The NCAR Auto-Nowcaster is a fuzzy logic based short-term thunderstorm prediction system. One to two hour nowcasts of thunderstorm initiation, growth and decay are routinely produced every 6 minutes. The NCAR Auto-Nowcaster has been running over the Dallas / Fort Worth terminal and surrounding area since early 2005 for the Forecaster Over the Loop demonstration for the National Weather Service. For the past two years, the nowcasts produced by the Auto-Nowcaster have been available to both the NWS Fort Worth office and the Fort Worth Center Weather Service Unit located at the ARTCC. Prior to the spring of 2006, the Auto-Nowcaster was run using one predefined set of fuzzy logic membership functions, weights, and list of predictor fields that were ingested into the system. The system had no flexibility to change the fuzzy logic rules in real-time to respond to diverse synoptic regimes experienced through the spring, summer and fall. For the spring of 2006, the Auto-Nowcaster was modified to allow the NWS forecaster to select from seven different predefined fuzzy logic rule sets to tailor the Auto-Nowcaster processing to the current synoptic regime. The seven new rule sets use different combinations of predictor fields, membership functions, and predictor weights. This paper will examine the change in performance in the Auto-Nowcaster with NWS forecaster selected rule sets versus a single hard-wired rule set using archive cases containing diverse synoptic regimes.