Identifying and quantifying the impact of air pollution source areas by nonparametric trajectory analysis.

摘要

In order to improve air quality, it is necessary to identify and quantify the sources of airborne pollution. Local emissions are more easily to control compared to regional emissions since multiple agencies and states are not involved in the regulatory process.;Generally two types of air quality models, source -- oriented models and receptor -- oriented models, are used to evaluate the impact of emission on air quality on a local, regional, and global scale. Source -- oriented models require detailed information on emission composition, rates and local meteorological data. Therefore, they are not suitable for sources of fugitive emissions and intermittent or temporary emissions, which cannot easily be quantified. On the other hand, receptor models need chemical composition data to identify and quantify sources affecting the monitoring sites. However, pollutants without distinguishable "fingerprints", such as SO 2, O3 cannot be apportioned by this method.;A new hybrid source -- receptor model was previously develop and is called Nonparametric Trajectory Analysis (NTA). It is based on nonparametric kernel smoothing and backtrajectory analysis. NTA was developed to identify and quantify local sources of species measured on a very short time scale, i.e., minute, and it has gotten some encouraging results. However, NTA sometimes produces artifacts areas that appear to be sources but not, this is especially true for sources very close to the receptor. A major objective of this study is to address this difficulty.;The NTA gives a map of the average concentration at the receptor when the air passes over each point on the map. This NTA map is obviously related to the local sources affecting the receptor, but it is not a map of the sources. One way to extend the NTA method is the Point Source Response (PSR) method. The NTA map can be considered a linear combination of responses to number of point sources. The NTA map for a point source at each point on a grid is calculated. Next, the weighted sum of the PRS maps that best fits the NTA map for the real data is estimated by principal components regression. In this way, the size and location of source affecting the receptor are estimated.;This method is illustrated by application to 1-minute SO2 data from Long Beach, and 1-minute PM10 data for Rubidoux along with meteorological data from nearby monitoring stations in South Coast Air Basin of Southern California. The result identified the Long Beach harbor and transportation hubs close to the intersection of freeway 710 and freeway 405 and Long Beach Airport as major SO2 sources. For the Rubidoux area, aggregate, and asphalt factories, and construction sites are identified as source of PM10.