Li et al. (2025) Uncertainty analysis and parameter optimization enhance assessment accuracy in water yield modelling

Identification

• Journal: Journal of Hydrology Regional Studies
• Year: 2025
• Date: 2025-12-10
• Authors: Jiaqi Li, Wang Bei, Juntao Zhong, Jia Xu, Peijun Sun
• DOI: 10.1016/j.ejrh.2025.103011

Research Groups

• College of Geomatics, Xi’an University of Science and Technology, Xi’an, China
• Shaanxi Key Laboratory of Earth Surface System and Environmental Carrying Capacity, Northwest University, Xi’an, China
• College of Urban and Environmental Sciences, Northwest University, Xi’an, China

Short Summary

This study develops an integrated framework for uncertainty analysis, sensitivity analysis, and parameter optimization to enhance the accuracy of water yield modeling using the InVEST model in the Qinling-Daba Mountains region, demonstrating significant improvements in simulation reliability through optimal precipitation dataset selection and Markov Chain Monte Carlo optimization.

Objective

• To quantify parameter uncertainty and identify sensitive parameters in the InVEST water yield model using the Monte Carlo method and global sensitivity analysis.
• To compare different precipitation datasets to select the optimal one for the study region.
• To optimize sensitive parameters, specifically the plant evapotranspiration coefficient (Kc), using the Markov Chain Monte Carlo method to improve water yield simulation accuracy.
• To introduce and improve a comprehensive "UA-SA-parameter optimization" framework for systematically quantifying and effectively reducing parameter uncertainty in ecosystem services assessment.

Study Configuration

• Spatial Scale: The Qinling-Daba Mountains region in Shaanxi Province (QB-SX), China, covering approximately 61,900 square kilometers. Specific focus on the Yiluo River, Danjiang River, and Hanjiang River basins within QB-SX.
• Temporal Scale:
  • Water yield estimation period: 2002 to 2022.
  • Runoff data for parameter optimization: 2005 to 2015.
  • Runoff data for validation: 2016 to 2022.
  • Precipitation data screening and comparison: 2005 to 2018.
  • Parameter averages for sensitivity and uncertainty analysis: 2005 to 2022.

Methodology and Data

• Models used:
  • Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) water yield model (version 3.14.1).
  • Monte Carlo simulation for uncertainty quantification.
  • Extended Fourier Amplitude Sensitivity Test (EFAST) for global sensitivity analysis.
  • Markov Chain Monte Carlo (MCMC) method, specifically the Metropolis-Hastings algorithm, for parameter optimization.
  • SimLab 2.2 for sensitivity and uncertainty analysis.
  • Python 3.9 with libraries (numpy, osgeo, natcap.invest.hydropower.hydropowerwateryield) for model execution and analysis.
  • ArcGIS 10.8 for spatial data processing (resampling, reclassification, GIS delineation).
• Data sources:
  • Precipitation (P): ERA5 dataset (0.1° resolution), CRU dataset (0.1° resolution), WorldClim dataset (0.1° resolution), Monthly precipitation dataset for China (1 km resolution).
  • Reference evapotranspiration (ET0): Senay et al. (2008) (1 km resolution).
  • Land Use Land Cover (LULC): Zhong et al. (2015) (30 m resolution).
  • Plant Available Water Capacity (PAWC): Sheldon et al. (2017) (1 km resolution).
  • Soil Restrictive Layer Depth (Soil depth): Wei et al. (2013) (1 km resolution).
  • Digital Elevation Model (DEM): Uhe et al. (2025) (30 m resolution).
  • Root Depth: Allen et al. (1998).
  • Seasonality Factor (Z): Donohue et al. (2012).
  • Plant evapotranspiration coefficient (Kc): Allen et al. (1998) (initial values for cropland, forest, shrub, grassland).
  • Research area boundary: Liu et al. (2016).
  • Observed runoff data: Shaanxi Province Water Resources Bulletin (for Yiluo River, Danjiang River, Hanjiang River basins, 2005–2022).
  • Human water consumption data: Shaanxi Province Water Resources Bulletin.
  • Meteorological station observations: Five stations (Fengxiang, Jinghe, Huashan, Hanzhong, Ankang) within QB-SX (2005–2018) for precipitation dataset validation.

Main Results

• Precipitation (P), reference evapotranspiration (ET0), and plant evapotranspiration coefficient (Kc) are identified as the most sensitive parameters in the InVEST water yield model, with P being the primary source of uncertainty.
• The "Monthly precipitation for China" dataset was identified as the optimal precipitation source, exhibiting the smallest relative deviation (11.92%) compared to ground observations, significantly outperforming ERA5 (76.76% deviation).
• After parameter optimization, the optimized Kc values showed distinct variations across land use types: forest (0.91) > cropland (0.74) > shrubland (0.50) > grassland (0.47).
• The Nash-Sutcliffe Efficiency (NSE) coefficient for water yield simulation significantly improved from -54.29 (initial parameters) to 0.36 (optimized parameters), demonstrating a substantial enhancement in model accuracy.
• Uncertainty quantification revealed significant variability in water yield estimates across different precipitation datasets, with standard deviations ranging from 77.41 mm to 174.22 mm and coefficients of variation from 18.58% to 43.55%.

Contributions

• Introduced and improved a novel "UA-SA-parameter optimization" framework for systematically quantifying and effectively reducing parameter uncertainty in ecosystem services assessment, exemplified by the InVEST water yield model.
• Integrated multi-source precipitation datasets for optimal selection, a key innovation for improving model input accuracy.
• Applied the Markov Chain Monte Carlo (MCMC) method for robust parameter optimization, enhancing the reliability of simulation results.
• Demonstrated that the framework significantly improves the accuracy of water yield simulations and provides a transferable methodology for optimizing other ecosystem services models.
• Provided scientific support for regional ecosystem management and policymaking, particularly in data-scarce regions.

Funding

• Youth Project of Natural Science Basic Research Program of Shaanxi Province of China (Grant number: 2024JC-YBQN-0276)
• National Natural Science Foundation of China (Grant number: NSFC 42001263)
• Qinghai Provincial Natural Science Foundation Project (Grant number: 2022-ZJ-906)

Citation

@article{Li2025Uncertainty,
  author = {Li, Jiaqi and Bei, Wang and Zhong, Juntao and Xu, Jia and Sun, Peijun},
  title = {Uncertainty analysis and parameter optimization enhance assessment accuracy in water yield modelling},
  journal = {Journal of Hydrology Regional Studies},
  year = {2025},
  doi = {10.1016/j.ejrh.2025.103011},
  url = {https://doi.org/10.1016/j.ejrh.2025.103011}
}