Vertical Mass Distribution

Topics about the HYSPLIT dispersion model.
mingwang
Posts: 47
Joined: May 8th, 2019, 9:19 pm
Registered HYSPLIT User: No

Vertical Mass Distribution

Post by mingwang »

Dear all,

I used HYSPLIT (concentration dispersion component) to run a particle dispersal simulation for 25 hours. I then checked the vertical mass distribution in the message file.

After the 24th hour of the simulation, the proportion of the mass at each height level to the total mass remained in the atmosphere was shown below.

Index Height %Mass
11 3360.0 0.32
10 2755.0 2.49
9 2210.0 14.98
8 1725.0 30.96
7 1300.0 25.90
6 935.0 24.16
5 630.0 12.90
4 385.0 11.81
3 200.0 15.54
2 75.0 5.89
1 10.0 2.06

I summed all proportions and it turned out to be 147.01%. It does not make sense to me. Should the sum of all proportions be 100%?

It would be great if you could explain it. Your help would be much appreciated. Thanks. :)


Kind Regards,

Ming
alicec
Posts: 284
Joined: February 8th, 2016, 12:56 pm
Registered HYSPLIT User: Yes

Re: Vertical Mass Distribution

Post by alicec »

Can you please attach your CONTROL and SETUP.CFG file you used for the run so we can reproduce the problem.
Thanks.
mingwang
Posts: 47
Joined: May 8th, 2019, 9:19 pm
Registered HYSPLIT User: No

Re: Vertical Mass Distribution

Post by mingwang »

Dear alicec,

Thank you very much for your help.

I actually ran the simulation on HPC (HYSPLIT for LINUX), due to a large number of particles (28,800,000). It took almost 20 hours to finish.

I have attached my CONTROL and SETUP.CFG file. Please find the attached files. Your help would be much appreciated. Thanks. :)


Kind Regards,

Ming
Attachments
setup.txt
(551 Bytes) Downloaded 19 times
control.txt
(355 Bytes) Downloaded 19 times
alicec
Posts: 284
Joined: February 8th, 2016, 12:56 pm
Registered HYSPLIT User: Yes

Re: Vertical Mass Distribution

Post by alicec »

Could you please try running with a smaller number of particles to see if the problem persists.
I reproduced your run with 288,000 particles and was unable to reproduce the problem. See below.
Please let us know.

Index Height %Mass
11 3360.0 0.23
10 2755.0 1.77
9 2210.0 10.20
8 1725.0 21.00
7 1300.0 17.62
6 935.0 16.36
5 630.0 8.85
4 385.0 8.20
3 200.0 10.30
2 75.0 4.04
1 10.0 1.42
mingwang
Posts: 47
Joined: May 8th, 2019, 9:19 pm
Registered HYSPLIT User: No

Re: Vertical Mass Distribution

Post by mingwang »

Dear alicec,

Thank you very much for your help.

I ran with a smaller number of particles (288,000) and an emission rate of being 9600 on my computer (Windows). There was no such problem. I got 100.01%. Please see below and find the attached control and setup files.

Index Height %Mass
10 2755.0 0.02
9 2210.0 3.38
8 1725.0 15.53
7 1300.0 19.54
6 935.0 19.29
5 630.0 17.01
4 385.0 11.77
3 200.0 7.32
2 75.0 4.39
1 10.0 1.76

I am not sure if the problem was caused by a large number of particles (28,800,000) released. Because I previously ran 6 simulations with different starting time and 28,800,000 particles.

The sum of all proportions for each simulation was 100.01%, 100%, 147.84%, 100%, 126.22%, 127.51%, respectively. Three of them had the problem.

Would I be able to convert the propotion at each height level (in the simulation where the summed proportion more than 100% occurred) to a proportion of 100%?

For example, in simulation 3, I got the summed proportion 147.84%. At the height level 5 (630.0 m), the proportion was 12.90%.

The converted proportion = (12.90/147.84)*100 = 8.85%

Would that be OK, if I convert it in such a way?

Your help would be much appreciated. Thanks. :)


Kind Regards,

Ming
Attachments
smaller_number_particles_setup.txt
(547 Bytes) Downloaded 18 times
smaller_number_particles_control.txt
(353 Bytes) Downloaded 17 times
mingwang
Posts: 47
Joined: May 8th, 2019, 9:19 pm
Registered HYSPLIT User: No

Re: Vertical Mass Distribution

Post by mingwang »

Dear alicec,

I have also just run with a smaller number of particles (288,000) and an emission rate of being 9600 on HPC for 6 replications (the same starting time). There was no such problem for all 6 replications, but the outputs and dispersal plumes were slightly different among the 6 replications. Is that because the random numbers generated on HPC for each replication are different?

