Runoff Analysis with the Method of Robust Regression

Abstract

Runoff analysis for actual usage requires that a runoff model should be constructed at anyplace discharge data are measured, even so, the data are good or bad. If a model is strong for these bad data, the model is called ‘robust’. This paper proposes a new runoff analysis method, the RFMD method, using the robust multiple regression analysis. The method uses the Biweight method of the M estimation method. A new concept, RFMD, is introduced according to FMD. The method was applied to the Lies 21 km² dam basin in Yamagata Prefecture. This method gave good results in estimating the discharge. With a comparison with the FMD method, the RFMD method can give similar accuracy as the FMD method, but robustness in much stronger. For separation of the linear runoff, a component of linear runoff of the RFMD method is more independent to RFMD than that of the FMD method. From this point, the RFMD method is superior to the FMD method.