Strange values in the CMAQ-adjoint model

Hi,

I am running the CMAQv5.0-adjoint model to calculate the sensitivity of O3 to NOx. However, I encountered strange values at some time points when I run the model backwardly. Please see the enclosed figures. The sensitivity at this time point suddenly became such a large value (-10*10^10). May I know what could be the reason for this?
image

Thank you very much.
Jie

Hi Jie,

This does not look right. Can you please tell me more about your backward run including your cost function/receptor, model resolution, etc? How did the sensitivities look like up to this point? Could you please double check if your sensitivities to NOX calculations are all right?

Thanks,
Shunliu

Hi Shunliu,

Thanks for your response. I set the receptor region as the whole of Malaysia, and the cost function is the hourly O3 concentration. The model resolution is 30km. I tried to set different receptor regions but still encountered the problem at this time step. Enclosed is the O3 (ppmv) sensitivity to NO2 with 2 hours time interval. The sensitivity seems to be correct at hours 23, 21, and 19, but when the model runs backwards at hour 17, it shows a strangely large value (white colour) compared with the background. This value increases explosively when the models run backwards. I checked the emission and the WRF output but I didn’t find any abnormal data. Can you help me find the reason? Thank you very much.

Thanks,
Jie

Hi Jie,

Sorry for the delay in getting back to you. Can you please share the input data and the detailed setup with me via email so that I can do a test run on my end? It would greatly help if I can duplicate your problem. And we can start from there.

Thanks,
Shunliu

Hi Shunliu,

Thanks for your reply. I will I compile the input data with the setup files and send to your email soon. Thanks again for your help.

Best regards,
Hu Jie

Hi, Jie and Shunliu

I am running the CMAQ v5.0 adjoint model to calculate the sensitivity of NO2 to NOx. In my backward model results, I’m also seeing strange values for all variables like yours. My modeling domain is the South Korea. Have you resolved this issue? I would appreciate it if you could inform me about the cause.

Thank you,
Jeonghyeok