Multi-model approach to hydrologic impact of climate change

Multi-model approach to both downscaling and hydrologic modeling is proposed to assess the variability of climate change impact on streamflow regime in the Serpent River watershed in northeastern Canada. The proposed method includes three downscaling models, namely a statistical method (SDSM), a stochastic weather generator (LARS-WG) and a temporal neural network (TLFN) along with three hydrologic models, namely a physically based watershed model WATFLOOD, two lumped-conceptual modeling systems HBV and CEQUEAU. The downscaling models are used in parallel to downscale meteorological variables (total daily precipitation, daily maximum and minimum temperature) based on climate predictors derived from the Canadian Global Climate Model (CGCM) forced with the Intergovernmental Panel on Climate Change (IPCC) IS92a scenario. The competitive hydrologic models are validated with meteorological data from both the historical records and the downscaled predictor variables. The ensembles of flow simulations generated by the different hydrologic models demonstrate the possible range of future flow regime variability in the selected watershed. The results highlight the uncertainty due to the downscaling methods and the hydrological models, and emphasize the advantage of multi-model approach in estimating hydrologic impact of climate change at the watershed scale.