The proposed approach reduces computational cost while maintaining high predictive accuracy, making it suitable for large-scale applications

Accurate water level forecasting is crucial for effective water resources management. Recently, a group of researchers from Jeonbuk National University has proposed a novel clustering-based machine learning framework for accurate, scalable, and computationally efficient water level prediction, especially in data-scarce regions. These findings are expected to lead to improved flood mitigation, agricultural irrigation, environmental sustainability, and long-term climate adaptation policies.

Title: Towards Enhanced Water Level Forecasting

Caption: Researchers propose a novel machine learning technique for water level prediction in data-sparse regions.

Credit: The authors

License type: Original Content

Usage restrictions: Cannot be reused without permission

Reliable and scalable water level prediction is crucial in hydrology for effective water resources management, especially when considering challenges owing to climate change, urbanization, improper land use, and high-water demand. It directly impacts the availability and distribution of freshwater in rivers and reservoirs. Therefore, accurate forecasting via early warning systems is a highly useful technique for flood mitigation, agricultural irrigation, ecosystem and environmental sustainability, and numerous other applications. In this regard, physically-based hydrodynamic river models can be used. However, these tools require enormous amounts of data, making them less useful in data-scarce regions. 

Recently, scientists have successfully utilized data-driven approaches, especially advanced machine learning techniques to overcome these limitations. However, in river networks, monitoring stations often have uneven record lengths, and many have time series that are too short to effectively train AI models for water-level prediction, necessitating more innovative approaches that enable predictions at all available monitoring stations and support the development of reliable watershed-scale early warning systems.

In a new development, Assistant Professor SangHyun Lee and Professor Taeil Jang from the Department of Rural Construction Engineering at Jeonbuk National University, Republic of Korea, have introduced a clustering-based machine learning framework that can accurately forecast water levels across all available stations, even when many have limited records. Their novel findings were made available online on 27 January 2026 and have been published in Volume 198 of the journal Environmental Modelling & Software on 01 March 2026.

Notably, instead of training separate AI models for every station, the proposed method groups stations with similar hydrological behavior and trains only one model per cluster. Specifically, the researchers select the station with the longest historical record in each cluster, train the model using that station, and apply the trained model to the remaining stations within the same cluster, reducing computational cost while maintaining high predictive accuracy. This approach enables a scalable, data-efficient AI system capable of accurately predicting water levels across an entire watershed using only a few representative stations.

This research has immediate practical value for water resource managers, emergency planners, and agricultural stakeholders. “By providing accurate short-term water level forecasts even in areas with limited historical data, the framework can support flood early-warning systems, optimize reservoir and irrigation management, and improve decision-making during extreme weather events,” points out Prof. Lee.

Furthermore, since the method reduces computational demands and does not require long-term records for all monitoring networks, it can also help agencies expand forecasting coverage across the watershed. In particular, regions that lack long-term hydrological records could still benefit from reliable forecasts using only a few representative monitoring stations.

Over the next 5–10 years, this type of approach could fundamentally improve how societies prepare for water-related risks under increasing climate variability. As floods and droughts become more frequent and unpredictable, scalable and data-efficient forecasting systems could enable real-time water management, automated infrastructure operation, and more resilient watershed planning. The ability to generate reliable predictions with limited data also means that advanced forecasting technology could become accessible worldwide, including in developing countries. 

Prof. Jang concludes: “Ultimately, such systems could enhance public safety, support sustainable agriculture, protect ecosystems, and strengthen long-term climate adaptation strategies for communities that depend on reliable water resources.” 

[Reference]

Title of original paper: Advancing water level prediction using clustering-based machine learning techniques in data-scarce regions 

Journal: Environmental Modelling & Software 

DOI: 10.1016/j.envsoft.2026.106899
 

[PURE Author Profile] Dr. SangHyun

Dr. SangHyun Lee is an Assistant Professor in the Department of Rural Construction Engineering at Jeonbuk National University. His group, the Agrohydrology Lab, develops advanced modeling frameworks using AI and physics-based tools to better understand complex hydrological processes in agricultural landscapes. In 2022, SangHyun received his PhD in Agricultural and Biological Engineering from the University of Illinois at Urbana-Champaign.

[PURE Author Profile] Professor Taeil Jang

Professor Taeil Jang is a Professor in the Department of Rural Construction Engineering at Jeonbuk National University. His research group, the Agricultural Water Resources Lab, focuses on precision agriculture and watershed environmental management for nonpoint sources pollution reduction, based on advanced monitoring and modeling of agro-ecological systems. In 2009, Taeil received his PhD in Agricultural Engineering from Seoul National University.