Chemical mass balance analysis of air quality data when unknown pollution sources are present

Abstract

Abstract Chemical mass balance (CMB) analysis is a standard approach for apportioning observed pollutant concentrations to their various pollution sources. To use CMB analysis, the researcher must assume that all sources affecting the airshed are identifiable, and that the pollution source profile associated with each source can be speciated. We consider the performance of several solutions to the CMB equations for cases in which one or more solutions affecting the airshed are unknown. We demonstrate that the presence of unknown sources in the airshed can lead to substantial (and sometimes surprising) errors when estimating the known source contributions. A simple illustration of the effect of unknown sources on the problem is given and the vulnerability of iterative estimators (such as the effective variance estimator) in the presence of unknown sources is explained. Methods for detecting unknown sources are proposed and evaluated. We propose a test for detecting unknown sources that is based on an intercept term included in the CMB equations. The approaches considered are compared via computer simulation, and with an example using real PM 2.5 data from the San Joaquin Valley Air Quality Study. We find that when unknown sources affect the airshed, a modified weighted least squares approach is superior to all other methods (including the effective variance approach).