Contributions of non-tailpipe emissions to near-road PM2.5 and PM10: A chemical mass balance study.

As the importance of non-tailpipe particles (NTP) over tailpipe emissions from urban traffic has been increasing, there is a need to evaluate NTP contributions to ambient particulate matter (PM) using representative source profiles. The Brake and Tire Wear Study conducted in Los Angeles, California in the winter of 2020 collected 64 PM2.5 and 64 PM10 samples from 32 pairs of downwind-upwind measurements at two near-road locations (I-5 in Anaheim and I-710 in Long Beach). These samples were characterized for inorganic and organic markers and, along with locally-developed brake wear, tire wear, and road dust source profiles, subject to source apportionment using the effective-variance chemical mass balance (EV-CMB) model. Model results highlighted the dominance of resuspended dust in both PM2.5 (23-33%) and PM10 (32-53%). Brake and tire wear contributed more to PM2.5 than tailpipe exhausts (diesel + gasoline) for I-5 (29-30% vs. 19-21%) while they were comparable for I-710 (15-17% vs. 15-19%). For PM10, the brake and tire wear contributions were 2-3 times the exhaust contributions. Different fleet compositions on and near I-5 and I-710 appeared to influence the relative importance of NTP and exhaust sources. The downwind-upwind differences in source contributions were often insignificant, consistent with small and/or nearly equal impacts of adjacent highway traffic emissions on the downwind and upwind sites. The utility of sole markers, such as barium and zinc, to predict brake and tire wear abundances in ambient PM is evaluated.

Evidence of non-tailpipe emission contributions to PM2.5 and PM10 near southern California highways.

Particulate Matter (PM) concentrations near highways are influenced by vehicle tailpipe and non-tailpipe emissions, other emission sources, and urban background aerosols. This study collected PM2.5 and PM10 filter samples near two southern California highways (I-5 and I-710) over two weeks in winter 2020. Samples were analyzed for chemical source markers. Mean PM2.5 and PM10 concentrations were approximately 10-15 and 30 µg/m3, respectively. Organic matter, mineral dust, and elemental carbon (EC) were the most abundant PM components. EC and polycyclic aromatic hydrocarbons at I-710 were 19-26% and 47% higher than those at the I-5 sites, respectively, likely due to a larger proportion of diesel vehicles. High correlations were found for elements with common sources, such as markers for brake wear (e.g., Fe, Ba, Cu, and Zr) and road dust (e.g., Al, Si, Ca, and Mn). Based on rubber abundances, the contributions of tire tread particles to PM2.5 and PM10 mass were approximately 8.0% at I-5 and 5.5% at I-710. Two different tire brands showed significantly different Si, Zn, carbon, and natural rubber abundances.

Lagged meteorological impacts on COVID-19 incidence among high-risk counties in the United States-a spatiotemporal analysis.

BACKGROUND: The associations between meteorological factors and coronavirus disease 2019 (COVID-19) have been discussed globally; however, because of short study periods, the lack of considering lagged effects, and different study areas, results from the literature were diverse and even contradictory. OBJECTIVE: The primary purpose of this study is to conduct more reliable research to evaluate the lagged meteorological impacts on COVID-19 incidence by considering a relatively long study period and diversified high-risk areas in the United States. METHODS: This study adopted the distributed lagged nonlinear model with a spatial function to analyze COVID-19 incidence predicted by multiple meteorological measures from March to October of 2020 across 203 high-risk counties in the United States. The estimated spatial function was further smoothed within the entire continental United States by the biharmonic spline interpolation. RESULTS: Our findings suggest that the maximum temperature, minimum relative humidity, and precipitation were the best meteorological predictors. Most significantly positive associations were found from 3 to 11 lagged days in lower levels of each selected meteorological factor. In particular, a significantly positive association appeared in minimum relative humidity higher than 88.36% at 5-day lag. The spatial analysis also shows excessive risks in the north-central United States. SIGNIFICANCE: The research findings can contribute to the implementation of early warning surveillance of COVID-19 by using weather forecasting for up to two weeks in high-risk counties.

Nonuniform impacts of COVID-19 lockdown on air quality over the United States.

