Genotype-by-environment interaction (GxE) remains to be a major issue in plant and animal breeding as well as in evolutionary biology. Over the past decades, voluminous GxE literature has focused on the development of statistical methods to describe rather than to predict the GxE variability. In this presentation, we will describe a new geostatistical method to predict the GxE variability based on spatial and ecological variables that are easily available or derivable from the geography of environments and/or the origin of genotypes. The method will be illustrated using empirical data from a set of barley breeding trials and a set of pine provenance trials in western Canada.