Modified SWAT to Forecast Water Availability in Mediterranean Mountainous Watersheds with Snowmelt Dominated Runoff

The assessment of the hydrological response to projected changes in climatic variables is imperative for water resources management, especially in watersheds where snowmelt represents a significant source of runoff. In this study, we modify the source code of the snow accumulation and melting algorithm of the Soil and Water Assessment Tool (SWAT) model to improve runoff simulations in snow dominated basins. A sinusoidal snowmelt function under the degree-day factor method was adopted with its parameters calibrated based on historical data. River flow simulations were compared to measured data under the modified and unmodified SWAT models. Model differences in future predictions of river flows (2032- RCP 4.5) were also assessed. The results showed that the modifications improved runoff simulations by better capturing flow dynamics as represented by daily flows and corresponding variability during the snowmelt period. The modified model increased the Nash–Sutcliffe Efficiency (from 0.64 to 0.79; 0.60 to 0.80; 0.70 to 0.75) and the coefficient of determination (from 48 to 67%, 48 to 69%, 58 to 70%) at three gauging stations. While both models predicted a decrease of water availability in the basin, future simulations with the modified snowmelt algorithm predicted that the drop in water availability as compared to baseline year (2008) will be less dramatic (24%) compared to predictions from the unmodified SWAT (31%). We argue that the proposed source code modifications to the snowmelt algorithm of SWAT provide better insights about future water availability in snow-dominated watersheds.