Seasonal rainfall pattern using coupled neural network-wavelet technique of southern Uttarakhand, India.

Hydrological data is crucial for accurate forecasting of precipitation which can be used for water resources planning and management. The purpose of this study is to develop a seasonal rainfall forecast model, using a hybrid wavelet-artificial neural network (WANN) model based on regression analysis to predict seasonal rainfall in Almora, Lansdown, Kashipur and Mukteswar region in Uttarakhand (India).The statistical results shows that the mean maximum rainfall was found to be 746.82 mm, 1586.58 mm, 1060.53 mm and 964.43 mm for Almora, Lansdown, Kashipur and Mukteswar, respectively. The models WANN-03 (Network 4–8-1), WANN-10 (Network 4–7-1), WANN-10 (Network 4–7-1) and WANN-15 (Network 4–8-1) were found to be the most efficient models for Mukteswar, Lansdown, Kashipur and Almora, based on the high coefficient of determination (R²) and coefficient of efficiency (CE) values and low root mean square error (RMSE) values that were obtained using each model. For each season, four WANN modelshave been developed (total of sixteen models) by varying the number of hidden neurons. The results shows that only one WANN model was not sufficient to predict the rainfall of all stations. Every station has a specific networked model which could model the data more precisely preciously. The findings illustrated that the hybrid model of WANN having Network (4–7-1) was found most superior model (R² = 0.857, RMSE = 32.192 and CE = 0.846) for the Lansdown stations among all the stations. [ABSTRACT FROM AUTHOR]