Yang et al. (2026) Identifying dominant parameters in SWAT across subbasin and HRU scales using a two-step deep learning-assisted spatial sensitivity analysis

Identification

• Journal: Hydrology and earth system sciences
• Year: 2026
• Date: 2026-06-30
• Authors: Jing Yang, Jiangjiang Zhang, Tian Jiao, Yuanzhang Zhao, Manya Luo, Lei Wu, Ming Ye, Jinxi Song
• DOI: 10.5194/hess-30-4095-2026

Research Groups

• School of Land Engineering, Key Laboratory of Degraded and Unused Land Consolidation Engineering of the Ministry of Natural Resources, Shaanxi Key Laboratory of Land Consolidation, and Shaanxi Province Land Consolidation Engineering Technology Research Center, Chang'an University, Xi'an 710054, China
• Key Laboratory of Environmental Simulation and Ecological Health in the Yellow River Basin, College of Urban and Environmental Sciences, Northwest University, Xi'an 710127, China
• State Key Laboratory of Water Disaster Prevention, Hohai University, Nanjing 210098, China
• Yangtze Institute for Conservation and Development, Hohai University, Nanjing 210098, China

Short Summary

This study presents a two-step deep learning-assisted spatial sensitivity analysis framework to identify dominant parameters in the Soil and Water Assessment Tool (SWAT) across subbasin and Hydrologic Response Unit (HRU) scales. The framework efficiently diagnoses dominant hydrological controls in high-dimensional distributed models.

Objective

• Identify dominant parameters controlling runoff generation in SWAT using a two-step deep learning-assisted spatial sensitivity analysis framework.
• Investigate how parameter importance varies across spatial scales while preserving the native parameterization structure of distributed models.

Study Configuration

• Spatial Scale: Subbasin and HRU scales.
• Temporal Scale: Daily time step over 1962–1986.

Methodology and Data

• Models used: Soil and Water Assessment Tool (SWAT) and multilayer perceptron (MLP) surrogates.
• Data sources: Daily meteorological observations, monthly runoff records, and spatial datasets for the Jinghe River Basin.

Main Results

• The two-step framework efficiently identified dominant parameters controlling runoff generation in SWAT across subbasin and HRU scales.
• Sensitivity hotspots were consistently concentrated in gauge-proximal subbasins, while HRU-scale analysis further resolved localized controls.
• Robustness analyses demonstrated that the identified sensitivity patterns remained stable across different performance metrics, NSE-constrained posterior parameter distributions, alternative observation configurations, and varying numbers of screened parameters.

Contributions

• This study advances understanding of how distributed parameters jointly control hydrologic responses across spatial scales, enhancing the diagnostic capability, calibration efficiency, and predictive reliability of physically based hydrological models.
• The proposed two-step framework provides a transferable and computationally feasible approach for sensitivity assessment in high-dimensional distributed models.

Funding

• Not specified.

Citation

@article{Yang2026Identifying,
  author = {Yang, Jing and Zhang, Jiangjiang and Jiao, Tian and Zhao, Yuanzhang and Luo, Manya and Wu, Lei and Ye, Ming and Song, Jinxi},
  title = {Identifying dominant parameters in SWAT across subbasin and HRU scales using a two-step deep learning-assisted spatial sensitivity analysis},
  journal = {Hydrology and earth system sciences},
  year = {2026},
  doi = {10.5194/hess-30-4095-2026},
  url = {https://doi.org/10.5194/hess-30-4095-2026}
}