Belinska et al. (2026) Exploring D-Pol-InSAR Coherence Regions for Snow Water Equivalent Estimation

Identification

• Journal: IEEE Transactions on Geoscience and Remote Sensing
• Year: 2026
• Date: 2026-01-01
• Authors: Kristina Belinska, Georg Fischer, Konstantinos Papathanassiou, Irena Hajnsek
• DOI: 10.1109/tgrs.2026.3672275

Research Groups

• Microwaves and Radar Institute, German Aerospace Center (DLR), Wessling, Germany
• Institute of Environmental Engineering, Swiss Federal Federal Institute of Technology in Zurich (ETH), Zürich, Switzerland
• Environmental Earth Observation Information Technology GmbH (ENVEO) (for ground measurements)

Short Summary

This study proposes a Differential Polarimetric Interferometric SAR (D-Pol-InSAR) approach to enhance Snow Water Equivalent (SWE) change estimation by modeling and exploiting polarimetric coherence regions. The method demonstrates improved accuracy and partial resolution of the 2π phase ambiguity compared to conventional single-polarization D-InSAR, particularly for significant snowfall events.

Objective

• To improve the accuracy of Snow Water Equivalent (SWE) change retrieval by combining Differential Interferometric Synthetic Aperture Radar (D-InSAR) measurements with polarimetric information.
• To advance the theoretical understanding of temporal coherence regions influenced by snow accumulation and anisotropy.
• To develop and demonstrate a D-Pol-InSAR retrieval method that optimizes snow layer depth, density, and anisotropy by minimizing the difference between modeled and measured coherence regions.

Study Configuration

• Spatial Scale: Woergetal valley, Austrian Alps. Analysis performed at pixel level using a 9x9 window size for coherence estimation, with areas of local slopes greater than 50% masked out.
• Temporal Scale: Airborne SAR campaign conducted from 2021-03-02 to 2021-03-19. Two interferograms were analyzed: 2021-03-03/2021-03-06 (Inf1) and 2021-03-06/2021-03-19 (Inf2).

Methodology and Data

• Models used:
  • Differential Polarimetric Interferometric SAR (D-Pol-InSAR) model (proposed in this study).
  • Model for D-InSAR phase and SWE change based on [10].
  • Maxwell Garnett formula for effective permittivity of ice-air mixtures [42].
  • Models for temporal decorrelation [18] and additive noise decorrelation [47].
• Data sources:
  • Airborne SAR data: DLR's F-SAR system, providing fully polarimetric C-band (wavelength $\lambda = 5.65 \text{ cm}$) and L-band ($\lambda = 22.62 \text{ cm}$) data.
  • Ground measurements: Snow depth (using avalanche probes) and snow density (from snow pits) collected by ENVEO.
  • Phase calibration: A corner reflector with a side edge of 0.95 m, located at longitude 10.96°, latitude 47.21°.

Main Results

• Conventional single-polarization D-InSAR SWE change estimates showed significant differences between HH and VV polarizations, which increased with larger snowfall events (e.g., 0.0014 m for Inf1 and 0.0033 m for Inf2 at C-band), attributed to snow anisotropy.
• The proposed D-Pol-InSAR model successfully simulates coherence regions, demonstrating that snow anisotropy leads to a wider spread of phase values across different polarizations.
• An increase in snow depth primarily results in a phase change and an increase in the phase extent of the coherence region, while an anisotropy change primarily affects the phase extent.
• The D-Pol-InSAR retrieval algorithm, which optimizes snow depth, density, and anisotropy, consistently outperformed single-polarization D-InSAR in all tested scenarios.
• For the larger snowfall event (Inf2), the D-Pol-InSAR model significantly improved accuracy and was able to resolve all 2π phase wraps in L-band and some in C-band.
• Root Mean Square Error (RMSE) for SWE change estimates:
  • Inf1 (C-band): D-Pol-InSAR RMSE = 0.0041 m, compared to 0.0046 m (VV) and 0.0044 m (HH) for single-pol D-InSAR.
  • Inf2 (C-band): D-Pol-InSAR RMSE = 0.0322 m, compared to 0.0686 m (VV) and 0.0676 m (HH) for single-pol D-InSAR.
  • Inf1 (L-band): D-Pol-InSAR RMSE = 0.0044 m, compared to 0.0046 m (VV) and 0.0046 m (HH) for single-pol D-InSAR.
  • Inf2 (L-band): D-Pol-InSAR RMSE = 0.0143 m, compared to 0.0672 m (VV) and 0.0575 m (HH) for single-pol D-InSAR.

Contributions

• Introduction and validation of a novel D-Pol-InSAR approach that integrates interferometric and polarimetric information to significantly improve SWE change estimation.
• Development of a theoretical framework and model for D-Pol-InSAR coherence regions, elucidating the influence of snow depth, density, and anisotropy on interferometric phase and phase extent.
• Demonstration that exploiting polarimetric diversity can partially mitigate the 2π phase ambiguity limitation inherent in conventional D-InSAR, particularly for larger SWE changes.
• Quantitative evidence of substantial improvements in SWE change retrieval accuracy (reduced RMSE) compared to single-polarization D-InSAR, especially during significant snowfall events.
• Highlighting the critical importance of accounting for snow anisotropy to achieve accurate SWE change retrieval using SAR data.

Funding

• ESA’s Simulation of Hydroterra SAR System Performance in the Mediterranean and the Alps Based on Experimental Airborne SAR Data (SARSimHT-NG) study.

Citation

@article{Belinska2026Exploring,
  author = {Belinska, Kristina and Fischer, Georg and Papathanassiou, Konstantinos and Hajnsek, Irena},
  title = {Exploring D-Pol-InSAR Coherence Regions for Snow Water Equivalent Estimation},
  journal = {IEEE Transactions on Geoscience and Remote Sensing},
  year = {2026},
  doi = {10.1109/tgrs.2026.3672275},
  url = {https://doi.org/10.1109/tgrs.2026.3672275}
}