Augas et al. (2026) Monolayer or Multilayer Snow Model: Implications for the HYDROTEL Hydrological Model for Flow Modeling

Identification

• Journal: Water
• Year: 2026
• Date: 2026-04-07
• Authors: Julien Augas, Alain N. Rousseau, Etienne Foulon
• DOI: 10.3390/w18070884

Research Groups

• Department of Geography, San Diego State University, San Diego, USA
• Centre Eau Terre Environnement, Institut National de la Recherche Scientifique, Québec, Canada

Short Summary

This study evaluates the impact of replacing the HYDROTEL hydrological model's original monolayer snow module with a multilayer formulation on streamflow simulations across ten snow-dominated watersheds in Quebec, Canada. The multilayer model significantly improved low-flow simulations, particularly during the freshet's falling limb, and reduced bias in cumulative freshet volumes, without degrading overall annual performance.

Objective

• To evaluate whether replacing the HYDROTEL hydrological model's original monolayer snow module with a multilayer formulation (incorporating ice and air layers, and freezing rain processes) improves the accuracy of streamflow simulations, specifically assessing overall modeling performance, spring freshet dynamics (rising and falling limbs), and various hydrological indicators (e.g., annual flood cumulative volume, annual maximum discharge, and freshet timing).

Study Configuration

• Spatial Scale: Regional scale, covering ten watersheds in Quebec, Canada, with areas ranging from 855 square kilometers to 15,490 square kilometers.
• Temporal Scale: Calibration period from 1 October 1981 to 30 September 1989; validation period from 1 October 1990 to 30 September 2002 (with a gap for one watershed). Model evaluation was restricted to 15 March to 15 November of each year.

Methodology and Data

• Models used:
  • HYDROTEL (version 2.8.x-078-00-4.1.15.5551) physically based, continuous, semi-distributed hydrological model.
  • Two snow model configurations:
    • "Mo" (Monolayer with Bands): Original physically based monolayer snow model with altitudinal bands.
    • "Multi" (Multilayer): Enhanced multilayer snow model (based on Augas et al., 2024) incorporating ice and air layers, explicit freezing rain processes, and altitudinal bands.
  • BV3C module for vertical water balance.
  • Kinematic wave and modified kinematic wave equations for flow routing.
  • Penman–Monteith formulation for potential evapotranspiration.
  • OSTRICH optimization framework with the Pareto Archived Dynamically Dimensioned Search (PADDS) algorithm for multi-objective calibration (Kling–Gupta Efficiency and logarithmic Nash–Sutcliffe Efficiency).
• Data sources:
  • Meteorological data: Daily or sub-daily total precipitation, minimum and maximum air temperatures interpolated from Environment and Climate Change Canada meteorological stations.
  • Hydrometric data: Observed daily discharge from ten watersheds in Quebec, Canada.
  • Land cover/land use data: GlobCover project (2009).
  • Watershed physical characteristics: Derived using PHYSITEL GIS-based software.

Main Results

• Overall streamflow performance, as measured by Kling–Gupta Efficiency (KGE), remained comparable between the monolayer (Mo) and multilayer (Multi) snow model configurations.
• The multilayer snow model (Multi) consistently and significantly improved low-flow simulations (logarithmic Nash–Sutcliffe Efficiency, NSElog) during both calibration (median NSElog Mo: 0.80; Multi: 0.81; p = 0.049) and validation periods (median NSElog Mo: 0.73; Multi: 0.75; p = 3.9 × 10⁻³).
• During the freshet rising limb, the monolayer model slightly outperformed the multilayer model for low-flow conditions (NSElog) during the calibration period (median NSElog Mo: 0.79; Multi: 0.76; p < 0.05).
• During the freshet falling limb, the multilayer model showed a statistically significant improvement in low-flow simulations (NSElog) during the calibration period (median NSElog Mo: 0.82; Multi: 0.87; p = 2.0 × 10⁻³).
• The multilayer configuration reduced the relative bias of the cumulative freshet volumes from 13.5% to 8.83% and the annual maximum freshet discharge from 8.99% to 7.06%.
• The multilayer model dynamically discretized the snowpack into 1 to 8 layers during the peak accumulation period.

Contributions

• This study provides the first evaluation of the enhanced multilayer snow module within the HYDROTEL hydrological model for streamflow simulation.
• It demonstrates that increasing the physical realism of snowpack representation (multilayer structure, ice/air layers, explicit freezing rain processes) improves the temporal distribution of snowmelt contributions to streamflow, particularly for low-flow conditions and the falling limb of the freshet.
• The research highlights that while overall annual performance metrics (like KGE) may not show dramatic improvements, the multilayer approach significantly reduces structural uncertainty and bias in critical hydrological indicators such as cumulative freshet volumes, which is crucial for practical water management and flood risk assessment.
• It contributes to the understanding of the "complexity-performance plateau" in snow hydrology, emphasizing the value of improved process representation and reduced bias over solely optimizing peak flow metrics.

Funding

• Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Program (A.N. Rousseau; RGPIN/06757-2019).
• NSERC Applied Research and Development (ARD) and Collaborative Research & Development (CRD) programs via a partnership between the Yukon Energy Corporation (YEC), INRS, and Yukon College (CRDPJ/499954-2016).

Citation

@article{Augas2026Monolayer,
  author = {Augas, Julien and Rousseau, Alain N. and Foulon, Etienne},
  title = {Monolayer or Multilayer Snow Model: Implications for the HYDROTEL Hydrological Model for Flow Modeling},
  journal = {Water},
  year = {2026},
  doi = {10.3390/w18070884},
  url = {https://doi.org/10.3390/w18070884}
}