State-of-the-art artificial intelligence and hybrid model approaches for sediment load prediction in dynamic watersheds

Predicting sediment production is a complex, nonlinear, and dynamic process influenced by watershed geomorphology and spatiotemporal variability in hydrological drivers. Conventional modeling approaches often face limitations in accurately representing these processes, which has encouraged the growing use of artificial intelligence (AI) and machine learning techniques. This review presents a comprehensive overview of AI-based methods applied to sediment load and concentration prediction in studies published between 2000 and 2025. The reviewed approaches include artificial neural networks, fuzzy logic–based models, support vector machines, tree-based algorithms, deep learning architectures, and evolutionary optimization techniques. The paper summarizes commonly used model structures, input variables, temporal scales, and training algorithms reported in the literature, and highlights the role of hybrid modeling frameworks in enhancing predictive performance. Particular attention is given to studies addressing sediment load prediction under flood conditions, where rapid and accurate estimation is essential. In addition, the review discusses how historical hydrological data have been utilized to improve model reliability. Finally, the paper outlines current challenges and presents general recommendations to support future research and applications of AI-based sediment load prediction models.