Cai et al. (2025) Enhancing hydrological modeling in large basin with intensive human water use through hierarchical parameterization and bias-integrated calibration

Identification

• Journal: Water Cycle
• Year: 2025
• Date: 2025-10-18
• Authors: Kaikui Cai, Jincheng Li, Qingsong Jiang, Lian Hu, Jiaxing Fu, Man Zhang, Yifan Li, Yue Qin, Yong Liu
• DOI: 10.1016/j.watcyc.2025.10.003

Research Groups

• College of Environmental Sciences and Engineering, State Environmental Protection Key Laboratory of All Material Fluxes in River Ecosystems, Peking University, Beijing, China
• Water Security Research Group, Biodiversity and Natural Resources Program, International Institute for Applied Systems Analysis (IIASA), Laxenburg, Austria
• Southwest United Graduate School, Kunming, China
• Institute of Tibetan Plateau, Peking University, Beijing, China

Short Summary

This study developed a hierarchical parameterization and bias-integrated calibration method to improve hydrological modeling in large basins with intensive human water use, applying it to the Pearl River Basin using the Community Water Model (CWatM). The method significantly enhanced simulation performance, identifying Water Resource Zones of level 3 (WRZ3) as the optimal calibration scale and effectively reducing irrigation bias while maintaining streamflow accuracy.

Objective

• To enhance hydrological modeling in large basins with intensive human water use by developing and validating a hierarchical parameterization and bias-integrated calibration method.
• To identify optimal calibration unit scales for hierarchical parameterization strategies.
• To integrate irrigation water explicitly into the calibration framework for coupled irrigation and streamflow simulations.
• To assess the regional applicability of large-scale hydrological models and quantify the impact of input data uncertainty (meteorological and inflow data) on modeling robustness.

Study Configuration

• Spatial Scale: Pearl River Basin (PRB), southern China (102°E−116°E and 21°N–27°N), with a total drainage area of 4.5 × 10^5 km^2. Modeling was conducted at a spatial resolution of 0.1° (approximately 10 km). Calibration units were configured at four scales: WRZ1 (1 unit), WRZ2 (7 units), WRZ3 (15 units), and site-specific catchments (29 units).
• Temporal Scale: The simulation period spanned 2001–2019. This included a 5-year spin-up period (2001–2005), a 10-year calibration period (2006–2015), and a 4-year validation period (2016–2019). Daily streamflow data were used for calibration and validation.

Methodology and Data

• Models used:
  • Community Water Model (CWatM v1.08) for hydrological and water resources simulation.
  • NSGA-II genetic algorithm (implemented with DEAP Python package) for model calibration.
  • Variance-based Sobol method (implemented with SALib Python library) for parameter sensitivity analysis.
• Data sources:
  • Physiographic Data: Local drainage direction map from HydroSHEDS; land use data from Global Resources Data Cloud.
  • Water Resource Data: Irrigation efficiency and water demand data from the Water Resources Bulletin (WRB) of PRB.
  • Meteorological Data: China Meteorological Forcing Dataset (CMFD v2.0) as primary input; GSWP3-W5E5 (ISIMIP3a) for uncertainty assessment. Required inputs included precipitation, average/maximum/minimum 2 m temperatures, near-surface pressure, humidity, 10 m wind speed, and long/short-wave downward surface radiation fluxes.
  • Water Demand Data: Gridded (0.1°) water withdrawal and consumption data for domestic, industrial, and livestock sectors (2000–2019) were reconstructed using WRB of PRB, WorldPop, China Industrial Water Withdrawal (CIWW) dataset, and Gridded Livestock of the World (GLW3/GLW4).
  • Observed Streamflow Data: Daily streamflow records from 29 representative hydrological stations across the PRB, compiled from the National Hydrological Yearbooks of the PRB.

Main Results

• Hierarchical calibration significantly improved simulation performance compared to non-regionalized methods, with average modified Kling-Gupta Efficiency (KGE) and Nash-Sutcliffe Efficiency (NSE) values increasing by over 0.5.
• The Water Resource Zones of level 3 (WRZ3) was identified as the optimal calibration scale, providing a balance between simulation accuracy and computational efficiency (e.g., mean KGE decreased by only 0.02 compared to site-level, while computational cost was halved).
• Integrating irrigation simulation bias into a single-objective function enabled the simultaneous optimization of both streamflow and irrigation simulations, reducing irrigation bias from 327 % to 51 % with only a minor decrease in streamflow accuracy (mean KGE from 0.81 to 0.75).
• The effective irrigation weighting coefficient (approximately 0.1) for the composite objective function was found to align closely with the basin's overall irrigation-to-total-runoff ratio (0.096).
• The CWatM was confirmed as suitable for regional applications, but its performance is sensitive to input data quality; high-resolution CMFD v2.0 meteorological data significantly improved streamflow simulation compared to lower-resolution GSWP3-W5E5 inputs (mean KGE increased from 0.77 to 0.81 at WRZ3 level).
• Utilizing observed daily streamflow data as boundary inputs significantly improved downstream CDA unit performance compared to simulated inflow, highlighting the cascading effect of upstream errors in hierarchical calibration (average KGE increase of 0.04 across 16 CDA units).
• The regionally calibrated CWatM demonstrated superior daily streamflow simulations (KGE consistently exceeding 0.85) compared to uncalibrated global models (KGE values ranging from 0.02–0.74), and more accurately captured seasonal flow variability and peak flow magnitudes and timings.

Contributions

• Proposed and validated a novel hierarchical parameterization and bias-integrated calibration method to enhance hydrological modeling in large basins with intensive human water use.
• Systematically evaluated the impact of calibration unit scales on hierarchical parameterization performance and identified Water Resource Zones of level 3 (WRZ3) as the optimal scale for the Pearl River Basin.
• Developed a bias-integrated calibration framework that explicitly incorporates irrigation water use, enabling the joint optimization of streamflow and irrigation simulations, and proposed a novel criterion (regional irrigation-to-discharge ratio) for selecting the irrigation weighting factor.
• Quantified the impact of meteorological and inflow boundary data uncertainty on model performance within a hierarchical calibration strategy, emphasizing the critical need for high-quality regional input data.
• Provided a reproducible technical pathway and regional parameterization strategy for applying large-scale hydrological models (like CWatM) to specific regional contexts, demonstrating superior performance compared to global uncalibrated models.

Funding

• National Key R&D Program of China, China (2024YFD1701300)
• Yunnan Provincial Science and Technology Project, China (202302AO370015)
• National Natural Science Foundation of China (42471021 and 42277482)

Citation

@article{Cai2025Enhancing,
  author = {Cai, Kaikui and Li, Jincheng and Jiang, Qingsong and Hu, Lian and Fu, Jiaxing and Zhang, Man and Li, Yifan and Qin, Yue and Liu, Yong},
  title = {Enhancing hydrological modeling in large basin with intensive human water use through hierarchical parameterization and bias-integrated calibration},
  journal = {Water Cycle},
  year = {2025},
  doi = {10.1016/j.watcyc.2025.10.003},
  url = {https://doi.org/10.1016/j.watcyc.2025.10.003}
}