The accuracy of hydrologic modeling results relies heavily on the quality of the precipitation estimates or forecasts used to drive a rainfall-runoff model. For short lead-times (0-3 hours), short-term extrapolation is considered to have higher forecast skill as compared to other available methods. To provide forecasts with high spatial and temporal resolution, this paper presents the use of an algorithm for short-term quantitative precipitation forecasts (QPFs) of two NEXRAD-based data products (Stage III and WSI). The chosen extrapolation model is the MIT Lincoln Laboratory Growth and Decay Stormtracker (GDST). This model utilizes a scale separation filtering process to accurately track the large scale, envelope motion of the storm. Several different storm cases varying in spatio-temporal organization have been selected from archived radar data for the Arkansas-Red River Basin over the 1998-2001 period. Results from radar extrapolation using the GDST are evaluated using the Critical Success Index (CSI) as a base statistic. The CSI concept is expanded to use larger verification areas, as well as thresholds to more accurately demonstrate forecast performance. Analysis results show that the short-term forecasts are useful for hydrologic forecasting, with neighborhood CSI scores greater than 50% for a 2 hour forecast and greater than 65% for a 1 hour forecast.