Rainfall-Runoff Models using Adaptive Neuro-Fuzzy Inference System (ANFIS) for an Intermittent River

Abstract

In the present study an Adaptive Neuro-Fuzzy Inference system (ANFIS) has been used to model the relationship between rainfall and runoff with lumped data. Intermittent runoff river system namely Kanand river in Maharashtra state, India is taken as the case study. The river has runoff response only during monsoon months (June to November). After finding the cross correlation between the rain gauge stations and runoff station; cause effect, combined and time series ANFIS models were developed using lumped data. All the ANFIS models were trained by varying the number of membership functions from one to five and it is found that all the ANFIS models performed better for three membership functions. Out of 276 monthly data sets, it is found that 70 % (193 data sets) of length for training and 30 % (83 data sets) of length for validation captured the trend better than any other combinations. It is also found that pure cause-effect ANFIS models are performing better than the time series and combined ANFIS models for the same number of time lagged input data sets.