Table 1. Cross-validation performance results.

Model	k	RMSPE	PIC90	PIC95
OK	4	18.55	0.900	0.962
LM	49	18.20	0.882	0.946
SLM	37	17.65	0.886	0.948
LM-TF	67	17.41	0.883	0.944
SLM-TF	51	16.82	0.882	0.937
LASSO	72	18.04
LASSO-TF	134	16.98
RF		16.52	0.914	0.955
RFRK		16.41	0.905	0.960

k is the number of parameters from the model fit using the entire data set (i.e., all n = 1859 observations); PIC90 and PIC95 are the coverages of the 90% and 95% prediction intervals; OK is the ordinary-kriging model; LM-TF, SLM-TF, and LASSO-TF are the LM, SLM, and LASSO with covariate transformations, respectively; and other abbreviations are defined in the text. Computation of prediction intervals for LASSO is not supported by the glmnet package.