Initial Analysis of Water Budget in Koshi Basin, Nepal: Assumptions and limitations D.J. Penton a , L.E. Neumann a , M. Dolk b , H. Zheng a a CSIRO Land and Water, Australia, b University of Oxford, United Kingdom Email: dave.penton@csiro.au Abstract: A growing number of studies use hydrological and cryospheric models to understand the water budget of the Himalayan basins of Nepal. These studies are important because they inform strategic basin development and investment in infrastructure; e.g. the proposed storage-based hydropower schemes for the Dudh Koshi River, currently under assessment by the Asian Development Bank. For large infrastructure schemes, understanding the development options requires hydrological models to explain how one proposed scheme may affect others, and the cumulated impacts of proposed schemes on downstream riparian water access and ecological function. These models also need to provide insights on the resilience of the system to the changing climate, and the climate’s effect on water stored in glaciers. This paper provides an initial water budget of the Koshi Basin region. It discusses the assumptions that are made, and how estimated water availability under climate change is sensitive to these assumptions. The contribution of this paper is to highlight known model limitations and provide a reasonable synthesis of available hydrological information. The method for undertaking the analysis includes: modelling of rainfall and snow extents; developing a hydrologic model with snow and glacier components; regional calibration of snow parameters, and local calibration of hydrological parameters; and sensitivity analysis of selected parameters. We estimated daily rainfall at locations below 3000mASL using observations from 288 rainfall stations. Independent cross- validation showed average of 20% bias, which was sufficient for the regional scale analysis. However, the lack of data at higher elevations allowed erroneous trends to emerge, so we relied on global reanalysis datasets above 3000mASL (0.5 degree grid) because the outputs were constrained by the modelled physics. Despite being better, the global datasets have limitations of resolution and accuracy. MODIS snow cover products allowed the snow processes to be constrained in the hydrological modelling, but they suffered from cloud cover issues. These were partially addressed through the use of Hidden Markov Model approaches to filter noise and provide a daily time-series of snow extent. A GR4J model with additional snow and glacier melt model (GR4JSG) was applied to six alpine catchments (Nepal et al, 2017). The model was able to reflect stream-flow (median 10% bias) and snow-extent characteristics (median 18% bias). Some parameter values did not make conceptual sense for the catchment (low degree day factors, large conceptual soil storage ~ x1 , large groundwater exchange term ~ x 2 ) and probably represent poor input data. For the Tibetan Plateau, many assumptions were required about the correct model structure and parameterisation given poor input climate records and no streamflow measurements on the Plateau itself. Our initial analysis is strongly affected by assumptions of glacial change. This study estimated that snow contributes around 7% (3500 MCM) of the annual streamflow of the Koshi River as it exits Nepal at the border and ice melt contributes a further 3% (1500 MCM). The major carriers of water are the Arun (33%) and Sun Koshi Rivers (49%). Runoff is being generated predominantly at elevations between 2000mASL and 4000mASL – around the same elevation as snow fall is greatest. The greatest uncertainty remains the rainfall at high elevations and the appropriate parameterisation and conceptualisation of snow and glacier models. Keywords: Himalayan water budget, cryosphere, Koshi Basin 22nd International Congress on Modelling and Simulation, Hobart, Tasmania, Australia, 3 to 8 December 2017 mssanz.org.au/modsim2017 1766