Time Step and Concentration Calculation

Topics about the HYSPLIT dispersion model.
mingwang
Posts: 36
Joined: May 8th, 2019, 9:19 pm
Registered HYSPLIT User: No

Re: Time Step and Concentration Calculation

Post by mingwang » June 18th, 2019, 4:23 am

Please see the attached MESSAGE files. Thanks. :)
Attachments
6_source_locations_MESSAGE.txt
(107.02 KiB) Downloaded 18 times
1_source_location_MESSAGE.txt
(95.04 KiB) Downloaded 16 times

mingwang
Posts: 36
Joined: May 8th, 2019, 9:19 pm
Registered HYSPLIT User: No

Re: Time Step and Concentration Calculation

Post by mingwang » June 19th, 2019, 12:17 am

Dear alicec,

I am very sorry that I attached a wrong CONTROL file for the six source location simulation. Please find the attached correct one.

Your help would be much appreciated. Thanks.


Kind Regards,

Ming
Attachments
six_sources_control.txt
(393 Bytes) Downloaded 13 times

alicec
Posts: 152
Joined: February 8th, 2016, 12:56 pm
Registered HYSPLIT User: Yes

Re: Time Step and Concentration Calculation

Post by alicec » June 19th, 2019, 9:13 am

Try the following:

1. Repeat your runs with the deposition turned off.
2. Look and see how many particles HYSPLIT emitted.

I think what you are running into is that HYSPLIT takes NUMPAR, number of particles to release, as a guide.
When you go to the 24 hour emissions cycle, HYSPLIT wants to release the same number of particles per hour, but
10,000 is not evenly divisible by 24. So HYSPLIT releases slightly more particles.
HYSPLIT will also want to divide number of particles evenly between your sources. You may also need to end up setting the time step (DELT).

So here is my suggestion. You need to run without deposition until you find the right combination of numpar, emission rate, etc. etc. so that you
can see in the MESSAGE file that each particle has one unit mass.

Then you can turn deposition on.

mingwang
Posts: 36
Joined: May 8th, 2019, 9:19 pm
Registered HYSPLIT User: No

Re: Time Step and Concentration Calculation

Post by mingwang » June 20th, 2019, 5:28 am

Dear alicec,

Thank you very much for your help. As you suggested, it works. I would be able to get the exactly same number of deposited particles between the output file (the number of deposited particles) and the MESSAGE file (the number of deposited particles = total released particles - the number of particles left in the atmosphere).

I have run the same CONTROL and SETUP files on HPC (Linux system) today. The number of deposited particles is different from the number generated from a Windows laptop.

261 (Linux) vs. 191 (Windows)

There is a 27% difference between Linux and Windows for the exactly same simulation.

Is it because of two different operating systems and different number of processors?

Your help would be much appreciated. Thanks. :)


Kind Regards,

Ming

alicec
Posts: 152
Joined: February 8th, 2016, 12:56 pm
Registered HYSPLIT User: Yes

Re: Time Step and Concentration Calculation

Post by alicec » June 21st, 2019, 9:15 am

There will be differences because the random numbers generated on the two systems will not be the same.
To see this on the same system, you can set the SEED parameter in the SETUP.CFG (at least if you have a more recent version). https://ready.arl.noaa.gov/hysplitusersguide/S410.htm

Try running with different seeds on the same system and see how big a difference you get.
The difference will depend on meteorological conditions, but it could be fairly large.

I suggest you work through Section 7 of the tutorial as well.
https://ready.arl.noaa.gov/documents/Tu ... index.html

If you look just at grid cells which contain a high density of particles, the difference between the two simulations should be smaller than
if you look at differences in grid cells around the edges of the plume where the number of particles is small.

mingwang
Posts: 36
Joined: May 8th, 2019, 9:19 pm
Registered HYSPLIT User: No

Re: Time Step and Concentration Calculation

Post by mingwang » June 26th, 2019, 10:21 am

Dear alicec,

Thank you very much for your help. I really appreciated it. Thanks. :)


Kind Regards,

Ming

Post Reply