PM25 Speciation and 'unspeciated' PM25 in CMAQ

#1

Hi everyone,

I’ve been using the spec_def.conc file to aggregate PM2.5 and the ‘unspeciated’ portion, from CMAQv5.0.2 output. ‘PM25_UNSPEC1’ is defined in terms of PM25_TOT and aggregated PM25 categories (e.g. PM25_SO4, PM25_EC, etc.). It is not defined in terms of actual CMAQ output variables. However, it should still be possible to rearrange the output variables and create an exact ‘formula’ for ‘PM25_UNSPEC1’. The problem that I am running into is that when I do this, (1) I find some double-counting in the existing definitions, and (2) I cannot recreate PM25_UNSPEC1 directly from output variables to match the PM25_UNSPEC1 defined in terms of aggregated PM25 categories.

Is it proper to attempt to do this? If so, then some of PM25_UNSPEC1 is composed of CMAQ output variables which are actual species. Is it correct to assume that these variables aren’t ‘unspeciated’ or ‘other’ and can be used for comparison to speciated data (assuming these species are reported in the data)? This would have the effect of reducing the amount of PM25 reported as ‘unspeciated’ from CMAQ output. Has anyone done this before or created a ‘formula’ for PM25_UNSPEC1?

That is my primary question, but here are specific details about the double-counting in the spec_def.conc file (attached and available on github):

In the PM25_TOT definition, PM25 is the sum of ATOTI, J, and K. In these definitions, all the SOA species are added as OM without any scaling factors. In the calculation of PM_UNSPEC1, the SOA species are subtracted out of PM25_TOT with scaling factors applied. (As part of AOCIJ, which is part of PM25_OC). This seems to mean some >0 fraction of the SOA species are part of ‘PM25_UNSPEC1’, which would inflates the amount of ‘other’ when aggregating PM25 this way.

For the calculation of ‘PM25_UNSPEC1’, ASOILJ is subtracted out of PM25_TOT (as part of PM25_SOIL). However, ASOILJ is defined in the spec_def file and includes multiplicative factors >1 on its constituent species (AALJ, ASIJ, ACAJ, AFEJ, ATIJ). When these species are added to ATOTJ, they have a factor of 1. This seems to ‘double count’ these variables in PM25_SOIL, since they are added with a factor of 1 in the PM25_TOT definition, but added with factors >1 in the PM25_SOIL definition. If that’s the case, this would also artificially decrease the amount of ‘other’ in PM25 when it is aggregated this way. Is this the intended way to calculate PM25 ‘unspeciated’?

I appreciate any feedback to this wordy post!

Thank you,
James East

spec_def.conc.txt (8.7 KB)

0 Likes

#2

Hi James,

I need some more time to wrap my head around what you’re saying, but looking at your post I did have a question.

You state that PM25_TOT is the sum of ATOTI, ATOTJ and ATOTK, and those values are calculated with the SOA added without factors applied, which is true. However, PM25_TOT is calculated with factors applied to ATOTI, ATOTJ and ATOTK. Is your issue then that the factors are being applied to the summed values ATOTI, ATOTJ and ATOTK and not at the individual species level? I’m not sure this would make a difference in the calculation overall. Let’s assume not. We then calculate PM25_UNSPEC1 by taking PM25_TOT (factors applied) and subtract out the other species with factors applied. So, it at least seems we’re being consistent about manipulating values with factors applied.

Perhaps you can clarify where you see the issues with the factors? Maybe I’m just missing something, which is very possible. In the mean time, I’ll look over your post further and see if I can do a better job tracking down the inconsistency you’re running in to.

Wyat

1 Like

#3

Hi James,

Regarding the addition of ASOIL and ASOILJ in several of the equations, there indeed is an inconsistency, which I recall having a conversation about with a couple people here a while ago. I think we agreed we need to fix the issue, but I’m not 100% sure if we agreed what that fix was and if that fix has made its way to our latest spec_def files. Then there is always the question of whether we go back and update the older versions of the spec_def files or not.

Let me talk to a couple people here to check on the status of that issue and how we intend to pursue fixing it. I’ll get back to you on this one sometime next week.

