The decision-oriented mapping of pollution using hybrid models based on statistical learning theory and geostatistics is considered