Entropy-based snow network design for spring peak flow forecasting

Abstract:

In northern regions the dominant phase of precipitation is snow, this precipitation persists and accumulates throughout the winter season until freshet. Quantitative information on snow, such as snow water equivalent and snow cover extent, is essential for water resources management in northern regions. Due to the inaccessibility and remoteness of snow course locations, snow surveys are usually expensive. Therefore an efficient network design strategy is required to provide a maximum amount of information while also minimizing the network cost. In this study, an entropy-based multiobjective optimization method is applied to design a snow network by adding new stations to the existing network in the La Grande River Basin of Quebec, Canada. Three hydrologic models, Sacramento, HBV, and HSAMI, are calibrated to 12 subwatersheds in the La Grande River Basin. Pareto optimal networks are given by the multiobjective optimization by maximizing joint entropy and minimizing total correlation. Each of the potential optimal networks is then evaluated using the calibrated hydrologic models to determine the most appropriate network for spring peak flow forecasting. The proposed methodology provides useful information for designing snow network appropriate for spring peak flow forecasting, which is essential for reservoir operation.