Hydrological and climatic uncertainties associated with modeling the impact of climate change on water resources of small Mediterranean coastal rivers

摘要

This paper investigates the uncertainties associated with using regional climate models and one hydrological model calibrated from non-stationary hydroclimatic time series to simulate future water resources of six Mediterranean French coastal river basins. First, a conceptual hydrological model (the GR2M model) was implemented in order to reproduce the observed river discharge regimes. Climatic scenarios were then constructed from a set of Regional Climate Models (RCMs) outputs and fed into the hydrological model in order to produce water discharge scenarios for the 2071-2100 period. At last, an assessment of uncertainties associated with the hydrological scenarios is given. With respect to the 1961-1990 period, RCMs project a mean annual temperature increase of 4.3-4.5℃ (3.1-3.2℃) under the IPCC A2 (B2) scenario. Precipitation changes, although more variable, indicate a decrease between-10% and -15.6% for A2 and between -6.1% and - 11.6% for B2. As a result, the GR2M model simulates a general water discharge decrease between -26% (-14%) and -54% (-41%) for the A2 (B2) scenario, depending on the basin of interest. Sensitivity tests on the hydrological modelling revealed that the hydrological scenarios are sensitive to the choice of the PE formulation, although this climatic input is negligible in the model calibration. Also, a slight but significant drift between the modelled and observed time series was detected for most basins, indicating that the hydrological model fails to adapt to non-stationary discharge conditions. A simple correction method based on a dynamical parametrization of one model parameter with temperature data considerably reduces the model drift in half of the investigated basins. When extrapolated this new parametrization to the future climate scenarios, decrease of water discharge is found to be twice as great as estimated from the standard parametrization. Our results suggest that the uncertainties stemming from hydrological models with fixed parametrizations should be further addressed in any climate change impact study.