CMAQv5.2_DDM-3D Jacobian files for other chemical mechanisms

hml · May 7, 2019, 2:21am

Hello,all
I found jac_cb05tucl_ae6_aq.F was not imperfect when I compile CMAQv5.2_DDM-3D with cb05tucl_ae6_aq mechanism.

jac_cb05tucl_ae6_aq.F(2021): error #6351: The number of subscripts is incorrect. [C]
PD( CLNO2 ,H2NO3PIJ ) = RK( 212 ) * C( )
--------------------------------------------------------^
jac_cb05tucl_ae6_aq.F(2022): error #6351: The number of subscripts is incorrect. [C]
& + RK( 213 ) * C( )
--------------------------------------------------------^
jac_cb05tucl_ae6_aq.F(2023): error #6351: The number of subscripts is incorrect. [C]
PD( CLNO2 ,H2NO3PK ) = RK( 214 ) * C( )
--------------------------------------------------------^
jac_cb05tucl_ae6_aq.F(2102): error #6351: The number of subscripts is incorrect. [C]
& - RK( 212 ) * C( )
--------------------------------------------------------^
jac_cb05tucl_ae6_aq.F(2103): error #6351: The number of subscripts is incorrect. [C]
& - RK( 213 ) * C( )
--------------------------------------------------------^
jac_cb05tucl_ae6_aq.F(2110): error #6351: The number of subscripts is incorrect. [C]
& - RK( 214 ) * C( )
--------------------------------------------------------^
compilation aborted for jac_cb05tucl_ae6_aq.F (code 1)
make: *** [jac_cb05tucl_ae6_aq.o] Error 1
ERROR while running make command

Could you help me?
Thank you for your time.

lizadams · October 4, 2019, 6:07pm

Please try again using the CMAQv5.3 version

github.com

USEPA/CMAQ/blob/master/DOCS/Users_Guide/CMAQ_UG_ch10_HDDM-3D.md


<!-- BEGIN COMMENT -->

[<< Previous Chapter](CMAQ_UG_ch09_process_analysis.md) - [Home](README.md) - [Next Chapter >>](CMAQ_UG_ch11_ISAM.md)

<!-- END COMMENT -->

# 10. Decoupled Direct Method in Three Dimensions (CMAQ-DDM-3D)

## 10.1 Introduction

The Decoupled Direct Method in Three Dimensions (DDM-3D) provides CMAQ concentration and deposition sensitivity information for user specified model parameters.

In air quality modeling, sensitivities measure the response of a model output to a change in one or several predefined model parameters. In policy applications, the parameters of interest are usually emissions and the output of interest is pollutant concentrations. We may be interested in emissions from a particular geographical region, like an urban area, a group of states, or a country, and/or emissions from a particular source, such as wildfires, electricity generating units (EGUs), or light duty diesel trucks.

Emissions sensitivities can be calculated by simply running the air quality model twice – once with standard emissions inputs, and once with the emissions of interest adjusted in some way. The difference in outputs between the two runs in relation to the size of the adjustment then becomes the model sensitivity. While this process is fairly easy to implement and interpret, it quickly becomes computationally complex as the number of desired sensitivities increases. For example, calculating sensitivity to EGU emissions from 10 southeastern states in the U.S. would require 11 separate air quality model simulations.

Alternatively, model sensitivities can be calculated with CMAQ-DDM-3D. This is done by altering the existing model algorithms to allow for sensitivity propagation through every science module in CMAQ. While this process does require more computational resources than standard CMAQ, it scales favorably with the number of desired parameters.

Besides emissions, sensitivities to other model parameters can also be calculated. Currently, CMAQ-DDM-3D can be used for sensitivity to emission rates, boundary conditions, initial conditions, reaction rates, potential vorticity, or any combination of these parameters. Second order sensitivity calculations, or sensitivity of sensitivity, sometimes known as higher-order DDM-3D (HDDM-3D) are also available.  Note: second order sensitivity outputs for particulate matter species are still in development and should be considered as a research option.

This file has been truncated. show original

bkoo · October 9, 2019, 5:04pm

I don’t think DDM-3D is available for CMAQ v5.3. Please correct me if I’m wrong.

lizadams · November 8, 2019, 6:28pm

You are correct that DDM-3D was not released for CMAQv5.3.
Here are a few resources that may help:
CMAQv5.2 DDM-3D

github.com

lizadams/CMAQ/blob/5.2/DOCS/Instrumented_Docs/CMAQ_DDM.md

Original Release Date:

# Overview


The '''CMAQ Decoupled Direct Method in 3 Dimensions (CMAQ-DDM-3D)''' is a sensitivity analysis technique for computing sensitivity coefficients simultaneously while air pollutant concentrations are being computed (see References). The sensitivity coefficients represent the change in concentration, of any modeled species at any modeled time, associated with a change in a model input (e.g., an initial condition, boundary condition or emission rate) or a parameter (e.g., a reaction rate).

