Jeong et al. (2025) An integrated watershed modeling approach using soil and water assessment tool and graph convolutional long short-term memory

Identification

• Journal: Journal of Hydrology
• Year: 2025
• Date: 2025-11-20
• Authors: Dae Seong Jeong, Do Hyuck Kwon, Jin Hwi Kim, Kyung Hwa Cho, Seo Jin Ki, Jae-Ki Shin, Yongeun Park
• DOI: 10.1016/j.jhydrol.2025.134611

Research Groups

• Future and Fusion Lab of Architectural, Civil and Environmental Engineering, Korea University, Seoul, South Korea
• Department of Civil, Urban, Earth, and Environmental Engineering, Ulsan National Institute of Science and Technology, Ulsan, South Korea
• School of Civil, Environmental, and Architectural Engineering, Korea University, Seoul, South Korea
• Department of Geology, Faculty of Science, Chulalongkorn University, Bangkok, Thailand
• Department of Environmental Engineering, Gyeongsang National University, Jinju, South Korea
• Limnoecological Science Research Institute Korea, THE HANGANG, Gyeongnam, South Korea
• Department of Civil and Environmental Engineering, Konkuk University-Seoul, Seoul, South Korea

Short Summary

This study proposes an integrated watershed modeling approach combining the process-based Soil and Water Assessment Tool (SWAT) with the graph-based Graph Convolutional Long Short-Term Memory (GCLSTM) model to simulate streamflow and Total phosphorus (TP) load. The integrated approach significantly improved simulation accuracy compared to calibrated SWAT, demonstrating the GCLSTM's ability to capture complex spatiotemporal dependencies and aggregate upstream signals.

Objective

• To develop and evaluate an integrated watershed modeling approach that combines the process-based Soil and Water Assessment Tool (SWAT) with the graph-based Graph Convolutional Long Short-Term Memory (GCLSTM) model to simulate streamflow and Total phosphorus (TP) load across multiple regions within a watershed.

Study Configuration

• Spatial Scale: Yeongsan River watershed, Korea.
• Temporal Scale: 2017–2021 (5 years).

Methodology and Data

• Models used: Soil and Water Assessment Tool (SWAT), Graph Convolutional Long Short-Term Memory (GCLSTM).
• Data sources: SWAT simulation results (with default parameter values), meteorological data, watershed information, upstream hydrometeorological and land use signals.

Main Results

• Coupling uncalibrated SWAT with GCLSTM significantly increased streamflow R² from a range of 0.22–0.74 (calibrated SWAT) to 0.40–0.88.
• TP load R² increased from a range of 0.02–0.36 (calibrated SWAT) to 0.50–0.81.
• The performance gain reflected GCLSTM's ability to aggregate upstream hydrometeorological and land use signals across the river network, capturing nonlinear spatiotemporal dependencies.
• Network analysis revealed that upstream precipitation is the dominant driver of downstream streamflow, while upstream land use patterns govern TP load variability.

Contributions

• Proposes a novel integrated watershed modeling framework combining process-based (SWAT) and deep learning (GCLSTM) models, addressing the limitations of relying solely on either approach.
• Demonstrates significant improvements in streamflow and Total phosphorus load simulation accuracy by leveraging GCLSTM's ability to capture complex spatiotemporal dependencies and upstream-downstream interactions.
• Provides insights into the dominant drivers of streamflow (upstream precipitation) and TP load (upstream land use patterns) through network analysis, enhancing understanding of watershed dynamics.
• Offers a potential decision-support tool for devising streamflow regulation and nutrient reduction measures that faithfully reflect actual watershed conditions.

Funding

• [No specific funding projects or reference codes are mentioned in the provided text.]

Citation

@article{Jeong2025integrated,
  author = {Jeong, Dae Seong and Kwon, Do Hyuck and Kim, Jin Hwi and Cho, Kyung Hwa and Ki, Seo Jin and Shin, Jae-Ki and Park, Yongeun},
  title = {An integrated watershed modeling approach using soil and water assessment tool and graph convolutional long short-term memory},
  journal = {Journal of Hydrology},
  year = {2025},
  doi = {10.1016/j.jhydrol.2025.134611},
  url = {https://doi.org/10.1016/j.jhydrol.2025.134611}
}