Research on Thunderstorm Early Warning Based on Microwave Radiometer

Thunderstorm warning is significant for effective disaster risk reduction. This paper utilizes microwave radiometer, lightning locator and conventional sounding data of Qingdao station from 2023 to 2024. The parameters closely related to thunderstorm activities are selected as the forecasting factors to construct a thunderstorm warning model based on neural network algorithm. Meanwhile, the thunderstorm index measured by microwave radiometer is introduced as a dynamic criterion to realize secondary judgment for thunderstorm warning. Nineteen thunderstorm cases from July to August in 2024 are used for independent test. After introducing thunderstorm index as a criterion, the probability of detection is improved to $89.4 \%$, the false alarm rate is reduced to $21.7 \%$, the TS score increases to $70.8 \%$, and the missed alarm rate decreases to $10.6 \%$. The validation result indicates that dual-judgment mechanism effectively enhances the capability of thunderstorm early warning model. This study realizes the intelligent forecasting of thunderstorm weather at single station, which can be effectively applied in disaster prevention and mitigation operations.