Gaussian/Top Hat model parameters

Topics about the HYSPLIT dispersion model.
rotendd
Posts: 27
Joined: September 25th, 2016, 12:16 pm
Registered HYSPLIT User: Yes

Re: Gaussian/Top Hat model parameters

Post by rotendd » April 16th, 2018, 3:09 pm

First off, thanks again for your previous help!

I have a few more directed questions.

I am conducting a sensitivity analysis by comparing two different scenarios:
  • A 24-hour dispersion from a source of small area (Area is included in EMITIMES file)
  • The same source modeled as a point source (Area = 0m^2 in EMITIMES file, all other parameters do not change)
I'm using initd = 130. (Gaussian horizontal to Particle horizontal) in the SETUP file. Conversion takes place after 1 hour.
NUMPAR = -1500
MAXPAR = 100000

To compare, I grid each dispersion and compare the corresponding grid cells between the dispersions. However, when I sum the grid cells in each model, I get drastically different values, indicating that mass isn't always conserved.

When this occurs, I also receive the warning:
WARNING main: puff splitting turned off ... expand MAXPAR
or increase merge parameters in CFG namelist file

I am sampling several days (all are multiple 24-hr periods). Mass appears to be conserved on other days and I do not always receive this warning.

My questions are:
  • Does this warning also indicate that mass may not be conserved?
  • Is there a way that I can select initd, NUMPAR, MAXPAR, etc. that will work for all days?
Thanks again!

alicec
Posts: 75
Joined: February 8th, 2016, 12:56 pm
Registered HYSPLIT User: Yes

Re: Gaussian/Top Hat model parameters

Post by alicec » April 17th, 2018, 9:29 am

the warning "Puff splitting turned off. Increase MaxPar" means that the puffs are not able to split anymore
because you have MaxPar puffs. Puff splitting is a bit hard to predict as it will depend on meteorological conditions.

I don't understand what you are saying about mass conservation or how you are calculating it. Please give more explanation on what
you mean by:
To compare, I grid each dispersion and compare the corresponding grid cells between the dispersions. However, when I sum the grid cells in each model, I get drastically different values, indicating that mass isn't always conserved.
Are you looking at the concentration grid output (cdump) files?

Each particle / puff carries a certain amount of mass. This mass will stay the same unless you have wet and/or dry
deposition turned on. When the puff splits, the amount of mass will be split as well.

Please look at the MESSAGE files or attach them here.
They have information about the number of particles / puffs and the amount of mass in the rightmost columns.
See example of part of a MESSAGE file below.
The particle number starts at 252 and ends at 5040. The amount of mass starts at 0.2 and ends at 4.0.


NOTICE metinp: GDAS 1 1 10 10 181 139 57119040 8 8 8 0
NOTICE advrng: (kg ,xyr,xy1) - 1 10 10 181 139
NOTICE main: Initial time step (min) 3
NOTICE emspnt: emissions started
NOTICE main: 1 57118923 252 0.20000000
NOTICE main: 1 57118926 504 0.40000281
NOTICE main: 1 57118929 756 0.60000277
NOTICE main: 1 57118932 1008 0.79999900
NOTICE main: 1 57118935 1260 0.99999523
NOTICE main: 1 57118938 1512 1.2000065
NOTICE main: 1 57118941 1764 1.4000177
NOTICE main: 1 57118944 2016 1.6000290
NOTICE main: 1 57118947 2268 1.8000402
NOTICE main: 1 57118950 2520 2.0000515
NOTICE main: 1 57118953 2772 2.2000628
NOTICE main: 1 57118956 3024 2.4000740
NOTICE main: 1 57118959 3276 2.6000853
NOTICE main: 1 57118962 3528 2.8000965
NOTICE main: 1 57118965 3780 3.0001078
NOTICE main: 1 57118968 4032 3.2001190
NOTICE main: 1 57118971 4284 3.4001303
NOTICE main: 1 57118974 4536 3.6001415
NOTICE main: 1 57118977 4788 3.8001528
NOTICE condsk: 1 output at 8 7 23 0
NOTICE main: 1 57118980 5040 4.0001640

rotendd
Posts: 27
Joined: September 25th, 2016, 12:16 pm
Registered HYSPLIT User: Yes

Re: Gaussian/Top Hat model parameters

Post by rotendd » April 20th, 2018, 6:40 pm

In regards to the comparison, I'm essentially gridding the cdump files and adding the average concentrations in each grid cell. (The grids cells are uniform in size.) I'm using this as an analog for mass.

