The National Weather Service issues public warnings for severe thunderstorms and tornadoes when these storms appear imminent. A study of the warning process was conducted at the National Weather Service Forecast Office at Raleigh, North Carolina, from 1994 through 1996. The purpose of the study was to examine the decision process by documenting the types of information leading to decisions to warn or not to warn and by describing the sequence and timing of events in the development of warnings. It was found that the evolution of warnings followed a logical sequence beginning with storm monitoring and proceeding with increasingly focused activity. For simplicity, information input to the process was categorized as one of three types: ground truth, radar reflectivity, or radar velocity.
Reflectivity, velocity, and ground truth were all equally likely to initiate the investigation process. This investigation took an average of 7 min, after which either a decision was made not to warn or new information triggered the warning. Decisions not to issue warnings were based more on ground truth and reflectivity than radar velocity products. Warnings with investigations of more than 2 min were more likely to be triggered by radar reflectivity, than by velocity or ground truth. Warnings with a shorter investigation time, defined here as “immediate trigger warnings,” were less frequently based on velocity products and more on ground truth information. Once the decision was made to warn, it took an average of 2.1 min to prepare the warning text. In 85% of cases when warnings were issued, at least one contact was made to emergency management officials or storm spotters in the warned county. Reports of severe weather were usually received soon after the warning was transmitted—almost half of these within 30 min after issue. A total of 68% were received during the severe weather episode, but some of these storm reports later proved false according to Storm Data.
Even though the WSR-88D is a sophisticated tool, ground truth information was found to be a vital part of the warning process. However, the data did not indicate that population density was statistically correlated either with the number of warnings issued or the verification rate.
* Department of Marine, Earth and Atmospheric Sciences, North Carolina State University, Raleigh, North Carolina.
+Department of Statistics, North Carolina State University, Raleigh, North Carolina.
#National Weather Service Forecast Office, Raleigh, North Carolina.