Convective Initiation Nowcasting Over China From Fengyun-4A Measurements Based on TV-L1 Optical Flow and BP_Adaboost Neural Network Algorithms

Convective initiation (CI) nowcasting over China has a problem of a high false-alarm rate (FAR) due to the local convective processes, most of which do not produce severe weather. In order to focus on those CIs with severe weather, a new 0-2 h CI nowcasting system is developed in this article, named rapidly developed convection monitoring system (RDCMS), using the data from the advanced geosynchronous radiation imager onboard China Fengyun-4A(FY-4A) satellite. The RDCMS is not only used to identify CIs at the early stage, but also to prevent the relative high FAR. The key solutions of RDCMS are to introduce the total variation L ₁ norm (TV-L ₁ ) optical flow method for more tracking efficiency, and to use a supervised learning method named BP_Adaboost neural network for severe convection checking. Case studies show that the RDCMS' skills have been improved as expected in four regions of China, which are Qinghai-Tibet Plateau, the East China, Northeast China, and South China. In the southern China region, the CI lead time is 17-40 min, and the best probability of detection is as high as 0.80, with the FAR lower than 0.34.