Scale factors for anthropogenic emissions
This page describes the methodology of the interannual scaling factors used for anthropogenic emissions in GEOS-Chem.
- 1 Overview
- 2 Seasonal Scale Factors
- 3 Comparison of Emission Configurations
- 4 Diurnal Variation
- 5 Source code and data
- 6 References
- 7 Known issues
Philippe Le Sager / Aaron van Donkelaar wrote:
Global annual scaling factors are available for NOx, CO, and SOx from 1985-2006. The approach, as implemented by van Donkelaar et al. (ACP, 2008) builds upon the work of Bey et al. (2001) and Park et al. (2004). Scale factors are automatically applied to any inventory, if needed, to get as close as possible to simulated year conditions. This behavior can be overwritten. Scale factors have been derived from numerous sources and directly use government statistics, where available.
Trend data for the following regions is derived from:
Canada – Environment Canada National Pollutant Release Inventory Trends (http://www.ec.gc.ca/inrp-npri/)
United States – Industrial emissions are scaled according to reported trends from the EPA Acid Rain Program (http://camddataandmaps.epa.gov/gdm/). Other emissions are scaled according to National Emissions Inventory (NEI) Air Pollutant Emissions Trends Data (http://www.epa.gov/ttn/chief/trends/)
Europe – European Monitoring and Evaluation Program (EMEP; UNECE (2009) Present state of emission data, ECE/EB.AIR/97, http://www.emep.int/)
Asia – Regional Emission inventory in Asia (REAS; Ohara et al., ACP, 2007; http://www.jamstec.go.jp/frcgc/research/p3/emission.htm)
Beyond these regions, NOx, SOx and CO are scaled according to CO2 trends available from the Carbon Dioxide Information Analysis Center (http://cdiac.ornl.gov/). NOx, SOx and CO are scaled according to trends in total, solid and liquid CO2, respectively.
We scale all regional and global inventories from their respective base year to 2003, the last year of available statistics, unless its base year is after 2003. Our approach follows Bey et al. (2001) and Park et al. (2004). Emissions are scaled according to estimates provided by individual countries, where available. These countries/regions include the United States, Canada, Japan and Europe. NOx emissions of remaining countries are scaled proportional to changes in total CO2 emissions. SOx emissions are similarly scaled to solid fuel CO2 emissions and CO emissions to liquid fuel CO2 emissions. CO2 emission data are obtained from the Carbon Dioxide Information Analysis Center (CDIAC).
Seasonal Scale Factors
Seasonal scale factors are also applied to many anthropogenic emissions:
Europe – When using EMEP, European monthly variation of NOx, SOx, CO and NH3 is based upon input into the Open Source Unified EMEP model (http://www.emep.int/OpenSource/), that was produced by the GENEMIS project, coordinated by the Institute of Energy Economics and the Rational Use of Energy (IER) at the University of Stuttgart. These source-specific, country-level variations are combined with sector-level gridded EMEP emissions (www.emep.int) to estimate an overall species specific variation.
North America – Regional NA emission monthly variation of NOx is based upon the VISTAS inventory. All other species, with the exception of NH3, are based upon NEI99 seasonality. For NEI2005, NH3 seasonality is based upon Park et al., JGR, 2004. When using VISTAS or NEI2005, a year-varying scalar in additionally included to account for recent ozone regulations. These are based upon reported trends from the EPA Acid Rain Program (http://camddataandmaps.epa.gov/gdm/).
Asia – When using the Streets inventory, monthly variation is present for NOx and CO only. As of v9-01-02, there is also seasonality in Asian NH3 emissions (described here).
Canada – There is currently no seasonal variation in the CAC inventory.
Mexico – There is currently no seasonal variation in the BRAVO inventory.
Global – The seasonal variation of both EDGAR (Olivier and Berdowski, 2001) and GEIA is derived from the seasonality of GEIA.
Comparison of Emission Configurations
The overall impact of the anthropogenic scale factors upon total emissions of NOx, SOx and CO can be greatly affected by the choice of base emission inventory. The following figures below illustrate the possible range of annual equivalent global and North American emissions for 6 different emission setups, based upon GEOS-Chem v8-02-03 run at 4 x 5 degrees. Not all setups affects all species (see table), and lines of equivalent emissions are coloured to match. Note the large impact of switching between Streets 2000 and Streets 2006 from 2000 to 2001 upon global NOx and CO (see discussion below).
An additional factor causing this jump is the inclusion of biofuel CO from Streets 2006 in the fossil fuel CO emission diagnostics (ND29). This inventory does not separate fossil fuel from biofuel emissions and therefore cannot by separated by source. This should not effect CO concentrations, but has impacted the CO emission diagnositic over the Asian region.
