Shi et al. (2026) A Physics‐Guided Feature Selection Machine Learning Approach for Cloud Probability Retrieval and Detection From TEMPO Imagery

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

Identification

• Journal: Journal of Geophysical Research Atmospheres
• Year: 2026
• Date: 2026-08-24
• Authors: Xiaohang Shi, Yulong Fan, Kai Yang, Zhanqing Li, Xiong Liu, Jhoon Kim, Lin Sun, Shulin Pang, Chao Liu, Jing Wei
• DOI: 10.1029/2026jd046434

Research Groups

• Laboratory for Atmospheric and Space Physics (LASP), University of Colorado Boulder
• Department of Computer Science, University of Colorado Boulder
• National Center for Atmospheric Research (NCAR)

Short Summary

This study presents a physics-guided feature selection machine learning framework that improves cloud detection from hyperspectral observations using TEMPO data. The framework achieves high classification performance and outperforms existing methods.

Objective

• Develop an accurate cloud detection method for hyperspectral remote-sensing measurements

Study Configuration

• Spatial Scale: North America
• Temporal Scale: Hourly

Methodology and Data

• Models used: Physics-guided feature selection machine learning (PGML) framework with LightGBM architecture
• Data sources: TEMPO hyperspectral observations, GOES-R cloud masks, AERONET observations, GEMS data

Main Results

• The PGML framework effectively captures diurnal variations in cloud fraction and spatial structures.
• The resulting cloud masks show high classification performance against reference masks (BOA: 84.43%).
• Independent validation with AERONET observations confirms the improvement (BOA: 78.54%).

Contributions

• This study provides a novel approach to cloud detection from hyperspectral data, improving upon existing methods.
• The framework's adaptability and transferability are demonstrated through successful application to GEMS data.

Funding

• This research was supported by NASA's TEMPO project (NNX16AC97P) and the National Science Foundation (NSF) under grant number 1948441.

Citation

@article{Shi2026PhysicsGuided,
  author = {Shi, Xiaohang and Fan, Yulong and Yang, Kai and Li, Zhanqing and Liu, Xiong and Kim, Jhoon and Sun, Lin and Pang, Shulin and Liu, Chao and Wei, Jing},
  title = {A Physics‐Guided Feature Selection Machine Learning Approach for Cloud Probability Retrieval and Detection From TEMPO Imagery},
  journal = {Journal of Geophysical Research Atmospheres},
  year = {2026},
  doi = {10.1029/2026jd046434},
  url = {https://doi.org/10.1029/2026jd046434}
}