Streamflow Prediction in the Great Lakes Basin Using Two Watershed Discretization Schemes

English Abstract:

The North American Great Lakes are fundamental to the society, economy, and environment in the United States and Canada. Consequently, a proper management of the Great Lakes water system is crucial for both countries. Large-scale distributed physically-based models are one of the key tools used to inform policy makers for integrated and sustainable water and environmental resources management. Representation of the heterogeneity in nature remains one of the main challenges for these models and is accomplished primarily via watershed discretization. In this work two watershed discretization schemes are compared for the application of a semi-distributed Land Surface-Hydrological modelling system, MESH (Modélisation Environmentale–Surface et Hydrologie), in the Great Lakes Basin. One scheme is simpler neglecting the within pixel heterogeneity. The second scheme is more complex representing within pixel heterogeneity using spatial land cover info. These schemes are compared in terms of their skill to predict streamflow as well as their calibration cost. Results from this work enables MESH modellers to evaluate the gain in model performance by adding more info and computational time, and select the better one that suits their interest at the end.