The thinning and reduction of extent of Arctic sea ice increase the potential for ocean- atmosphere coupling, especially through surface energy exchanges and wind-induced mixing caused by storm events. The proposed project will address high-latitude storms and coupling with sea ice as simulated by two state-of-the-art models with regional high-resolution capabilities: the Department of Energy’s Energy Exascale Earth System Model (E3SM) and the Weather Research and Forecasting Model (WRF). The models’ storm characteristics and storm/sea ice coupling will be examined using simulations for three time frames: historical, 20th- Century with prescribed increases of greenhouse gases, and late-21st Century. The analysis will be based on output from DOE-archived E3SMv1 simulations and on simulations to be performed as part of the project using regionally refined configurations over the North Pacific/subarctic and the Southern Ocean. We will test hypotheses that summer storms increase the melt of sea ice by enhanced mixing and downwelling longwave radiation, that the loss of sea ice favors more intense atmospheric storms and a northward shift of storm activity, and that corresponding changes in the Antarctic will be limited primarily to the Southern Hemisphere winter season. The analysis of the model simulations will include evaluations of the large-scale modes of variability (Arctic Oscillation, El Nino/Southern Oscillation, Pacific Decadal Oscillation) known to be associated with variations of storm activity in both hemispheres. The Northern Hemisphere focus will be on the North Pacific sector and will include evaluation of cloud and precipitation properties, surface energy fluxes, and associated poleward moisture transports. The Southern Hemisphere analysis will span all longitudes because of the zonally symmetric land-sea configuration. Particular attention will be paid to the models’ ability to simulate smaller-scale cyclonic systems such as polar lows and tropopause polar vortices, especially the extent to which simulations of these systems are sensitive to model resolution and the choice of the physical parameterization schemes used in the models.