Please see below.

Replication 1:

Index Height %Mass
11 3360 0.22
10 2755 1.69
9 2210 10.23
8 1725 20.98
7 1300 17.65
6 935 16.3
5 630 8.85
4 385 8.11
3 200 10.46
2 75 4.02
1 10 1.47

Replication 2:

Index Height %Mass
11 3360 0.23
10 2755 1.7
9 2210 10.21
8 1725 21.16
7 1300 17.57
6 935 16.39
5 630 8.88
4 385 8.17
3 200 10.35
2 75 3.94
1 10 1.42

Replication 3:

Index Height %Mass
11 3360 0.23
10 2755 1.67
9 2210 10.2
8 1725 21.15
7 1300 17.62
6 935 16.39
5 630 8.85
4 385 8.14
3 200 10.24
2 75 4.02
1 10 1.47

Replication 4:

Index Height %Mass
11 3360 0.22
10 2755 1.75
9 2210 10.26
8 1725 21.15
7 1300 17.44
6 935 16.36
5 630 8.83
4 385 8.24
3 200 10.25
2 75 4.05
1 10 1.47

Replication 5:

Index Height %Mass
11 3360 0.22
10 2755 1.69
9 2210 10.31
8 1725 21.11
7 1300 17.57
6 935 16.35
5 630 8.92
4 385 8.08
3 200 10.32
2 75 4.01
1 10 1.42

Replication 6:

Index Height %Mass
11 3360 0.22
10 2755 1.77
9 2210 10.27
8 1725 20.96
7 1300 17.54
6 935 16.33
5 630 8.89
4 385 8.25
3 200 10.43
2 75 3.91
1 10 1.44

The sum of all proportions for each replication are 99.98%, 100.02%, 99.98%, 100.02%, 100%, and 100.01%, respectively.

It looks like that the problem might be caused by a large number of particles released.

Thank you very much for your help, again. Your help would be much appreciated. Thanks. :)


Kind Regards,

Ming
alicec
Posts: 284
Joined: February 8th, 2016, 12:56 pm
Registered HYSPLIT User: Yes

Re: Vertical Mass Distribution

Post by alicec »

I have reproduced your problem with the large number of particles and am in the
process of tracking down the cause.

WIth only 2,880,000 you can see below that the mass (most right column) increases with
time until 6 hours which it should be doing as you specified a 6 hour emission time period.

Code: Select all

NOTICE emspnt: emissions started
  NOTICE   main:            1    55365130       80000   770095.7
  NOTICE   main:            1    55365140      160000   1513306.
  NOTICE   main:            1    55365150      240000   2238102.
  NOTICE   main:            1    55365160      320000   2955152.
  NOTICE   main:            1    55365170      400000   3662226.
  NOTICE   main:            1    55365180      480000   4370622.
  NOTICE advrng: (kg ,xyr,xy1) -  1   10   10   54   18
  NOTICE   main:            2    55365195      600000   5443726.
  NOTICE   main:            2    55365210      720000   6532004.
  NOTICE   main:            2    55365225      840000   7633086.
  NOTICE   main:            2    55365240      960000   8738619.
  NOTICE advrng: (kg ,xyr,xy1) -  1   10   10   54   18
  NOTICE   main:            3    55365255     1080000   9845495.
  NOTICE   main:            3    55365270     1200000  1.0970443E+07
  NOTICE   main:            3    55365285     1320000  1.2104691E+07
  NOTICE   main:            3    55365300     1440000  1.3247584E+07
  NOTICE advrng: (kg ,xyr,xy1) -  1   10   10   54   18
  NOTICE   main:            4    55365315     1560000  1.4395510E+07
  NOTICE   main:            4    55365330     1680000  1.5547104E+07
  NOTICE   main:            4    55365345     1800000  1.6701367E+07
  NOTICE   main:            4    55365360     1920000  1.7878998E+07
  NOTICE advrng: (kg ,xyr,xy1) -  1   10   10   54   18
  NOTICE   main:            5    55365375     2040000  1.9049466E+07
  NOTICE   main:            5    55365390     2160000  2.0223650E+07
  NOTICE   main:            5    55365405     2280000  2.1398812E+07
  NOTICE   main:            5    55365420     2400000  2.2574502E+07
  NOTICE advrng: (kg ,xyr,xy1) -  1   10   10   54   18
  NOTICE   main:            6    55365435     2520000  2.3750900E+07
  NOTICE   main:            6    55365450     2640000  2.4927738E+07
  NOTICE   main:            6    55365465     2760000  2.6104594E+07
  NOTICE   main:            6    55365480     2880000  2.7281366E+07
However with the 28,800,000 particles the total mass stops increasing after
4 hours and stays pegged at the 1.67e7 value through the rest of the run.
This is incorrect. You are using a much larger number of particles than we test
with or use in research. I will post more when I determine the cause of the behavior.
You might want to consider if you need this many particles as well. As you note, it makes
for rather long run times.

