Lightning identification based on multiple weather radar product data

Lightning, with its extremely high energy level, can cause fires and power outages and pose a significant risk to people and infrastructure. However, predicting lightning remains challenging. The occurrence and development of lightning are closely related to thunderstorms, and weather radar is a cru...

Full description

Saved in:
Bibliographic Details
Main Authors: Mingyue Lu, Tongtong Dong, Min Chen, Manzhu Yu, Hui Liu, Caifen He, Jingke Zhang, Yongwei Mao
Format: Article
Language:English
Published: Taylor & Francis Group 2025-08-01
Series:International Journal of Digital Earth
Subjects:
Online Access:https://www.tandfonline.com/doi/10.1080/17538947.2025.2498604
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Lightning, with its extremely high energy level, can cause fires and power outages and pose a significant risk to people and infrastructure. However, predicting lightning remains challenging. The occurrence and development of lightning are closely related to thunderstorms, and weather radar is a crucial tool for detecting thunderstorm characteristics. Therefore, conducting deep learning research on lightning based on radar data can reveal hidden relationships between lightning and thunderstorm characteristics, laying the foundation for lightning prediction on the basis of thunderstorm features. In this context, this paper proposes a lightning identification model, MLDYOLO, which integrates Mixed Local Channel Attention, Large Separable Kernel Attention, and Dynamic Head modules into the You Only Look Once version 8 model. Using radar-detected thunderstorm characteristic data as input, the model identifies lightning by recognizing the environment in radar data. In this model, five types of radar product data are fused into a multi-feature radar image through a Principal Component Analysis Weighted Average Fusion method, which serves as the model’s input. To validate the proposed model, it is compared with advanced models. Results demonstrate the significant potential of the proposed approach in identifying lightning and supporting future lightning predictions.
ISSN:1753-8947
1753-8955