Monday, 5 November 2012: 5:15 PM
Symphony I (Loews Vanderbilt Hotel)
Chad M. Shafer, Univ. of South Alabama, Mobile, AL; and M. W. Stanford, M. Richman, L. Leslie, C. A. Doswell III, and A. E. Mercer
Several severe weather diagnostic variables (SWDVs) used to diagnose severe weather environments are unphysical (EHI, SCP, STP, etc.). Therefore, it is unclear how a particular magnitude of such a SWDV should be interpreted and compared to other magnitudes. This study evaluates SWDVs by computing the frequencies with which practically perfect (PP) probabilities of severe weather of any type or of specific types are exceeded for given magnitudes of the SWDVs in a 40x40 km horizontal Lambert conformal grid encompassing the conterminous United States (CONUS). North American Regional Reanalysis (NARR) data from 20012010 are collected for analysis. PP probabilities are computed for varying time periods (24-h, 6-h, 3-h, and 1-h samples) to determine sensitivities to temporal resolution. In addition, the SWDVs are evaluated using threshold conditional criteria (e.g., observed convective precipitation at the analyzed grid point, the presence of at least one severe report of the analyzed type within the CONUS domain during the time period of interest, etc.) and unconditionally for the 10-y period.
The SWDVs are evaluated by identifying the grid points for which a predetermined magnitude is equaled or exceeded. For each of these grid points, if a selected PP probability is equaled or exceeded, a hit is counted. Otherwise, the grid point is counted as a false alarm. For each grid point in which the SWDV magnitude is below the selected threshold, if the selected PP probability is equaled or exceeded, a miss is counted. Otherwise, the grid point is counted as a correct null. Contingency statistics are computed for various combinations of SWDV magnitudes and PP probabilities. This technique provides insight into the PP probability for which a given SWDV magnitude has maximum accuracy or skill.
Preliminary findings suggest (1) several SWDVs exhibit considerable skill in identifying locations with >0.01 PP probabilities of severe weather of any type or of a specific type, (2) in general, skill decreases with increasing PP probability and increasing SWDV magnitude, (3) skill decreases with increasing temporal resolution, (4) skill is lower for unconditional PP probabilities, particularly for single severe weather report types, owing to an increased number of false alarms, and (5) most SWDVs are highly correlated, resulting in statistically similar accuracy and skill in identifying PP probability exceedance regions.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner