Research on a long-term salinity intrusion forecasting model for the Dongjiang River Delta based on machine learning approaches
XIE Yuhang
LIAO Zijin
WANG Chennai
ZHOU Zhe
WANG Jingjing
Abstract：Long-term hydrological forecasting is a key area in hydrology and water resources management.It provides essential support for water resources planning,flood warning,agricultural irrigation,and urban water supply by predicting hydrological variables over extended time periods.Focusing on the key challenges in water resources allocation under salinity intrusion,this study investigates the application of machine learning-based salinity intrusion forecasting models in the Dongjiang River Delta.Emphasis is placed on the selection of characteristic factors and the medium-to long-term prediction of salinity intrusion on an interannual scale under varying conditions.A salinity intrusion impact evaluation system is developed to identify critical characteristic factors.Random forest,gradient boosting tree,and ensemble models are employed for forecasting.The results indicate that non-flood season rainfall,end-of-flood season storage in Longtan Reservoir,diversion ratio at Sanshui Station,and end-of-flood season storage in Xinfengjiang Reservoir are the key factors influencing salinity intrusion.The ensemble model constructed with these critical factors enhances the accuracy and stability of salinity intrusion forecasting.Additionally,predictions conducted under multiple practical scenarios provide scientific support for water resources management and salinity intrusion warning in the Dongjiang River Delta.
Keywords：salinity intrusion evaluation systemmachine learningcharacteristic factorssalinity intrusion forecasting modelDongjiang River Delta
Publication Date：2025-07-30
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：10( 56-65 )