The WSR-88D Operational Support Facility (OSF) Operations Training Branch (OTB) has trained National Weather Service (NWS) forecasters and interns on operational usage of the WSR-88D for more than a decade. In addition to the Distance Learning Operations Course (DLOC; OSF/OTB, 1999), the OSF/OTB trains NWS employees about warning decision making (WDM; Quoetone and Huckabee, 1995) and how to work more efficiently during severe weather shifts.

During the most recent DLOC delivery, NWS interns took part in residence courses (WDM I) that focused on making warning decisions using the Advanced Weather Information Processing System (AWIPS; Friday, 1994). Students issued hundreds of warnings during severe weather scenarios that simulated using AWIPS in a displaced real-time (DRT) mode. These courses, as well as more advanced WDM workshops (WDM II), were the first chance for NWS forecasters and interns to learn about the WDM process using AWIPS as a training tool. These workshops provided us with data about NWS training that is unique in that they allow for a standardized evaluation of forecaster performance.

This study had two goals, both of which rely on the warnings from the most recent WDM I. The first goal was to determine how successful NWS interns were at warning issuance using AWIPS and determine if any changes in training were necessary for future delivery. The second goal was to evaluate two different warning verification methods: a county-by-county method (currently used by the NWS) and a polygon method (possible for the first time using AWIPS). Although similar work is under way by the NWS Southern Region Headquarters (SRH; Smith, 2000), this study will focus on the training-related issues related to these verification methods.

During the three offerings of WDM I, the students issued over 300 severe thunderstorm and tornado warnings for three severe weather scenarios. For each scenario, the class was divided into groups (usually of three) who worked at one of nine workstations. For one of the scenarios, each group acted as an individual "office." For the other two scenarios, three groups worked in tandem as an "office". In this latter situation, warning-issuance responsibility was divided between sections of the county warning area (CWA).

Evaluating these warnings is an ongoing process. At this moment, basic analysis is complete for the current (county-by-county) verification method. Analysis for two of the three classes using the proposed (polygon) verification method is also complete. For each scenario, each "office" was scored using the standard NWS verification variables (Polger et al., 1994) for both verification methods: probability of detection (POD), false-alarm rate (FAR), critical skill index (CSI), and warning leadtime. In some cases, the warning polygon points were accidentally deleted from participants' warnings. Only the county verification statistics are available for those groups.

Detailed descriptions of the county and polygon verification methodologies can be found in Polger et al. (1994) and Smith (2000), respectively. Put simply, the county verification method utilizes the counties listed in the warning text to define the warned area; the polygon method utilizes the latitude and longitude points listed at the bottom of each warning to define the warning area. The rest of this section will focus only on the aspects of these methodologies that are different from one another or are specifically pertinent to this study.