Most of the state governments in United States (U.S.) issued lockdown or business restrictions amid the COVID-19 pandemic in March 2020, which created a unique opportunity to evaluate the air quality response to reduced economic activities. Data acquired from 28 long-term air quality stations across the U.S. revealed widespread but nonuniform reductions of nitrogen dioxide (NO2) and carbon monoxide (CO) during the first phase of lockdown (March 15-April 25, 2020) relative to a pre-lockdown reference period and historical baselines established in 2017-2019. The reductions, up to 49% for NO2 and 37% for CO, are statistically significant at two thirds of the sites and tend to increase with local population density. Significant reductions of particulate matter (PM2.5 and PM10) only occurred in the Northeast and California/Nevada metropolises where NO2 declined the most, while the changes in ozone (O3) were mixed and relatively minor. These findings are consistent with lower transportation and utility demands that dominate NO2 and CO emissions, especially in major urban areas, due to the lockdown. This study provides an insight into potential public health benefits with more aggressive air quality management, which should be factored into strategies to reopen the U.S. and global economy.

Spatial and temporal variability of brown carbon in United States: implications for direct radiative effects.

A newly developed dataset from the Interagency Monitoring of PROtected Visual Environments (IMPROVE) observation network, combined with a 3-D chemical transport model, is used to evaluate the spatial and temporal variability of brown carbon (BrC) in the United States. The model with BrC emitted from biomass burning and biofuel emissions agrees with the seasonal and spatial variability of BrC planetary boundary layer (PBL) absorption aerosol optical depth (AAOD) observations within a factor of 2. The model without whitening, the tendency for absorption to decrease with aerosol aging, overestimates the observed BrC PBL AAOD, and does not reflect the measured BrC PBL AAOD spatial variability. The model shows higher absorption direct radiative effects (DRE) from BrC at northern high latitudes than at mid-latitudes in spring and summer, due to boreal fire emissions, long whitening lifetimes and high surface albedos. These findings highlight the need to study BrC over the Arctic region.

Health Benefits from Upgrading Public Buses for Cleaner Air: A Case Study of Clark County, Nevada and the United States.

Public transit buses, which move more than 5 billion passengers annually in the United States (U.S.), can contribute substantially to the environmental health burden through emitted air pollutants. As a leader in transforming to cleaner bus fleets, the Regional Transport Commission of Southern Nevada (RTC) has been transitioning from diesel to compressed natural gas (CNG) transit buses since 1999. By 2017, ~75% of RTC's buses operating in Clark County, Nevada were CNG-powered. This study assesses the health benefits of the venture using the US Environmental Protection Agency's (EPA) Co-Benefits Risk Assessment (COBRA) model, considering the emission and exposure changes from the 2017 baseline for two hypothetical scenarios: (1) no transition (CC_D) and (2) complete transition (CC_N). The CC_D scenario shows realized health benefits, mostly due to avoided mortality, of $0.79⁻8.21 million per year for 2017 alone, while CC_N suggests an additional $0.88⁻2.24 million annually that could be achieved by completing the transition. The wide range of estimates partly reflects uncertainties in determining diesel bus emissions under business-as-usual. These health benefits were not limited locally, with ~70% going to other counties. Two national-scale scenarios, US_D and US_N, were also constructed to explore the health impact of transitioning from diesel to CNG buses across the U.S. As of 2017, with CNG powering only ~20% of transit bus mileages nationwide, there could be massive unrealized health benefits of $0.98⁻2.48 billion per year including 114⁻258 avoided premature deaths and >5000 avoided respiratory and cardiovascular illnesses. Taking into account the health benefits, economic costs, and the inter-state nature of air pollution, expanding federal assistances to accelerate a nationwide transition to cleaner bus fleets is highly recommended.

Separation of brown carbon from black carbon for IMPROVE and Chemical Speciation Network PM2.5 samples.

