Daily gauge-discharge estimation using ANN and wavelet-ANN models for Muri station, Jharkhand

The artificial neural networks (ANNs) and wavelet-based artificial neural networks (WANN) models were considered for time series modeling of daily discharge for Muri site situated at Subernrekha river basin of Jharkhand. The observed daily time series for gauge and discharge data from 2009 to 2013 were used for simulation of input and output data in which 70% of data were used for train the model while 30% were used for test the model. The input for gauge discharge modeling was selected from different combinations made from lagging the Gauge and Discharge data by using the gamma test. The best input combination with least gamma value was used as input for both the models. The correlation coefficient (r) and mean square error (MSE, cumec) was adopted to evaluate the model's performance. For the best WANN model, the value for 'r' and MSE was 0.9667 and 0.000206 respectively For the best ANN model, the value for 'r' and MSE is 0.94892. and 0.001960 respectively. Based on the comparison of both algorithms, the WANN model has superior performance than the ANN model in estimating daily discharge.