Code: Select all

 NOTICE   main:            1    55365130      800000   770281.0
  NOTICE   main:            1    55365140     1600000   1514304.
  NOTICE   main:            1    55365150     2400000   2237937.
  NOTICE   main:            1    55365160     3200000   2949301.
  NOTICE   main:            1    55365170     4000000   3655445.
  NOTICE   main:            1    55365180     4800000   4362111.
  NOTICE advrng: (kg ,xyr,xy1) -  1   10   10   54   18
  NOTICE   main:            2    55365195     5562111   5435770.
  NOTICE   main:            2    55365210     6762111   6524240.
  NOTICE   main:            2    55365225     7962111   7625643.
  NOTICE   main:            2    55365240     9162111   8738376.
  NOTICE advrng: (kg ,xyr,xy1) -  1   10   10   54   18
  NOTICE   main:            3    55365255     9938376   9861727.
  NOTICE   main:            3    55365270    11138376  1.0994698E+07
  NOTICE   main:            3    55365285    12338376  1.2135175E+07
  NOTICE   main:            3    55365300    13538376  1.3280758E+07
  NOTICE advrng: (kg ,xyr,xy1) -  1   10   10   54   18
  NOTICE   main:            4    55365315    14480758  1.4430858E+07
  NOTICE   main:            4    55365330    15680758  1.5584026E+07
  NOTICE   main:            4    55365345    16880758  1.6739972E+07
  NOTICE   main:            4    55365360    18080758  1.6777216E+07
  NOTICE advrng: (kg ,xyr,xy1) -  1   10   10   54   18
  NOTICE   main:            5    55365375    19097738  1.6777216E+07
  NOTICE   main:            5    55365390    20297738  1.6777216E+07
  NOTICE   main:            5    55365405    21497738  1.6777216E+07
  NOTICE   main:            5    55365420    22697738  1.6777216E+07
  NOTICE advrng: (kg ,xyr,xy1) -  1   10   10   54   18
  NOTICE   main:            6    55365435    23739484  1.6777216E+07
  NOTICE   main:            6    55365450    24939484  1.6777216E+07
  NOTICE   main:            6    55365465    26139484  1.6777216E+07
  NOTICE   main:            6    55365480    27339484  1.6777216E+07
[\code]
alicec
Posts: 284
Joined: February 8th, 2016, 12:56 pm
Registered HYSPLIT User: Yes

Re: Vertical Mass Distribution

Post by alicec »

You are correct. There will be some differences between simulations made with different numbers of particles or started with a different random seed. You can find some discussion of the expected uncertainties and how to estimate how many particles you need in this publication section 2.4. https://www.mdpi.com/2073-4433/11/12/1369
For most situations you can configure the model so the differences due to the random seed are quite small.
mingwang
Posts: 47
Joined: May 8th, 2019, 9:19 pm
Registered HYSPLIT User: No

Re: Vertical Mass Distribution

Post by mingwang »

Dear alicec,

Thank you very much for your help.

I have reduced the number of particles released to 15,000,000 and run some simulations. The sum of each proportion at the different height levels is 100% now, which is good. But I have found another issue.

After 6 hours of emission, the number of particles released did not reach 15,000,000. It was always between 10,000,000 and 14,000,000.

In one run, it was even below 10,000,000, which was 9,289,945. After the simulation, I checked the number of particles suspended in the atmosphere and it was 4,002,104, as well as the number of particles deposited on the ground which was 10,985,705.

4,002,104 + 10,985,705 = 14,987,809 (~15,000,000)

The total number of particles matches with NUMPAR. But there were actually only 9,289,945 particles released after the 6th hour (shown in the message file).

The number should be about 9,289,945, if I sum the number of particles suspended in the atmosphere and the number of particles deposited on the ground, right?

I could not figure out what it is going on. Please find the attached files. Your help would be much appreciated. Thanks. :)


Kind Regards,

Ming
Attachments
MESSAGE.txt
(23.68 KiB) Downloaded 17 times
setup.txt
(551 Bytes) Downloaded 15 times
control.txt
(361 Bytes) Downloaded 16 times
alicec
Posts: 284
Joined: February 8th, 2016, 12:56 pm
Registered HYSPLIT User: Yes

Re: Vertical Mass Distribution

Post by alicec »

HYSPLIT calculates number of particle it releases on numpar as well as a few other factors
such as number of release locations, number of release cycles, number of types of computational particles.
It will often not release exactly numpar particles.
Post Reply