Prediction of Flood Using Hybrid ANFIS-FFA Approaches in Barak River Basin

Abstract

Flood is a natural hydrological phenomenon caused by various meteorological events such as unusual water surge and intense or prolong rainfall leading to destruction of both life and property every passing year. In flood-prone regions, rapid and precise flood forecasting is vital. Hence, development of appropriate flood prediction models is important for decision-makers which will permit abundant time for inhabitants in perilous areas to take early precaution against severity of damage. In recent years, researchers have been working on various approaches based on artificial intelligence models such as artificial neural network (ANN) which are being widely utilized for modelling many hydrological processes, including flood models. This chapter describes application of feedforward back propagation neural network (FFBPN), adaptive neuro-fuzzy inference systems (ANFIS), and integration of ANFIS with firefly algorithm (ANFIS-FFA) to forecast monthly flood at two gauge stations of Barak River basin in Barak valley, situated in Assam, India. Performances of the models were evaluated with MSE, RMSE, and WI. Obtained results exhibited suitable agreement amid predicted and real hydrological records. ANFIS-FFA model performed best; however, it requires a huge number of constraints. Performances of ANFIS and FFBPNN models are also good, yet it entails lengthier computation period and extra modelling parameters.