Liu et al. (2026) Representing canopy interception in a distributed snow model to improve snow accumulation and melt timing

Identification

• Journal: Journal of Hydrology
• Year: 2026
• Date: 2026-10-23
• Authors: Yijing Liu, J. Michelle Hu, Alvaro Robledano, S. McKenzie Skiles
• DOI: 10.1016/j.jhydrol.2026.136669

Research Groups

• School of Environment, Society & Sustainability, University of Utah, Salt Lake City, UT, USA

Short Summary

This study integrated forest canopy snow processes into the widely used iSnobal model to improve snow accumulation and melt timing in subalpine forests. The enhanced model, Canopy-iSnobal, significantly improved snow depth simulations and melt timing, particularly during late melt, by accounting for intercepted snowfall and canopy snow loss.

Objective

• Quantify coniferous forest canopy-intercepted snowfall in the East River Watershed.
• Compare snow simulation differences between the default iSnobal and the canopy-enabled Canopy-iSnobal.
• Evaluate improvements in model skill when forest canopy effects are considered in iSnobal.

Study Configuration

• Spatial Scale: East River Watershed (ERW), Upper Colorado River Basin, with elevations ranging from 2440 m to 4330 m. The study focused on approximately 290 km² of coniferous forests, modeled at a 50 m spatial resolution.
• Temporal Scale: Five water years (WY 2018–WY 2022), with hourly simulations. Each water year spans from 1 October to 30 September of the following year.

Methodology and Data

• Models used:
  • iSnobal: A process-based, spatially distributed snow energy and mass balance model.
  • Canopy-iSnobal: The modified iSnobal implementation incorporating canopy processes.
  • Hedstrom-Pomeroy (HP98) model: Used for quantifying snow interception by the forest canopy.
  • Roesch et al. (2001) unloading scheme: Applied to quantify snow mass loss from the canopy due to wind and temperature.
  • Spatial Modeling for Resource Framework (SMRF): A preprocessing tool for distributing meteorological forcings.
  • Katana: Downscales HRRR wind fields using the WindNinja model.
  • Automated Water Supply Model (AWSM): Manages daily iteration and time-step execution of SMRF and iSnobal.
  • Random Forest models: Employed for statistical analysis to assess the influence of environmental and canopy process variables on model performance.
• Data sources:
  • Meteorological forcing: High-Resolution Rapid Refresh (HRRR) numerical weather prediction model outputs (3 km resolution).
  • Elevation: USGS 3DEP (10 m resolution, resampled to 50 m).
  • Vegetation cover/type: LANDFIRE (30 m resolution, resampled to 50 m).
  • Leaf Area Index (LAI): Estimated from Landsat 8 Operational Land Imager (OLI) surface reflectance (30 m resolution).
  • Snow depth for validation: Lidar-based snow depth measurements from Airborne Snow Observatories, Inc. (50 m spatial resolution).

Main Results

• Annually, 50–150 mm of snowfall was intercepted by canopies, representing approximately 10% to 30% of cumulative snowfall.
• Within the canopy, snow melting was the dominant loss mechanism, ranging from 80 mm to 180 mm in mid-elevation areas, while wind-driven unloading was generally limited (below 5 mm monthly in most cases) except at higher elevations (exceeding 80 mm).
• Relative to baseline iSnobal runs, Canopy-iSnobal simulated decreased snow depths (mean reduction of 0.09–0.11 m), delayed snow onset dates (SOD) (by 3–13 days), and shortened snow cover duration (SDD advanced by 3–6 days).
• The largest mean snow depth differences between the two implementations occurred in March (–0.19 ± 0.09 m), April (–0.21 ± 0.09 m), and May (–0.19 ± 0.08 m).
• Validation against lidar-mapped snow depth showed that Canopy-iSnobal improved model performance in 39.4% to 57.3% of pixels, with reduced errors (mean absolute error and root mean square error decreased by 0.11 m to 0.17 m on three dates), particularly during the late melt period (May and June).
• Random forest analysis indicated that intercepted snowfall, delayed SOD, and LAI were the primary controls on snow depth differences between models (R² = 0.93). Aspect and October snowfall were the most important factors for delayed SOD (R² = 0.67), while cumulative snowfall, elevation, and slope were key for advanced SDD (R² = 0.56).

Contributions

• This study presents the first integration of forest canopy snow processes (interception, melt, and unloading) into the widely used, process-based distributed snow energy balance model iSnobal.
• It provides a comprehensive quantification of canopy snow processes and their significant impact on snow accumulation, onset, and depletion timing in a mountainous, partially forested basin.
• The research demonstrates substantial improvements in iSnobal's skill for simulating snow depth and melt timing, particularly during the critical late melt season, through the explicit representation of canopy effects.
• The developed Canopy-iSnobal framework offers a basis for more reliable snowmelt inputs for operational water supply forecasting and enhanced water resource management in snow-dominated forested regions.
• The findings underscore the importance of accounting for forest canopy processes in snow models for accurate snow cover responses to climate change and improved climate modeling studies.

Funding

• Bureau of Reclamation (Grant No. R24AC00061)
• Cooperative Institute for Research to Operations in Hydrology (CIROH) (NOAA Cooperative Institute Program award NA22NWS4320003)

Citation

@article{Liu2026Representing,
  author = {Liu, Yijing and Hu, J. Michelle and Robledano, Alvaro and Skiles, S. McKenzie},
  title = {Representing canopy interception in a distributed snow model to improve snow accumulation and melt timing},
  journal = {Journal of Hydrology},
  year = {2026},
  doi = {10.1016/j.jhydrol.2026.136669},
  url = {https://doi.org/10.1016/j.jhydrol.2026.136669}
}