Talbot et al. (2025) Enhancing physically based and distributed hydrological model calibration through internal state variable constraints

Identification

• Journal: Hydrology and earth system sciences
• Year: 2025
• Date: 2025-11-20
• Authors: Frédéric Talbot, Jean‐Daniel Sylvain, Guillaume Drolet, Annie Poulin, Richard Arsenault
• DOI: 10.5194/hess-29-6549-2025

Research Groups

• Hydrology, Climate and Climate Change Laboratory, École de technologie supérieure, Université du Québec, Montréal, Canada
• Direction de la recherche forestière, Ministère des Ressources naturelles et des Forêts, Québec, Canada

Short Summary

This study evaluates the impact of incorporating groundwater recharge constraints into the calibration of the physically-based Water Balance Simulation Model (WaSiM). It finds that while streamflow-only calibration yields higher Kling-Gupta Efficiency, adding groundwater recharge constraints improves the representation of internal hydrological processes and seasonal runoff patterns.

Objective

• To investigate how incorporating groundwater recharge constraints into the calibration of the physically-based Water Balance Simulation Model (WaSiM) affects the representation of internal hydrological variables across 34 catchments in Southern Québec.

Study Configuration

• Spatial Scale: 34 catchments in Southern Québec, Canada, ranging from 525 km² to 6840 km². Model simulations were performed at 250 m and 1000 m spatial resolutions, derived from an initial 30 m (resampled to 50 m) Digital Surface Model.
• Temporal Scale: Daily time step for meteorological and streamflow data. Meteorological data spanned 1981–2020, observed streamflow 1981–2010. Calibration period was 2000–2009, validation period 1990–1999, with a 5-year spin-up period before each simulation.

Methodology and Data

• Models used:
  • Water Balance Simulation Model (WaSiM) version II, utilizing the process-oriented Richards approach for subsurface flow.
  • Three model configurations:
    • Baseline (BL): Standard streamflow-only calibration, conceptual groundwater flow.
    • Physical Groundwater Model (GW): Streamflow-only calibration, physically-based groundwater flow module activated.
    • Physical Groundwater with Recharge Calibration (GW-RC): Multi-objective calibration (streamflow and groundwater recharge), physically-based groundwater flow module activated.
  • Calibration algorithm: Dynamically Dimensioned Search (DDS) algorithm.
  • Objective functions: Kling-Gupta Efficiency (KGE) for BL and GW; modified KGE incorporating groundwater recharge mean and standard deviation for GW-RC.
• Data sources:
  • Meteorological: ECMWF’s Reanalysis v5 (ERA5) for daily total precipitation and mean temperature.
  • Streamflow: Hydroclimatic Atlas of Southern Québec (MDDELCC, 2022) for daily observed streamflow.
  • Elevation: NASA Shuttle Radar Topography Mission version 3.0 Global 1 (SRTM-DSM) at 30 m resolution, resampled to 50 m, then 250 m and 1000 m for modeling.
  • Soil type: SIIGSOL 100 m database.
  • Land use: 2015 North American Land Change Monitoring System (NALCMS) 30 m land cover dataset.
  • Groundwater recharge: Estimates from the "Projets d’acquisition de connaissances sur les eaux souterraines (PACES)" and other regional studies (ranging from 50 to over 500 mm yr⁻¹).

Main Results

• All configurations achieved Kling-Gupta Efficiency (KGE) values above 0.5, with calibration and validation performances deviating by less than 5 %.
• During validation, BL (median KGE = 0.824) and GW (median KGE = 0.830) configurations showed higher KGE values for streamflow compared to GW-RC (median KGE = 0.770). However, GW-RC demonstrated more consistent KGE values between calibration and validation periods.
• GW-RC simulated higher surface runoff (21 %) and baseflow (17 %) but lower interflow (20 %) proportions compared to BL and GW during the validation period.
• BL exhibited the highest actual evapotranspiration (47 %) and lowest groundwater recharge and baseflow.
• GW-RC showed a lower dynamic range of groundwater recharge during snowmelt but consistently higher recharge rates during winter, summer, and autumn, peaking in fall.
• GW-RC's simulated seasonal recharge rates (e.g., Matane catchment: winter 50 mm, spring 54 mm, summer 60 mm, fall 72 mm) aligned well with documented regional estimates.
• BL struggled to simulate recharge rates exceeding 250 mm yr⁻¹, while GW and GW-RC better captured spatial trends in groundwater recharge, with GW-RC providing estimates more consistent with PACES data.

Contributions

• Demonstrates that incorporating groundwater recharge as a constraint in hydrological model calibration significantly improves the representation of internal hydrological processes and seasonal runoff patterns, even if it leads to a slightly lower streamflow-only performance metric (KGE).
• Provides a more physically realistic simulation of groundwater dynamics and the balance between surface runoff and interflow, particularly during snowmelt, which is crucial for snow-dominated catchments.
• Offers a robust approach to mitigate equifinality in hydrological modeling by moving beyond streamflow-only calibration, enhancing the reliability of simulated internal variables for applications like climate change impact assessment and water resource management.
• Highlights that effective multi-variable calibration does not necessarily require highly precise initial recharge estimates, as even a modest emphasis on recharge within the objective function can lead to substantial improvements in model realism.

Funding

• Ministère des Ressources naturelles et des Forêts (Quebec, Canada) – Project number 112332187 (Direction de la recherche forestière, led by Jean-Daniel Sylvain).
• Ministère des Ressources naturelles et des Forêts (Quebec, Canada) – Forest research service contract number 3322-2022-2187-01 (obtained by Richard Arsenault).

Citation

@article{Talbot2025Enhancing,
  author = {Talbot, Frédéric and Sylvain, Jean‐Daniel and Drolet, Guillaume and Poulin, Annie and Arsenault, Richard},
  title = {Enhancing physically based and distributed hydrological model calibration through internal state variable constraints},
  journal = {Hydrology and earth system sciences},
  year = {2025},
  doi = {10.5194/hess-29-6549-2025},
  url = {https://doi.org/10.5194/hess-29-6549-2025}
}