Cross-validation error by resolution. Mean RMSE of linear regression (black), universal kriging (blue), and Bayesian kriging (red) models for T
(B), and precipitation (C). Mean RMSE is significantly different between models at all resolutions for T
(Bonferroni-corrected, paired t test, α = 0.05). T
mean RMSE is significantly lower for the linear regression model at all resolutions. Universal and Bayesian kriging models are significantly different at 3,000, 5,000, and 10,000 meters. Precipitation mean RMSE was significantly higher for linear regression models than for universal or Bayesian kriging models, which were not significantly different at any resolution.