Yang et al. (2026) A Convective Initiation Nowcasting Algorithm Based on FY-4B Satellite AGRI and GHI Data

Identification

• Journal: Atmosphere
• Year: 2026
• Date: 2026-04-08
• Authors: Zongxin Yang, Zhigang Cheng, Wenjun Sang, Wen Zhang, Yu Huang, Yuwen Huang, Z. Wang
• DOI: 10.3390/atmos17040380

Research Groups

• College of Aviation Meteorology, Civil Aviation Flight University of China, Chengdu, China
• China Meteorological Administration Aviation Meteorology Key Laboratory, Chengdu, China
• Mianyang Flight College, Civil Aviation Flight University of China, Mianyang, China
• Chengdu Municipal Meteorological Department, Chengdu, China

Short Summary

This study develops a convective initiation (CI) nowcasting algorithm for Sichuan Province, China, by integrating high-resolution FY-4B AGRI and GHI satellite data with optical flow compensation. The algorithm significantly improves CI detection probability and lead time compared to existing methods, offering enhanced short-term warnings for regional convective weather.

Objective

• To develop a physical convective initiation nowcasting algorithm using FY-4B AGRI and GHI observations.
• To improve early detection through high-temporal-resolution data and motion-compensated cloud analysis.
• To evaluate the algorithm’s performance against existing methods using radar observations and statistical metrics.

Study Configuration

• Spatial Scale: Algorithm developed using data from Mianyang radar station coverage (30.5° N to 32.5° N latitude, 103.6° E to 106° E longitude). Validated in the remaining areas of the Chengdu radar scan area (28.5° N to 32.5° N latitude, 102° E to 106° E longitude), excluding a 230 km radius around Mianyang Airport.
• Temporal Scale: Data collected between 22 July and 9 August 2023. Nowcasting lead times up to 34.2 minutes on average. Satellite data temporal resolutions: AGRI (15 minutes), GHI (approximately 1 minute). Radar data temporal resolution: approximately 6 minutes.

Methodology and Data

• Models used:
  • Proposed Convective Initiation (CI) nowcasting algorithm (physical, rule-based).
  • Farneback optical flow method for motion compensation.
  • Compared against: AGRI-only method (variant of proposed), SATCAST algorithm.
• Data sources:
  • Satellite:
    • Fengyun-4B (FY-4B) satellite Advanced Geostationary Radiation Imager (AGRI) L1 data (15 minutes temporal resolution, 0.5 km to 4 km spatial resolution).
    • Fengyun-4B (FY-4B) satellite Geostationary High-speed Imager (GHI) 250 m L1 data (approximately 1 minute temporal resolution, 0.25 km to 2 km spatial resolution).
  • Observation (Radar):
    • Doppler radar echo base data from Mianyang Airport and Chengdu Meteorological Bureau (approximately 6 minutes temporal resolution). CI defined as first occurrence of composite reflectivity ≥ 35 dBZ.
  • Auxiliary: Global Precipitation Measurement (GPM) mission precipitation data (2-hour cumulative precipitation).

Main Results

• The proposed algorithm achieved a Probability of Detection (POD) of 83.1% (95% CI: 77.5–88.6%), a False Alarm Ratio (FAR) of 33.0% (95% CI: 26.3–39.6%), and a Critical Success Index (CSI) of 58.9% (95% CI: 52.6–65.4%).
• The average lead time for correctly predicted CI events was 34.2 minutes, which is 4.6 minutes longer than the AGRI-only algorithm (29.6 minutes) and 7.9 minutes longer than the original SATCAST algorithm (26.3 minutes).
• The 25th percentile threshold for algorithm criteria provided the best balance between detection and false alarms, yielding an F1 score of 0.765. The algorithm demonstrated low sensitivity of CSI to threshold variations, indicating robustness.
• For individual-type CI events, the average lead time was 36.69 minutes (95% CI: 33.57–39.84), while for merging-type CI events, it was 27.47 minutes (95% CI: 23.45–31.82).
• The algorithm showed particular advantages in detecting small-scale, weak, or slowly developing convection, and in warm-cloud-convection environments prevalent in Sichuan during summer.

Contributions

• Developed a novel physical convective initiation nowcasting algorithm for the Sichuan region, effectively integrating FY-4B AGRI and GHI satellite data.
• Enhanced early detection capabilities through the incorporation of high-temporal-resolution GHI observations and motion-compensated cloud analysis using the Farneback optical flow method.
• Demonstrated significant improvements in key performance metrics (POD, CSI) and extended average lead times (5–8 minutes) compared to existing AGRI-only and SATCAST algorithms.
• Established and validated CI determination thresholds specifically tailored for the warm cloud convection climatic background of Sichuan.
• Provides valuable new insights and tools for short-term warnings of regional convective weather, particularly for challenging small-scale and subtle convective events.

Funding

• National Natural Science Foundation of China (Grant No. 42105087)
• Open Foundation of China Meteorological Administration Key Laboratory for Aviation Meteorology (Grant No. HKQXM-2024015)
• Fundamental Research Funds for the Central Universities (Grant No. 25CAFUC04049)

Citation

@article{Yang2026Convective,
  author = {Yang, Zongxin and Cheng, Zhigang and Sang, Wenjun and Zhang, Wen and Huang, Yu and Huang, Yuwen and Wang, Z.},
  title = {A Convective Initiation Nowcasting Algorithm Based on FY-4B Satellite AGRI and GHI Data},
  journal = {Atmosphere},
  year = {2026},
  doi = {10.3390/atmos17040380},
  url = {https://doi.org/10.3390/atmos17040380}
}