Uncertainty of glacier runoff projection from precipitation data for High Mountain Asia

It is widely accepted that glacier runoff will increase initially due to accelerated glacier melting under warming climate, but will decrease as the stored glacier reduces. Rivers originating in glaciers at High Mountain Asia are highly dependent on those melt water. The temporal variation of glacier runoffs have the potential to influence downstream water supply in the long term. However, the glacier runoffs under climate change have only been addressed for relatively limited glaciers in HMA. Furthermore, current model-based projections are affected by considerable uncertainties. One of the largest uncertainties originates from underestimation of input precipitation data due to poor gauge network in mountainous regions. It hampers effort to simulate the glacier runoff peak timing and magnitude. Here, we developed a precipitation dataset focusing on correction of topographic bias to simulate glacier runoff by a glacier model. We showed glacier runoff peak timing and its magnitude from the simulation for all glaciers in HMA. We developed a precipitation dataset for the period 1948-2015 with a daily temporal and 0.25 spatial resolution for HMA. The long-term mean of precipitation was based on satellite dataset (TRMM2A25) and interpolated gauge precipitation dataset (APHRODITE). A correction for gauge under-catch and orographic effects was introduced by inferring catchment-average precipitation from streamflow observations and a precipitation-elevation regression. The temporal variability of our dataset was determined by APHRODITE. The glacier runoffs were calculated by the glacier model forced by our developed precipitation data and observed temperature data over historical period and forced by eight GCMs over future period under the RCP4.5 and RCP8.5. The projection of glacier runoffs for all glaciers in HIMA reveals configurations and its geographical distribution. The uncertainty of glacier runoff projection from input precipitation data will be also discussed.