Li et al. (2026) High Spatio-Temporal Resolution CYGNSS Reflectivity Reconstruction via TCN for Enhanced Freeze/Thaw Retrieval

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Remote Sensing
• Year: 2026
• Date: 2026-04-01
• Authors: Xiangle Li, Wentao Yang, Dong Wang, Weiwei Li, Dandan Wang, Lei Yang
• DOI: 10.3390/rs18071056

Research Groups

Not explicitly stated in the provided text. The Cyclone Global Navigation Satellite System (CYGNSS) is a NASA mission.

Short Summary

This paper proposes a Partial Convolution–Time Convolutional Network (PTCN) to reconstruct high-resolution Cyclone Global Navigation Satellite System (CYGNSS) data, significantly improving spatial and temporal coverage for freeze/thaw (F/T) state retrieval while maintaining accuracy.

Objective

• To develop a novel model (PTCN) for reconstructing high-resolution (3 km) Cyclone Global Navigation Satellite System (CYGNSS) data to enhance freeze/thaw (F/T) state retrieval, particularly in data-sparse high-altitude regions.

Study Configuration

• Spatial Scale: 3 km (reconstruction resolution), 9 km, 36 km (F/T retrieval validation resolutions).
• Temporal Scale: Daily (reconstructed data), improved by 256% compared to original observations.

Methodology and Data

• Models used: Partial Convolution–Time Convolutional Network (PTCN), integrating partial convolution and a time convolutional network (TCN).
• Data sources: Cyclone Global Navigation Satellite System (CYGNSS) reflected signals (original and reconstructed), Soil Moisture Passive–Active (SMAP) data (for gap filling comparison), ground-based F/T retrievals (for validation).

Main Results

• The proposed PTCN model successfully reconstructs CYGNSS data at a 3 km resolution without relying on auxiliary data.
• The coverage of the reconstructed data is six times that of the original observational data.
• Quantitative evaluation using simulated missing data shows an R-value of 0.92 and a root mean square error (RMSE) of 2.7 for the reconstructed data.
• The temporal resolution of the reconstructed data is improved by 256%, successfully filling 92% of observational gaps in Soil Moisture Passive–Active (SMAP) data.
• Freeze/thaw (F/T) retrieval accuracy after reconstruction is comparable to that before reconstruction at both 36 km and 9 km resolutions.
• Compared with ground-based F/T retrievals, the reconstructed F/T accuracies are 87.71% at 36 km and 82.3% at 9 km.

Contributions

• Proposes a novel Partial Convolution–Time Convolutional Network (PTCN) for reconstructing high-resolution CYGNSS data, which does not require auxiliary data.
• Introduces, for the first time, a simulation of missing data for the quantitative evaluation of observational data reconstruction.
• Significantly enhances the spatial (3 km) and temporal (256% improvement) resolution and coverage (6 times) of CYGNSS data for freeze/thaw (F/T) state retrieval, particularly beneficial for high-altitude regions.

Funding

Not explicitly stated in the provided text.

Citation

@article{Li2026High,
  author = {Li, Xiangle and Yang, Wentao and Wang, Dong and Li, Weiwei and Wang, Dandan and Yang, Lei},
  title = {High Spatio-Temporal Resolution CYGNSS Reflectivity Reconstruction via TCN for Enhanced Freeze/Thaw Retrieval},
  journal = {Remote Sensing},
  year = {2026},
  doi = {10.3390/rs18071056},
  url = {https://doi.org/10.3390/rs18071056}
}