The replacement of the Desert Research Institute (DRI) model 2001 with model 2015 thermal/optical analyzers (TOAs) results in continuity of the long-term organic carbon (OC) and elemental carbon (EC) database, and it adds optical information with no additional carbon analysis effort. The value of multiwavelength light attenuation is that light absorption due to black carbon (BC) can be separated from that of brown carbon (BrC), with subsequent attribution to known sources such as biomass burning and secondary organic aerosols. There is evidence of filter loading effects for the 25% of all samples with the highest EC concentrations based on the ratio of light attenuation to EC. Loading corrections similar to those used for the seven-wavelength aethalometer need to be investigated. On average, nonurban Interagency Monitoring of PROtected Visual Environments (IMPROVE) samples show higher BrC fractions of short-wavelength absorption than urban Chemical Speciation Network (CSN) samples, owing to greater influence from biomass burning and aged aerosols, as well as to higher primary BC contributions from engine exhaust at urban sites. Sequential samples taken during an Everglades National Park wildfire demonstrate the evolution from flaming to smoldering combustion, with the BrC fraction increasing as smoldering begins to dominate the fire event. IMPLICATIONS: The inclusion of seven wavelengths in thermal/optical carbon analysis of speciated PM2.5 (particulate matter with an aerodynamic diameter ≤2.5 µm) samples allows contributions from biomass burning and secondary organic aerosols to be estimated. This separation is useful for evaluating control strategy effectiveness, identifying exceptional events, and determining natural visibility conditions.

Atmospheric deposition of particles at a sensitive alpine lake: Size-segregated daily and annual fluxes from passive sampling techniques.

Lake Tahoe, a North American alpine lake long appreciated for its clear water and geographic setting, has experienced a trend of declining water clarity due to increasing nutrient and particle inputs. Contributions from atmospheric deposition of particulate matter (PM) could be important, yet they are inadequately quantified. This study established a yearlong deposition monitoring network in the northern Lake Tahoe Basin. Dry deposition was quantified on surrogate surfaces while wet deposition was based on particles suspended in precipitation at 24-hour resolution. The particle size ranges by these passive techniques were 1-64µm and 0.5-20µm in diameter for dry and wet deposition, respectively. Dry deposition of submicrometer (0.5-1µm) particles was also estimated by extrapolation of a lognormal size distribution. Higher daily number deposition fluxes (NDFdry and NDFwet) were found at a near-shore site, confirming substantial impacts of commercial and tourist activities. The two more isolated sites indicated a uniform regional background. On average, daily NDFdry is about one order of magnitude lower than daily NDFwet. Dry deposition velocities increased rapidly with particle size, as evidenced by collocated measurements of NDFdry and ambient particle number concentrations, though it seems less so for wet deposition due to different scavenging mechanisms. Despite fewer "wet" days than "dry" days during the monitoring period, wet processes dominated seasonal particle deposition, particularly in winter and spring when most precipitation occurred. Adopting sediment (insoluble, inorganic) particle fraction estimates from the literature, this study reports an annual particle flux of 2.9-5.2×1010#m-2yr-1 for sediment particles with 1-20µm diameter and 6.1-11×1010#m-2yr-1 for those with 0.5-20µm diameter. Implications of these findings to the current knowledge of atmospheric deposition in the Lake Tahoe Total Maximum Daily Load (TMDL) are discussed.

Inter-annual variability of wintertime PM2.5 chemical composition in Xi'an, China: Evidences of changing source emissions.

Chemical characteristics of PM2.5 in Xi'an in wintertime of 2006, 2008, and 2010 were investigated. Markers of OC2, EC1, and NO3(-)/SO4(2-) ratio were calculated to investigate the changes in PM2.5 emission sources over the 5-year period. Positive matrix factorization (PMF) model was used to identify and quantify the main sources of PM2.5 and their contributions. The results showed that coal combustion, motor vehicular emissions, fugitive dust, and secondary inorganic aerosol accounted for more than 80% of PM2.5 mass. The importance of these major sources to the PM2.5 mass varied yearly: coal combustion was the largest contributor (31.2% ± 5.2%), followed by secondary inorganic aerosol (20.9% ± 5.2%) and motor vehicular emissions (19.3% ± 4.8%) in 2006; the order was still coal combustion emissions (27.6% ± 3.4%), secondary inorganic aerosol (23.2% ± 6.9%), and motor vehicular emissions (20.9% ± 4.6%) in 2008; while coal combustion emission further decreased (24.1% ± 3.1%) with fugitive dust (19.4% ± 5.5%) increasing in 2010. The changes in PM2.5 chemical compositions and source contributions can be attributed to the social and economic developments in Xi'an, China, including energy structure adjustment, energy consumption, the expansion of civil vehicles, and the increase of urban construction activities.

