EPA/NEI05 North American emissions
Philippe Le Sager / Aaron van Donkelaar wrote:
The EPA/NEI2005 data were processed by Aaron van Donkelaar onto 1 x 1 and 1/2 x 2/3 degree grids, with 5 vertical levels. Original gridded data on a 4 km x 4km grid as provided as input for WRFCHEM (available from ftp://aftp.fsl.noaa.gov/divisions/taq/emissions_data_2005). Four (4) related masks have been created by Philippe Le Sager. These emissions were first implemented into GEOS-Chem v8-02-04.
Others should feel free to add to this description ...
The source code for the EPA/NEI2005 emissions is located in module nei2005_anthro_mod.f.
For more information about the data, please see the following README file: GEOS_1x1/NEI2005_200910/readme.txt
--Bob Y. 13:34, 24 February 2010 (EST)
Original validation by Philippe Le Sager
Philippe Le Sager wrote:
- I have made the modifications in the current code to deal with NEI2005, based on the preliminary work of Aaron. I had to rewrite the PBL mixing of the NOx/SOx/NH3 emissions, since these emissions are given distributed into 5 model levels. I had to create several masks for correct overlaps, since the inventory covers both Mexico and Canada.
- I made a one month run and simply run the benchmark comparison with the last version of the code (v8-02-03, which used the older EPA/NEI99 emissions).
- The main difference is in CO (note that the reference v8-02-03 uses Rynda correction to EPA99 for CO and NOx). Here are the emissions totals that change:
ANTHROPOGENIC Tracer nei05 v8-02-03 nei05 - v8-02-03 ===================================================== NOx 1.957 1.927 0.029 Tg N CO 31.460 28.455 3.005 Tg ALK4 1.392 1.966 -0.573 Tg C ACET 0.054 0.065 -0.012 Tg C MEK 0.045 0.047 -0.002 Tg C PRPE 0.591 0.667 -0.076 Tg C C3H8 1.720 1.853 -0.133 Tg C C2H6 1.054 1.142 -0.087 Tg C SO2 4.280 4.323 -0.043 Tg S SO4 0.090 0.096 -0.006 Tg S BIOFUELS Tracer nei05 v8-02-03 nei05 - v8-02-03 ===================================================== NOx 0.188 0.186 0.002 Tg N CO 14.707 14.547 0.160 Tg ALK4 0.064 0.063 0.001 Tg C ACET 0.021 0.021 0.000 Tg C MEK 0.111 0.110 0.001 Tg C ALD2 0.055 0.054 0.001 Tg C PRPE 0.531 0.525 0.006 Tg C C3H8 0.078 0.077 0.001 Tg C CH2O 0.101 0.100 0.001 Tg C2H6 0.173 0.171 0.002 Tg C ACETONE SINKS Tracer nei05 v8-02-03 nei05 - v8-02-03 ===================================================== ACETol 0.902 0.913 -0.011 Tg C CO SOURCES Tracer nei05 v8-02-03 nei05 - v8-02-03 ===================================================== COanth 31.460 28.455 3.005 Tg CObf 14.707 14.547 0.160 Tg NOx SOURCES Tracer nei05 v8-02-03 nei05 - v8-02-03 ===================================================== NOx-an 1.957 1.927 0.029 Tg N NOx-bf 0.188 0.186 0.002 Tg N SO2 SOURCES Tracer nei05 v8-02-03 nei05 - v8-02-03 ===================================================== SO2-an 4.280 4.323 -0.043 Tg S SO4 SOURCES Tracer nei05 v8-02-03 nei05 - v8-02-03 ===================================================== SO4-an 0.090 0.096 -0.006 Tg S
- I put all the plots into PDF files. I think this is fine and we can go forward.
