Comparative evaluation of neural network training algorithms and deep learning models for long-term streamflow forecasting using climate indices

Using Niño3.4 lags, ANN-LM, ANN-BR, ANN-SCG and LSTM predicted monthly Goulburn River streamflow at 1-, 3- and 6 month leads; performance was assessed by R , RMSE, MAE and MAPE; ANN LM performed best, confirming ENSO influence up to 6 months.