Mass reconstruction methods for PM2.5: a review.

Major components of suspended particulate matter (PM) are inorganic ions, organic matter (OM), elemental carbon (EC), geological minerals, salt, non-mineral elements, and water. Since oxygen (O) and hydrogen (H) are not directly measured in chemical speciation networks, more than ten weighting equations have been applied to account for their presence, thereby approximating gravimetric mass. Assumptions for these weights are not the same under all circumstances. OM is estimated from an organic carbon (OC) multiplier (f) that ranges from 1.4 to 1.8 in most studies, but f can be larger for highly polar compounds from biomass burning and secondary organic aerosols. The mineral content of fugitive dust is estimated from elemental markers, while the water-soluble content is accounted for as inorganic ions or salt. Part of the discrepancy between measured and reconstructed PM mass is due to the measurement process, including: (1) organic vapors adsorbed on quartz-fiber filters; (2) evaporation of volatile ammonium nitrate and OM between the weighed Teflon-membrane filter and the nylon-membrane and/or quartz-fiber filters on which ions and carbon are measured; and (3) liquid water retained on soluble constituents during filter weighing. The widely used IMPROVE equations were developed to characterize particle light extinction in U.S. national parks, and variants of this approach have been tested in a large variety of environments. Important factors for improving agreement between measured and reconstructed PM mass are the f multiplier for converting OC to OM and accounting for OC sampling artifacts.

Long-term trends in visibility and at Chengdu, China.

Long-term (1973 to 2010) trends in visibility at Chengdu, China were investigated using meteorological data from the U.S. National Climatic Data Center. The visual range exhibited a declining trend before 1982, a slight increase between 1983 and 1995, a sharp decrease between 1996 and 2005, and some improvements after 2006. The trends in visibility were generally consistent with the economic development and implementation of pollution controls in China. Intensive PM2.5 measurements were conducted from 2009 to 2010 to determine the causes of visibility degradation. An analysis based on a modification of the IMPROVE approach indicated that PM2.5 ammonium bisulfate contributed 27.7% to the light extinction coefficient (bext ); this was followed by organic mass (21.7%), moisture (20.6%), and ammonium nitrate (16.3%). Contributions from elemental carbon (9.4%) and soil dust (4.3%) were relatively minor. Anthropogenic aerosol components (sulfate, nitrate, and elemental carbon) and moisture at the surface also were important determinants of the aerosol optical depth (AOD) at 550 nm, and the spatial distributions of both bext and AOD were strongly affected by regional topography. A Positive Matrix Factorization receptor model suggested that coal combustion was the largest contributor to PM2.5 mass (42.3%) and the dry-air light-scattering coefficient (47.7%); this was followed by vehicular emissions (23.4% and 20.5%, respectively), industrial emissions (14.9% and 18.8%), biomass burning (12.8% and 11.9%), and fugitive dust (6.6% and 1.1%). Our observations provide a scientific basis for improving visibility in this area.

Fine particulate matter and visibility in the Lake Tahoe Basin: chemical characterization, trends, and source apportionment.

Speciated PM2.5 (particulate matter with an aerodynamic diameter

PM2.5 source apportionment: reconciling receptor models for U.S. nonurban and urban long-term networks.

