Modelling Lake Ice Phenology with an Examination of Satellite-Detected Subgrid Cell Variability

Lake ice was simulated for the province of Quebec, Canada, for both contemporary and future climate conditions using a one-dimensional thermodynamic ice model. The model was forced with NARR data (32 km) and both the daily IMS product (4 km) and the MODIS snow product (500 m) were assessed for their utility at determining lake ice phenology at the subgrid cell level (based on the 32 km NARR grid). Both products were useful for detecting ice-off; however, the MODIS product was advantageous for detecting ice-on, mainly due to the finer resolution and resulting spatial detail. The subgrid cell variability in ice-on/off dates was typically less than 2% of the mean, although it ranged up to 10% for some grid cells. The simulations were found to be within the satellite-detected subgrid cell variability: 62% of the time for ice-off and 80% of the time for ice-on. Forcing the model with future climate scenarios from the Canadian Regional Climate Model predicts the regional ice cover durations will decrease by up to 50 days from the current 1981-2010 means to the 2041-2070 means and decrease from 15 to nearly 100 days shorter from the current means to the 2071-2100 means.