Grantee Research Project Results
2008 Progress Report: Evaluation of Regional Scale Receptor Modeling
EPA Grant Number: R832156Title: Evaluation of Regional Scale Receptor Modeling
Investigators: Lowenthal, Douglas H. , Chen, Lung-Wen Antony , Watson, John L. , Koracin, Darko
Institution: Desert Research Institute
EPA Project Officer: Chung, Serena
Project Period: January 1, 2005 through December 31, 2007 (Extended to December 31, 2009)
Project Period Covered by this Report: January 1, 2008 through December 31,2008
Project Amount: $436,687
RFA: Source Apportionment of Particulate Matter (2004) RFA Text | Recipients Lists
Research Category: Air Quality and Air Toxics , Particulate Matter , Air
Objective:
Evaluate multivariate and trajectory-based receptor models for regional source apportionment relevant to the U.S. Environmental Protection Agency Regional Haze Rule. Document models currently in use, including classical factor analysis, Positive Matrix Factorization (PMF), UNMIX, and Trajectory Mass Balance Regression (TMBR). Review previous model applications and critically evaluate the results. Apply receptor models to synthetic data generated with an air quality model for two eastern IMPROVE sites: Brigantine National Wildlife Refuge (BRIG), NJ, and Great Smoky Mountains National Park (GRSM), TN. Document the approach required to reproduce the known regional contributions to sulfate. Perform a “blind” test on a second simulated data set using this guidance. Apply models to real-word IMPROVE data at these sites. Finalize guidance for systematic application and validation of these models in future regional-scale applications.
Progress Summary:
Positive Matrix Factorization (PMF)
The CMAQ-MM5 modeling system was used to generate synthetic IMPROVE-style data sets at Brigantine National Wildlife Refuge (BRIG), TN, and Great Smoky Mountains National Park (GRSM), TN, during summer (July-September), 2002. Chemically speciated PM2.5 concentrations data were created by associating 43 source profiles with 101 PM source categories, which accounted for at least 95 percent of criteria pollutant emissions in the model domain. CMAQ estimated contributions of haze-producing sulfate aerosols from seven regions representing Regional Planning Organizations and subdivisions thereof: 1) MANE-VU; 2) VISTAS-EAST (E); 3) VISTAS-WEST (W); 4) VISTAS-SOUTH (S); 5) MIDWEST; 6) CENRAP-NORTH (N); and 7) CENRAP-SOUTH (S) using a partial “emissions-in” and “emissions-out” approach. The modeled (“true”) average regional sulfate contributions at BRIG and GRSM are shown in Figure 4 along with regional SO2 emissions from the NEI inventory. CMAQ assigned the largest contributions of ambient PM2.5 sulfate to the local regions at BRIG and GRSM, i.e., MANE-VU (52%) and VISTAS-W (31%), respectively. With respect to receptor modeling, the primary result was that PMF was unable to distinguish regional sources of sulfate (Lowenthal et al., 2008). However, at both sites, PMF identified a single “sulfate factor,” which was not correlated with any of the sources of primary PM2.5 used to construct the data sets.
Figure 1. Modeled contributions to the average sulfate concentration at BRIG and GRSM and annual regional SO2 emissions from the 2002 NEI.
The remaining PMF factors represented individual source types or mixtures of sources whose source profiles were used to generate the data. The identified sources were similar to those reported in previous PMF studies at the two sites: soil, industrial activity, mobile, coal power plant, oil power plant, and residential wood combustion. Regional sources were not resolved because their regional source profiles, i.e., ratios of their respective chemical components to primary PM2.5, were much more similar to one another than were the emissions source profiles used to construct the data. The divergence between the true and PMF factor contributions to sulfate at BRIG and GRSM is shown clearly in Table 1.
Given the inability of the PMF model to address regional-scale source apportionment with IMPROVE-style data, a comprehensive analysis was conducted to quantitatively assess PMF performance as applied to synthetic data (Chen et al., 2009). Due to the complexity of contributing sources and meteorology, factors derived from receptor models such as PMF in an ideal case will represent pure sources or unique combinations of sources rather than individual sources. A diagnostic index, D2, was developed to evaluate the ability of PMF to estimate factors that represent sources with known primary PM2.5 contributions, as is the case with our synthetic data. A value of D2 = 0 signifies that PMF creates unique groupings of sources into factors, i.e., a particular source is associated with only one factor although other sources may also be associated with that factor. This level of uniqueness was not achieved with the synthetic data; D2 was always greater than 0. However, D2 was minimized by adjusting number of factors, the number of observations, factor rotation with the FPEAK parameter in PMF, and uncertainties that are used as weighting factors in the model. Increasing the number of factors (K) does not necessarily result in added source resolution. Considering the minimization of D2, the best choice for the number of factors occurs when R2 is approximately equal to 0.95. Increasing K further does not produce more meaningful factors that correspond to actual sources.