Chemical mass balance (CMB) and trajectory receptor models were applied to speciated particulate matter with aerodynamic diameter < or =2.5 microm (PM2.5) measurements from Speciation Trends Network (STN; part of the Chemical Speciation Network [CSN]) and Interagency Monitoring of Protected Visual Environments (IMPROVE) monitoring network across the state of Minnesota as part of the Minnesota PM2.5 Source Apportionment Study (MPSAS). CMB equations were solved by the Unmix, positive matrix factorization (PMF), and effective variance (EV) methods, giving collective source contribution and uncertainty estimates. Geological source profiles developed from local dust materials were either incorporated into the EV-CMB model or used to verify factors derived from Unmix and PMF. Common sources include soil dust, calcium (Ca)-rich dust, diesel and gasoline vehicle exhausts, biomass burning, secondary sulfate, and secondary nitrate. Secondary sulfate and nitrate aerosols dominate PM2.5 mass (50-69%). Mobile sources outweigh area sources at urban sites, and vice versa at rural sites due to traffic emissions. Gasoline and diesel contributions can be separated using data from the STN, despite significant uncertainties. Major differences between MPSAS and earlier studies on similar environments appear to be the type and magnitude of stationary sources, but these sources are generally minor (<7%) in this and other studies. Ensemble back-trajectory analysis shows that the lower Midwestern states are the predominant source region for secondary ammoniated sulfate in Minnesota. It also suggests substantial contributions of biomass burning and soil dust from out-of-state on occasions, although a quantitative separation of local and regional contributions was not achieved in the current study. Supplemental materials are available for this article. Go to the publisher's online edition of the Journal of the Air & Waste Management Association for a summary of input data, Unmix and PMF factor profiles, and additional maps.

Regional source identification using Lagrangian stochastic particle dispersion and HYSPLIT backward-trajectory models.

The main objective of this study was to investigate the capabilities of the receptor-oriented inverse mode Lagrangian Stochastic Particle Dispersion Model (LSPDM) with the 12-km resolution Mesoscale Model 5 (MM5) wind field input for the assessment of source identification from seven regions impacting two receptors located in the eastern United States. The LSPDM analysis was compared with a standard version of the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) single-particle backward-trajectory analysis using inputs from MM5 and the Eta Data Assimilation System (EDAS) with horizontal grid resolutions of 12 and 80 km, respectively. The analysis included four 7-day summertime events in 2002; residence times in the modeling domain were computed from the inverse LSPDM runs and HYPSLIT-simulated backward trajectories started from receptor-source heights of 100, 500, 1000, 1500, and 3000 m. Statistics were derived using normalized values of LSPDM- and HYSPLIT-predicted residence times versus Community Multiscale Air Quality model-predicted sulfate concentrations used as baseline information. From 40 cases considered, the LSPDM identified first- and second-ranked emission region influences in 37 cases, whereas HYSPLIT-MM5 (HYSPLIT-EDAS) identified the sources in 21 (16) cases. The LSPDM produced a higher overall correlation coefficient (0.89) compared with HYSPLIT (0.55-0.62). The improvement of using the LSPDM is also seen in the overall normalized root mean square error values of 0.17 for LSPDM compared with 0.30-0.32 for HYSPLIT. The HYSPLIT backward trajectories generally tend to underestimate near-receptor sources because of a lack of stochastic dispersion of the backward trajectories and to overestimate distant sources because of a lack of treatment of dispersion. Additionally, the HYSPLIT backward trajectories showed a lack of consistency in the results obtained from different single vertical levels for starting the backward trajectories. To alleviate problems due to selection of a backward-trajectory starting level within a large complex set of 3-dimensional winds, turbulence, and dispersion, results were averaged from all heights, which yielded uniform improvement against all individual cases.

Quality assurance and quality control for thermal/optical analysis of aerosol samples for organic and elemental carbon.

Accurate, precise, and valid organic and elemental carbon (OC and EC, respectively) measurements require more effort than the routine analysis of ambient aerosol and source samples. This paper documents the quality assurance (QA) and quality control (QC) procedures that should be implemented to ensure consistency of OC and EC measurements. Prior to field sampling, the appropriate filter substrate must be selected and tested for sampling effectiveness. Unexposed filters are pre-fired to remove contaminants and acceptance tested. After sampling, filters must be stored in the laboratory in clean, labeled containers under refrigeration (<4 °C) to minimize loss of semi-volatile OC. QA activities include participation in laboratory accreditation programs, external system audits, and interlaboratory comparisons. For thermal/optical carbon analyses, periodic QC tests include calibration of the flame ionization detector with different types of carbon standards, thermogram inspection, replicate analyses, quantification of trace oxygen concentrations (<100 ppmv) in the helium atmosphere, and calibration of the sample temperature sensor. These established QA/QC procedures are applicable to aerosol sampling and analysis for carbon and other chemical components.

