Updates to the NCAR Auto-nowcaster for the 2004 convective weather season

The NCAR Auto-nowcaster runs as part of the FAA Aviation Weather Research Program's (AWRP) Convective Weather PDT's research demonstration on convective initiation in the Chicago area. This area has very congested air traffic routing, typically very large delays during thunderstorms, and exhibits a great deal of convective initiation by fronts, lake breezes, colliding outflows, and elevated forcing mechanisms. The region covered by the Auto-nowcast System is a sub-section of this larger domain and consists of northeastern Illinois, most of Indiana, and western Ohio. The Auto-nowcaster has been running during the summer months over this region in 2002, 2003, and 2004. The NCAR Auto-nowcaster's unique human-computer interface allows forecasters to enter boundaries and regions of potential storm initiation. Forecaster-entered predictor fields are used along with predictors derived from NWP, observations, extrapolation, and feature-detection algorithms to provide detailed forecasts of convective initiation, growth and decay with 30 and 60 minute lead times to help improve aviation safety and efficiency.

Several updates were implemented for the 2004 operations in response to limitations identified in the previous year's operations. The primary goal of these updates was to increase the sensitivity to better capture convective initiation and also minimize the number of obvious false alarm forecasts. In order to accomplish this goal improved methods for using more large scale information were incorporated into the forecast logic (primarily through the use of RUC data). The use of these large-scale data led to more consistent forecasts and helped capture storm initiation while limiting the number of false alarms.

A new algorithm was developed that allows optimal use of satellite data in the real-time operations. Satellite data is of minimal use for convective initiation forecasts in regions where there is contamination due to large areas of high cloud that obscure the signals for cumulus cloud types and/or cooling IR temperatures that imply vertical development. The algorithm automatically suspends satellite data input to the forecast when its use is not appropriate.

These modifications result in the initiation component of the forecasts being presented as likelihoods rather than deterministic forecasts. This is much more representative of the predictability of these initiation events. A detailed description of these recent updates will be provided along with sample forecasts and statistical analyses of the results.