Évaluation de la performance de prévisions hydrologiques d'ensemble issues de prévisions météorologiques d'ensemble
|Abstract:||Hydrological forecasts consist in the evaluation of future streamflow. For the time being, current deterministic forecasts do not give any information concerning the uncertainty, which might be a problem in a decision-making process. Ensemble forecasts are expected to fill the gap. To do so, different input parameter sets are produced by adding a small deviation to the original set in order to evaluate the uncertainty. It results in a probability distribution of the output. Since 1996, Environnement Canada produces an 8-member meteorological ensemble forecasts (PME). In july 2007, an improved model now produces a 20-member PME on a 100 km grid. Within the framework of this study, this new PME is used for the first time so as to produce hydrological ensemble forecasts (PHE). The studied watershed is the Lièvre River (Quebec) where 4 locations are set to evaluate the PHE. The 17-day study in October 2007 contains a flood event, and forecasts are produced daily with a 3-day forecast horizon. The deterministic forecast is also available and is compared with the PHE. The results are positive. Even though the deterministic forecast behaves slightly better for a 48h forecast, the PHE is better for a 72h forecast, especially for 1 site. Meanwhile, an other site appears to behave strangely and its results must be taken carefully. The uncertainty applied to the input set is under-dispersed and biaised, which means that the performance of the PHE might be increased with an appropriate calibration. In order to generalize these results, more similar experiments have to be done, by working on other watersheds, with other model and/or other time periods.|
|Document Type:||Mémoire de maîtrise|
|Open Access Date:||13 April 2018|
|Collection:||Thèses et mémoires|
All documents in CorpusUL are protected by Copyright Act of Canada.