h-block crossvalidation is a useful modification of the usual k-fold method when training set data are autocorrelated.