Automatic Recognition Of Convective Initiation Based On Satellite Imagery

Atmospheric short-term strong convective activity, usually arising from severe weather, poses serious threats on the production and life of human beings. To respond to severe weather, using data collected by satellite and radar to monitor and forecast short-term convective activities is a trend. For characteristics of short-term convective activity, the appearance of Convective Initiation (CI) is the important signs of the strong convective weather events, e.g. hails and tornadoes. Recognizing and forecasting CI have therefore become an important research field of short-term forecasting.One of the difficulties on CI forecasting is that traditional atmospheric monitoring instruments are unable to provide enough information due to the low temporal and spatial resolution. The geostationary satellites is capable of monitoring a wide range as compared with other monitoring instruments, and can provide information in the infrared band, thus has been a common means of monitoring and forecasting convective activities. Through the analysis of images of visible, infrared channels obtained by the satellite, the relevant characteristics of atmospheric convections can be extracted. As to achieve the purposes of atmospheric convection, forecasting short-term convective activity will be obtained.This paper studied the methods for automatic recognition of CI through satellite imagery analysis and processing. Based on the atmospheric characteristics of CI activity, interpolation and mask processing technology are used to pre-process satellite images. In accordance with the characteristics of the atmosphere CI activity, design the corresponding checker pattern classifier. By put pixel level processes on the post-processed image, to get the recognition results of CI activity prediction of the monitoring region. Finally, to verify the feasibility of the system, the data of rainfall collected by weather monitoring radar are used to validate forecasting results. The satellite and radar data of a number of convective events in the Beijing-Tianjin region were used to evaluate the proposed methods. The result of experiments shows the efficiency of the proposed methods.