The secret assumption of transfer functions: problems with spatial autocorrelation in evaluating model performance [An article from: Quaternary Science Reviews]