Improving runoff forecasting by input variable selection in Genetic Programming

Abstract

Determining the relationship between rainfall and runoff for a watershed is one of the most important problems faced by hydrologists and engineers. This relationship is known to be highly complex with strong correlation between the model parameters. In any model development process, the selection of appropriate model inputs is extremely important. Many authors in the past have attempted to address the issue of selecting the most relevant parameters of a given data set based on sensitivity analysis, yet the effect of interaction of variables is not clearly expatiated. In this study, we use the Group Method of Data Handling (GMDH) technique for selecting the significant variables to be used as input to Genetic Programming, which leads to improved runoff forecasting. The main advantage of GMDH technique is that it considers the interaction amongst the variables while selecting the ones that are significant. Copyright ASCE 2004.