2009 Progress Report: Improving Particulate Matter Source Apportionment for Health Studies: A Trained Receptor Modeling Approach with Sensitivity, Uncertainty and Spatial AnalysesEPA Grant Number: R833866
Title: Improving Particulate Matter Source Apportionment for Health Studies: A Trained Receptor Modeling Approach with Sensitivity, Uncertainty and Spatial Analyses
Investigators: Russell, Armistead G. , Klein, Mitchel , Mulholland, James , Sarnat, Stefanie Ebelt , Sarnat, Jeremy , Tolbert, Paige
Current Investigators: Russell, Armistead G. , Klein, Mitchel , Marmur, Amit , Mulholland, James , Sarnat, Stefanie Ebelt , Sarnat, Jeremy , Tolbert, Paige
Institution: Georgia Institute of Technology , Emory University
EPA Project Officer: Ilacqua, Vito
Project Period: December 1, 2008 through November 30, 2012 (Extended to November 30, 2013)
Project Period Covered by this Report: December 1, 2008 through December 31,2009
Project Amount: $899,956
RFA: Innovative Approaches to Particulate Matter Health, Composition, and Source Questions (2007) RFA Text | Recipients Lists
Research Category: Health Effects , Particulate Matter , Air
The main objectives of this research are to test four hypotheses derived from ongoing source apportionment (SA)-based epidemiologic and air quality modeling studies:
- A receptor-based approach, trained using an ensemble of model results (including receptor and emissions-based models), can be developed that neither introduces excessive nor inhibits an appropriate level of day-to-day variability.
- The method can be applied to long-term data sets for use in acute health effect studies.
- The method can be used to temporally interpolate between observations (e.g., for data available every third day) and spatially interpolate between urban and rural monitors.
- Uncertainties can be propagated from SA model inputs to health analysis outputs, with outputs most sensitive to source profile inputs.
To test the hypotheses, a chemical mass balance (CMB) approach has been developed for particulate matter (PM) source apportionment (SA) that utilizes an ensemble of both source- and receptor-based approaches to train a CMB method for use in longer term applications. This approach may be particularly useful in acute effects health studies that seek to ascertain associations between PM2.5 sources and health outcomes. The method utilized in the study involved three steps:
- Averaging SA results, using weights based on method uncertainty, from four receptor models and one chemical transport model, the community multiscale air quality (CMAQ) model, to develop ensemble-based source impacts.
- Using the weighted source impacts (from Step 1) in an application of CMB with the Lipschitz Global Optimizer (CMB-LGO) to calculate nine ensemble-based source profiles (EBSPs): The source profiles developed include gasoline vehicles (GV), diesel vehicles (DV), dust (DUST), biomass burning (BURN), coal combustion (COAL), secondary organic carbon (SOC), SULFATE, NITRATE, and AMMONIUM.
- Using the EBSPs on a longer term data set of observations to develop improved source impacts.
We have successfully developed and tested the ensembling approach involving three steps:
Step 1: Ensemble averaging. Ensemble average results were developed using the five source apportionment methods over the two periods where all five SA methods had available results (one summer and one winter period). Ensembling reduced day-to-day variation in source impacts and had less bias between observed and estimated PM2.5 mass (Lee, et al., 2009).
Step 2: Ensemble-based source profiles. Because the computational requirements to run CMAQ for long time periods are extensive, and molecular marker-based CMB results also are not available over long time periods, ensemble results are developed over shorter periods. In this case, two periods, one in the summer and one in the winter, had results available from the five methods we employ. Ensemble results were used to obtain two seasonally based new source profiles using observations for July 2001 (to represent summer) and January 2002 (to represent winter). The EBSPs for GVs showed a distinct seasonal split with an organic carbon/elemental carbon (OC/EC) ratio of 3.5 in winter and 1.5 in the summer. The COAL profiles also showed a seasonal split in the amount of sulfate, OC and EC emitted. This result was surprising because emissions from coal combustion operations are not expected be different from season to season. Nevertheless, the COAL EBSPs led to a reduction in zero impact days. This will be evaluated further in future work.
Step 3: New source impacts. The EBSPs then were used to apportion PM2.5 sources using a CMB-LGO application for a 12 month data set of daily PM2.5 measurements at the Atlanta, GA, Jefferson Street (JST) site. Compared to using CMB-LGO with MBSPs, ensemble-trained CMB approaches reduce zero-impact days from sources that are known to be existent, implying that results can be improved using this method. Subsequently, the EBSPs have been applied to a 9.5 year data set (8/31/98 – 12/31/07) of measurements at JST. In both the 2002 and 9.5 year data sets, the ensemble method resulted in similar changes in source impacts. In summer, biomass burning and road dust impacts increased while gasoline vehicle and SOC impacts decreased when using EBSPs. In winter, biomass burning, road dust, and coal combustion impacts increased while Other OC impacts decreased.
The CMB-LGO model using both EBSPs and MBSPs predicted zero impacts from DV, DUST, BURN, COAL, and SOC. Preliminary results show that DV zero impact days had an obvious day-of-week pattern. The weekend accounted for 43% and 42% of the zero impact days for DV when using EBSPs and MBSPs, respectively. The other sources that had zero impact days did not exhibit any specific day of week pattern.
A publication describing the method and results has been published (Lee, et al., 2009).
Our future work will focus on addressing the following issues: improving estimates of uncertainty, assessing variability, applying this method to other locations, and its use in epidemiologic modeling. We currently are working to develop a method to determine uncertainty in the ensembling method (Step 1) as well as refining the uncertainties of each method used in the ensemble. We also will investigate using a central-difference metric as a measure of variability, which will be applied for the 9.5 year data set for SA results using both EBSPs and MBSPs. We also will conduct time series filtering by using a Fourier transform method to better understand variability. We will be applying the procedure to a simulated JST data set that mimics other data sets that typically only have speciated PM2.5 data every three or six days and develop a method to interpolate data for days without measurements. Subsequently, we will apply the entire ensemble method to a data set in St. Louis, MO, to assess regional differences and demonstrate applicability to other locations. Finally, we will use the results in epidemiologic modeling.
Journal Articles on this Report : 1 Displayed | Download in RIS Format
|Other project views:||All 30 publications||20 publications in selected types||All 18 journal articles|
||Lee D, Balachandran S, Pachon J, Shankaran R, Lee S, Mulholland JA, Russell AG. Ensemble-trained PM2.5 source apportionment approach for health studies. Environmental Science & Technology 2009;43(18):7023-7031.||