Grantee Research Project Results
2011 Progress Report: Improvements in Emissions Inventories using Semi-Continuous Monitoring Data and Concentrations Field Analysis
EPA Grant Number: R834557Title: Improvements in Emissions Inventories using Semi-Continuous Monitoring Data and Concentrations Field Analysis
Investigators: Schauer, James J. , Turner, Jay R. , deFoy, Benjamin
Institution: University of Wisconsin - Madison , Washington University , Saint Louis University - Main Campus
Current Institution: University of Wisconsin - Madison , Saint Louis University - Main Campus , Washington University
EPA Project Officer: Chung, Serena
Project Period: June 1, 2010 through May 30, 2013 (Extended to May 30, 2014)
Project Period Covered by this Report: June 1, 2011 through May 30,2012
Project Amount: $499,777
RFA: Novel Approaches to Improving Air Pollution Emissions Information (2009) RFA Text | Recipients Lists
Research Category: Air Quality and Air Toxics , Air
Objective:
This project focuses using yearlong datasets from St. Louis, Milwaukee and Los Angeles to map sources of air pollutants that do not have well-developed emissions inventories. The modeling efforts are directed at improving emissions inventory data for black carbon, ultrafine particle number concentrations and fine particle organic carbon using data from the EPA-funded St. Louis Supersite; speciated mercury compounds in the Milwaukee region using data from an EPA STAR Project; and fine particle carbonaceous particulate matter and associated precursor gases in Los Angeles. Concentration Field Analysis (CFA) and related mapping tools are being used to map emissions sources and identify unknown or poorly identified source regions using stochastic backward and forward particle trajectories with a temporal resolution finer than 1 hour and spatial resolution finer than 5 km during the yearlong study periods. The integration of high-quality monitoring data with multiple 3D modeling approaches is being used to assess existing emissions inventories and improve the understanding and representation of the temporal distribution of emissions, spatial distributions of emissions, missing sources, and inaccurate emissions estimates for point sources, mobile sources and area sources.
The goal of this project is to couple high-resolution meteorological modeling with existing high time resolution atmospheric pollutant datasets to assess and improve emissions inventories.
CFA is being used to identify probabilistic source regions from the measurements independent of the emissions inventory data. Forward Lagrangian modeling will then be used to evaluate individual transport events. Cluster analysis will link yearlong trends with the hour-long episodes to assess the statistical relevance of the conclusions. Uncertainties due to the simulation of vertical dispersion will be constrained by comparing measurements and particle transport with forward Eulerian models.
Progress Summary:
Efforts in Year 2 of the project have been directed as the analysis of data from all three study locations. The first paper covering the mapping of source of elemental mercury is current in review and a second mercury paper is in preparation. Likewise, papers from the LA analysis and the St. Louis analysis should be submitted for publication in the early part of Year 3. A summary of the three analysis efforts are summarized below:
Inverse modeling of Gaseous Elemental Mercury emissions using Milwaukee data
For Milwaukee, we are analyzing measurements of Mercury from April 2003 to March 2004 at Devil’s Lake State Park and from July 2004 to May 2005 in Milwaukee. We have performed nested mesoscale meteorological simulations using the Weather Research & Forecast model (WRF) for the entire time periods. We are using the North American Regional Reanalysis (NARR) project as meteorological boundary and initial conditions for the simulations. Sensitivity tests were performed of the model parameters for a more recent winter and a summer episode. The simulations used three nested domains with grid resolution of 27, 9 and 3 km. Backward trajectories were calculated for the two measurement sites for every hour of the measurements using WRF-FLEXPART, and were converted to gridded fields of “Residence Time Analysis.”
Concentration Field Analysis was carried out for the measurements of gaseous elemental mercury, reactive mercury and particulate mercury for both sites, revealing transport patterns and possible source regions. Two weaknesses of the CFA method were encountered: (1) CFA does not account for time varying emissions, and (2) CFA does not differentiate sources along radial lines when the winds tend to be uniform. We have therefore used forward Eulerian modeling to deal with biomass burning and are working on implementing a more sophisticated inversion method to evaluate the emission strengths.
