Predicting natural streamflows in regulated snowmelt-driven watersheds using regionalization methods
Abstract. Predicting streamflows in snow-fed watersheds in the Western United States is important for water allocation. Since many of these watersheds are heavily regulated through canal networks and reservoirs, predicting expected natural flows and therefore water availability under limited data is always a challenge. This study investigates the applicability of the flow duration curve (FDC) method for predicting natural flows in gauged and ungauged snow-fed watersheds. Point snow observations, air temperature, precipitation, and snow water equivalent, are used to simulate snowmelt process with SNOW-17 model and extended to streamflow generation by a FDC method with modified current precipitation index. For regulated (ungauged) watersheds, a parametric regional FDC method is applied to reconstruct natural flow. For comparison, a simplified Tank Model is used as well. The proximity regionalization method is used to generate streamflow using the Tank Model in ungauged watersheds. The results show that the FDC method can produce acceptable natural flow estimates in both gauged and ungauged watersheds under data limited conditions. The performance of the FDC method is better in watersheds with relatively low evapotranspiration (ET). Multiple donor data sets including current precipitation index are recommended to reduce uncertainty of the regional FDC method for ungauged watersheds. In spite of its simplicity, the FDC method can perform better than the Tank Model under minimal data availability.