Development of hybrid baseflow prediction model by integrating analytical method with deep learning

In recent years, the success of deep learning in many different fields of Engineering has attracted attention. Baseflow separation is one of the Engineering problems which remains difficult due to different hydro-climatic circumstances. In this study, we proposed a hybrid baseflow prediction model by combining analytical methods and deep learning algorithms. Six analytical methods were chosen and their performance was compared by different metrics. Baseflow-Lyne and Hollick algorithm (BFLOW-LHA) outperforms the others in terms of R², Mean Absolute Error (MAE), BIAS, Nash–Sutcliffe Efficiency (NSE), and Root Mean Squared Error (RMSE) metrics. The proposed model was trained using streamflow and baseflow data generated by the BFLOW-LHA with the Dawa Melka Guba dataset and then tested on prediction for the basin's remaining three watersheds. The experimental results show that the proposed model improves the prediction of baseflow as compared with BFLOW-LHA and can be used for watersheds with similar characteristics.