Particulate emission factors for mobile fossil fuel and biomass combustion sources.

PM emission factors (EFs) for gasoline- and diesel-fueled vehicles and biomass combustion were measured in several recent studies. In the Gas/Diesel Split Study (GD-Split), PM(2.5) EFs for heavy-duty diesel vehicles (HDDV) ranged from 0.2 to ~2 g/mile and increased with vehicle age. EFs for HDDV estimated with the U.S. EPA MOBILE 6.2 and California Air Resources Board (ARB) EMFAC2007 models correlated well with measured values. PM(2.5) EFs measured for gasoline vehicles were ~two orders of magnitude lower than those for HDDV and did not correlate with model estimates. In the Kansas City Study, PM(2.5) EFs for gasoline-powered vehicles (e.g., passenger cars and light trucks) were generally <0.03 g/mile and were higher in winter than summer. EMFAC2007 reported higher PM(2.5) EFs than MOBILE 6.2 during winter, but not during summer, and neither model captured the variability of the measured EFs. Total PM EFs for heavy-duty diesel military vehicles ranged from 0.18±0.03 and 1.20±0.12 g/kg fuel, corresponding to 0.3 and 2 g/mile, respectively. These values are comparable to those of on-road HDDV. EFs for biomass burning measured during the Fire Laboratory at Missoula Experiment (FLAME) were compared with EFs from the ARB Emission Estimation System (EES) model. The highest PM(2.5) EFs (76.8±37.5 g/kg) were measured for wet (>50% moisture content) Ponderosa Pine needles. EFs were generally <20 g/kg when moisture content was <20%. The EES model agreed with measured EFs for fuels with low moisture content but underestimated measured EFs for fuel with moisture content >40%. Average EFs for dry chamise, rice straw, and dry grass were within a factor of three of values adopted by ARB in California's San Joaquin Valley (SJV). Discrepancies between measured and modeled emission factors suggest that there may be important uncertainties in current PM(2.5) emission inventories.

Evaluation of regional-scale receptor modeling.

The ability of receptor models to estimate regional contributions to fine particulate matter (PM2.5) was assessed with synthetic, speciated datasets at Brigantine National Wildlife Refuge (BRIG) in New Jersey and Great Smoky Mountains National Park (GRSM) in Tennessee. Synthetic PM2.5 chemical concentrations were generated for the summer of 2002 using the Community Multiscale Air Quality (CMAQ) model and chemically speciated PM2.5 source profiles from the U.S. Environmental Protection Agency (EPA)'s SPECIATE and Desert Research Institute's source profile databases. CMAQ estimated the "true" contributions of seven regions in the eastern United States to chemical species concentrations and individual source contributions to primary PM2.5 at both sites. A seven-factor solution by the positive matrix factorization (PMF) receptor model explained approximately 99% of the variability in the data at both sites. At BRIG, PMF captured the first four major contributing sources (including a secondary sulfate factor), although diesel and gasoline vehicle contributions were not separated. However, at GRSM, the resolved factors did not correspond well to major PM2.5 sources. There were no correlations between PMF factors and regional contributions to sulfate at either site. Unmix produced five- and seven-factor solutions, including a secondary sulfate factor, at both sites. Some PMF factors were combined or missing in the Unmix factors. The trajectory mass balance regression (TMBR) model apportioned sulfate concentrations to the seven source regions using Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) trajectories based on Meteorological Model Version 5 (MM5) and Eta Data Simulation System (EDAS) meteorological input. The largest estimated sulfate contributions at both sites were from the local regions; this agreed qualitatively with the true regional apportionments. Estimated regional contributions depended on the starting elevation of the trajectories and on the meteorological input data.

Toward effective source apportionment using positive matrix factorization: experiments with simulated PM2.5 data.

