Integration of a Parsimonious Hydrological Model with Recurrent Neural Networks for Improved Streamflow Forecasting

Ye Tian,Yue-Ping Xu,Guoqing Wang,Zongliang Yang,Qian Zhu

doi:10.3390/w10111655

Abstract

This study applied a GR4J model in the Xiangjiang and Qujiang River basins for rainfall-runoff simulation. Four recurrent neural networks (RNNs)—the Elman recurrent neural network (ERNN), echo state network (ESN), nonlinear autoregressive exogenous inputs neural network (NARX), and long short-term memory (LSTM) network—were applied in predicting discharges. The performances of models were compared and assessed, and the best two RNNs were selected and integrated with the lumped hydrological model GR4J to forecast the discharges; meanwhile, uncertainties of the simulated discharges were estimated. The generalized likelihood uncertainty estimation method was applied to quantify the uncertainties. The results show that the LSTM and NARX better captured the time-series dynamics than the other RNNs. The hybrid models improved the prediction of high, median, and low flows, particularly in reducing the bias of underestimation of high flows in the Xiangjiang River basin. The hybrid models reduced the uncertainty intervals by more than 50% for median and low flows, and increased the cover ratios for observations. The integration of a hydrological model with a recurrent neural network considering long-term dependencies is recommended in discharge forecasting.

Highlights

Hydrological models are used to predict streamflows, ground water content, evapotranspiration, and other hydrological variables
The statistical indices for both basins show that the NARX and long short-term memory (LSTM), which both have long-term memory in their architecture, provide the most accurate estimates, followed by echo state network (ESN)
The NARX and LSTM are more accurate than GR4J for the Xiangjiang basin

Summary

Introduction

Hydrological models are used to predict streamflows, ground water content, evapotranspiration, and other hydrological variables. Hydrological models can present rainfall-runoff processes and broaden our understanding of rainfall-runoff. Errors and uncertainties in hydrological modeling are inevitable due to limited knowledge of the hydrological system, forcing and response data, and simplification of the real world [1,2]. Prediction errors are often presented as underestimation or overestimation in the discharge predicted by a hydrological model. Errors can measure the performance of a model and serve as bases to improve the prediction accuracy against observation by regression analysis. Errors contain information about both observations and models; they can be used to improve forecasting [3,4,5].

Objectives

Methods

Results

Discussion

Conclusion