Emission within these figures are configured as:
Descriptor Global Inventory CAC NEI VISTAS ICARTT EMEP BRAVO Streets "GEIA" GEIA NO NO NO NO NO NO NO "EDGAR" EDGAR NO NO NO NO NO NO NO "RGN-NEI05" EDGAR YES 2005 NO NO YES YES YES "RGN-NEI99" EDGAR YES 1999 NO NO YES YES YES "RGN-VISTAS" EDGAR YES 1999 YES NO YES YES YES "RGN-ICARTT" EDGAR YES 1999 NO YES YES YES YES
A diurnal variation is applied to all NOx. It is derived from EDGAR hourly variations sc(k,H) for each source k and hour H, spatially weighted by the sources. In other words:
Σ sc(k,H) * NOx(I,J,k) ScaleFactor(H,I,J) = -------------------------- Σ NOx(I,J,k)
Philippe Le Sager wrote:
Note that the scale factors in van Donkelaar et al, 2008] have been updated to go up to 2005, and are based on REAS data now for South East Asia: Relative changes in the REAS inventory (Ohara et al., ACP, ) over East Asia have been used.
This should be a good improvement as REAS emissions are gridded, rather than national scale emissions, giving us much better spatial detail. Also, these scalars are now based on actual NOx, SOx and CO emission estimates, not strictly an assumed proportionality between total, solid and liquid CO2 emissions.
Source code and data
The source code for the anthropogenic scale factors module is scale_anthro_mod.f.
For more information about the data, please see the following README file: GEOS_1x1/anth_scale_factors_200911/README
- Anthropogenic emissions inventories in GEOS-Chem v8-02-03 (and higher versions)
- Anthropogenic emissions inventories used in GEOS-Chem v8-01-04 thru v8-02-02
- Anthropogenic emissions inventories prior to GEOS-Chem v8-01-01
- Bey, I., D. J. Jacob, R. M. Yantosca, J. A. Logan, B. Field, A. M. Fiore, Q. Li, H. Liu, L. J. Mickley, and M. Schultz, Global modeling of tropospheric chemistry with assimilated meteorology: Model description and evaluation, J. Geophys. Res., 106, 23,073–23,096, 2001. PDF
- Ohara, T., H. Akimoto, J. Kurokawa, N. Horii, K. Yamaji, X. Yan, and T. Hayasaka, An Asian emission inventory of anthropogenic emission sources for the period 1980–2020, Atmos. Chem. Phys.,, 7, 4419-4444, 2007. ACP site
- Park, R. J., D. J. Jacob, B. D. Field, R. M. Yantosca, and M. Chin, Natural and transboundary pollution influences on sulfate-nitrate-ammonium aerosols in the United States: implications for policy, J. Geophys. Res., 109, D15204, 10.1029/2003JD004473, 2004. PDF
- van Donkelaar, A., R. V. Martin, W. R. Leaitch, A.M. Macdonald, T. W. Walker, D. G. Streets, Q. Zhang, E. J. Dunlea, J. L. Jimenez, J. E. Dibb, L. G. Huey, R. Weber, and M. O. Andreae, Analysis of Aircraft and Satellite Measurements from the Intercontinental Chemical Transport Experiment (INTEX-B) to Quantify Long-Range Transport of East Asian Sulfur to Canada, Atmos. Chem. Phys., 8, 2999-3014, 2008. PDF
--Bob Y. 09:41, 23 February 2010 (EST)
Apparent differences in anth_scale_factors_200911 data
Claire Carouge wrote:
- I'm benchmarking GEOS-Chem v8-03-01 and I have some questions about the scaling factors for the emissions. You are saying that the new scaling factors are the same than before but were extended to 2006.
- I know there have been changes in Streets and EMEP emissions. So I've ran v8-03-01 and v8-02-04 without Streets and EMEP for July 2005. And I get some small changes in the total anthropogenic emissions:
ANTHROPOGENIC Tracer v8-03-01 v8-02-04 v8-03-01 - v8-02-04 ===================================================== NOx 2.175 2.144 0.031 Tg N CO 26.001 25.756 0.245 Tg ALK4 1.740 1.809 -0.069 Tg C ACET 0.052 0.054 -0.002 Tg C MEK 0.060 0.062 -0.002 Tg C PRPE 0.681 0.670 0.011 Tg C C3H8 1.680 1.721 -0.041 Tg C C2H6 1.142 1.170 -0.028 Tg C SO2 5.416 5.650 -0.233 Tg S SO4 0.118 0.124 -0.006 Tg S NH3 4.862 4.862 0.000 Tg
- So I had a look at the scale factors and they are definitely different, see attached plot for 2005. The "old" scale factors are the ones in anth_scale_factors_200905/ directory, and the new ones are in the anth_scale_factors_200911/ directory.
