the model.
We can view the splitting process by going to: ADVANCED→VIEW
MESSAGES
Looking at the second right most column we can see the number of
puffs. In this case the puff splitting took place at 83 09 25 22 or 5
hours into the simulation.
Here, we see we started with 1 particles and the splitting goes: 5, 7,
17, 29, 65, 113, 237. *Recall the splitting occurs for 5^(# of splits).
Why? Not all puffs grow at same rate and some may leave domain,
so it is difficult to quantify the amount of puffs at each ∆t.
*There is a limit to the amount of puffs though in a given simulation
(MAXPAR) and the model will slow down and eventually terminate as it
approaches it. This could result in a substantial concentration under-prediction
and can only be addressed by increasing the value of MAXPAR.
25