Utilisation des données du MRCC15 pour la détermination de la distribution spatiale optimale du réseau d’observations météorologiques en modélisation hydrologique distribuée et globale

Abstract

The objective of this study is to determine the optimal density of a weather station network when being used for hydrological modelling. Since there are no dense enough networks in sparsely populated areas, such as in Northern Canada, to perform such a study, a virtual world was used. Data from the Canadian Regional Climate Model at 15 km resolution (CRCM15) was used to create a virtual network of stations with long and complete series of meteorological data over the Toulnustouc River basin. Three hydrological models were used in this study. Two are lumped (HSAMI and HMETS) while the last is distributed (Hydrotel). The weather stations to be fed to the models were selected in order to minimize the number of stations while maintaining the best hydrological performance possible. A multi-objective genetic algorithm was put in place to determine which stations were to be used, and by the same occasion, where the stations should be located. It was shown that the number of stations making up the network on the virtual Toulnustouc River basin should be at least two (2) but not higher than five (5), no matter what hydrological model is chosen. The optimization algorithm showed that combinations of two or three stations can result in better hydrological performance in calibration than if a high density network was fed to the models. Furthermore, it was shown that a high number of stations will definitely reduce the variance related to the selection of the stations to be used. Two conclusions can be reached through this study: 1) If a basin must be fitted with a weather network, it is possible to install a relatively small amount of stations in strategic locations and still get as good hydrological model performance in calibration mode as if a high density network was used; 2) If stations must be positioned randomly on the river basin and if financial resources are plentiful, it is advantageous to build a network as dense as possible to reduce model output variability.

Date	4 May 2012
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	François Brissette (Supervisor)