To elucidate the relationship between factors resolved by the positive matrix factorization (PMF) receptor model and actual emission sources and to refine the PMF modeling strategy, speciated PM2.5 (particulate matter with aerodynamic diameter < 2.5 microm) data generated from a state-of-the-art chemical transport model for two rural sites in the eastern United States are subjected to PMF analysis. In addition to chi2 and R2 used to infer the quality of fitting, the interpretability of PMF factors with respect to known primary and secondary sources is evaluated using a root mean square difference analysis. For the most part, factors are found to represent imperfect combinations of sources, and the optimal number of factors should be just adequate to explain the input data (e.g., R2 > 0.95). Retaining more factors in the model does not help resolve minor sources, unless temporal resolution of the data is increased, thus allowing more information to be used by the model. If guided with a priori knowledge of source markers and/or special events, rotation of factors leads to more interpretable PMF factors. The choice of uncertainty weighting coefficients greatly influences the PMF modeling results, but it cannot usually be determined for simulated or real-world data. A simple test is recommended to check whether the weighting coefficients are suitable. However, uncertainties in the data divert PMF solutions even when the optimal weighting coefficients and number of factors are in place.

Evaluation of Regional-Scale Receptor Modeling.

The ability of receptor models to estimate regional contributions to fine particulate matter (PM2.5) was assessed with synthetic, speciated datasets at Brigantine National Wildlife Refuge (BRIG) in New Jersey and Great Smoky Mountains National Park (GRSM) in Tennessee. Synthetic PM2.5 chemical concentrations were generated for the summer of 2002 using the Community Multiscale Air Quality (CMAQ) model and chemically speciated PM2.5 source profiles from the U.S. Environmental Protection Agency (EPA)'s SPECIATE and Desert Research Institute's source profile databases. CMAQ estimated the "true" contributions of seven regions in the eastern United States to chemical species concentrations and individual source contributions to primary PM2.5 at both sites. A seven-factor solution by the positive matrix factorization (PMF) receptor model explained approximately 99% of the variability in the data at both sites. At BRIG, PMF captured the first four major contributing sources (including a secondary sul-fate factor), although diesel and gasoline vehicle contributions were not separated. However, at GRSM, the resolved factors did not correspond well to major PM2.5 sources. There were no correlations between PMF factors and regional contributions to sulfate at either site. Unmix produced five- and seven-factor solutions, including a secondary sulfate factor, at both sites. Some PMF factors were combined or missing in the Unmix factors. The trajectory mass balance regression (TMBR) model apportioned sulfate concentrations to the seven source regions using Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) trajectories based on Meteorological Model Version 5 (MM5) and Eta Data Simulation System (EDAS) meteorological input. The largest estimated sulfate contributions at both sites were from the local regions; this agreed qualitatively with the true regional apportionments. Estimated regional contributions depended on the starting elevation of the trajectories and on the meteorological input data.

Toward Effective Source Apportionment Using Positive Matrix Factorization: Experiments with Simulated PM2.5 Data.

To elucidate the relationship between factors resolved by the positive matrix factorization (PMF) receptor model and actual emission sources and to refine the PMF modeling strategy, speciated PM2.5 (particulate matter with aerodynamic diameter <2.5 µm) data generated from a state-of-the-art chemical transport model for two rural sites in the eastern United States are subjected to PMF analysis. In addition to χ-2 and R 2 used to infer the quality of fitting, the interpretability of PMF factors with respect to known primary and secondary sources is evaluated using a root mean square difference analysis. For the most part, factors are found to represent imperfect combinations of sources, and the optimal number of factors should be just adequate to explain the input data (e.g., R 2 > 0.95). Retaining more factors in the model does not help resolve minor sources, unless temporal resolution of the data is increased, thus allowing more information to be used by the model. If guided with a priori knowledge of source markers and/or special events, rotation of factors leads to more interpretable PMF factors. The choice of uncertainty weighting coefficients greatly influences the PMF modeling results, but it cannot usually be determined for simulated or real-world data. A simple test is recommended to check whether the weighting coefficients are suitable. However, uncertainties in the data divert PMF solutions even when the optimal weighting coefficients and number of factors are in place.