Evaluation of the Improved YOLO-Based Severe Thunderstorm Wind Intelligent Recognition Algorithm in Guangdong
Abstract：To evaluate the ability of the intelligent recognition algorithm (YOLO) for severe thunderstorm winds operated by the Guangdong Meteorological Observatory to identify high-disaster-risk extreme thunderstorm wind events and its support capability for the province-wide thunderstorm wind warning business, an application evaluation was conducted based on the extreme wind data from the dense automatic weather stations, lightning location, and radar observation data in Guangdong Province from 2023 to 2024. The results show that in two consecutive years of operational applications, the hit rate of thunderstorm wind identification was higher than 0.900, the false alarm rate was lower than 0.450, and the critical success index was higher than 0.520. The algorithm showed stable performance for thunderstorm winds of different intensities, with a hit rate of 0.941 and a critical success index of 0.521 for extreme winds. The algorithm also demonstrated good identification capabilities for thunderstorm winds under different weather patterns, with better performance for systematic thunderstorm winds.
Keywords：Thunderstorm windIntelligent recognition algorithmEffect evaluation
Publication Date：2025-10-20
Online Publishing Date：2025-10-30(First online date of this platform, not the publication date of the document)
Pages：5( 81-84,90 )