Comparative application of multiple receptor methods to identify aerosol sources in northern Vermont.

This study applies and compares results of four receptor modeling techniques to a common set of speciated fine particle measurement data collected ata remote site in northwestern Vermont between 1988 and 1995. Two multivariate mathematical models, positive matrix factorization and UNMIX, were applied to the measurement data and identified seven "common" sources that had similar compositions and similar fine mass contributions in both models. Two ensemble backward trajectory techniques, potential source contribution function and residence-time analysis, were also applied to evaluate and interpret the mathematical model results. The trajectory techniques indicate a strong regional character to the upwind locations associated with aerosol contributions from most of the sources identified independently by the mathematical models and help in the interpretation of those results. The process of model comparison provides insights on the strengths and limitations of the individual and combined source attribution techniques. Convergent results among the multiple methods provide a degree of confidence that each of the receptor methods may represent useful tools for future air quality management. Divergent or inconsistent results among the models can help identify limitations of the individual models and of the underlying aerosol and meteorological data sets.