Rainfall-Runoff modeling by using M5 model trees technique: an example of Tigris catchment area in Baghdad, Middle of Iraq

Abstract

This paper investigates the applicability of the M5 model trees technique to emulate rainfall-runoff transformation of Tigris catchment area in Baghdad city, Middle of Iraq. For building M5 model, a free of charge open –leading machine learning and data mining weka software is used. Four models are build firstly to study the interdependency among the input variables and to select the effective variables. The applicability of this technique is studied by predicting runoff (discharge) of Tigris River one and two months ahead. The results show the high accuracy of the M5 technique to identify low values and some of high values of flow with very high accuracy, but most of the high flows were underestimated. M5 model tree and other data-driven models could be used alone or corporation with physically-based models such as HEC-HMS to manage water resourcesof Iraq after a detailed monitor hydrological programming surveys are employed