Ma et al. (2026) Divergence or Convergence? A Comparison of InVEST and SWAT in Simulating Water Conservation Patterns and Drivers

Identification

• Journal: Water Resources Management
• Year: 2026
• Date: 2026-03-01
• Authors: Li Ma, Jianling Yang, Yingjuan Han, Zhenhua Di, Xilai Zhang, Yazhen Wu, Liping Ma, Xuemei Yang
• DOI: 10.1007/s11269-026-04549-8

Research Groups

• Key Laboratory for Monitoring, Early Warning and Risk Management of Meteorological Disasters for Characteristic Agriculture in Arid Regions, China Meteorological Administration, Yinchuan, 750002, China
• Ningxia Hui Autonomous Region Institute of Meteorological Science, Yinchuan, 750002, China
• Beijing Normal University, Beijing, 100875, China
• Ningxia Meteorological Information Center, Yinchuan, 750002, China

Short Summary

This study quantitatively compares the InVEST and SWAT models in simulating water conservation patterns and drivers in the Liupan Mountain region from 2003 to 2022, finding both models show an increasing trend and consistent spatial heterogeneity, but with complementary strengths for different application contexts.

Objective

• To quantitatively evaluate regional water conservation capacity from 2003 to 2022 using the SWAT model, the InVEST model, and a regional water balance approach in the Liupan Mountain region.
• To systematically examine the key climatic and environmental drivers underlying the spatiotemporal variation in water conservation capacity.
• To provide methodological and empirical guidance for model selection, collinearity treatment, and multi-scale analysis in complex terrain.

Study Configuration

• Spatial Scale: Liupan Mountain region, northwestern China (spanning Ningxia, Gansu, and Shaanxi Provinces), part of the Loess Plateau. The study area was divided into 71 sub-basins for SWAT, and all inputs were resampled to 30 meter raster resolution.
• Temporal Scale: 2003 to 2022 (20 years), analyzed at annual and monthly scales. Model calibration period: 2003–2016; validation period: 2017–2022.

Methodology and Data

• Models used:
  • SWAT (Soil and Water Assessment Tool) hydrological model
  • InVEST (Integrated Valuation of Ecosystem Services and Tradeoffs) Water Yield module
  • Regional water balance approach
• Data sources:
  • Vector boundaries
  • Digital Elevation Model (DEM)
  • Precipitation data
  • Evapotranspiration data
  • Land use data
  • Soil properties data
  • Runoff data (from Jingheyuan and Fenggeling hydrological stations)
  • Temperature, relative humidity, and sunshine duration data

Main Results

• Both SWAT and InVEST models revealed a statistically significant upward trend in water conservation capacity in the Liupan Mountain region from 2003 to 2022 (p < 0.05). The SWAT model showed more stable interannual variability (mean annual capacity: 5.86 mm, mean amount: 5.92 × 10^8 m^3), while InVEST exhibited greater fluctuations (mean annual capacity: 3.92 mm, mean amount: 3.45 × 10^8 m^3).
• Both models consistently indicated a clear spatial gradient in water conservation capacity, with higher values in southern and eastern areas and lower values in northern and western areas. High-value zones were mainly concentrated in forested and valley regions within the Weihe and Jinghe River basins.
• Simulation results from both models demonstrated a strong positive relationship between water conservation capacity and precipitation, confirming precipitation as the dominant climatic driver. Partial correlation analysis further showed precipitation (partial r = 0.626) as the most influential factor, with relative humidity (partial r = 0.426) and sunshine duration (partial r = -0.257) playing secondary roles, and temperature having a limited direct effect (partial r = 0.141).
• The SWAT model, characterized by stable temporal performance and higher sensitivity to climatic variability, is well suited for investigating hydrological processes and supporting long-term water resource planning. In contrast, the InVEST model, requiring fewer input parameters and offering higher computational efficiency, is more appropriate for rapid spatial assessments and evaluations of ecological conservation effectiveness.
• SWAT model calibration and validation showed strong performance (e.g., Jingheyuan Station: R^2 = 0.82/0.85, NSE = 0.81/0.82, KGE = 0.88/0.86, PBIAS = 8.0%/6.32%). InVEST model validation against measured runoff showed low relative errors (Jingheyuan: -2.8%; Fenggeling: -5.0%).

Contributions

• Provides the first systematic comparison of the SWAT and InVEST models in the Liupan Mountain region, an ecologically fragile arid-to-semi-arid transition zone.
• Offers a dual-model framework that helps reduce model-specific bias and enhances the robustness and reliability of water conservation assessments.
• Extends water conservation analysis to seasonal and monthly scales, providing a more detailed understanding of temporal dynamics.
• Strengthens climate attribution by employing partial correlation analysis to effectively resolve collinearity among key climatic drivers.
• Addresses the scale mismatch between sub-basins and pixels, improving the accuracy and applicability of multi-scale analyses.
• Offers practical methodological and empirical guidance for model selection, collinearity treatment, and multi-scale analysis in complex hydrological and ecological environments.

Funding

• Joint Funds of the National Natural Science Foundation of China (U22A20577)
• China Meteorological Administration Climate Change Special Project (QBZZ202509)
• Ningxia Hui Autonomous Region Science and Technology Benefiting the People Project (2025CMG03053)

Citation

@article{Ma2026Divergence,
  author = {Ma, Li and Yang, Jianling and Han, Yingjuan and Di, Zhenhua and Zhang, Xilai and Wu, Yazhen and Ma, Liping and Yang, Xuemei},
  title = {Divergence or Convergence? A Comparison of InVEST and SWAT in Simulating Water Conservation Patterns and Drivers},
  journal = {Water Resources Management},
  year = {2026},
  doi = {10.1007/s11269-026-04549-8},
  url = {https://doi.org/10.1007/s11269-026-04549-8}
}