Use of Spatial Patterns and Their Uncertainty Estimates in the Model Evaluation Process

摘要

Regional model evaluation is a new arena in which traditional residuals analysis has too little to offer. New techniques are needed that address quantification of regional pattern comparison, accounting for uncertainty and spatial correlation. Some of the more promising techniques are discussed. One technique, kriging, reasonably addresses the uncertainty in regional patterns and provides for an objective comparison. Other techniques, while useful, require that the comparison be based on subjective judgments. More work is needed in this area of regional pattern comparison.