A Comparative Analysis of Multiple Machine Learning Methods for Flood Routing in the Yangtze River

Abstract

Obtaining more accurate flood information downstream of a reservoir is crucial for guiding reservoir regulation and reducing the occurrence of flood disasters. In this paper, six popular ML models, including the support vector regression (SVR), Gaussian process regression (GPR), random forest regression (RFR), multilayer perceptron (MLP), long short-term memory (LSTM) and gated recurrent unit (GRU) models, were selected and compared for their effectiveness in flood routing of two complicated reaches located at the upper and middle main stream of the Yangtze River. The results suggested that the performance of the MLP, LSTM and GRU models all gradually improved and then slightly decreased as the time lag increased. Furthermore, the MLP, LSTM and GRU models outperformed the SVR, GPR and RFR models, and the GRU model demonstrated superior performance across a range of efficiency criteria, including mean absolute percentage error (MAPE), root mean square error (RMSE), Nash–Sutcliffe efficiency coefficient (NSE), Taylor skill score (TSS) and Kling–Gupta efficiency (KGE). Specifically, the GRU model achieved reductions in MAPE and RMSE of at least 7.66% and 3.80% in the first case study and reductions of 19.51% and 11.76% in the second case study. The paper indicated that the GRU model was the most appropriate choice for flood routing in the Yangtze River.

Review of Machine Learning Methods for River Flood Routing

“…The idea of SVR is that an entire sample set can be characterized by a small number of support vectors [30]. In SVR, a kernel function is used to create a linearly divisible space by converting the sample space, and then predictive analysis on the new samples is performed using the hyperplane with the largest margin and support vectors [14]. The learning ability of a SVR model is highly affected by the selected kernel function.…”

“…In this study, the dynamics of the studied floods were captured by applying the SVM to observed data, and the model showed good performance for flood routing modeling. Katipoglu and Sarigol [27] compared the performance of various ML models, including SVM, for flood routing prediction in Eskisehir, Sivas, and Ankara, and Zhou and Kang [14] used a linear kernel in a SVR for flood routing in the Yangtze River.…”

“…This study showed that the feedforward ANN is a promising alternative in Muskingum routing after comparing the FMLP to the nonlinear Muskingum models using other reported methods such as GA [37], nelder-mead simplex (NMS) algorithm [38], and broyden-fletcher-goldfarb-shanno (BFGS) method [39] in solving the parameter estimation. In addition, Zhou and Kang [14] applied a MLP with one hidden layer, including 100 hidden neurons, for the flood routing.…”

“…It has been reported that GRU is simpler than LSTM but has similar learning abilities regarding longterm dependencies in time series data [17]. Zhou and Kang [14] constructed a GRU model for flood routing in the Yangtze River and compared the performance of the GRU with other machine learning models such as SVR, gaussian process regression (GRP), MLP, random forest regression (RFR), and LSTM. It was demonstrated that the GRU model showed superior performance compared to other models.…”

“…ML methods can deal with more complex hydrological processes by mapping nonlinear relationships between hydrological variables. Commonly applied ML methods in river flood routing computation include support vector regression (SVR) [13,14], artificial neural networks (ANNs) [15,16], multilayer perceptron (MLP) [14], gated recurrent units (GRUs) [14,17], long short-term memory (LSTM) [14,18], and so on. Machine learning methods have been widely applied in hydrological fields such as rainfall-runoff models and flood forecasting.…”