Table 1. Comparison of “true” regional sulfate contributions with PMF sulfate apportionments. Percentages are sorted in descending order.
The effect of measurement uncertainties on PMF results was simulated by random-normally perturbing the simulated concentrations to different degrees. This evaluation confirmed that the mean weighted c2 could attain values near unity (the expected value) with the proper choice of species uncertainty and K. These perturbations changed the resulting factor profiles and contributions but did not always degrade their relationships with sources. Even with accurate formulation of uncertainty and the optimal K, the PMF results became more unstable as the uncertainty increased. The effects of higher random measurement uncertainties on the PMF solution can be reduced by increasing the number of observations (N) for the same reason that the standard error of the mean decreases as the square root of N.
Trajectory Mass Balance Regression
By contrast, the TMBR analysis proved a more useful approach for resolving regional source contributions to sulfate. The most recent TMBR analysis (Lowenthal et al., 2008) contrasted regional apportionments with different meteorological inputs (MM5, EDAS) to HYSPLIT trajectories for different sample averaging times (6 and 24 hours). Table 2 summarizes the overall fits (AAE (%) = the average absolute difference between true and predicted regional sulfate contributions with respect to the average true contribution) with different trajectory starting elevations. The AAE varied with trajectory starting elevation and meteorological data source. Comparisons based on 24-hr samples were nearly always worse than those based on 6-hr samples. This is a significant finding since IMPROVE sample durations are 24 hours. Comparisons based on integrated (100-3000 m) trajectories were also worse than results based on discrete starting elevations. The AAE considering all regions was ≤ 27 percent at BRIG for starting elevations between 100 and 500 m using EDAS data. AAEs at BRIG using MM5 data ranged from 35-46 percent over the six starting elevations. At GRSM, AAEs ranged from 28-41 percent and 22-41 percent over the six starting elevations using EDAS and MM5 data, respectively. Considering the three strongest sulfate sources at each location improved the comparisons significantly. Contrary to expectation, the high-resolution MM5 input data did not produce better results than the coarser EDAS data.
|
|
|
||||||
|
BRIG
|
GRSM
|
||||||
|
|
|
|
|
||||
|
EDAS
|
MM5
|
EDAS
|
MM5
|
||||
Starting
|
|
|
|
|
|
|
|
|
Elevation (m)
|
6 hr
|
24 hr
|
6 hr
|
24 hr
|
6 hr
|
24 hr
|
6 hr
|
24 hr
|
|
|
|
|
|
|
|
|
|
100
|
19 (6)
|
64
|
46 (34)
|
45
|
36 (30)
|
37
|
38 (22)
|
49
|
200
|
23 (13)
|
66
|
36 (20)
|
60
|
33 (28)
|
38
|
34 (19)
|
46
|
500
|
27 (20)
|
77
|
39 (28)
|
58
|
40 (37)
|
40
|
22 (11)
|
37
|
1000
|
57 (48)
|
89
|
35 (26)
|
50
|
36 (39)
|
42
|
37 (32)
|
52
|
1500
|
67 (52)
|
95
|
45 (38)
|
75
|
42 (35)
|
53
|
40 (45)
|
46
|
3000
|
51 (44)
|
95
|
42 (30)
|
63
|
52 (41)
|
53
|
41 (32)
|
59
|
Integrated from 100-3000 m
|
51
|
97
|
47
|
67
|
47
|
50
|
34
|
46
|
Table 2. Comparison (AAE %) of TMBR results at BRIG and GRSM with EDAS and MM5 as inputs to HYSPLIT. AAEs for 6-hr samples based on the three strongest regional sources at each site are shown in parentheses.
Comparison between the Lagrangian Stochastic Particle Dispersion Model in Inverse Mode and HYSPLIT Back Trajectories
In order to enhance the back-trajectory analysis, we used a Lagrangian random particle model (Koracin et al., 2007; 2008) in an inverse mode to estimate the most probable source areas starting from receptor locations. By using inverse modeling, we are able to establish probabilities for the receptor-source relationship between a single receptor and many source elements. The inverse modeling leads to explicit calculation of a source-receptor relationship in terms of linear transformations (i.e., in a matrix form) to describe the relative importance of specific subsets of the source to the impact at a specific receptor site. The elements of the receptor-source matrix simply represent the residence time spent in the respective source grid cell by the particles released from the receptor.
