Streamflow estimation in ungauged catchments using regionalization techniques

Abstract

Continuous streamflow regionalization in ungauged catchments is contemplated as a challenging task. In a developing country like India where the subject of prediction in ungauged basins (PUB) is not prevalent, thirty-two catchments were specified in order to analyze continuous streamflow estimation. Spatial proximity (Inverse Distance Weighted, Kriging, and global mean), regression and physical similarity were the regionalization approaches implemented in conjunction with SWAT (Soil and Water Assessment Tool) for streamflow estimation in each catchment treated as ungauged in turn. Kriging and IDW were the two methods that produced superior results than other applied techniques in terms of Nash-Sutcliff Efficiency (NSE), RMSE-observations standard deviation ratio (RSR) Percentage bias (PBIAS) and Peak percent threshold statistics (PPTS). Physical similarity and regression approaches, which were based on catchment attributes exhibited better results than global mean approach. Sequential Uncertainty Fitting (SUFI-2) tool analyzed the uncertainty associated with regionalization techniques in terms of 95% prediction uncertainty (95PPU). The comparative assessment proposes that presence of well-gauged catchments in proximity with the ungauged catchment is more beneficial than catchments having resemblance between them in terms of physiographic attributes for continuous streamflow prediction.