I am trying to run CMAQ DDM, but I have an issue that the sensitivity coefficients in ASENS outputs are all zero. I tried to use CMAQ v5.3.3 at first but I got a suggestion from this forum that this version might not support DDM so I move to version 5.4. I successfully run the base simulation using v5.3.3 already so I think there might not be any problem with the inputs. Is there anyone who had experienced such an issue? I would like to know how to fix my run. Thanks!
Note: ACONC, CONC are all successfully simulated with reasonable values on all variables.
Hi,
I also encountered zero sensitivity coefficients (O3 to NOx or VOCs emssions) when I ran CMAQv5.3.2 and v5.3.3 while I got too high second-order sensitivity coefficients using CMAQv5.4 as reported in CMAQv5.4 HDDM gives too high second order sensitivity. I would be appreciate if someone could help us to solve the problem
For the “all-zeros” behavior, I am sorry I have never been able to replicate that on our test cases. However, 2nd order output in 5.4 does seem to be giving incorrectly high values. We are working on a cause and solutions.
Hi Sergey,
Thanks for your response. I wouldn’t expect the ‘‘all-zero’’ issue, it is not supposed to happen in such a full-tested model. I assume I made some mistakes in the run script or control file (SEN_INPUT). I modified these files based on the template from the source code based on the manual guide. May I have any success templates of these files to check if I made anything wrong in mine, basically I would like to check if there is any wrong setting. Looking forward to your help.
If you have not done so already, you might consider working through the v5.4 DDM benchmark tutorial and compare your output to the benchmark output. Then use this as a starting place to double check your scripts and inputs for your application: CMAQ/CMAQ_UG_tutorial_DDM3D.md at main · USEPA/CMAQ · GitHub
sensinput.dat.txt (195 Bytes)
Hi Sergey, thanks for you response. I attached my sensinpu.dat here as well as my run script. I am wondering if there is any settings in the scripts is wrong that cause all zero sensitivity coefficient. run_cctm_ddm.2016.csh.txt (37.5 KB)
Your sensinput files looks file and I don’t see anything immediately wrong with your script. You got zeros for all species in the ASENS output file? Have you tried reproducing the benchmark output ask Kristen Foley suggested above?
I rebuilt the IOAPI with the “large” version and increase the memory I request from the nodes and now it seems to come out with reasonable values. Thanks for your help.
Hi,wry2233
Did you fix the issue with running version 5.3.2 as I had a similar issue recently. If so, can you tell me how it was solved. I would be appreciate it.
Lotso
Hi Sergey,
I encountered another issue on this run. I successfully run the simulation but after the second day in the loop, the sensitivity coefficient in the ASENS file turns out to be extremely large (more than E+10). I am using same benchmark script as provided in the V5.4 tutorial. Do you know how to check if this run is wrong?
To understand what you are doing correctly - you tried reproducing the v5.4 benchmark case and got large values on the second day? Is that right? Were all the species large or certain ones? Did the first day’s output match what is provide with the release?
Thanks for your reply. I tried the V5.4 benchmark and all results seem reasonable (no extreme values) so I move to run my own case but the extreme values came to all species.
I check the benchmark run and compare it with the release, results are the same.
Are you using the sensinput.dat you provided above or something different? There is a known issue with a fix for second order sensitivities being large.
After looking back your your sensinput.dat, I may have a solution for you. Please, compare your sensinput.dat file with the one provided with the benchmark. You will see that the species list is specified on the single comma delimited line instead of separate lines in response to people wanting to simulate long lists of species. See Chapter 10 of the User’s Guide on github for more information.
The model should have crashed with the way you have it specified, so I do need to double check the error handling for this type of situation.
Thanks for your suggestion. I tried several time and use the correct sensinput.dat but still got problems. The simulation is done successfully and all ACON and CCON file looks well. From the log files, they all show program done successfully. But the problem is the ASENS file is in error, the file size is only 11k, it is not generated successfully. And the data in SENGRID are all zero again. I have no idea what causes this problem. Could you help me to figure out how to solve this problem?
Hi Sergey, Thanks for your fast response. I attached the sensinput.dat here. Sorry for didn’t make it clear. I am using v5.4-DDM to run my jobs and the benchmark run was successfully done.
I checked all log files and there is no Errors but there are some WARNINGS such as:
CTM_SENS_1 :/scratch/general/vast/copycky/data2/cmaq_out/soberanes/saprc.ddm/CCTM_SENGRID_v54_DDM3D_gcc_Soberanes.12km_20160720.nc
>>--->> WARNING in subroutine OPEN3
File not available.
Could not open CTM_SENS_1 file for update - try to open new
Timestep written to CTM_SENS_1 for date and time 2016203:000000
It seems like these warnings are not the reason of the problems I got. sensinput.dat (2).TXT (50 Bytes)
. I suggest you use v5.4.