If I only change a single parameter (release height, release temp, etc.), I can use the same method but get a drastically different summation than previous. I'm curious as to why the value is so different.

Everytime this happens, I get a "Puff splitting" error in one of the two runs that I'm comparing.

alicec
Posts: 75
Joined: February 8th, 2016, 12:56 pm
Registered HYSPLIT User: Yes

Re: Gaussian/Top Hat model parameters

Post by alicec » April 24th, 2018, 9:14 am

I think I understand what you are doing. The cdump files are already gridded though. So I assume you are just adding up the concentrations
in each grid cell and the two runs have identical concentration grids.
A few thoughts

1. How big is your concentration grid? It could be that some of the puffs/particles are moving off the concentration grid (either horizontally or
vertically).

2. Can you compare the MESSAGE file output from the two runs to see what it says about the mass and number of particles / puffs?
This would be very helpful in diagnosing what is going on. You could attach them here if you have trouble figuring out the format.

Working with puffs can be tricky. Puff splitting is often very sensitive to initial conditions. What is the maxpar that you are working with
and what are the approximate transport distances and times you are interested in?

Cheers,
Alice

rotendd
Posts: 27
Joined: September 25th, 2016, 12:16 pm
Registered HYSPLIT User: Yes

Re: Gaussian/Top Hat model parameters

Post by rotendd » May 20th, 2018, 5:24 pm

Excellent! My apologies for a slow response. It's the end of a semester :-)

In regards to your additional questions:

1. How big is your concentration grid? It could be that some of the puffs/particles are moving off the concentration grid (either horizontally or vertically).

I receive no warnings about puffs or particles leaving the grid when I'm having this issue. I also plot the two dispersions using HYSPLIT and both are well contained in the grid. My ceiling is set at the upper levels of the met data I'm using. (Unless the grid containing the concentration data is set up differently than the displayed dispersions)

2. Can you compare the MESSAGE file output from the two runs to see what it says about the mass and number of particles / puffs?
This would be very helpful in diagnosing what is going on. You could attach them here if you have trouble figuring out the format.


I have recently archived my MESSAGE and SETUP files so I will dig through these once again. I recently found where the total mass by level is reported. I will compare these.

In the mean time, perhaps you can affirm or correct my intuition here:
Is it safe to say that using the hybrid functionality of gaussian dispersions (for area emissions scenarios) depends heavily on the meteorology at a particular time? I'm trying to run several days in a batch run but I only get the 'maxpar' error (and mass differences) on certain days. Is this because the number of puffs released and maxpar need to be tuned depending on the meteorology for the particular instance?

Thanks so much!

alicec
Posts: 75
Joined: February 8th, 2016, 12:56 pm
Registered HYSPLIT User: Yes

Re: Gaussian/Top Hat model parameters

Post by alicec » May 24th, 2018, 11:29 am

Is it safe to say that using the hybrid functionality of gaussian dispersions (for area emissions scenarios) depends heavily on the meteorology at a particular time? I'm trying to run several days in a batch run but I only get the 'maxpar' error (and mass differences) on certain days. Is this because the number of puffs released and maxpar need to be tuned depending on the meteorology for the particular instance?


Yes, this is correct. When a puff becomes large enough to cover more than one meteorological grid cell, it must split.
How quickly the number of puffs increases is sensitive to meteorological conditions.
https://www.ready.noaa.gov/hysplitusersguide/S622.htm

alicec
Posts: 75
Joined: February 8th, 2016, 12:56 pm
Registered HYSPLIT User: Yes

Re: Gaussian/Top Hat model parameters

Post by alicec » May 31st, 2018, 11:25 am

With the puffs you may also need to make your concentration grid finer.
Some of your puffs may not be counted if they go between grid cell centers.
See:

https://www.ready.noaa.gov/hysplitusersguide/S622.htm
A mixed-mode may be selected to take advantage of the more accurate representation of the 3D particle approach near the source and the smoother horizontal distribution provided by one of the hybrid puff approaches at the longer transport distances. In a long-range or regional puff simulation, where the concentration grid may be rather coarse, puffs may pass between concentration sampling nodes during the initial stages of the transport, a stage when the plume is still narrow.

Post Reply