Waring et al. (2025) A Deep Learning Approach to Downscaling Microwave Land Surface Temperatures for a Clear-Sky Merged Infrared-Microwave Product

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

Identification

• Journal: Remote Sensing
• Year: 2025
• Date: 2025-11-30
• Authors: Abigail Marie Waring, Darren Ghent, David Moffat, Carlos Jimenez, John Remedios
• DOI: 10.3390/rs17233893

Research Groups

[Information not available in the provided text.]

Short Summary

This study presents the first validated clear-sky merged land surface temperature (LST) product for the USA by combining downscaled passive microwave (PMW) data with MODIS thermal infrared (TIR) observations using a modified U-Net, demonstrating improved spatial coverage and temporal completeness over single-sensor products.

Objective

• To develop and validate a clear-sky merged land surface temperature (LST) product for the USA by fusing passive microwave (PMW) and thermal infrared (TIR) satellite data, thereby overcoming the limitations of cloud gaps in TIR and coarse resolution/uncertainty in PMW.

Study Configuration

• Spatial Scale: Continental USA, 5 km resolution.
• Temporal Scale: 18 years (2004–2021).

Methodology and Data

• Models used: Modified U-Net deep learning network.
• Data sources: Downscaled passive microwave (PMW) data from AMSR-E and AMSR2; MODIS thermal infrared (TIR) observations.

Main Results

• The merged LST product covers 2004–2021 at 5 km resolution, providing a compromise between spatial detail and robustness.
• The deep learning model achieved strong performance with R² values of 0.80 for daytime and 0.75 for nighttime LST.
• The merged time series shows a marked reduction in cloud-contamination artefacts compared to MODIS and AMSR signals.
• The product is spatially consistent across sensor transitions (AMSR-E to AMSR2) and reduces artefacts from TIR cloud contamination.
• Validation against ground stations yielded accuracy between that of TIR and PMW products, with better accuracy at night and moderate positive biases influenced by land cover and terrain.
• The merged product enhances spatial coverage over AMSR alone and temporal completeness over MODIS alone.
• Moderate residual temporal and seasonal biases were observed, linked to land cover, emissivity error propagation, and sampling differences.
• Strong positive biases persist over complex terrain.

Contributions

• Presents the first validated clear-sky merged land surface temperature (LST) product for the USA.
• Demonstrates a novel application of a modified U-Net deep learning network for fusing satellite LST data from different sensor types (PMW and TIR).
• Provides a spatially consistent LST record with enhanced coverage and completeness, suitable for long-term environmental and climate monitoring, overcoming limitations of single-sensor products.

Funding

[Information not available in the provided text.]

Citation

@article{Waring2025Deep,
  author = {Waring, Abigail Marie and Ghent, Darren and Moffat, David and Jimenez, Carlos and Remedios, John},
  title = {A Deep Learning Approach to Downscaling Microwave Land Surface Temperatures for a Clear-Sky Merged Infrared-Microwave Product},
  journal = {Remote Sensing},
  year = {2025},
  doi = {10.3390/rs17233893},
  url = {https://doi.org/10.3390/rs17233893}
}