CMAQ-DDM-3D generates concentration outputs that are essentially identical to normal CMAQ results, while simultaneously computing sensitivity coefficients for any species concentration to a change in initial conditions, boundary conditions, or emission rates.  We have attempted to adhere to CMAQ programming conventions.  Some changes have been made to the input and output of data to keep the sensitivity output files to a reasonable size and to facilitate the necessary passing of information from mother domains to daughter domains for nested runs (see Output of Data).

Current DDM-3D implementation is available for version 5.2 of the Community Multiscale Air Quality (CMAQ) model.  This version of the code has been the work of Sergey L. Napelenok, and is based on previous contributions from Wenxian Zhang, Daniel Cohan, Ted Russell, Yueh-Jiun Yang, Amir Hakami and James Boylan.  This code can be compiled and executed in parallel mode identically to the base CMAQ model to take advantage of available multiprocessing capabilities.

This guide is intended to assist users of our implementation of DDM sensitivity analysis into CMAQ.  We acknowledge that our implementation is a work in progress and list cautionary notes (see Shortcomings and Unimplemented Features) that should be considered when using CMAQ-DDM-3D. However, we have tested that for ozone chemistry and PM processes CMAQ-DDM-3D gives results in good agreement with sensitivities calculated by differencing multiple brute-force runs of CMAQ, at a significant savings of computational time.

CMAQ-DDM-3D is released as part of public release version 5.2.  Updates to the code include the migration to the most recent base model science (version 5.2). CMAQ 5.0.2 updates included 2nd order sensitivity calculation for particulate matter species, and improved flexibility in the control file.

Implementation approach/methodology remains largely unchanged, so only the major changes are described in this document.

## Implementation Highlight:  CMAQv5.2

CMAQ-DDM-3D for model release verion 5.2 followed the same science updates and code restructuring as the base model.  No major functionality in the sensitivity model was added at this update.

This file has been truncated. show original

CMAQv5.0.2 DDM
https://www.airqualitymodeling.org/index.php/CMAQv5.0.2_Direct_Decoupled_Method

Successful testing was performed with the cb6r3_ae6nvPOA_aq mechanism:
https://www.airqualitymodeling.org/index.php/CMAQ_version_5.2_DDM_Technical_Documentation

sergey · November 12, 2019, 1:12pm

We are currently working on implementing a more robust mechanism support for the next release of CMAQ-DDM. In the current system, any slight changes to the base mechanism will likely break the subsequent DDM algorithms. Possibly, this is what happened here.

I am currently running the mechanism “cb6r3_ae6nvPOA” with no issues, so I suggest you try that one. If you application must use cb05tucl_ae6, I can probably come up with something, but it would for the non-volatile version.

mqin · October 14, 2020, 6:58am

Do you have the cb05tucl_ae6nvPOA version that works?

erica · May 29, 2022, 8:06am

Sorry to disturb, when will the next version will release and could you give me some help on correct running the DDM-3D with a revised mechanism based on the cb6r3_ae6nvPOA chemical mechanism? Thanks!

erica · June 4, 2022, 2:28pm

Did you get the jac_cb05tucl_aq6_aq.f file that can run normally? Could you please share a copy with me? Thank you very much for your help!

erica · June 7, 2022, 6:17am

@lizadams Sorry to disturb. I am wondering when the next version will be released. And could you give me some help on correctly running the DDM-3D for other chemical mechanisms? I want to use the cb05 chemical mechanism, and I have tried to revise the jac_xxx.f file by myself, but the results seem wrong. Specifically, the senseitivity of NOx to Ozone was zero. Do you know if jac_xxx.f is the only file needed to revise when using another chemical mechanism? Thanks!

sergey · June 23, 2022, 1:48pm

Hi Erica,

Sorry, I completely missed this threat the entire time. The next version of CMAQ is coming out in the fall. I am not exactly sure on the date, but it should be fairly soon. In that version, the need for jac_*.f is going away, so there should be no more issues with incompatible changes to chemistry.

However, it doesn’t sound like your current issue is with the jacobian. If your jacobian is wrong for the mechanism, the model would crash and not produce output. Since you report getting zeros in your output, there is likely a different issue.

