Cheng et al. (2025) A Multi-Stage Deep Learning Framework for Multi-Source Cloud Top Height Retrieval from FY-4A/AGRI Data

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

Identification

• Journal: Atmosphere
• Year: 2025
• Date: 2025-11-12
• Authors: Yinhe Cheng, Shen Liu, Jia-Wei Zhang, Hongjian He, Xiaomin Gu, Shengxiang Wang, Tinghuai Ma
• DOI: 10.3390/atmos16111288

Research Groups

Institutions associated with the development and application of Fengyun-4A satellite data, likely within China's meteorological research sector. Institutions involved in the CALIPSO mission (for providing reference data).

Short Summary

This study proposes a multi-stage deep learning framework to enhance the accuracy of Cloud Top Height (CTH) retrieval from Fengyun-4A (FY-4A) satellite data. The model significantly improves CTH retrieval accuracy, reducing the Mean Absolute Error by 49.12% to 2.03 km compared to the official FY-4A product, and successfully characterizes CTH spatial distribution in complex weather systems.

Objective

• To enhance the accuracy of Cloud Top Height (CTH) retrieval from Fengyun-4A (FY-4A) satellite data using a multi-stage deep learning framework.

Study Configuration

• Spatial Scale: Large scale, covering regions observable by a geostationary meteorological satellite (e.g., regional to continental). Applied to typhoon regions.
• Temporal Scale: Event-specific (e.g., double typhoon event on 10 August 2019) with implications for continuous monitoring by geostationary satellites.

Methodology and Data

• Models used: Multi-stage deep learning framework, multi-source data fusion neural network model.
• Data sources:
  • Fengyun-4A (FY-4A) satellite data.
  • FY-4A/AGRI (Advanced Geosynchronous Radiation Imager) Level 1 calibrated scanning imager radiance data (multi-channel).
  • Cloud parameters: Cloud Top Temperature (CTT) and Cloud Top Pressure (CTP) (likely derived from FY-4A).
  • Reference data: Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) satellite CTH measurement data.

Main Results

• The proposed multi-stage model significantly improves CTH retrieval accuracy.
• Compared to the official FY-4A CTH product, the Mean Absolute Error (MAE) was reduced by 49.12% to 2.03 km.
• The Pearson Correlation Coefficient (PCC) reached 0.85.
• The model successfully characterized the spatial distribution of CTH within typhoon regions during a double typhoon event on 10 August 2019.
• The results are consistent with National Satellite Meteorological Centre (NSMC) reports and clearly reveal the different intensity evolutions of the two typhoons.

Contributions

• Proposes an innovative multi-stage deep learning framework for progressive refinement of cloud parameter estimation.
• Provides an effective solution for high-precision retrieval of high-level cloud CTH at a large scale using geostationary meteorological satellite remote sensing data.
• Demonstrates significant improvement in CTH retrieval accuracy over existing official products (quantified by MAE and PCC).
• Validates the model's applicability under complex weather conditions, such as typhoons.

Funding

Not specified in the provided text.

Citation

@article{Cheng2025MultiStage,
  author = {Cheng, Yinhe and Liu, Shen and Zhang, Jia-Wei and He, Hongjian and Gu, Xiaomin and Wang, Shengxiang and Ma, Tinghuai},
  title = {A Multi-Stage Deep Learning Framework for Multi-Source Cloud Top Height Retrieval from FY-4A/AGRI Data},
  journal = {Atmosphere},
  year = {2025},
  doi = {10.3390/atmos16111288},
  url = {https://doi.org/10.3390/atmos16111288}
}