Daily runoff forecasting by deep recursive neural network

Abstract In recent years, deep Recurrent Neural Network (RNN) has been applied to predict daily runoff, as its wonderful ability of dealing with the high nonlinear interactions among complex hydrology factors. However, most existing studies focused on model structure and computational load, without considering impact from selection multiple input variables prediction. This article presents a study evaluate this influence provides method identifying best meteorological for run off model. Rainfall data have considered Principal Component Analysis (PCA) contrast, reduce dimensionality redundancy within data. Two different RNN models, long-short term memory (LSTM) gated recurrent unit (GRU) model, were comparatively runoff these inputs. study, Muskegon river Pearl taken examples. The results demonstrate that great predictions made using while is shown achieve higher accuracy than rainfall alone. PCA can improve effectively it reflect core information by classifying original into several comprehensive variables.