Stochastic Modeling of the Space‐Time Structure of Atmospheric Chemical Deposition

A multivariate space‐time stochastic model suitable for the analysis of weekly atmospheric chemistry wet deposition measurements is described. The model is hierarchical, with weekly ion concentration fields represented as the sum of a persistent long‐term mean field, and yearly and weekly variation fields. A simple method of moments estimation scheme is proposed which exploits the hierarchical nature of the model to separate spatial structure at weekly, yearly, and persistent time scales. Estimation of both isotropic and anisotropic covariance functions are considered. The model was applied to precipitation, sulfate concentration, and pH measurements made in the Northeastern United States during 1980–1981. While significant spatial correlation was found at weekly and longer (long‐term mean) time scales, there was little or no temporal correlation. The observed spatial correlation was substantially anisotropic, particularly for precipitation and sulfate concentration fields. The anisotropy, as well as correlation length scales, showed strong seasonal variations. Substantial differences were apparent in the spatial structure of pH and sulfate, with sulfate more closely linked to precipitation. An application of a multivariate version of the model showed that the spatial structure of the sulfate fields is not due solely to the spatial structure of the precipitation fields.