Improved regionalisation scheme for the regionalization of parameters of rainfall runoff models

抄録

The prediction of stream flow at ungauged sites is one of the fundamental challenges which are often met through a process of parameter regionalisation. The reliable estimation of continuous stream flow in ungauged catchment has remained a largely unsolved problem so far (Wagener et al., 2004). The performance of five different types of regional model i.e., multiple linear regressions (MLR), multiple polynomial regression (MPR), artificial neural network (ANN) , partial least square regression (PLSR) and indirect calibration method to regionalize the parameters are investigated in this study. In addition, the study proposes the methodology to constrain the result of regionalization scheme with an assumption that if sufficient constraints on parameters are imposed the inconsistencies between different regionalization schemes can be reduced with better performance.