Abstract:The identification, tracking and early warning system for thunderstorm cloud clusters based on Doppler weather radar measurements is one of the most i mportant weather prediction techniques in weather nowcasting service. A newly de veloped automatic tracking and recognition system for thunderstorm cloud cluste rs (TRACER) presented in this study is based on pattern recognitions of cloud c lusters to identify and topoloze their borderlines to establish their lifecycles , and then to make short term predictions of their movements. This system incl udes three primary components: (1) identification of thunderstorm cloud clusters based on their bor derlines; (2) establishment of the ancestry of each thunderstorm cloud cluster using six distinct indices: quad tree, overlap, size, external re ctangle covering the cloud, pattern similarity, partial pattern similarity; and (3) linear extrapolation prediction of their propagation, speed, size, maximum e cho center and status (e.g., strengthening or weakening, and expending or shrink ing). Preliminary results demonstrate that this system can accurately identify and predict thunderstorm cloud clusters. In cooperated with a GIS system, TRACE R can display the geo location, trajectory, evolution of any targeted cloud clu ster and its predicted status. In addition, this system could be used to quanti tatively analyze and evaluate the predictions of any cloud clusters.