I am currently running the 2016v2 platform as input for CMAQ and am noticing some large differences in emissions and concentrations compared to a previous 2016_beta-CMAQ set up.
I first noticed model grid cells with airports having erroneously high NO emissions but then found (2017 Emissions Modeling Platform - CMASWIKI) which stated airport inventory from the 2017NEI was overestimating emissions at airports. Hopefully using the updated 2017gb emissions will resolve this issue.
However, I am also modeling high PM2.5 concentrations overall, especially near industrial point sources. I have not found documentation suggesting issues with the ptnonipm sector emissions but I am curious if anyone has any suggestions as to what is causing these elevated PM levels.
The 2016v2 platform has a lot of updates since the 2016 beta platform, and the 2016v3 platform has been released since 2016v2. Technical documentation for the platform and links to summaries are here: https://www.epa.gov/air-emissions-modeling/2016v2-platform
There are many emissions data summaries that should help you understand which sectors are causing high PM in certain areas here: Index of /Air/emismod/2016/v2/reports these include county-daily and county-monthly summaries. As you noted, the airport emissions are reduced in 2016v2 compared to 2016 beta. We aren’t sure why you would switch to 2017gb emissions - that shouldn’t be necessary as the reductions are already implemented in 2016v2.
Thank you for you response. After digging into the documentation a bit more I realize I created emissions using the aero 7 module as input into CMAQv5.2, which is using the previous aero 6 module. I believe this is likely the cause for excess PM2.5 in the simulation. Is there documentation available on how I can retrofit emissions using aero7 to work in CMAQv5.2? Thank you.
Using AERO7 emissions in AERO6 would likely cause 30% of your monoterpene emissions to be lost and I would expect you would underestimate PM2.5. So, there are definitely biogenic emission adjustments needed to switch between ae6/7, but it doesn’t sound like that is your major issue.
Here is the difference in the biogenic profiles: The corresponding speciation profile codes are B10C6 and BC6E7, respectively (shortened from B10C6AE7 related to the SMOKE profile code length issue we looked at recently). The main difference is that the AE7 profile outputs acetic acid and formic acid as their own species (AACD/FACD) instead of lumping with other species.
We think that the change in the emissions inputs between the two cases is the more likely cause of the change. Have you compared the total emissions by pollutant and sector to better understand the differences between the emissions inputs for the two cases?
If you are still having issues with overestimation, please provide some maps and quantitative information about where and when the overestimates occurred.
For a bit more quantitative context: I have compared my August county-level emissions with the EPA monthly reports for the 2016v2 platform (https://gaftp.epa.gov/Air/emismod/2016/v2/reports/county_monthly/). I found that my modeled emissions for Lake and Porter Counties for Indiana were within a 2% of the EPA’s reports which is great. However, when I run CMAQ with those simulated emissions, I notice PM2.5 in a grid cell in Porter, IN (near a steel mill) is quite high and averaging concentrations above 180 ug/m3 for the month.
I’ve attached a figure showing the CMAQ monthly (August 2018) average for PM2.5_TOT output on the left for Porter County and the simulated gridded 1.3 km resolution emissions for PTEGU, PTNONIPM, and CMV_C1C2 on the right. PTNONIPM contributes the most PM species to the grid cells in question. I show these emission sectors as they have large contributions to emission in those grid cells. Air quality sensors in the area suggest these PM2.5 values are quite high relative to observed concentrations. Do you have any suggestions on other metrics I could investigate to determine the validity of the simulated emissions or pollutant concentrations? Thank you in advance.
Thank you for providing this additional information, it is helpful context.
We note that there are a couple of large nonEGU facilities in Lake and Porter counties, including the steel mill you mention. Are you saying that they didn’t see a hotspot here in beta platform but are now seeing one in the v2 platform?
The point source emissions in those areas didn’t change much between beta and v2 platforms, but the plume rise methodology changed. In beta platform, when ptnonipm emissions were split into separate 2-D (layer 1) and inline files, half of the emissions from the Porter County steel mill were 2-D and half were inline. In the v2 platform, all emissions from the source are treated as inline and therefore are subject to plume rise above layer 1. This could result in some emissions going into layer 1 and others going above layer 1. It is possible that this is a reason for some differences in result. We also note that emissions changed in many other sectors changed between platforms, but those other sectors’ PM2.5 emissions are nowhere near as large as the nonEGU sources in this area of NW Indiana.
We are wondering whether you can tell if the proportions of the PM components that make up the 180 ug/m3 monthly average changed between the versions.
Thank you for insight into how ptnonipm sources have changed between platforms, Alison.
I have attached a comparison of PM25_TOT components from the max PM grid cell in Porter County (using emissions 2016beta and 2016v2 as input). PM25_SOIL appears to be the largest contribution of PM2.5 in the 2016v2 simulation. PM25_SOIL, being comprised of metal particulates, furthers my suspicions that ptnonipm (steel mills) is contributing to the excess simulated PM.
My question at this point is what is the a path of least resistance to resolve this PM hot spot issue?
Do you think it is a reasonable decision to manually reduce the ptnonipm sector metal particulate emissions in the inln_mole files for these few grid cells? I appreciate all of your help in navigating this question.
We note that the total PM2.5 emissions / stack parameters / coordinates did not change between these platforms, aside from negligible differences. In addition to the plume rise methodology, although it is possible the speciation profiles for the SCCs at the steel mill 7376511 / ARCELORMITTAL BURNS HARBOR LLC changed in addition to the plume rise methodology. However, if the emissions at the steel mill are large, it would have already been in the inln file in the beta platform, so that probably wasn’t the issue.
We don’t have any suggestions at this time for how to change the emissions modeling in this area to reduce the overestimation. Although we may review some information in this area from other platforms.
I have been digging into ptnonipm emissions in Indiana’s Lake and Porter county. From the PM25-PRI FF10 files, I found that there are several facilities with zero stack heights producing a lot of particulate mass. These facilities are co-located with the grid cells simulating high PM2.5. The SCCs suggest the highest PM is coming from paved and unpaved roads associated with primary metal production. Do you think the change in plume methodology for these emissions could be contributing to the high PM2.5 totals? Thank you for all of your help.
There are some large PM sources in those counties. I’ve pasted a picture of a filtered facility summary that is available with the 2016v3 reports, although I’m not sure if you can see it. The columns ending in fj are from 2016v2 and the ones ending in gf are from 2016v3.
These very large PM source could be the reason for the elevated emissions in certain grid cells.
I think we have narrowed down the problem! I have one last question to better understand the ptnonipm PM emissions.
I’ve attached a photo of some of the largest ptnonipm PM25-PRI emitters in Indiana’s Lake and Porter County from the ‘2016v2 2016fj_nonegu_additions_from_2017_2018_17jun2021_nf_v1.csv’ file. The two largest emitters are from a paved and unpaved road. Did the PM from these SCCs at the ARCELORMITTAL BURNS HARBOR LLC facility’s apply a transport fraction adjustment?
We reviewed the SCCs for the high-emitting sources and found that the high emitting SCCs referenced here for Arcelor Mittal are all related to vehicles on paved and unpaved roads at an iron and steel manufacturing facility (303015*). So it sounds correct that the release points for these SCCs should be fugitives (erptype=1).
The 10200704 SCCs are process gas from blast furnaces. At US Steel Gary some of these SCCs are associated with stacks and some are fugitive, and at least some of the processes with the SCC at Gary are fugitive.
We don’t normally apply a transport fraction to point source emissions, but you may decide to do so for your own modeling. You may want to review whether the facilities are in wooded areas or more open areas to determine the amount of the reduction that would be appropriate.
