I am wondering whether there is a major change in calculating PM10 in CMAQV5.4 compared to CMAQV5.3.1. With exactly the same configurations and same emission DESID I get very low PM10 in CMAQV5.4 compared to CMAQv5.3. I found the fraction of PM10 in coarse mode is quite low. There are new options in cmaq run script:
setenv IC_AERO_M2USE T
setenv BC_AERO_M2USE T
These options were set to F inititallly, however when I set them as T ( the default value), it would give higher PM10 values but still lower than CMAQV5.3.1. I am wondering how this change would be explained.
There were substantial changes to the representation of aerosol dry deposition in CMAQv5.4 for both the M3Dry and STAGE dry deposition schemes. Generally, these changes caused increased aerosol dry deposition and therefore lower concentrations. In addition, the changes were generally more pronounced in M3Dry than STAGE, though the effects vary spatially, temporally, and with PM size range. Please see the following CMAQv5.4 release documentation for additional information:
CMAQ Release Notes: Dry Deposition Air Surface Exchange: M3DRY
CMAQ Release Notes: Dry Deposition Air Surface Exchange: Surface Tiled Aerosol and Gaseous Exchange (STAGE)
What differences should I expect in my model results with v5.4 compared to v5.3.3?
thank you very much for your reply and clarification.
However, my problem is only for PM10 while PM2.5 is not much lower compared to v5.3. PM10 is almost 50% lower when I use:
But when set them as (T), PM10 would be higher but PM2.5 becomes lower. I use M3DRYscheme for depostion. In both of the above cases I get significant underestimation against measurement. Do you have any idea of why it behaves like this? Thanks
The fact that the setting of IC_AERO_M2USE and BC_AERO_M2USE has such a big impact on your model domain suggests that a) there are inconsistencies between the aerosol representation in the model used to generate your initial and boundary conditions and the aerosol module used in CMAQ and b) you are running for a fairly short time period only (such that initial conditions have an important impact on your results) and/or for a fairly small domain only (such that boundary conditions have an important impact on your results). The treatment of inconsistencies in aerosol properties when reading in the initial and boundary conditions was updated in CMAQv5.4:
Release Notes - Update Aerosol Size Distribution Check for ICs and BCs
Without knowing more specifics about your model setup and analysis (how were the initial and boundary conditions generated, what is the size of your modeling domain and duration of your simulation and analysis period, where is your domain located, what period are you analyzing, how many PM25 and PM10 monitors are you using in your analysis, etc.), it is difficult to provide more specific answers.
Hi, Thanks for your reply. I run the model for the whole year (2019) and it covers the whole UK for the finest domain at 2km resolution. IC and BC come from Hemispheric CMAQ and I compare the annual averages with measurements at more than 50 sites.
I see IC_AERO_M2USE and BC_AERO_M2USE are introduced in the new version. How this was treated in the previous versions?
Thanks for this additional information. So ICs should not really be an issue, BCs apparently do have an impact. To clarify, are you running a single 2 km domain covering the UK driven by H-CMAQ BCs, or are you running multiple nested domains with only the outer one driven by H-CMAQ BCs and the inner one(s) by BCs generated from the outer one(s)?
For H-CMAQ, did you work with the old 2016 seasonal average file, or the newer daily average 108NHEMI EQUATES files available for 2002 - 2019 in the CMAQ_108NHemi directory?
I have 3 nested domains at 50km, 10km, and 2km. The fisrt domain is quite big and covers most of the europe. I use the old HCMAQ one. I did not have any problem with v5.3 and the results and model performace was reasonable.
But now i got these issues with the new version, using exactly the same inputs for met and emissions, and ICs/BCs
Thanks. And for the 10 km grid, you run BCON to generate BC files from the 50 km run, and then for the 2 km run, you run BCON to generate BC files from the 10 km run, correct? And the comparisons to observations are from the 2 km run?
yes, this is exactly what I do
Thanks. And in your new v54 setup, you rerun all three domains with the latest model version, correct? Did you play with the IC_AERO_M2USE and BC_AERO_M2USE options for the three different domains, or only your innermost domain?
yes I ran all of the domains with the new version, and also played around with those options in all of them as well
Thank you for sharing this extended information with us about your simulation. Just as a bit of background on this issue:
In the CMAQv5.3 series, the boundary condition routine assumed that aerosol surface area from the boundaries reflected the dry size distribution. This is true when nesting from larger CMAQ domains, but isn’t always true. So we added the capability to assume the boundary condition aerosol data reflected wet distributions. We also added the M2USE variable so that a default internal diameter could be applied. While we expect both of these options to have an effect, we don’t expect them to be as large as the impact of deposition updates, as Christian pointed out.
(1) Could you give us some quantitative sense of the magnitude of the impacts you’re seeing on PM10 and PM2.5 mass when you toggle the aerosol boundary condition options? Since you are nesting from CMAQ domains, we recommend using the settings marked as default, with the ‘M2USE’ variables set to True and the ‘M2WET’ variables set to False.
(2) Our evaluations indicated substantial differences between the M3DRY deposition option and STAGE. We recommend recompiling and rerunning with the STAGE deposition option to get a sense for the impact of this choice. In our hemispheric CMAQ evaluation comparing v5.4 to v5.3.3, for example, using the M3DRY option led to decreases in PM10 in Europe, while using STAGE led to increases. If you use STAGE, we recommend setting:
Thanks @Ben_Murphy and @hogrefe.christian for your help.
Here are the comparison of using M2USE = T (case1) and M2USE = F (case 2).
mean PM2.5 in case1 would be 5-6% lower compared to that in case2, however mean PM10 in case1 would be 30-35% higher compared to that in case2.
I’ve kept everything else the same between these two cases.