CMAQv5.3 ISAM - difference with reference data

CMAQ-ISAM
1

Hi Dr. Napelenok et al.,

I enjoyed the 10/8/2019 CMAQv5.3 webinar, Dr. Napelenok mentioned “sum of target species concentrations in ISAM should be very close to the concentration of this species from basic CMAQ output”. I then tested CMAQv5.3 ISAM with benchmark

TAG CLASSES     |SULFATE, OZONE

!!! The following are source definition text blocks in the format:
!!!   TAG NAME        |Three character text string
!!!   REGION(S)       |Keyword EVERYWHERE or variable names from the region file (multiple regions need to be comma delimited)
!!!   FILENAME(S)     |Emissions labels (multiple labels need to be comma delimited)


TAG NAME        |GR1
REGION(S)       |EVERYWHERE
EMIS STREAM(S)  |GRIDDED_EMIS

ENDLIST eof

I then summed up O3_OTHER O3_ICON O3_BCON O3_GR1 and compared the data with O3 from original CMAQv5.3 output. scatter plot and domain average (total_O3_ISAM - O3_orig) look fine
scatter
bar difference

However, I saw some large differences in certain grid-cells
absolute difference (up to 16 ppb in some grids)

percent difference (up to 267%)

Are these large differences normal for CMAQv5.3 ISAM? BTW, I’ve checked my benchmark run with reference data without ISAM, and even in the two-weeks simulation the difference < 0.5%.

Thank you

Joey

2

Hi Joey,

The CMAQ benchmark case comes with ISAM output as well. Do your SA_ACONC files compare well with what is distributed with the release?

Sergey

3

Hi Sergey,

The absolute differences are not significant

NCDAQ_reference_diff
NCDAQ_reference_diff1312×773 493 KB

NCDAQ_reference_diff
NCDAQ_reference_diff1312×770 36.2 KB

but if we look into percent differences, the numbers are huge in some areas, I believe these huge differences are due to small number in denominator

NCDAQ_reference_diffpercent
NCDAQ_reference_diffpercent1312×773 545 KB

Let me know if you have any thoughts,

Cheers,

Joey

4

Hi Joey,

There is the expectation that the sum of the tags will diverge from the bulk concentration even for nonreactive species due to the non-linear nature of the transport processes in the model. Blowing around pieces of a thing should give you different results from blowing around the entire thing. The main concern would be if the error accumulates and increases in time. I don’t really see that in your case.

However, we are rewriting specifically the ozone apportionment algorithms and will release an updated version.