The Spatial Variability of Prediction Errors

Abstract

This paper is about the estimation of error in spatial problems, how we do it and how we might do it. It is also about our methods of interpreting error estimates to our scientific audience. We have two specific objectives in mind. The first objective is to elucidate the role of averaging or pooling, i.e., the role of statistics, in the reporting of error estimates. Our discussion of pooling is largely philosophical and is focused on the proper interpretations that should be given to error estimates. The second objective is to emphasize the importance of viewing errors as a spatially varying field. Here we will have some specific suggestions which affect how we usually estimate errors in geostatistical problems.