Knowing that climate change will modify the quantity of available water resources and their management, more and more climate simulations are being made available to researchers. This growing number makes climate change impact studies more complex. Thus, a selection of representative climate simulations before the hydrological modelling stage would speed up the process. In this project, methods for identifying climate simulations that can be used to properly sample the field of hydrological indicators will be presented.
Two methods for selecting climate simulations, the correlation method as a predictor and the clustering method, are tested and compared to a random selection method. These two methods are applied to one hundred North American watersheds. Since climate indicators affecting the hydrology of a watershed can vary from one climate to another, 79 different climate indicators are used as a predictor of 34 hydrological indicators for each watershed. Three global and conceptual hydrological models (GR4J, HMETS, MOHYSE), calibrated using the KGE criterion, are used for watershed hydrology modeling. The climate data (daily minimum temperature, daily maximum temperature and daily precipitation) used come from climate models from three different projects (ClimEx, CMIP5 and CORDEX) at various scale resolutions, all following the same greenhouse gas emission scenario (RCP8.5). A total of 57 or 107 climate members are used, depending on the location of the watershed. Before being provided to the hydrological models, the output data from these simulations are post-processed by a bias correction method (DBC).
A correlation coefficient is calculated between each combination of climate and hydrological indicator for the two selection methods. The climate indicator obtaining the best correlation for each watershed is retained, called the best predictor. The modal indicator for these predictors was inferred and applied to all other watersheds. The similarity between the samples of hydrological indicator, obtained using the different selection methods, and the complete distributions of hydrological indicators is calculated using the Kolmogorov-Smirnov test.
The results show that the climate indicator to be used depends greatly on the hydrological model and the selection method. However, a consensus is visible for several of them. Over the annual period, for example, total precipitation is the climate indicator that best predicts the mean discharge.
The hydrological model plays a large role in the results obtained. In addition, in order to properly represent all hydrological events, a large number of climate members should be used, since the members used to obtain the best predictor are not the same from one hydrological indicator to another.
The use of modal climate indicators significantly reduces the calculation time of the hydrological modeling chain when the analysis is performed on specific hydrological indicators.
| Date | 23 Sept 2020 |
|---|
| Original language | French |
|---|
| Awarding Institution | - École de technologie supérieure
|
|---|
| Supervisor | Richard Arsenault (Supervisor) & David Huard (Co-supervisor) |
|---|
Vandal, M. (Author),
Arsenault (Supervisor) & Huard (Co-supervisor),
23 Sept 2020Student thesis: Master's thesis › Master in Engineering: Construction Engineering