Using CloudSat-CALIPSO ice water, cloud fraction, and radiation; Clouds and the Earth’s Radiant Energy System (CERES) radiation; and long-term station-measured surface air temperature (SAT), we identified a substantial underestimation of the total ice water path, total cloud fraction, land surface radiative flux, land surface temperature (LST), and SAT during Northern Hemisphere winter in Coupled Model Intercomparison Project Phase 5 (CMIP5) models. We perform sensitivity experiments with the National Center for Atmospheric Research (NCAR) Community Earth System Model version 1 (CESM1) in fully coupled modes to identify processes driving these biases. We found that biases in land surface properties are associated with the exclusion of downwelling longwave heating from precipitating ice during Northern Hemisphere winter. The land surface temperature biases introduced by the exclusion of precipitating ice radiative effects in CESM1 and CMIP5 both spatially correlate with winter biases over Eurasia and North America. The underestimated precipitating ice radiative effect leads to colder LST, associated surface energy-budget adjustments, and cooler SAT. This bias also shifts regional soil moisture state from liquid to frozen, increases snow cover, and depresses evapotranspiration (ET) and total leaf area index in Northern Hemisphere winter. The inclusion of the precipitating ice radiative effects largely reduces the model biases of surface radiative fluxes (more than 15Wm-2), SAT (up to 2-4K), and snow cover and ET (25-30%), compared with those without snow-radiative effects.