Rainfall-runoff modelling of a watershed

Abstract

In this study an adaptive neuro-fuzzy inference system was used for rainfall-runoff modelling for the Nagwan watershed in the Hazaribagh District of Jharkhand, India. Different combinations of rainfall and runoff were considered as the inputs to the model, and runoff of the current day was considered as the output. Input space partitioning for model structure identification was done by grid partitioning. A hybrid learning algorithm consisting of back-propagation and least-squares estimation was used to train the model for runoff estimation. The optimal learning parameters were determined by trial and error using gaussian membership functions. Root mean square error and correlation coefficient were used for selecting the best performing model. Model with one input and 91 gauss membership function outperformed and used for runoff prediction.

Keywords: Rainfall, runoff, modelling, ANFIS