Stream flow simulation is often challenging in mountainous watersheds because of irregular topography and complex hydrological processes. Rates of change in precipitation and temperature with respect to elevation often limit the ability to reproduce stream runoff by hydrological models. Anthropogenic influence, such as water transfers in high altitude hydropower reservoirs increases the difficulty in modeling since the natural flow regime is altered by long term storage of water in the reservoirs. The Soil and Water Assessment Tool (SWAT) was used for simulating stream flow in the upper Rhone watershed located in the south western part of Switzerland. The catchment area covers 5220 km2, where most of the land cover is dominated by forest and 14 % is glacier. Stream flow calibration was done at daily time steps for the period of 2001–2005, and validated for 2006–2010. Two different approaches were used for simulating snow and glacier melt process, namely the temperature index approach with and without elevation bands. The hydropower network was implemented based on the intake points that form part of the inter-reservoir network. Sub-basins were grouped into two major categories with glaciers and without glaciers for simulating snow and glacier melt processes. Model performance was evaluated both visually and statistically where a good relation between observed and simulated discharge was found. Our study suggests that a proper configuration of the network leads to better model performance despite the complexity that arises for water transaction. Implementing elevation bands generates better results than without elevation bands. Results show that considering all the complexity arising from natural variability and anthropogenic influences, SWAT performs well in simulating runoff in the upper Rhone watershed. Findings from this study can be applicable for high elevation snow and glacier dominated catchments with similar hydro-physiographic constraints.