Tracking and Short-Term Forecasting of Typhoon Structure

The tracking of wind field circulation is an important and difficult problem in typhoon forecasting. This paper studies how to catch the shape and intensity deformation characteristics of this type of severe weather system, which includes the wind field estimation and typhoon structure prediction in a short period of time. A correlation matching approach is used to retrieve typhoon wind field without the estimation of typhoon center. Firstly, an energy function is designed to dynamically restrain the tracking area instead of using a constant one. Then for each grid at previous time step, a corresponding one with maximum correlation coefficient at current time step is computed. The collection of these computed grid pairs reflects the movement of circulation structure and thus the wind field is obtained. Finally, a quality control is performed with total variation reconstruction to smooth the global wind field by eliminating local error data and filling missing data points with the average velocity in the neighborhood. With the computed wind field, the short-term forecasting of typhoon structure that represented by radar echo data is realized by the semi-Lagrangian advection approach. Experimental result shows that the proposed method could retrieve accurate and smooth typhoon wind field compared with observation data, or data computed with optical flow algorithm, and could predict relative accurate typhoon structure in 1 hour.