An atmosphere-ocean-land coupled global climate model at high horizontal resolution (∼0.5° land mesh) with a five-layer, 4.0 m deep soil was evaluated as a tool to simulate changes in the distribution of frozen ground and subsurface hydrothermal regimes under the global warming scenario. The land scheme explicitly treats soil freezing/thawing processes, surface energy exchange, and water fluxes, including snow cover effects. Modeled soil temperature showed large cold biases in the cold seasons, especially in high latitudes, which likely resulted from shallow and simplified soil column with zero heat flux condition at the bottom and insufficient snow representation. Two types of frozen ground were classified according to monthly soil temperatures: “permafrost” for regions with the maximum active layer thickness less than 4 m and “seasonally frozen ground.” Simulated present-day (1980–1999) distribution is in good agreement with observational estimates for both. Under climatic warming forcing, projected change in land and subterranean hydrothermal regimes shows amplification in higher latitudes. Approximately 60% of the present-day permafrost regions would degrade into seasonally frozen ground by 2100. In the circum-Arctic basins, increased precipitation would lead to a 13.7% increase in freshwater discharge into the Arctic Ocean. The limitations of the current model configuration and the implied reliability of the results are discussed, and the future improvements and potentials of such modeling approach are also presented. A medium-resolution (∼2.8° mesh) ensemble produces qualitatively similar results for the hydrothermal regimes in the cold regions, although some local features were inevitably smoothed out, giving different quantitative estimates.