Lake ice phenology parameters (dates of ice onset and thaw) provide an integrative climatic description of autumn to springtime conditions. Interannual variations in lake ice duration and thickness allow estimates of local climatic variability. In addition, long-term changes in lake ice phenology may provide a robust indication of climatic change. The relationship between lake ice and climate enables the use of process-based models for predicting the dates of freeze-up and thaw. LIMNOS (Lake Ice Model Numerical Operational Simulator) is one such model, which was originally designed to simulate the ice phenology of several lakes in southern Wisconsin. In this study, LIMNOS is modified to run globally on a 0.5¿ by 0.5¿ latitude-longitude grid using average monthly climate data. We initially simulate the ice phenology for lakes of 5- and 20-m mean depths across the northern hemisphere to demonstrate the effects of lake depth, latitude, and elevation on ice phenology. To evaluate the results of LIMNOS we also simulate the ice phenology of 30 lakes across the northern hemisphere which have long-term ice records. LIMNOS reproduces the general geographic patterns of ice-on and ice-off dates, although ice-off dates tend to occur later in the model. Lakes with extreme depths, surface areas, or precipitation are simulated less accurately than small, shallow lakes. This study reveals strengths and weaknesses of LIMNOS and suggests aspects which need improving. Future investigations should focus on the use of geographically extensive lake ice observations and modeling to elucidate patterns of climatic variability and/or climate change. ¿ 1998 American Geophysical Union