87th AMS Annual Meeting

Tuesday, 16 January 2007: 9:30 AM
A principal component analysis of tornado outbreaks.
216AB (Henry B. Gonzalez Convention Center)
Andrew E. Mercer, CIMMS/Univ. of Oklahoma, Norman, OK; and C. M. Shafer, M. B. Richman, C. A. Doswell III, and L. Leslie
Tornado outbreaks are common to locations across the central and eastern United States. Although these events are well documented, no study has attempted to distinguish between tornado outbreaks and non-tornado severe weather outbreaks. The goal of this work is to determine the amount of synoptic-scale signal present to discriminate between tornadic and non-tornadic outbreaks. The study uses synoptic scale NCEP/NCAR 2.5 latitude-longitude reanalysis data and the analysis technique uses a principal component analysis on five reanalysis variables: u and v wind components, temperature, relative humidity, and height, at all 17 of the reanalysis pressure levels. 50 case studies of tornado outbreaks and 50 studies of severe weather outbreaks were used in the creation of the composites. All cases are based on 0000 UTC severe weather maxima, with reanalysis variables extracted 24 hours, 48 hours, and 72 hours prior to the maxima of the event. The calculation of a spatial covariance for the principal component analysis has artificial bias introduced owing to the convergence of longitude lines toward the North Pole. To minimize this bias, the reanalysis data were interpolated onto an equally spaced grid. The principal component analysis of the five reanalysis variables is used to create three-dimensional composite fields for each of the five parameters. These composites will be used for the initialization of meoscale numerical model simulations. The goal of the research is to assess how well numerical models so initialized can distinguish between a tornado outbreak and a non-tornado severe weather outbreak using covariates derived from the model variables, since the model will not simulate tornadoes directly.

Supplementary URL: