A normalized regression based regional model for generating flows at ungauged basins

Revision of existing methodologies for generating monthly-flow series at ungauged basins based on multivariate nonlinear correlation has led to a simple two-parameter model. While the existing methodology used hydrological, meteorological and geomorphologic input data, the proposed model requires hydrological and geomorphologic input data only. The proposed methodology requires formation of separate pools of donor catchments for model parameter estimates. The proposed two-parameter model and improvement in the sphere of homogeneous region identification were verified using 195 runoff data sets from hydrologic stations in Serbia in the 1961–2005 period, divided into three non-overlapping 15-year periods. Nash-Sutcliffe's model efficiency coefficient (NSE) was used to assess the: (1) quality of proposed model with identified model parameters; (2) quality of a nonlinear multivariate equation for standard normal variables estimation with identified model parameters; (3) quality of proposed model with model parameter estimates. Generated time-series statistics and nonlinear multivariate equation quality are also evaluated. Five model calibration and validation results are shown. Generated flow series variation coefficient is the best replicated statistics with relative absolute error less than 10%.