Carletti et al. (2025) Multitemporal analysis of Sentinel-1 backscatter during snowmelt using high-resolution field measurements and radiative transfer modelling

Identification

• Journal: ˜The œcryosphere
• Year: 2025
• Date: 2025-11-12
• Authors: Francesca Carletti, Carlo Marín, Chiara Ghielmini, Mathias Bavay, Michael Lehning
• DOI: 10.5194/tc-19-5579-2025

Research Groups

• WSL Institute for Snow and Avalanche Research SLF, Davos, Switzerland
• School of Architecture, Civil and Environmental Engineering, École Polytechnique Fédérale de Lausanne EPFL, Lausanne, Switzerland
• Institute for Earth Observation, Eurac Research, Bolzano, Italy

Short Summary

This study presents a unique high-resolution dataset of wet-snow properties to validate Sentinel-1 backscatter links to snowmelt stages and investigate scattering mechanisms using a radiative transfer model, revealing the dominant influences of liquid water content and surface roughness.

Objective

• To collect a comprehensive ground reference dataset on melting snow tailored for Synthetic Aperture Radar (SAR) applications and utilize it with the Snow Microwave Radiative Transfer (SMRT) model to better understand the key processes governing Sentinel-1 backscatter signatures.

Study Configuration

• Spatial Scale: Weissfluhjoch field site near Davos, Switzerland, with manual measurements conducted within a 20 m x 20 m area, corresponding to the final spatial resolution of the post-processed Sentinel-1 images.
• Temporal Scale: Two consecutive snow seasons (2022–2023 and 2023–2024), with Sentinel-1 acquisitions having a repeat cycle of 12 calendar days (due to Sentinel-1B failure).

Methodology and Data

• Models used:
  • Snow Microwave Radiative Transfer (SMRT) model (v1.0)
  • Electromagnetic model: Symmetrized Strong-Contrast Expansion (SymSCE)
  • Permittivity models: Microwave Emission Model for Layered Snowpacks (MEMLSv3), Debye-like model modified by Hallikainen et al. (1986) (H-86)
  • Interface model: Integral Equation Model (IEM)
• Data sources:
  • Satellite: Sentinel-1 (S1) C-band (5.405 GHz, 5.5 cm wavelength) SAR imagery, VV co-polarization, 20 m x 20 m resolution.
  • Field measurements (manual snow pits): Snow temperature (10 cm or 5 cm vertical resolution, ±0.2 °C uncertainty), density (3 cm vertical resolution, 100 cm³ cutter, 5–10% uncertainty), specific surface area (SSA) (4 cm vertical resolution, InfraSnow sensor, 15% RMSE), liquid water content (LWC) (2, 5, or 10 cm vertical resolution, Denoth capacitive sensor, New Capacitive Snow sensor (NCS), melting calorimetry, ±0.5% to ±1% uncertainty), surface roughness (root mean square of heights (RMSH), correlation length (CL) from digital photography), snow water equivalent (SWE) (cylinder cutter, 10% uncertainty).
  • Automatic measurements: Runoff (lysimeter, sub-hourly resolution), SWE (SSG1000 snow scale, sub-hourly resolution, ±10% error), air temperature (meteorological sensors, sub-hourly resolution).

Main Results

• The observed monotonous increase in Sentinel-1 backscatter following its local minimum during snowmelt is physically attributed to the development of surface roughness, rather than the gradual disappearance of snow cover.
• Multitemporal Sentinel-1 backscatter time series identify two main scattering regimes: one dominated by liquid water content (LWC) in early melt stages (moistening and ripening phases) and another dominated by surface roughness later in the season (runoff phase).
• Simulations generally showed a negative bias of approximately 5 dB with respect to Sentinel-1 recordings across both seasons, with biases more pronounced in 2024.
• Sentinel-1 backscatter signal saturates at approximately -22.4 dB to -23.7 dB, while SMRT simulations saturate at much lower values, around -30 dB, suggesting overestimated absorption loss in existing permittivity formulations.
• C-band radar backscatter sensitivity strongly depends on LWC for lower surface roughness (RMSH ≤ 3 mm); this dependence weakens as surface roughness increases.
• A change in surface RMSH from 2 mm to 3 mm, for relatively low LWC (approximately 1%), can cause a significant backscatter increase of about 6 dB, which is comparable to the observed bias between S1 and SMRT simulations.
• For smooth surfaces (1–2 mm RMSH) and LWC ≥ 1.5%, the difference in backscatter between 30° and 40° incidence angles can exceed 2 dB, indicating significant angular dependence. This angular dependence decreases with increasing surface roughness, dropping below 1.0 dB for RMSH ≥ 10 mm (fully formed suncups).

Contributions

• Presents the first comprehensive, high-resolution ground reference dataset of wet-snow properties specifically tailored for SAR applications.
• Advances the understanding of Sentinel-1 radar backscatter interaction with wet snow, particularly the effects of spatiotemporal LWC variability and the impact of surface roughness on backscatter signatures.
• Demonstrates that the monotonic increase in backscatter after the local minimum during snowmelt is due to surface roughness development, a previously hypothesized phenomenon.
• Redefines the interpretation of multitemporal Sentinel-1 backscatter, suggesting it identifies LWC-dominated and roughness-dominated scattering regimes rather than traditional melting phases.
• Highlights critical challenges in radiative transfer modeling of melting, layered snowpacks, including the need for fully validated permittivity, microstructure, and roughness models for wet snow at C-band frequencies.
• Suggests the potential to estimate surface roughness from Sentinel-1 backscatter and surface LWC, which could inform physics-based snow models for hydrological applications.

Funding

• Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung (Swiss National Science Foundation)
• Autonomous Province of Bolzano
• Project: “SnowTinel — Sentinel-1 SAR assisted catchment hydrology: toward an improved snow-melt dynamics for alpine regions” (grant no. 20021-205190)

Citation

@article{Carletti2025Multitemporal,
  author = {Carletti, Francesca and Marín, Carlo and Ghielmini, Chiara and Bavay, Mathias and Lehning, Michael},
  title = {Multitemporal analysis of Sentinel-1 backscatter during snowmelt using high-resolution field measurements and radiative transfer modelling},
  journal = {˜The œcryosphere},
  year = {2025},
  doi = {10.5194/tc-19-5579-2025},
  url = {https://doi.org/10.5194/tc-19-5579-2025}
}