The NCAR Convective Storm Nowcast System provides 0-1 hr high resolution, time and space specific nowcast of convective storms. The expert system uses fuzzy logic to combine predictor fields based on radar, satellite, sounding, mesonet, and a cloud scale numerical model and its adjoint. This paper useds data collected during June and July of 2001 in the Denver, Colorado area to evaluate (a) a new gridded reflectivity and precipitation rate nowcast field, (b) nowcast differences obtained using TDWR and WSR-88D, and (c) nowcast differences based on automated and nowcaster-entered boundaries.

Preliminary results show that accurate and consistent boundary detection and extrapolations increase in importance as the nowcast period increases. For 0-30 min nowcasTS the position of current storms and developing cumulus constrain the area of the nowcast. However, as the nowcast lenth increases, the importance on trends in storm and cumulust development becomes less important. With increasing time the reliance increases on boundary-layer kinimatic fields such as veritcial velocity, boundary relative shear, boundary relative steering flow, boundary collision, and boundary-storm collision. Accurate measures of these fields requires good boundary detections. Results show that automated boundary detection techniques provide more Consistant and reliable boundary nowcast using TDWR then WSR-88D data. Further automated and human inserted boundaries show little statistical differences in 30 min validations however at 60 min forecaster entered boundaries vs. automated can make a significant difference to the nowcasts.