We used daily air and water temperatures from 14 lakes in Ontario to develop and test a simple method for constructing lake-specific empirical models for predicting daily littoral water temperatures. Data requirements for prediction are modest (average air temperature, day of the year) and the method is robust and practical, requiring only a few (15–20) well-spaced water temperature observations to construct a single-year model that can generate reasonably accurate predictions for an entire ice-free season. The ability of a singe-lake model to predict several years of temperature data is significantly improved by explicitly including information on ice-out date in the model. Our multiyear model for Lake Opeongo described most (18 of 22) years well. Years that were not well described were usually coincident with El Nino Southern Oscillation events. Used with caution, the method can be an effective tool for supplementing direct monitoring of littoral water temperatures and for generating historical water temperature estimates when direct estimates are lacking. These capabilities should be of particular use to fisheries biologists studying or managing populations of fish species with critical life stages that are affected by littoral water temperatures.