We evaluate two landscape-scale microhabitat temperature prediction models suitable for the mountain pine beetle, Dendroctonus ponderosae Hopkins (coleoptera: scolytidae). Both models are based on maximum and miniature air temperatures measured at meteorological stations. The first, 'lapse' model employs temperature observations for a single nearby weather station, adjusting maximum and minimum temperatures based on elevation differences and appropriate historical vertical lapse rates. The second, 'geographic trend surface' model is based on the locations, elevations, and daily temperature measurements of surrounding weather stations. Predicted air temperatures are adjusted with a radiance-based exposure index to estimate sub-cortical phloem temperatures. Model parameters were estimated using original field measurements at three sites, and using 25 years of regional temperature measurements for sites in the western United States. Stand air temperature and phloem predictions were validated in comparisons with four withheld weather stations, and against one year of independently measured temperatures from four forest stands. Mean errors (observed minus predicted) of daily maximum stand air temperature ranged from -3.9 to 1.5°C, while mean prediction errors for daily minimum air temperatures ranged from -6.3 to 1.7°C. The geographic trend surface model performed slightly better across a range of sites, both for maximum and minimum air temperatures. Phloem temperature predictions were generally more variable, particularly for maximum temperatures on south sides of trees. Standard deviations in prediction errors were generally lower for minimum temperatures and did not differ by model form or tree exposures.