Environmental Thematic Map Prediction and Easy Probabilistic Classification within GRASS
Environmental observations are usually sampled at irregularly spaced points, but, in most cases, are representative of continuous phenomena; therefore, to restore the original information a continuous thematic map should be interpolated from the available observations. To interpolate the thematic map, the stochastic approach is often the best method to apply: it is based on the hypothesis that the observed phenomenon is the sum of a deterministic trend and a stochastic process. Usually, the stochastic approaches implemented in GIS software require a priori knowledge of the deterministic trend: a simplified and automated approach to the trend model selection has been studied. One of the main advantages of the stochastic prediction approaches is that they make available both the prediction map (thematic map) and a related prediction error map; starting from these two information layers a procedure has been implemented to verify if in some areas the predicted data are within an assigned range or exceed a threshold value at some certain degree of probability. The procedures were developed by creating GRASS commands calling upon and integrating the Gstat statistics program, and commands which allow the probabilistic computing of risk areas.
Document Type: Research Article
Affiliations: Politecnico di Milano Polo Regionale di Como
Publication date: 2004-04-01