- Is there any reason for this change? The antarctica scale factor increased from 1 to about 1.5? I know the global change is not that big, but since you said the scale factors should be the same, I'm just wondering why we see these differences.
Aaron van Donkelaar wrote:
- It looks as though all the changes you've noticed relate to countries that rely on the CDIAC CO2 estimates for their interannual change. CDIAC's latest estimates not only extended their temporal range by a year, but also incorporate improved estimates for previous years. As a result, minor changes can occur to all years. This is one of the reasons I changes formats in my updates, to ensure that consistent CO2 estimates were being used.
- As the scalars show a relative change, small absolute changes to small sources can show up as a larger change in scalar (i.e. last Sudan 2005 scalar estimate was 2898 ton /1509 ton = 1.9205, whereas the new Sudan 2005 scalar estimate is 3000 ton / 1509 ton = 1.9881).
- It would surprise me if these changes result in significant changes to the total emissions, but as you noticed, small changes can certainly occur. All that said, it is still true that I have done nothing different with the new scalars, only extended the dataset by a year and incorporated the latest estimates.
--Bob Y. 13:59, 22 April 2010 (EDT)
Large jump in emissions in Asia from 2005 to 2006
Jennifer Logan wrote:
- I've inadvertently discovered what seems to be a problem with the time dependent emissions, that likely relates to when it switches from one base inventory to another. For CO, there is a huge difference between emissions in Asia for 2005 and 2006 - see Junhua's plots below:
- I am guessing that this is because 2005 is scaled forward from the Streets inventory for 2000 (his latest for China I hope, not the original TRACE-P one), while 2006 is taken from the new Streets inventory for INTEX-B.
Jennifer Logan wrote:
- You are right for Streets. There is no backward scaling. So for 2006 and after, we use Streets 2006 and for 2005 and before we use Streets 2000.
- I guess you would prefer to take the data from the closest inventory instead. This looks like relatively easy to implement in the code but there is still the question of the scaling factors.
Aaron van Donkelaar wrote:
- Lok Lamsal (cc'd) also found a jump with the NOx emissions. We indeed traced to back to the REAS inventory, which is based upon projections past 2003 that significantly underestimates regional growth.
- I have already provided updated code to Bob (amongst the modifications for the nested NA and nested EU simulations that will go into GEOS-Chem v8-03-01) which include 2006 scalars and backscale from the 2006 Streets inventory. This, of course, really only addresses the problem for simulations of recent time periods, as there will now be a jump at 2000, but is probably sufficient for most users. Lok has written to the Streets group requesting their estimates of annual changes, but I don't know if he has heard anything back.
Jennifer Logan wrote:
- One more thing. I looked at the Zhang et al. (2009) paper on the INTEX-B inventory. They have also redone their 2001 inventory using the improved methodology of the INTEX work, so we shouldn't even be using the old TRACE-P inventory for 2000, but their revised inventory for 2001. They are different by 27% for NOx according to Table 5 in Zhang et al., 2009.
- How do you backscale from 2006? Which scaling factors do you use? If the REAS ones are not good going forward, they will be no good going backwards.
- Also, we should definitely backscale from 2006 for Asian VOC emissions.
- For pre-2000, it might be best to backscale from the improved 2001 Streets inventory, as least for Asia in the 1990s.
Aaron van Donkelaar wrote:
- I think Yuxuan Wang implemented the Streets inventory, so she may know differently, but I don't think the Streets inventory has been updated to the INTEX-B methodology from the 2009 paper. CO emissions, however, were updated at some point to 2001 to include improved estimates.
- The scalars are the same as before, just extended by a year....The scalars are generally in the right direction, just underestimated, so they should be more accurate than using the original inventory if we are relatively near the base inventory year.
- There are no scalars for VOC emissions...
- If the Street inventory is active, emissions over Asia are presently backscaled from the 2000/2001 emissions for pre-2000.
Lok Lamsal wrote:
- My concern was that emissions trend (for NOx) in Asia is very small (~5% from 2003 to 2006 in China) mainly because of the use of REAS projections. Recently I have obtained the emissions for 2005 from Ohara which indicates a trend of about 9%/year. This suggests that we need to update our implementation in GEOS-Chem. I am going to obtain Streets inventory for 2003 (that uses the INTEX-B methodology) from Qiang Zhang in early March. These are the data we will have to implement in GEOS-Chem.
--Bob Y. 15:12, 17 March 2010 (EDT)
Update for nested model
This fix was mistakenly left out for the nested model at least until v9-01-02. It will be included in a future release, version TBD.
--Jenny Fisher, 31 January 2011
Problem in GEIA scale factors over S. Africa
This update was tested in the 1-month benchmark simulation v9-01-02h and approved on 12 Aug 2011.
--Bob Y. 12:16, 22 February 2012 (EST)