Fix for seasonal scaling factors
NOTE: The code used by Philippe Le Sager to create the emission totals and plots listed above had a minor error. The seasonality imposed on the EPI/NEI05 emissions species (see file nei05_anthro_mod.f) was an annual average instead. However, due to the way the EPA/NEI05 emissions were constructed, the emsissions species needed to be scaled to "ozone season" (i.e. August). This discrepancy has now been fixed in the GEOS-Chem v8-02-04 mainline source code. Please follow the email trail below for more information:
Dylan Millet wrote:
- The NEI documentation (e.g. ftp://aftp.fsl.noaa.gov/divisions/taq/emissions_data_2005/Weekday_emissions/readme.txt) states that emissions are provided for a "typical summer / August" conditions. Saturday, Sunday, and Weekday files are provided, which I assume Aaron has weighted and averaged to generate the bpch files being read into GEOS-Chem v8-02-04.
- In the nei2005_anthro_mod.f that Philippe gave me, NEI-99 seasonal scaling is applied to the non-NOx tracers, and for NOx the VISTAS seasonal scaling and then an O3-season scaling is applied according to the ARP files in the VISTAS directory. I guess this is b/c the VISTAS seasonality does not account for recent summertime reductions?
- So as I understand it the seasonality will be wrong for NOx and non-NOx tracers as well.
- How about leaving the input files alone and making the following simple fix in the nei2005 module:
- For non-NOx tracers, the seasonality factor is changed to (NEI99 current month) / (NEI99 august) instead of (NEI99 current month) / (NEI99 annual average)
- For NOx, first apply an ozone season scaling (ARP current month / ARP august) and then a seasonal scaling of (VISTAS current month / VISTAS august).
- P.S. One other point:
- "The area inventory does not include wildfire or prescribed forest or rangeland burns. It does, however, include non-forest agricultural fire estimates from each state, which are usually at a minimum during the summer months."
- So there may be a bit of double counting when we combine this with a biomass burning inventory. A small effect?
Aaron Van Donkelaar replied:
- Thanks for catching this (and for offering to help correct it).
- The data files as I received them contain only weekday emissions, so there is no weighted average being applied. nei2005_mod.f applies the weekend/weekday ratios from nei99.
- The VISTAS inventory was specific to 2002, so it includes some reduction associated with ozone season, but is prior to the more full reductions presently employed. To that end, during the VISTAS implementation, I created scalars that accounted for changes in NOx emissions relative to the 2002 levels for each month and year (these are the ARP scalars).
- The scalings you suggest for both NOx and non-NOx species sound appropriate.
Bob Yantosca replied:
- Here are the results of emissions-only runs, with and without the NEI fix that Dylan & Aaron sent:
ANTHROPOGENIC Tracer with fix w/o fix with fix - w/o fix ===================================================== NOx 23.630 23.414 0.215 Tg N CO 371.149 373.066 -1.917 Tg ALK4 16.688 16.705 -0.018 Tg C ACET 0.631 0.631 -0.000 Tg C MEK 0.532 0.532 -0.000 Tg C PRPE 7.058 7.063 -0.004 Tg C C3H8 20.544 20.545 -0.001 Tg C C2H6 12.414 12.416 -0.002 Tg C
- See this PDF file for plots. I think the fix is in the good direction, NOx goes up, and CO & VOC's decrease slightly (not a huge effect).
Dylan Millet replied:
- Looks reasonable to me. A fairly big effect for NOx over the Ohio River Valley, but less important elsewhere and for other tracers.
Aaron Van Donkelaar replied:
- I think one of the larger complications is that we don't have NEI99 seasonality of NH3, which may explain some of the excessively high nitrate values I've been seeing in winter. This will need to be addressed at some point. I've attached an investigation I came across into the seasonality of these emissions (pg. 15, Inverse modeling to estimate seasonal NH3 emission estimates, Gilliland et al, CMAQ Model Peer Review Meeting, 17 Dec 2003. (Click here for PDF file.)
Lin Zhang replied:
- Thanks for the suggestion. I was looking through the old NH3 emission seasonality in the model, which was compared well with Gilliland's estimates according to Park et al. 2004. I can get the scalar numbers for each month, and with these numbers the annual US NH3 emissions are 2.3 Tg yr-1, compared with 4.3 Tg yr-1 without the scaling.