The mercury time series is clearly impacted by episodic biomass burning events. We have obtained biomass burning emission inventories based on MODIS fire detection for North America from Christine Wiedinmyer of NCAR. We used these to simulate mercury impacts at the measurement sites using CAMx, a 3D Eulerian grid model that uses the same meteorological simulations as FLEXPART. This shows that some of the mercury peaks detected are clearly due to biomass burning episodes.
A new nested inverse model is being developed, which can handle the backward particle trajectories and the forward biomass burning time series at the same time. This is based on a least squares inversion of a matrix that includes the residence time analysis fields and the CAMx time series. Ideally, the finer the grid used for the residence time analysis the higher the resolution of the results will be. However, the more grid points there are, the greater the degrees of freedom in the inversion method. This is constrained by the number of data points available: if there are too many degrees of freedom, then the algorithm would be able to obtain an artificial fit that is not physically meaningful. We are therefore using polar grids with fine resolution close to the site and exponentially increasing grid spacing along the radials. This is providing stable results. We are testing the method with pseudo-data obtained from forward model simulations as well as with sulfur dioxide concentrations from the measurement sites.
The purpose of the numerical analysis is to evaluate existing emission inventories. We have therefore analyzed the National Emission Inventory (NEI), the Toxic Release Inventory (TRI) and the Wisconsin state emission inventory. The TRI is released annually and has significant evolutions in mercury emissions. For the NEI, we analyzed the 2002 inventory and are now looking at changes with the latest inventory (2008). Results of the inverse method are being compared with these. In addition to source strengths, the inverse method provides an estimate of the impacts at the site due to different sources. This provides useful additional information on which sources are the most important when considering Milwaukee air quality.
The parallel analysis of reactive mercury measurements performed in Milwaukee and a removal site in central Wisconsin are now under way.
Concentration Field Analysis for Los Angeles using WRF-Flexpart simulations
As part of a California Air Resource Board funded project, 24-hour average carbonaceous aerosols and molecular markers were measured every day for a full year. These measurements were used in a number of source apportionment models to refine the understanding of the primary and secondary source of OC and EC. The daily source apportionment results were used in a preliminary inverse mapping to understand source regions of the primary and secondary sources of carbonaceous aerosol. As part of the current project, a more detailed meteorological analysis is being used to map the sources quantified by the source apportionment model. In addition, the mapping is integrating measurements of gas-phase volatile organic compounds (VOCs) that are expected to the precursors of the anthropogenic and biogenic secondary organic aerosol sources quantified by the model. Efforts are currently focusing on secondary organic aerosol, mobile source and biomass source mapping.
Inverse modeling of Elemental and Organic Carbon at the St. Louis Supersite
The goal of this analysis is to better understand the emissions of elemental and organic carbon aerosols using measurements made at the East St. Louis Supersite. We apply an inverse model using a combination of backward particle trajectories and forward Eulerian simulations. The results are compared with the products of simpler analysis methods to check for consistency. Hourly measurements were obtained from April 2001 to July 2003. Meteorological data were available from instruments at the site itself as well as from surrounding airports.
A windrose analysis was performed to identify dominant wind patterns at the surface and wind directions and diurnal patterns associated with high concentrations. This showed that high concentrations of carbonaceous particles were clearly associated with calm winds and with winds from the south-southeast. The diurnal profile showed lower concentrations during the day with maxima at night.
A time series analysis was carried out using a Least-Squares Linear Model on all the data available, including data from the EPA’s Air Quality System (AQS) network. This showed that elemental carbon was strongly associated with black carbon only, with an R2 coefficient for the fit of 0.87. In contrast, the black carbon time series was associated with elemental carbon, carbon monoxide, organic carbon, PM2.5 and nitric oxide in order of decreasing importance. Organic carbon was associated with black carbon, PM2.5, carbon monoxide, nitrogen dioxide, ozone, elemental carbon and nitric oxide. This suggests that whereas black carbon and elemental carbon do follow each other, black carbon is influenced by other pollutants. Organic carbon shares some sources with elemental carbon but appears to be more influenced by mobile sources (CO and NO2) as well as more regional impacts (ozone).
