Streamflow Prediction with Deep Learning

Abstract

In this study, some part of streamflow modelling and data analysis carried out within the frame of a comprehensive project on the web-based development of a watershed information system is reported. Flowrate prediction is a challenging work because it involves nonlinear, chaotic, multidimensional, instantaneous and continuous processes. The study basically aims to present the daily discharge predictions from the actual discharge using Recurrent Neural Networks (RNNs) as a deep learning approach. RNN is back ended by the LSTM (Long Short-Term Memory) and improved by an Adam optimization algorithm. The initial results are found promising compared to those of conventional Artificial Neural Network (ANN) models.