REVIEW article

Front. Water

Sec. Water and Artificial Intelligence

A Comparative Assessment of Machine and Deep Learning Applications in River Modeling

  • 1. Shandong University, Jinan, China

  • 2. National University of Sciences and Technology, Islamabad, Pakistan

  • 3. Institut za zastitu i ekologiju Republike Srpske, Banja Luka, Bosnia and Herzegovina

  • 4. Sveuciliste Hercegovina, Mostar, Bosnia and Herzegovina

  • 5. Mersin Universitesi, Mersin, Türkiye

  • 6. Changchun University of Science and Technology, Changchun, China

The final, formatted version of the article will be published soon.

Abstract

River-system modeling is essential for understanding hydrological processes, managing water resources, and mitigating risks associated with floods, sediment transport, and water-quality degradation. Traditional physically based and statistical models may face limitations in representing the complex, nonlinear, and dynamic behavior of river systems, particularly under changing climatic and anthropogenic conditions. Machine learning (ML) and deep learning (DL) have therefore emerged as increasingly important data-driven approaches for modeling and predicting river processes. This review provides a comprehensive and comparative synthesis of ML and DL applications in river-system modeling. Following a structured review methodology guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework, 52 peer-reviewed studies published between 2015 and 2025 were analyzed. The review focuses on river discharge prediction, flood forecasting, sediment-load estimation, and water-quality assessment, while comparing the performance, advantages, limitations, and application requirements of widely used approaches, including Random Forest, Support Vector Machines, Artificial Neural Networks, and Long Short-Term Memory networks. The findings indicate that conventional ML models are well-suited to small-to-moderate structured datasets because of their computational efficiency and comparatively high interpretability, whereas DL models are more suitable when sufficiently large, continuous, and information-rich temporal or spatial datasets are available. Model performance is strongly influenced by data availability, temporal and spatial resolution, basin characteristics, prediction objectives, and evaluation protocols. Recent developments in large-sample and regional LSTM modeling have expanded hydrological learning toward multi-basin and transferable applications, while hybrid, physics-informed, and differentiable process-based approaches increasingly integrate data-driven learning with hydrological process understanding. Important challenges remain in data scarcity, interpretability, uncertainty quantification, transferability, and reliable evaluation under nonstationary conditions. Overall, this review supports a data- and application-dependent model-selection strategy and highlights emerging pathways toward more reliable, interpretable, transferable, and operationally useful river-modeling systems.

Summary

Keywords

deep learning, flood forecasting, hydrology, machine learning, River modeling, sediment transport, Water Quality

Received

05 December 2025

Accepted

29 September 2026

Copyright

© 2026 Jat Baloch, Ali, Ilić, NEŠKOVIĆ MARKIĆ, STOJANOVIĆ BJELIĆ, Usman and Nigar. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Muhammad Yousuf Jat Baloch; Anam Nigar

Disclaimer

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.

Outline

Share article

Article metrics