Résumé

Every day, some decisions must be made by Hydro-Québec about the quantity of hydroelectricity produced. These decisions are based on forecasts of water supplies in watersheds using hydrological models. These models are influenced by several factors, including the presence or absence of snow on the ground. Indeed, this information is critical during the spring melt to anticipate future contributions. It is therefore necessary for forecasters to monitor snow cover daily. Methods for mapping snow cover are being used at the Hydro-Québec Research Institute (IREQ), but they have some shortcomings.

The objective of this project is to use passive microwave data (GTV) with a statistical approach to produce snow mapping. To do so, a Bayesian probabilistic model of mixture of distributions has been developed to classify the GTV in “Snow” or “No snow” and to quantify the uncertainty of this classification.

Mylène Teasdale