Daily averaged 2D water temperature model for the St. Lawrence river

A daily averaged two‐dimensional water temperature model has been developed for the freshwater part of the St. Lawrence River, between Lake St. Louis and Trois‐Rivières (Québec, Canada). The model was first calibrated and validated for the area of Lake St. Pierre, a natural enlargement of the river subject to strong lateral and longitudinal thermal variations. Forecasts from the Global Environmental Multiscale model were used in preference to observations from meteorological stations for model inputs, both to increase the spatial resolution and ultimately to allow the water temperature model to be used in predictive mode. The resulting model provided daily water temperature estimates with an overall root mean square error (RMSE) of 1.18 °C and a Nash–Sutcliffe coefficient of 0.44. Comparisons between Landsat images and simulations demonstrated that the model not only simulated accurate water temperature values but also showed the adequacy of the model in general. It not only simulated local water temperature relatively accurately but also provided a good representation of the spatial water temperature patterns within the study area. The error varied between deep and shallow water areas. In deeper water, the overall RMSE is 0.41 °C, and the modified Nash coefficient rises up to 0.92. Because shallow water areas are subject to greater variations, longer, more spatially dense data sets will be needed to refine the hydrodynamic and thermal budget models for those specific areas.