Automatic Detection of Linear Mesoscale Convective Systems

The Linear Mesoscale Convective System (MCS) is a vital weather system that causes severe weather such as heavy rain, hail, thunderstorms, and tornadoes. The traditional manual operation cannot give an effective means for timely observation and forecasting of weather systems due to the increase number in meteorological data. Thus, this paper presents an automatic approach for identifying linear MCSs from radar reflectivity' data using support vector machine (SVM). First, a multi-scale segmentation method is used to segment a linear MCS from radar reflectivity images. Then, the linear MCS is standardized to an 8x8 region, and a 64D feature is used to describe the linear MCS. Finally, the detection of linear MCSs is performed by SVM. It is shown in this paper that the linear MCSs can be well automatically identified, and the automatic recognition results are potentially useful for accurate nowcasting.