Hiérarchie des incertitudes sur les étiages futurs liées au climat et à l'hydrologie | Hierarchy of climate and hydrological uncertainties in low flow projections

This works aims at providing a understanding of the different types of modelling uncertainties associated with future low-flow projections. Hydrological projections come indeed at the bottom of the well-known cascade of uncertainty and therefore potentially integrate all types of modeling uncertainties related to both climate and hydrological. The specificity of low flow characteristics is their tight dependency on catchment processes, suggesting a non negligible contribution of the hydrological modelling step to the overall uncertainty. This study builds on an extensive top-down climate impact approach set up within the R2D2-2050 project. This integrated assessment project aims at informing water resource adaptation strategies for the Durance river basin located in the Southern French Alps, by confronting future hydroclimate projections to prospective scenarios of water demand and water use. Future water resource estimates have been based on 11 runs from 4 GCMs from the ENSEMBLES project under the A1B emissions scenario. These large-scale projections have been further statistically downscaled with 3 probabilistic methods based on the k-nearest neighbours resampling approach, leading to 330 spatially distributed climate projections. Downscaled projections have then been used as forcings to 6 diverse hydrological models, from global conceptual to physically-based fully distributed, in order to derive transient hydrological projections from 1961 to 2065 for up to 25 stations over the basin. Most of the Durance subcatchments are under the influence of both the Alpine and Mediterranean climates, which leads to two distinct periods of low flows, one in winter when precipitation is stored as snowpack and one in late summer after the snowmelt. Analyses have therefore been applied separately to the two periods to infer the future evolution of low flows characteristics and its consistency among the range of available projections. Several analysis of variance (ANOVA) frameworks have further been applied to quantify the respective contributions of climate and hydrological modeling uncertainties on the results and their evolution over time. Results first show contrasted evolutions according to the season and the catchment, with little evolution in winter low flows for high-elevation catchments, and a dramatic decrease in summer low flows over the whole basin. The contribution of the hydrological modelling step to the overall uncertainty is moreover limited in winter, but high in summer with a pronounced increase over time. This last result critically suggests a divergence of climate change responses between the hydrological models in terms of summer low flows, potentially due to difference in the implementation of snowmelt processes.