Study Of Forecasting Convective Initiation Base On Geostationary Meteorological Satellite

Convective weather can have a huge impact on flight safety,causing strong bumps and turbulence on aircraft,while convective weather can lead to flight delays,cancellations or flight diversions,causing economic losses and inconvenience to airlines and passengers.Therefore,it is of great significance to have early warning of convective weather for short-time catastrophic weather protection.In this thesis,a set of convective primordial warning algorithm is designed for Guangdong region using the satellite data provided by the Himawari-8 geostationary meteorological satellite for July-September 2021 and the Guangdong radar echo puzzle data for the corresponding time period.The core of the article is divided into the following parts: the establishment of convective incipient case set,cumulus mask,the selection of warning indicators and the determination of the corresponding thresholds,and the minimum number of warning indicators required to satisfy.(1)The infrared channel data are first normalized using the inverse distance weighted difference,and then the infrared and visible images are fused using the CL(Chui-Lian)multiwavelet with directional contrast,which enriches the cloud map information by this step and also paves the way for the subsequent cumulus cloud mask.(2)Using radar echo puzzles and satellite fusion cloud maps,we analyze a typical convective incipient activity process to illustrate how this thesis collects convective incipient cases to build up a case set,and then construct a convective incipient warning indicator set based on the physical significance of Sunflower 8 single channel and channel combination.(3)In this thesis,the relevant threshold values of cumulus mask are adjusted,and the results show that the cumulus mask technique can eliminate a large number of irrelevant pixels such as ground and mature cumulus clouds,which can improve the accuracy and effectiveness of the subsequent convective warning algorithm.(4)The indicators in the set of early warning indicators are ranked by the importance of Gini impurity in random forest,and the 8 indicators most relevant to convective primordial are screened out,and the corresponding indicator thresholds are obtained by specific statistical analysis of these 8 indicators,and then the minimum number of indicators required to satisfy a certain hit rate is calculated by the pod(Probability Of Detection)function.The conclusion shows that at least 6 of the 8 indicators can achieve more than 70% hit rate to build the complete algorithm.The overall hit rate of the final algorithm is 80.35%,the false alarm rate is 40.07%,and the advance time basically varies from 10-80 minutes.(5)The algorithm is validated by three examples and compared with the improved SATCAST(Satellite Convection Anayl Sis and Tracking)algorithm by Guo Wei,a Chinese scholar.The results show that the algorithm in this thesis can better achieve the early warning of convective primaries,and the warning effect is better than the improved SATCAST algorithm.