Meteorological simulations were performed on nested grids starting with a 27 km resolution down to 3 km resolution using the Weather Research & Forecast (WRF) model. This was run for 18 months in simulations of 6 days initialized with the North American Regional Reanalysis (NARR) wind fields. Wind roses and diurnal profiles of the meteorological simulations were comparable to those from the airport meteorological observations.
The high resolution winds were used to simulate particle back-trajectories from the measurement site for each hour of the time series using WRF-FLEXPART. In the first configuration, 100 stochastic particles were released every hour. Positions were stored every hour over a period of 2 days. In the second configuration, 1,000 particles were released per hour and were tracked for 6 days. The particle trajectories were mapped onto grids corresponding to the WRF grids to perform a Residence Time Analysis. This showed that there were preferred transport directions from the south and from the north along the river valley directions. This is in contrast to the preferred transport directions aloft, which are very clearly from the west as would be expected from the direction of the prevailing winds.
Concentration Field Analysis was obtained using the grids from the Residence Time Analysis by multiplying the grid from each hour with the concentration measured at that hour. This shows the regions whose air masses are associated with high levels of pollutants, and are therefore potential source regions of those pollutants. For the St. Louis supersite, this shows a signal from the St. Louis metropolitan region. Even stronger than this however is a signal from the southeast. Current work suggests that this might be local sources, for example, a large rail yard that is less than 1 mile away. However, Concentration Field Analysis is much better at resolving the direction of source areas than the distance. For this reason, the inverse model will be used to provide a better way of identifying different source regions and source types.
The inverse model developed for the mercury data in Milwaukee was applied to the St. Louis supersite. The main sources of elemental carbon were as expected in the vicinity of the measurement site, from the southeast and southwest quadrant. From the preliminary analysis of the data, we could clearly see that the site was influenced by calm winds. On the one hand, these are particularly challenging for a numerical simulation, and on the other they are indicative of local sources. We therefore performed forward simulations of local mobile and point sources using CAMx, an Eulerian grid model. We used the emissions inventory for NOx as a proxy for aerosols. Although this has the disadvantage of not being a perfect match for different aerosol sources, it has the advantage of being a mature and well-understood inventory. Forward simulations were performed for different periods of the day so that the model would determine the diurnal profile as an output of the calculation rather than as an a priori. We also performed forward Eulerian simulations of forest fires for the entire United States split by regions to identify emission ratios as well as the contribution of these sources to the pollutant loading in an urban area.
The inverse model therefore solved for different types of sources: local mobile and point sources using the NEI NOx inventory, national forest fire sources using the FINN model and local to regional sources using grids of back-trajectories. This work is currently ongoing. Preliminary conclusions suggest that local sources are the main influence at the supersite but that we can identify the forest fire signal and hence determine emission factors from the data.
For St. Louis, we are progressing along similar lines. WRF simulations were carried out for 1 year of measurements, and FLEXPART trajectories are being tested. We are evaluating the meteorological simulations by comparing specific episodes with observations and analyzing the behavior of the trajectories. We are currently in the process of analyzing the aerosol measurements to begin the Concentration Field Analysis.
The mapping of EC, OC and particle number sources is under way and will continue into Year 3 of the project.
Explanation of Expenditures: There were no major changes in Year 2 expenditures from those laid out in the budget submitted with the original proposal.
Journal Articles on this Report : 1 Displayed | Download in RIS Format
Other project views: | All 17 publications | 15 publications in selected types | All 15 journal articles |
---|
Type | Citation | ||
---|---|---|---|
|
de Foy B, Wiedinmyer C, Schauer JJ. Estimation of mercury emissions from forest fires, lakes, regional and local sources using measurements in Milwaukee and an inverse method. Atmospheric Chemistry and Physics 2012;12(19):8993-9011. |
R834557 (2011) R834557 (2012) R834557 (Final) |
Exit Exit Exit |
Supplemental Keywords:
National Emissions Inventory, Toxic Release InventoryProgress and Final Reports:
Original AbstractThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.