Over the past decade the United States has seen drastic increases in air traffic delays resulting in enormous economic loses. Analysis shows that more then 50% of air traffic delays are due to convective weather. In response the FAA has assembled scientific and engineering teams from MIT Lincoln Laboratory, NCAR. NSSL, FSL and several universities to develop convective weather forecast systems to aid air traffic managers in delay reduction. A user-needs study conducted by Lincoln Laboratory identified that a major source of air traffic delay was due to line thunderstorms (Forman et al., 1999). Recognizing that the line storm envelope motion was distinct from the local cell motion was the impetus for developing the Growth and Decay Storm Tracker' (Wolfson et al., 1999). The algorithm produces forecasts by extracting large-scale features from two dimensional precipitation images. These images are tracked, using either correlation techniques (Terminal Convective Weather Forecast or TCWF) or centroid techniques (National Convective Weather Forecast or NCWF). In TCWF, the track vector field is used to advect the current precipitation images formed to produce a series of forecasts into minute increments up to 60 minutes. The TCWF forecasts are highly skilled for large scale persistent line storms. However, detailed performance analysis of the algorithm has shown that in cases dominated by airmass storms, the algorithm occasionally performed poorly (Theriault et al., 2001). In this paper we describe the sources of error discovered in the TCWF algorithm during the Memphis 2000 performance evaluation, and describe recent enhancements designed to address these problems.