Eight 7-day episodes were simulated with the Lagrangian stochastic model using high-resolution (12-km) MM5 fields as meteorological input. These eight cases represented conditions for dominant impact of various regions at each of the two receptors (BRIG and GRSM). For each case, Lagrangian particles were continuously released from the receptor in inverse mode. A region with the greatest residence time was identified, and a most probable source area was selected. During the first run (20 July, 2002 backward to 13 July), the most probable source impacting BRIG was identified in Indiana. As a test and a first-guess evaluation, the Lagrangian stochastic model was run in a forward mode from the Indianapolis urban area. As expected, there is a noticeable similarity, which confirms that the selected source had an impact according to the simulation results (Figure 2).
Figure 2. Top-view distribution of the simulated particles after 168 hours from the backward run starting on 20 July 2002 at 00 UTC (left panel) and the forward run starting on 13 July 2002 at 00 UTC.
A quantitative analysis, including HYSPLIT, Lagrangian model, and CMAQ-derived sulfate concentrations, was conducted to examine differences in these two methodologies for identifying sources in backward mode. Figure 3 shows histograms for the same case (Figure 2) of normalized average residence time computed from the Lagrangian model and HYSPLIT results compared with the normalized CMAQ-derived sulfate concentrations for each region. The HYSPLIT back trajectory started at 100 m elevation concurrent with the Lagrangian model simulation. The sampling of residence time was done for hourly positions, and an associated region was identified using ArcGIS. Regarding the Lagrangian particles, sampling was completed for each particle trajectory for every hourly position.
Figure 3. Histograms of normalized average residence times of the Lagrangian model (blue) and HYSPLIT (purple) vs. normalized CMAQ-derived sulfate concentrations (yellow) for all seven regions for the case study (July 20->13, 2002).
One of the conclusions from this comparison is that the HYSPLIT results tend to underestimate regions near the receptor and overestimate regions that are further away. This is caused by a rapid pass through the near-receptor area and too much weight on a particular direction limited by the one-dimensional coverage of the back trajectory. The Lagrangian model appears to capture both of these effects due to multiple particle trajectories and dispersion covering larger area as the transport time increases.
Future Activities:
This project has been extended to December 31, 2009, under a no-cost-extension to retain funds for publication charges. Additional analysis will compare CMAQ sulfate source apportionments for summer and winter and corresponding apportionments using PMF and TMBR results. PMF will be applied to real-world IMPROVE summer concentrations from 2000-2006 at BRIG and GRSM, and the results will be compared with those obtained using synthetic data.
Journal Articles on this Report : 2 Displayed | Download in RIS Format
Other project views: | All 10 publications | 7 publications in selected types | All 6 journal articles |
---|
Type | Citation | ||
---|---|---|---|
|
Koracin D, Panorska A, Isakov V, Touma JS, Swall J. A statistical approach for estimating uncertainty in dispersion modeling: an example of application in southwestern USA. Atmospheric Environment 2007;41(3):617-628. |
R832156 (2008) R832156 (Final) |
Exit Exit Exit |
|
Watson JG, Chen L-WA, Chow JC, Doraiswamy P, Lowenthal DH. Source apportionment: findings from the U.S. Supersites Program. Journal of the Air & Waste Management Association 2008;58(2):265-288. |
R832156 (2007) R832156 (2008) R832156 (Final) |
Exit Exit |
Supplemental Keywords:
RFA, Scientific Discipline, Air, Ecosystem Protection/Environmental Exposure & Risk, particulate matter, Air Quality, Environmental Chemistry, Monitoring/Modeling, Environmental Monitoring, Atmospheric Sciences, Environmental Engineering, atmospheric dispersion models, atmospheric measurements, model-based analysis, area of influence analysis, source receptor based methods, source apportionment, chemical characteristics, emissions monitoring, environmental measurement, airborne particulate matter, air quality models, air quality model, air sampling, particulate matter mass, analytical chemistry, modeling studies, real-time monitoring, aerosol analyzers, chemical speciation sampling, particle size measurementRelevant Websites:
http://www.adim.dri.eduProgress 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.