The impact of uncertainty in climate change scenarios on projections of future water supply from the Asian water towers

The water towers of Asia, referred to also as the Third Pole for their importance in terms of cryospheric processes and magnitude of water resources from snow and ice, provide water resources to sustain the lives of nearly two billion people and the source of water for ten major river basins. Assessment of future changes in the water resources from this area is crucial because of the importance of these water sources for population needs and because of the already existing pressure on water availability and increasing water scarcity. Quantifying the response of such catchments to current and future changes in climate is complicated by two main factors: i) the scarcity of ground data, especially those on the cryosphere, which hinders both proper understanding of processes and calibration of models; ii) difficulty in capturing the large variability in climate over short horizontal distances. In addition most predictions of changes in water resources in the region are based on a few GCM scenarios used to force hydrological models. Given the complexity of the climatic mechanisms in the region, and of the monsoon in particular, it is essential that impact studies examine the entire range of expected changes in future climate scenarios. In this study we therefore examine the full range of CMIP3 GCMs to assess the robustness of projected trends in precipitation, air temperature and other variables relevant for cryospheric processes. We do this analysis for the upstream(defined as all area higher than 2000 meter), and downstream areas of the 10 large river basins in Asia that originate in the Himalayas, adjacent mountain ranges and the Tibetan plateau. Secondly, we investigate how this uncertainty in GCM projections translates into the hydrological response of the Hunza catchment in the Karakoram mountains in Pakistan. A physically-based and distributed hydrological model is set-up and calibrated based on observed discharges and MODIS snow cover maps and subsequently forced with the full range of (downscaled) CMIP3 GCMs. Downscaling is carried out using a stochastic approach that permits to establish a range of statistically probable future scenarios and provides information about the local uncertainty in the final result. In this way, we quantify uncertainty in discharge, stream flow composition, glacier evolution and snow cover which is associated with both the range of future GCMs predictions and downscaling technique. Results show that forcing hydrological impact models with only a selection of GCMS can lead to misleading predictions of changes in water resources, and errors can be as large as those associated with the uncertainty in model parameters.