Sandoval et al. (2025) Towards water resilience: A multi-stage calibration framework for large-scale integrated surface–subsurface hydrological models

Identification

• Journal: Journal of Hydrology
• Year: 2025
• Date: 2025-11-04
• Authors: Leonardo Sandoval, Alberto Guadagnini, Laura E. Condon, Mònica Riva
• DOI: 10.1016/j.jhydrol.2025.134527

Research Groups

• Dipartimento di Ingegneria Civile e Ambientale, Politecnico di Milano, Italy.
• Department of Hydrology and Atmospheric Sciences, The University of Arizona, USA.

Short Summary

This study presents a multi-stage calibration framework for large-scale, high-fidelity integrated surface water–groundwater models using sensitivity analysis and Gaussian Process Regression surrogates. The approach resulted in the first robustly calibrated integrated model of the Po River District (87,000 km²), effectively capturing complex 3D subsurface dynamics and river discharge.

Objective

• To develop and implement a computationally efficient methodological framework for the calibration of large-scale, physically-based, fully coupled surface water–groundwater (SW–GW) models.
• To identify the most influential hydraulic parameters and enhance model reliability for water resource management under climate and anthropogenic pressures.

Study Configuration

• Spatial Scale: Po River District, Northern Italy (87,000 km²), utilizing a 2 km × 2 km horizontal grid and a 225 m deep vertical domain discretized into six layers.
• Temporal Scale: 10-year simulation period (2009–2018) with hourly meteorological forcing and steady-state initialization.

Methodology and Data

• Models used: ParFlow-CLM (integrated 3D variably saturated subsurface flow and 2D overland flow) and Gaussian Process Regression (GPR) for surrogate-assisted optimization.
• Data sources:
  • Topography: HydroSHEDS digital elevation model.
  • Land Cover: MODIS MCD12Q1.
  • Meteorology: COSMO-REA6 reanalysis.
  • Evapotranspiration: GLASS ET dataset.
  • Subsurface Architecture: 3D probabilistic hydrostratigraphic reconstruction and SoilGrids.
  • Observations: Water table depth (WTD) from 285 piezometers and daily discharge from 31 gauging stations (EStreams dataset).
  • Anthropogenic: Regional groundwater abstraction and irrigation datasets.
• Calibration Strategy:
  1. Local sensitivity analysis (one-at-a-time) to identify key parameters.
  2. Phase 1: Steady-state calibration of hydraulic conductivities ($K$) for six geomaterial types using GPR surrogates against WTD data.
  3. Phase 2: Transient calibration of Manning roughness coefficients ($\eta$) for river channels against discharge data.

Main Results

• Sensitivity: Clay hydraulic conductivity ($k4$) is the dominant parameter driving groundwater table dynamics in the valley, while channel Manning roughness ($\eta2$) is the primary control for river discharge.
• Groundwater Performance: The calibrated model achieved absolute errors in WTD below 10 m for 90% of observation points and below 2.5 m for 45% of points.
• Surface Water Performance: The optimal Manning roughness for channels was identified as $7.5 \times 10^{-6}$ h m⁻¹/³ (equivalent to $0.027$ s m⁻¹/³).
• Efficiency Metrics: 87% of the gauging stations achieved a Kling–Gupta Efficiency (KGE) $\ge -0.41$, with a median KGE of 0.34 across the basin.
• Parameter Values: Calibrated vertical hydraulic conductivities ranged from $1.52$ m h⁻¹ for gravel to $1.18 \times 10^{-2}$ m h⁻¹ for clay.

Contributions

• Provides the first robustly calibrated, high-fidelity, fully integrated SW–GW model at a supra-regional scale (87,000 km²).
• Introduces a synergistic workflow (sensitivity analysis + GPR surrogates + multi-stage calibration) that overcomes the prohibitive computational costs of calibrating large-scale physically-based models.
• Explicitly accounts for 3D subsurface heterogeneity and unsaturated-saturated flow dynamics, moving beyond simplified "static reservoir" groundwater representations.

Funding

• European Union’s Horizon 2020 research and innovation program (RECYCLE project, Marie Skłodowska-Curie Grant Agreement 872607).
• Water Alliance - Acque di Lombardia.
• European Union Next Generation EU (National Recovery and Resilience Plan - NRRP, RETURN Extended Partnership, PE0000005).
• ISCRA (access to GALILEO100 and LEONARDO supercomputers at CINECA).

Citation

@article{Sandoval2025Towards,
  author = {Sandoval, Leonardo and Guadagnini, Alberto and Condon, Laura E. and Riva, Mònica},
  title = {Towards water resilience: A multi-stage calibration framework for large-scale integrated surface–subsurface hydrological models},
  journal = {Journal of Hydrology},
  year = {2025},
  doi = {10.1016/j.jhydrol.2025.134527},
  url = {https://doi.org/10.1016/j.jhydrol.2025.134527}
}

Generated by BiblioAssistant using gemini-3-flash-preview (Google API)