search

CEMPD / SMOKE

Create emissions inputs for multiple air quality modeling systems with unmatched speed and flexibility
https://www.cmascenter.org/smoke/
45 stars 21 forks source link

EGU hourly FF10 processing: HAP calculation error #97

Open callen52 opened 1 month ago

callen52 commented 1 month ago

We have started processing EGU emissions with hourly FF10 inventories based on CEMs, in which pollutants other than NOX and SO2 are based on a new variable in the inventories called HOURACT. With this processing method, HAP emissions are not being calculated properly. I'm attaching a test case (scripts and inputs) to illustrate.

The test case includes two INVTABLEs, invtable_2017_NATA_CMAQ_15jun2023_v9.txt and invtable_2017_NATA_CMAQ_07may2024_nf_v13.txt. The v9 INVTABLE is our regular one, which includes some HAPs which have a factor of less than 1. For example, this HAP is in our ptegu inventory, and the factor 0.4406 is applied to all CAS=7738945 emissions from the annual inventory: CHROMHEX 7738945 Y 0.4406 Y N tons/yr Chromium VI, PM

In the v13 INVTABLE, the factors for all HAPs - including 7738945 - are set to 1.

If the test case is run with the v9 INVTABLE, and then run a second time with the v13 INVTABLE, it is expected that CHROMHEX emissions will be different between the two runs, while all other pollutants should have identical emissions in both runs. However, what I see is that HAPs other than CHROMHEX end up with lower emissions when using the v9 INVTABLE compared to the v13 INVTABLE. My theory is that with this type of hourly processing, SMOKE is getting the factors from the INVTABLE mixed up and applying the wrong factors to the wrong HAPs.

SMOKE_ptegu_hourly_FF10_test_case_for_UNC.zip

hnqtran commented 1 month ago

Could not replicate the reported issue.

Descriptions on conducted test cases:

Case 1: Running Annual_point_onetime_ptegu_12US1_2021hb_cb6_21k_20240520091405.csh using invtable_2017_NATA_CMAQ_15jun2023_v9.txt (v9 INVTABLE)

Case 2: Running Annual_point_onetime_ptegu_12US1_2021hb_cb6_21k_20240520091405.csh using invtable_2017_NATA_CMAQ_07may2024_nf_v13.txt (v13 INVTABLE)

Case 3: Running Annual_point_onetime_ptegu_12US1_2021hb_cb6_21k_20240520091405.csh using a modified invtable_2017_NATA_CMAQ_15jun2023_v9.txt with doubling CHROMHEX CHROMHEX 7738945 Y 2.0 Y N tons/yr Chromium VI, PM (v9x2 INVTABLE)

Case 4: Running Annual_point_onetime_ptegu_12US1_2021hb_cb6_21k_20240520091405.csh using a modified invtable_2017_NATA_CMAQ_15jun2023_v9.txt with reducing CHROMHEX CHROMHEX 7738945 Y 0.0001 Y N tons/yr Chromium VI, PM (v9x0.0001 INVTABLE)

Case 5: Running Daily_point_summer_ptegu_12US1_2021hb_cb6_21k_20240520091645.csh using invtable_2017_NATA_CMAQ_15jun2023_v9.txt (v9 INVTABLE)

Case 6: Running Daily_point_summer_ptegu_12US1_2021hb_cb6_21k_20240520091645.csh using invtable_2017_NATA_CMAQ_07may2024_nf_v13.txt (v13 INVTABLE)

Comparison method: Outputs (pnts*.ncf for annual script; pthour*.ncf for daily script)from smkinven in each case were compared using IOAPI m3diff tool for overall comparison of max, min and mean value. Additionally, the netcdf outputs were dumped into ascii files for line-by-line comparison

Findings:

Case 1 vs Case 2: CHROMHEX emission in Case 1 is smaller than in Case 2 (expected result); NOX and SO2 identical between two cases

Case 1 vs Case 3: CHROMHEX emission in Case 1 is smaller than in Case 3 (expected result); NOX and SO2 identical between two cases

Case 1 vs Case 4: CHROMHEX emission in Case 1 is higher than in Case 4 (expected result); NOX and SO2 identical between two cases

Case 5 vs. Case 6: CHROMHEX emission in Case 5 is smaller than in Case 6 (expected result); NOX and SO2 identical between two cases

I noticed that the run script provided by Christine call emf scripts smk_pt_annual_onetime_emf.csh and smk_pt_daily_emf.csh from smoke4.6. My tests were conducted with scripts from smoke5.0. It is unclear if the difference in version of emf script has any significant in the findings.

callen52 commented 1 month ago

The smoke4.6 scripts are identical to the smoke5.0 scripts, so that should be fine.

When comparing the different cases, did you check pollutants other than CHROMHEX, NOX, and SO2? In my testing, NOX and SO2 are the same (as expected), but there are several HAPs which are different but should be the same, e.g. LEAD, ARSENIC, XYLENES, and ETHYLBENZ.

If you're seeing no differences, I wonder if the issue is somewhere downstream of Smkinven, perhaps in Temporal? If you were to go all the way through Smkmerge and generate model-ready emissions (which is what I had been doing in my testing), how do those compare?

hnqtran commented 1 month ago

I'm seeing LEAD, ARSENIC, XYLENES and ETHYLBENZ identical between v9 and v13. I have not looked pass smkinven since I though HAP processing happens in smkinven. I will try subsequent SMOKE program and update with the results.