A significant discrepancy in simulation results


I followed the instructions in the CMAQ v5.0 Adjoint User’s Manual and run the model successfully.
I chose the one hour concentration of ozone in a grid (94.5ppb) as the cost function, and then used reverse simulation to calculate the contribution of nitrogen oxides (NOx) and volatile organic compounds (VOCs) to ozone for 48 hours. After when I analyzed the results, I found that the contribution of NOx was 4.3ppb and VOC was 0.4ppb, with a total of only 4.7ppb. However, the ozone concentration I selected was 94.5ppb, and there was a significant deviation in the simulation results. How should I solve this problem?

I want to find some ways to solve question. May I ask how to solve the problem?

Best regards

Hi Hake,

The adjoint produces the sensitivities of a cost function with respect to concentrations and emissions. I wonder how you calculate those contributions from NOx and VOCs using the sensitivities in your test. Also, you might need more than 1 pair of data for cell-by-cell comparisons.


Hi zhaoshunliu,

I only calculated the sensitivity results for emissions using the outputs from the model. The specific formula for the calculation is shown in the following figure. After calculating the results for each grid separately, I added up the results for all grids within 48 hours in the model.

Can you explain in detail what “Also, you might need more than 1 pair of data for cell-by-cell comparisons” specifically means?