- So based on your modification, So we may change within GET_NEI99_SEASON to
ELSEIF ( TRACER .eq. IDTNH3 ) THEN NH3_SCALE = (/0.226d0, 0.236d0, & 0.279d0, 0.381d0, 0.626d0, 0.768d0, 1.007d0, 1.d0, 0.837d0, & 0.470d0, 0.303d0, 0.236d0/) AS(:,:,:) = 1.d0 * NH3_SCALE(ThisMN) RETURN ENDIF
Aaron Van Donkelaar replied:
- That generally seems like a good approach to use. I've made two modifications:
- The scalars should be adjusted relative to August, so I divide the scalar by the August value.
- Directly using the scalars from the NEI99 implimentation assumes that we have the same bias (i.e. 4.3 Tg/yr should be 2.3 Tg/yr). I'm not prepared to make such an adjustment yet. NEI2005 collection methods have advanced since the NEI99 inventory, so hopefully there is not an inherant bias in the emissions. To that end, I've normalized the scalars such that the total annual adjustment is 1.
- These changes play out to the following (as applied in the attached):
ELSEIF ( TRACER .eq. IDTNH3 ) THEN NH3_SCALE = (/0.426d0, 0.445d0, & 0.526d0, 0.718d0, 1.179d0, 1.447d0, 1.897d0, 1.884d0, 1.577d0, & 0.886d0, 0.571d0, 0.445d0/) AS(:,:,:) = NH3_SCALE(ThisMN) / NH3_SCALE(8) RETURN ENDIF
A 1-month (July 2005) emissions run with the seasonal scaling fixed as described above yields these totals:
ANTHROPOGENIC Tracer with fix v8-02-03 with fix - v8-02-03 ===================================================== NOx 1.929 1.927 0.002 Tg N CO 31.095 28.455 2.641 Tg ALK4 1.401 1.966 -0.565 Tg C ACET 0.054 0.065 -0.012 Tg C MEK 0.045 0.047 -0.002 Tg C PRPE 0.592 0.667 -0.075 Tg C C3H8 1.721 1.853 -0.132 Tg C C2H6 1.054 1.142 -0.087 Tg C SO2 4.250 4.323 -0.073 Tg S SO4 0.090 0.096 -0.007 Tg S NH3 4.040 4.027 0.013 Tg BIOFUELS Tracer with fix v8-02-03 with fix - v8-02-03 ===================================================== NOx 0.188 0.186 0.002 Tg N CO 14.707 14.547 0.160 Tg ALK4 0.064 0.063 0.001 Tg C ACET 0.021 0.021 0.000 Tg C MEK 0.111 0.110 0.001 Tg C ALD2 0.055 0.054 0.001 Tg C PRPE 0.531 0.525 0.006 Tg C C3H8 0.078 0.077 0.001 Tg C CH2O 0.101 0.100 0.001 Tg C2H6 0.173 0.171 0.002 Tg C SO2 0.023 0.023 0.000 Tg S NH3 0.160 0.160 0.000 Tg CO SOURCES Tracer with fix v8-02-03 with fix - v8-02-03 ===================================================== COanth 31.095 28.455 2.641 Tg CObf 14.707 14.547 0.160 Tg NOx SOURCES Tracer with fix v8-02-03 with fix - v8-02-03 ===================================================== NOx-an 1.929 1.927 0.002 Tg N NOx-bf 0.188 0.186 0.002 Tg N SO2 SOURCES Tracer with fix v8-02-03 with fix - v8-02-03 ===================================================== SO2-an 4.250 4.323 -0.073 Tg S SO4 SOURCES Tracer with fix v8-02-03 with fix - v8-02-03 ===================================================== SO4-an 0.090 0.096 -0.007 Tg S
- Totals for GEOS-Chem v8-02-03 were taken from the 1-month benchmark output. This version used the older EPA/NEI99 emissions.
- Note that the differences in NOx is much smaller than without the seasonality fix because we are scaling to the August "ozone season" values. The difference between July and August is smaller than the difference between July and the annual average.
--Bob Y. 09:53, 17 December 2009 (EST)
NEED CITATIONS PLEASE!