The National Center for Atmospheric Research (NCAR) Auto-Nowcast (AN) system provides 30 and 60 min time and place specific forecasts of thunderstorm initiation, growth and dissipation on a local-scale (~200 x 200 km grid with 1 km resolution). This expert system uses fuzzy logic methodology and incorporates forecast predictors derived from years of field research, numerical modeling studies, and theory. The system makes use of operational data sets including radar, satellite, surface stations, lightning, profilers and radiosondes. Also included are diagnostic and forecast information from numerical models that are run in real time. Development of the AN system is sponsored by the; Federal Aviation Administration (FAA) Weather Research Program as part of the Convective Weather PDT, Army Test and Evaluation Command (ATEC), National Weather Service (NWS) as part of the System for Convection Analysis and Nowcasting (SCAN), and National Science Foundation under the U.S. Weather Research Program.
In this paper, AN forecast from two operational sites, the NWS Weather Forecast Office (WFO) at Sterling, Virginia and the Forecast Office at White Sands Missile Range are statistically compared to extrapolation forecast. Summary statistics for the 30 and 60 min forecast show that the AN system has increased skill in terms of Probability of Detection (POD) and Critical Success Index (CSI). Although the False Alarm Ratio (FAR) varies, generally it is similar for both extrapolation and AN systems. As expected the differences between extrapolation and AN forecast CSI scores were relatively small when examined over long time periods. The small differences are partially due to large multi-cellular systems where an extrapolation forecast works very well. These large systems tend to dominate the statitics because of their area coverage. However from an user-standpoint accurate forecast of change, e.g. storm initiation, growth and dissipation, are extremely important.
To better understand the forecast capabilities various weather situations are reviewed. They include; initiation of squall line at location of boundary collision, initiation along quasi-stationary boundary, initiation and growth associated with a quasi-stationary boundary on an active storm day, initiation of storms as a boundary moves into a field of cumulus clouds, storm growth at the location of storms and boundary intersections and storm-storm mergers, extrapolation and formation of bow-echo, initiation and growth of severe thunderstorms and dissipation of a line of thunderstorms. These case examples are encouraging because they show the AN ability to improve over extrapolation forecast. The improvement is primarily associated with the AN ability to correctly forecast new storm development.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner