Author(s): Zhenjie Mo; Jue Wang; Zhengzheng Zhou; Shuguang Liu

Linked Author(s):

Keywords: Saltwater intrusion forecast; Random Forest; Support Vector Machine; Recurrent Neural Network

Abstract: Due to the extreme high temperature and drought events in the Yangtze River Basin in 2022, the Yangtze River Estuary experienced significant and prolonged saltwater intrusion. Despite the considerable challenges posed by the ever-changing river dynamics and numerous influencing factors, saltwater intrusion forecasting in the Yangtze River Estuary remains a challenging problem. To examine the viability of machine learning algorithms in saltwater forecasting in the Yangtze River Estuary, this study employed Random Forest (RF), Support Vector Machine (SVM) and Recurrent Neural Network (RNN) to develop saltwater prediction models. Based on the hydrological and meteorological data from the representative observation stations in the estuary, the various predictors including salinity, tide level, flow at Datong station and wind speed from the ERA5 reanalysis dataset were utilized in the constructing the prediction models. The Root Mean Square Error (RMSE), Determination Coefficients (R2) and Nash-Sutcliffe Efficiency coefficient (NSE) were used to evaluate the model performance. The results demonstrate that the machine learning methods exhibit satisfactory predictive ability in saltwater intrusion forecasting in this area, and the RF model overperforms yielding an RMSE of 2.167, NSE of 0.376, and R2 of 0.901.

DOI:

Year: 2024