Identification of homogeneous regions for runoff estimation in ungauged basins

Abstract

This paper addresses the problem of defining appropriate characteristic which leads to homogeneous regions identification, required for application of multiple nonlinear regression methodology for interpolation of hydrologic data from gauged to ungauged basins. The underlying methodology of multiple nonlinear regression is based on theory of standardization and normalization of correlation. In order to obtain synthetic runoff series at ungauged basins, one of the key steps is estimation of spatial correlation function. There are theoretical and empirical correlation functions. While empirical correlation function is estimated according to correlation coefficients of runoff series at gauged basins and their mutual distance, its pair - theoretical correlation function is merely a shift of empirical correlation function. This shift is made to achieve correlation coefficient of one for zero distance. However, this shift is acceptable only if performed in homogeneous region comprising of basins which behave in a hydrologically similar way. A basin characteristic that describes hydrological behavior similarity is examined among morphologic characteristic, basin proximity, and runoff time series statistics. There are two criteria for homogeneity control. They refer to empirical and theoretical spatial correlation function, and eventually identify donor catchments for ungauged basin. Study area for research is the territory of Republic of Serbia without provinces of Vojvodina and Kosovo. Basins up to area of 1000 km2 are taken into consideration. Within the period 1961-2005, three sub-periods of 15 years each are examined for mean monthly runoff time series simulation potential at ungauged basins.