This paper compares high-elevation surface temperatures based on the Global Historical Climate Network/Climatic Research Unit (GHCN/CRU) and snow telemetry (SNOTEL) datasets, with simultaneous free-air equivalent temperatures, interpolated from NCEP–NCAR reanalysis. Mean monthly temperature anomalies from 1982 to 1999 are examined for 60 SNOTEL and 296 GHCN/CRU sites at elevations over 500 m with relatively homogenous records. The surface/free-air temperature difference ΔT (Ts − Ta) is calculated for both the SNOTEL and GHCN/CRU datasets. Topography influences the correlation between surface and free-air temperature anomalies. Physically realistic diurnal and seasonal changes in ΔT\E are illustrated. Systematic secular trends in surface temperatures, free-air temperatures, and ΔT are revealed, but the sign and magnitude of change depends on location, meaning that regional signals are weak. The Ts trends are positive for most GHCN and CRU sites, and for SNOTEL sites at night. Daytime cooling in the SNOTEL network reduces the mean daily warming trend. The Ta trends are consistently positive for both networks and are often larger than Ts. Thus mean ΔT trends are negative for both datasets. The smaller sample size in the SNOTEL dataset means that error estimates for regional signals are much wider than for the GHCN/CRU dataset. Trend difference maps identify potentially anomalous SNOTEL records. Trends show no correlation with elevation and topography. Surface trends show higher variability and account for most of the uncertainty in ΔT trends. Sensitivity of trends to time period is also discussed. Such changes in the free-air/surface temperature difference may indicate change in the energy balance of mountain areas.