Analysis of patterns and trends in atmospheric deposition and surface water quality

A focal point of the Clean Air Act from its first incarnation in 1970 to the 1990 amendments has been the effects of air pollution on terrestrial and aquatic ecosystems. In response to evidence that anthropogenic emissions of sulfur and nitrogen oxides were resulting in increased precipitation acidity and subsequent damage to surface water and vegetation, the Clean Air Act Amendments of 1990 (CAAA) created the acid rain program to limit emissions of these acid precipitation precursors. To assess the potential effectiveness of this legislation the trends in sulfate deposition and the association between sulfate deposition and surface water sulfate concentrations were modeled using a Belief network approach.; Spatial and temporal modeling of sulfate deposition data from the National Atmospheric Deposition Program using Multivariate Adaptive Regression Splines revealed that embedded in a overall weak downward trend in sulfate precipitation concentration is a year-latitude interaction effect. This year-latitude interaction effect indicates a stronger downward trend in the southern latitudes while the northern latitudes show virtually no trend.; Spatial models of sulfate deposition were developed from National Atmospheric Deposition Program data and linked to watershed sulfate enrichment models developed from surface water concentration data from the Long Term Monitoring Project sites in the Adirondack Mountains of New York. This modeling was conducted within a Bayesian framework employing Markov Chain Monte Carlo simulations to assess the impact of deposition prediction uncertainty on the conclusions drawn from the model of sulfate reduction.; Deposition predictions from a classical spatial modeling approach, ignoring spatial covariance uncertainty, were found to have greater variability than predictions under the sulfate enrichment belief network. Two sulfate enrichment model specifications were examined. Uncertainty in sulfate enrichment model parameters was found to be slightly greater for belief network posterior distributions. Choice of model specification did not impact the conclusion that the Adirondack lakes analyzed in this study were slightly reducing. However, model specification did have a dramatic impact on anticipated lake response to a 40% reduction in sulfur dioxide emissions.