I am currently using Process Analysis in CMAQ version 5.3.2 to obtain IPR contribution values. In my study, I adjusted BVOC emissions during a heatwave episode and compared the resulting changes in ozone concentrations between the BASE case and sensitivity scenarios. I used Process Analysis to investigate which physical and chemical processes contributed to the changes in ozone concentrations.
However, since the PA/IPR values are signed quantities, I am concerned that directly subtracting the IPR values between scenarios to calculate process-level differences may lead to misleading interpretations.
Therefore, I would like to ask for your advice on how CMAQ IPR results should be analyzed when comparing different scenarios. In particular, I would appreciate your opinion on whether it is more appropriate to compare the absolute IPR contributions for each scenario separately, or whether there is a recommended way to interpret process-level changes between scenarios.
Thank you very much for your time and consideration.
This is a very open-ended question. Ultimately, this problem comes up a lot in other domains as well. For example, the difference in bias, etc. The right answer always depends on the number of elements you’re trying to compare, the graphics you’re showing, the actual differences, and your focus. Where possible, it is best to show at least the original and difference or values from case A, B and difference.
The best way to decide how to show the comparison is to look at the existing PA literature. My work built on earlier work by Jeffries and Tonnesen and Jang et al., and Kimura et al, and many others.
Process Analysis in a 3D Eulerian grid (like CMAQ) has the potential to have millions of individual elements if you include all 3d data. In that case, it is hard to show all the elements for cases A, B and difference. The literature uses “analysis volumes”, which allow you to synthesize conceptual entities – like an airshed, a plume, or an urban area.
In my PA work (Henderson et al 2010, 2011), I have used “analysis volumes” to synthesize cases across space and show the IPR values as a function of time. This allows you to create single values that are representative of an volume instead of individual cells. The design of the analysis volume is very important to represent a conceptual Lagrangian or Eulerian perspective. After synthesizing the analysis volume for both cases, you’ll have two time-series that are easy to graphically compare in an A, B and difference format.
Hope that helps.