Extended Abstract (156K)J2.5
Developing a Severe Weather Alert System
Ron Holmes, NOAA/NWSFO, State College, PA; and R. H. Grumm
A severe weather alert system for operational forecasters is being developed using historical severe weather events and the North American Regional Reanalysis (NARR) data. The events were derived from both the National Climatic Data Centers Storm Data and the Storm Prediction Center local storm reports. To match the North American Regional Reanalysis data, events were limited to the period from 1 January 1979 through 31 December 2007.
The top severe weather days were chosen based on the total number of severe weather reports. The events were then stratified to determine the largest number of tornado, hail, and thunderstorm wind events. All severe weather days were used to extract critical information from the North American Regional Reanalysis data. Standard severe weather parameters such as CAPE, helicity, and shear were extracted at representative grid points. Anomalies of select fields such as 850 hPa temperatures, 850 hPa winds, mean sea level pressure and precipitable water; related to severe weather were used. In addition, NCAPE was also estimated using NARR data to provide a better correlation between the type of severe weather event and the depth of convection.
These data were fed into a data mining application called WEKA to determine key predictors for the different types of severe weather. Preliminary results suggest that grid point data from the North American Regional Reanalysis data and thus a forecast model, such as the NMM, could be used as input to data mining or artificial intelligence algorithms to alert forecasters to the threat of severe weather.
Joint Session 2, Applications of artificial intelligence methods in the context of interactive information processing systems
Tuesday, 22 January 2008, 8:15 AM-9:45 AM, 206
Previous paper Next paper
Recorded presentation