Spatial prediction from networks
This article defines a random-field model that can be used for the prediction of pollutants at locations where no data are available, based on data taken from a spatial network of monitoring sites. Acid deposition data collected from the UAPSP network in 1982 and 1983, are analyzed in two stages. Bias-resistant and outlier-resistant techniques are used to determine the spatial dependence; then a spatial model is built that is made up of a quadratic trend surface and the spatially correlated error. Spatial sampling plans and optimal designs for selecting monitoring sites are summarized and discussed. The question of the location of additonal sites (and deletion of existing ones) is also addressed.
Please refer to publisher version or contact your library.