Wyat

0 Likes

#4

Hi Wyat,

Thanks for both replies. I may be missing something as well, so I’ll try to clarify a bit regarding the SOA factors.

I think the issue is the difference between AOCIJ and AOMIJ. AOMIJ is in the sum for PM25_TOT. All species that are summed to create AOMIJ are multiplied by a factor of 1. To calculate PM25_UNSPEC1, those species are subtracted out of PM25_TOT when AOCIJ is subtracted out. In AOCIJ, all species are multiplied by factors <1. So, a fraction of those species remains in ‘PM25_UNSPEC1’. Is this intentional?

For example, the first 3 species in AOMIJ:
AXYL1J+AXYL2J+AXYL3J (No factors)

The first 3 species in AOCIJ:
(AXYL1J+AXYL2J+AXYL3J)/2.0 (Factor of 1/2)

So I think that 50% of the mass of these species would be included in the mass of PM25_UNSPEC1, if that’s correct.

Perhaps I’m misunderstanding and it’s not an inconsistency. If I’m using an outdated spec_def file that could be the problem as well.

Thank you for looking into this.

James

0 Likes

#5

Hi James,

I hope I won’t muddy the waters further, but yes, in my understanding it is intentional that the non-carbon portion of OM in included in the definition of PM25_UNSPEC1.
For the definition of the IJ mode species rather than the sharp cutoff PM2.5 species, you’ll see that two different definitions of “unspeciated” are used, AUNSPEC1IJ and AUNSPEC2IJ. The former is defined analogously to PM25_UNSPEC1 while the latter excluded the non-carbon portion of OM, i.e. the difference between OM and OC, and would be a “lower bound” of unspeciated IJ mass.
It seems that in this version of the model and species definition file, no corresponding definition of PM25_UNSPEC2 was included, but if desired, it should be possible to do that following the approach in the portion of the file that defines the aggregated IJ mode species.

I believe the logic for including the non-carbon portion of OM in the model estimate of “unspeciated” PM2.5 is that you could then compare this estimate against an observation-derived estimate where you subtract the sum of the explicit species that are typically measured (and modeled) (i.e. sulfate, nitrate, ammonium, elemental carbon, organic carbon [measurements typically are for OC, not OM], sodium, chloride, and a few crustal species) from a measurement of total PM2.5.

1 Like

#6

Hi Christian and Wyat,

Thanks – that helps clarify the OM/OC difference in the unspeciated definition. The reasoning for how PM25_UNSPEC1 is helpful as well.

Thanks,
James

0 Likes

#7

Christian/James,

Thanks for your help clarifying Christian. Seems like your response was more helpful than mine, which is was. The issue I thought we had with the SOIL calculation isn’t an error, it’s more to do with how we want to compare CMAQ output to observations of trace elements, and using the soil equation from IMPROVE probably isn’t the best way to do that. However, there’s no error present in the definition of those values.

Wyat

0 Likes

#8

Hi Wyat,

So, just to clarify: using ‘ASOILJ’ to define PM25_UNSPEC1 but not in the definition of PM25_TOTAL is correct? Does this depend on measurement we are using for comparison?

Thanks,
James

0 Likes

#9

Hi James,

This is the apparent inconsistency I had hinted at earlier. The formulation for PM25_UNSPEC1 is consistent with the formula for AUNSPEC1IJ. After a long discussion with Christian, we concluded that ASOILJ was likely used in place of the individual cation species to be consistent with the calculation of measured unspeciated mass (perhaps this was necessary in the past, I don’t recall).

Given that both the IMPROVE and CSN networks currently measure all the individual cation species, Christian and I agreed that it would be simpler going forward to just use the individual cation species instead of the IMPROVE soil formula, which just complicates the comparisons among observed and modeled species. We can maintain the IMPROVE formula in the species definition file for those that might use it, but probably rename it to something more descriptive and less confusing in name with CMAQ’s SOIL species. We’ll look to update future species definition files accordingly.

Let me know if that helps at all or if you have any other questions.

Wyat

0 Likes

#10

Hi Wyat,

Thanks for the follow up. This has all been helpful.

James

0 Likes