The province of Quebec in Canada has abundant fresh water resources. These resources represent a considerable interest for the province, including regarding the production of hydroelectricity. The production of such energy involves understanding well the hydrological cycle in regions and river basins of interest. The study of the hydrological cycle is usually carried out through the analysis of meteorological and hydrological data, measured with weather stations, and/or with the help of hydrological modelling. However, the density of weather stations is very irregular over the province of Quebec. In fact, several stations cover the south of the province, whereas they are very few in the north.
Nevertheless, there are gridded alternative datasets that cover regions of interest entirely with gridded data, such as gridded observational datasets and reanalyses. However, observational gridded datasets may reveal questionable data in regions where interpolations have been performed with a very few weather stations. Reanalyses are generated similarly as weather forecast model outputs, with the difference that they are not updated during the production data process. Moreover, reanalyses provide datasets covering the recent past (until the present day for some of them) from the assimilation of observation data measured by different devices including satellites, boats, radio soundings, aircrafts, buoys or weather stations.
The objectives of this study are to evaluate the potential of reanalyses to represent the terrestrial hydrological cycle in the province of Quebec, and to assess their performance in hydrological modelling over river basins of the province. The analysis of the hydrological cycle of the four reanalyses CFSR, NARR, MERRA and ERA-Interim is performed according to three different spatial scales: (1) over the entire province, (2) according to the five climatic regions of Bukovsky over the province, and (3) at the watershed scale, according to six hydrological regions defined beforehand. Results show that each reanalysis has its own strengths and weaknesses. Globally, ERA-Interim seems to provide the most reliable precipitation, whereas NARR presents the less conclusive results. Contrary to the three other reanalyses, MERRA shows a closed water budget. However, this reanalysis does not distribute the precipitation into runoff and evaporation according the expected amounts of water. This last aspect is well respected by CFSR, but this reanalysis also reveals the highest precipitation over the province of Quebec, among the four reanalyses.
Following this study, the performance of the four reanalyses in hydrological modelling has been evaluated with the global conceptual HSAMI model. The performance has been assessed using the Nash-Sutcliffe criteria over 106 river basins of the province of Quebec, during the 1979-2008 period, and according to the six hydrological regions. Thirty (30) calibrations have been performed, as well as one validation with the parameters that provided the best Nash-Sutcliffe criteria in calibration for each river basin. Results indicate globally a good capacity of reanalyses to reproduce the observed river discharges, including in northern regions of the province, where the Nash-Sutcliffe criteria varies from 0.6 to 0.9. Among the four reanalyses, ERA-Interim presents the highest Nash-Sutcliffe value distributions, although its superiority is not statistically significant. On the other hand, the questionable precipitation of NARR over the province of Quebec influences the performance of this reanalysis downward, and thus reveals the lowest performance among the four reanalyses. MERRA and CFSR show quite similar results over the six hydrological regions of the province.
In conclusion, reanalyses show promising results both in their potential to represent the hydrological cycle and in their performance in hydrological modelling. Moreover, in order to validate their applicability regarding hydrological studies, it is recommended to strengthen these results with additional comparisons between reanalyses data and weather stations data, especially over the northern regions of the province. Finally, it would be interesting in future studies to assess the hydrological performance of reanalyses using different types of hydrological models (distributed models, and/or physical basis) and other assessment criteria, as the RMSE or criteria based on extreme events.
| Date | 28 Sept 2015 |
|---|
| Original language | French |
|---|
| Awarding Institution | - École de technologie supérieure
|
|---|
| Supervisor | Annie Poulin (Supervisor) |
|---|