An Evaluation of an Automatic Cell Detection and Tracking Algorithm

Book excerpt: A storm tracking algorithm designed to detect and track fine structure in digitized radar data is evaluated. These fine structures are defined by regions containing values within 3 dB of peaks in reflectivity factor. The algorithm describes storm structure and evolution by correlating these peak regions in time and space. The evaluation consists of a comparison of the algorithm output with raw data and with output from an AFGL algorithm which detects and tracks three-dimensional reflectivity weighted centroids defined by a preselected threshold. It is concluded that the algorithm cannot reliably detect and track significant structures within storms when applied to data sets with a temporal resolution of aprox. 6 min and a spatial resolution of 1.0 deg in azimuth and 0.7 deg in elevation. The significance of tracking 3 dB peaks is questioned and the implication of defining a larger peak threshold is discussed. The algorithm does track the large features of storms with results similar to the AFGL algorithm. However, it does not run in real time and is not modular, unlike the AFGL algorithm.