Now, you say that you are looking at sensitivity of NOx to O3. Do you actually mean the sensitivity of O3 to NOx? The sensitivity of NOx to O3 emissions probably should be zero in most applications, because there is no O3 emissions.

Sergey

erica · June 24, 2022, 5:39am

Hi Sergey,
Thank you so much for your response !!!
You are right, and I have calculated the sensitivity of O3 to NOx emissions. The sensinput file, the run script, and the final results are attached. I think I have set the sensitivity options correctly, but I still get a lot of zero values. I am hoping you can give me some advice. Thank you again!
sensinput.cmas.dat.ENV1.txt (221 Bytes)
run_cctm_pbs20.csh (23.6 KB)

sergey · June 27, 2022, 1:59pm

Erica,

Your control file and run script look fine. So, the next thing to check would be if you are getting reasonable output for the base concentrations. Does your O3 output look ok?

Also, are all sensitivity outputs zero? What does NO_EN look like? Is the zero only in the first time step or the entire day?

Finally, does this file:

set EMISfile = EM_{CFG}_{yyyymmdd_emis}.nc #> Offline 3d emissions file name
setenv EGRIDFILE1 EMISpath/EGRID1/EM_{CFG}_${yyyymmdd_emis}.nc

look ok? This is your emissions files that I think you are trying to find the sensitivity to.

Sergey

erica · June 27, 2022, 3:06pm

Hi Sergey, @sergey
Thank you for your reply!
The O3 output looks fine, and the O3 distribution and magnitude are consistent with my previous simulations.
All sensitivity outputs are zero, except for NO2_EN and NO_EN. It is strange that the NO_EN result has abnormally high values while the emission file is continuous at the same time in the same area and there are no high values in emissions. Could you tell me what this means?
The results are zero for all times of the day for sensitivity output like O3_EN.
I used the emission files from my base simulation as well as the emission files required for the sensitivity simulations, and after inspection, the emission file looks correct.
I’m sorry to disturb you. Have you even encountered similar errors? I hope you can give me some advice.

Thanks,
Erica

erica · June 29, 2022, 8:56am

@sergey On the way, I got similar zero values with the cb05e51 chemical mechanism using the benchmark file given by the CMAQ website. As a result, I believe that some variables may be mismatched throughout the DDM process. I’m hoping you can give me some advice.

sergey · June 29, 2022, 12:14pm

I am not really sure why get such behavior. You said you revised the jacobian file. It is possible that is the cause. Does the standard cb6r3_ae6nvPOA jacobian not work for you?

Sergey

erica · June 29, 2022, 12:28pm

Thank you for responding! The standard cb6r3_ae6nvPOA can be used properly, but the aforementioned error will occur if the mechanism is altered or a new reaction is introduced. I found the jacbian file for the cb05e51 mechanism in the /src/gas/ directory. Which file do you think went wrong, and maybe I can test it further? Thank you very much!

Erica

sergey · July 5, 2022, 2:42pm

My guess would be that the file you altered is what is caused the issue. If the standard jacobian is working properly, what are you trying to do that requires changing it?

Sergey

erica · July 6, 2022, 2:02am

Hi Sergey, @sergey
Thank you for your reply!
I want to use the CB05 mechanism and change some of the chemical reactions in the existing mechansims. When I don’t alter the jacobian file copied from the /src/gas directory, it still doesn’t work.

Erica

sergey · July 7, 2022, 5:46pm

Erica,

Now I think I finally understand what is happing.

The jacobian used in the DDM calculations is specific to the chemical mechanism. Since you modified the mechanism, it should no longer work properly. This is is the behavior you are experiencing. You could look inside the jac*.f file and make the analogous modification there to the rows and columns of the matrix it is defining based on whatever changes you made to the mechanism.

Alternatively, the next CMAQ release will have the capability to generate this matrix automatically during runtime.

Sergey

erica · July 13, 2022, 9:06am

Sergey,
Thanks for your reply. I’ll keep trying.
Erica

Topic		Replies	Views
DDM Jacobian files for other chemical mechanisms CMAQ-DDM-3D	4	1047	March 24, 2022
CMAQ-DDM Error: number of in Jacobian does not match N_GC_SPC CMAQ-DDM-3D	2	795	June 21, 2019
CMAQ-DDM saprc07tc species run error CMAQ-DDM-3D	15	1529	June 4, 2022
CMAQ-DDM for v5.3.3 CMAQ-DDM-3D	4	950	July 17, 2022
Error in subroutine LOAD_CSQY_DATA in CMAQV5.4 CMAQ	2	23	August 29, 2024

CMAQv5.2_DDM-3D Jacobian files for other chemical mechanisms

Related Topics