The use of wavelet - artificial neural network and adaptive neuro fuzzy inference system models to predict monthly precipitation

Abstract

Precipitation forecasting due to its random nature in space and time always faced with many problems and this uncertainty reduces the validity of the forecasting model. Nowadays nonlinear networks as intelligent systems to predict such complex phenomena are widely used. One of the methods that have been considered in recent years in the fields of hydrology is use of wavelet transform as a modern and efficient method to analysis of signals and time series.In this study, wavelet analysis combined with artificial neural network and compared with fuzzy inference system-adaptive neural for forecasting rainfall in Vrayneh station in the Nahavand. For this purpose, the original time series using wavelet theory decomposed to multi time sub-signals, then these sub-signals as input data to the neural network was used to predict monthly flow.Obtained results showed that hybrid wavelet - neural network model outperformed than fuzzy inference system - adaptive neural model and cant used for prediction of short and long term precipitation. Also the results showed that the hybrid model of wavelet - neural network acts well in estimating the extent points.