Pollution source apportionment using a priori information and positive matrix factorization

Abstract The use of a priori information in positive matrix factorization (PMF) is examined in the context of pollution source apportionment. The impact of PMF's general run control settings is evaluated and simulation experiments are employed to illustrate the relative advantages and hazards associated with different uses of a priori information. Pulling source profile elements to zero appears to be uniformly beneficial when using data with low measurement error and no contamination from unknown sources. However, the benefit of F element pulling is less pronounced when data are subject to higher degrees of measurement error and when some elements are erroneously pulled to zero. The use of source profile targeting shows much promise, both for incorporating well-established knowledge about pollution sources and as a tool for incremental exploratory analysis of the data. A data analysis of the latter type is illustrated using PM2.5 data from the St. Louis Supersite.