Yin et al. (2026) A Two‐Dimensional Deep Learning Scheme With One Predicator Only to Parametrize Global Lightning

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

Identification

• Journal: Geophysical Research Letters
• Year: 2026
• Date: 2026-04-02
• Authors: Ming Yin, Yuyi Wang, Yun Fang, Feng Li, Xiushu Qie, Yeying Wang, Daoming Wei, Huihui Yuan, Yafeng Ji, Yu Zhang, Jie Liu
• DOI: 10.1029/2025gl120502

Research Groups

Not specified in abstract.

Short Summary

This study introduces a novel two-dimensional artificial intelligence-based global lightning scheme utilizing convective available potential energy (CAPE) as a single predictor, achieving significant improvements in global lightning simulation performance and effectively mitigating the underestimation of extreme lightning density.

Objective

• To develop and evaluate a novel two-dimensional artificial intelligence-based global lightning scheme using a single predictor (CAPE) to enhance global lightning simulation performance, particularly for extreme lightning events.

Study Configuration

• Spatial Scale: Global, two-dimensional gridded (involving neighboring and multiple grid cells).
• Temporal Scale: Not explicitly specified, but implied to be continuous for global simulation and event-based for extreme lightning density.

Methodology and Data

• Models used: Two-dimensional artificial intelligence-based global lightning scheme, lightweight deep neural network.
• Data sources: Convective available potential energy (CAPE) as a single predictor. The specific source of CAPE data is not detailed in the abstract.

Main Results

• The new scheme significantly improved global lightning simulation performance, achieving a determination coefficient (R²) of 0.89.
• This represents a 24% increase in R² over an existing machine learning-based global lightning scheme.
• It achieved a 41% reduction in absolute bias and a 38% decrease in root mean square error (RMSE).
• The scheme effectively alleviated the underestimation of extreme lightning density, a common issue in current lightning schemes.
• This improvement is attributed to the incorporation of nonlocal feature information, suggesting extreme lightning simulation depends on convective system movement from neighboring grid cells or the presence of large convective systems spanning multiple grid cells.

Contributions

• Presentation of a novel two-dimensional artificial intelligence-based global lightning scheme.
• Demonstrated significant improvements in global lightning simulation performance using only a single predictor (CAPE).
• Crucially addressed and alleviated the long-standing issue of underestimation of extreme lightning density in simulations.
• Provided insight into the importance of nonlocal feature information (convective system dynamics) for extreme lightning simulation.
• Developed a lightweight deep neural network with potential for implementation in global climate models.

Funding

Not specified in abstract.

Citation

@article{Yin2026TwoDimensional,
  author = {Yin, Ming and Wang, Yuyi and Wang, Yuyi and Fang, Yun and Li, Feng and Qie, Xiushu and Wang, Yeying and Wang, Yeying and Wei, Daoming and Yuan, Huihui and Ji, Yafeng and Zhang, Yu and Liu, Jie},
  title = {A Two‐Dimensional Deep Learning Scheme With One Predicator Only to Parametrize Global Lightning},
  journal = {Geophysical Research Letters},
  year = {2026},
  doi = {10.1029/2025gl120502},
  url = {https://doi.org/10.1029/2025gl120502}
}