New Technologies for Source ApportionmentEPA Grant Number: R832157
Title: New Technologies for Source Apportionment
Investigators: Henry, Ronald C. , Fine, Philip M.
Institution: University of Southern California
EPA Project Officer: Chung, Serena
Project Period: January 20, 2005 through January 19, 2008 (Extended to January 19, 2009)
Project Amount: $450,000
RFA: Source Apportionment of Particulate Matter (2004) RFA Text | Recipients Lists
Research Category: Air , Air Quality and Air Toxics , Particulate Matter
The proposed three-year project will develop new receptor-oriented methods of source apportionment based on the source information in wind speed and direction and the periodic variations of concentrations as extracted from the data by nonparametric regression and Fourier analysis. These methods will be applied to airborne particulate and other air quality data sets nationwide and to data from a particulate health-effects study to be carried out in Long Beach, CA during the summer and fall of 2004 and run by the USC Keck School of Medicine with additional resources provided by the Southern California Particle Center and Supersite. The source apportionment aspects of this field study will be strengthened by additional analysis of particulate samples for organic and inorganic species funded by this project. The newly developed receptor models will be augmented with the chemical mass balance (CMB) model and the Unmix multivariate receptor model.
CMB and other receptor models based on compositional data alone have difficulty in identifying sources that do not have unique chemical fingerprints or separating local from regional sources. Nonparametric regression methods using wind speed and direction have been shown to accurately locate nearby sources by triangulation. This project will combine nonparametric regression methods with the joint probability distribution of wind speed and direction determined by kernel density estimation methods that, like nonparametric regression, make no assumptions about the distribution of the data. The result is a Nonparametric Source Apportionment Model (NSAM) that will be used to estimate the contribution of local sources to airborne particles and precursor gases. NSAM makes good use of new measurement methods that give one hour or shorter time average concentrations. For example, in the Long Beach study particle measurements will be taken at a central site and at 10 nearby sites where PM2.5 will be measured by nephelometers with 10-minute time resolution. Finally, this project will quantitatively examine the periodic nature of pollutant concentrations using the Lomb-Scargle periodogram developed for astrophysical research. Standard, finite fast Fourier analysis methods must have evenly spaced data with no missing data. Few, if any, air quality data sets of any length satisfy these requirements and thus these powerful methods have seldom been applied to air quality data. The Lomb-Scargle periodogram is an estimate of the power spectrum of a time series with missing data. Source variability information in the power spectrum will be used for source identification and apportionment.
Commercial traffic at the Long Beach Airport and the harbor has grown greatly in past years. By examining historical data, it may be possible to predict the air quality impacts of future growth of these facilities or evaluate regulations on their operations. NSAM, CMB, Unmix, and the Lomb periodogram will identify and quantify the contribution of such local sources as refineries, major power plants, some of the most heavily traveled highways in the nation, the nation's busiest seaport, and a growing commercial airport. In addition to these site-specific results, the project will develop general methods whereby the strength and shape of daily, weekly, and seasonal variations can be determined by automated methods. This can be applied to removing periodicities from time series and may improve detection of long-term trends. The project will determine how to optimally use nonparametric regression techniques for source apportionment using wind speed and direction. Thus, the vast amount of historical routine air quality monitoring data can be used for source apportionment, making it possible to use forensic methods to make more efficient regulations. A secondary goal of the project is to promote the use of nonparametric regression and the Lomb periodogram in air quality studies in particular and environmental studies in general.