Overwintering habitat for Arctic freshwater fish is essential, such that understanding the distribution of winter habitat quality at the landscape-scale is warranted. Adequate dissolved oxygen (DO) is a major factor limiting habitat quality in the Arctic region where ice cover can persist for 8 months each year. Here we use a mixed-effect model developed from 20 lakes across northern Alaska to assess which morphology and landscape attributes can be used to predict regional overwintering habitat quality. Across all lakes, we found that the majority of the variations in late winter DO can be explained by lake depth and littoral area. In shallow lakes (<4 m), we found evidence that additional variables such as elevation, lake area, ice cover duration, and snow depth were associated with DO regimes. Low DO regimes were most typical of shallow lakes with large littoral areas and lakes that had high DO regimes often were lakes with limited littoral areas and deeper water. Our analysis identifies metrics that relate to late winter DO regimes in Arctic lakes that can aid managers in understanding which lakes will likely provide optimum DO for overwintering habitat. Conversely, lakes which predicted to have marginal winter DO levels may be vulnerable to disturbances that could lower DO below critical thresholds to support sensitive fish. In regions where lakes are also used by humans for industrial winter water supply, such as ice-road construction for oil and gas development, these findings will be vital for the management of resources and protection of Arctic fish.