Warning Verification
|
Warning Verification |
|
|
Currently, the Tulsa National Weather Service Forecast Office issues severe thunderstorm
and tornado warnings for 25 counties in Eastern Oklahoma and 7 in Northwest Arkansas.
As a measure of success, verification scores for each office's County Warning Area (CWA).
False Alarm Rate, (FAR), and Probability of Detection (POD) are the three scores used for verification.
Before any scores can be tabulated, we need to know a few things. First, did the storm produce
hail(dime size or greater), wind gusts in excess of 50 knots, or spawn a tornado. Property damage
resulting from a storm also completes the first step in the verification process. Verbal
acknowledgment (from a host of sources including trained spotters, law enforcement, and the
general public) of the any of these criteria counts as a verified warning, or warned event.
(Damaging tornadoes in most cases require a NWS employee to view the damage first hand.) No verbal
acknowledgment or acknowledgment not meeting criteria counts as an unverified warning. A missed
event is recorded when a warning is prompted by any of the verification criteria mentioned above.
Total warnings = warned events + unverified warnings Total events = warned + missed CSI = warned events/(total events + unverified warnings) This tells us how well our issued warnings verify. A CSI score of 1 is optimal. We strive to have 0 unverified warnings, thus warned events would equal total events. FAR = unverified warnings/total warnings We only want to issue warnings for storms that pose a threat to our warning area, i.e. severe/tornadic. We look for as low of an FAR as possible. Since we want unverified warnings to equal 0, the equation will always give the optimal score of 0. CSI vs. FAR is an inverse relationship. High CSI = low FAR, high FAR = low CSI. FAR tells us the rate at which the issued warnings remain unverified. POD = warned events/total events We also want to issue warnings based on meteorological and radar principles, not reports that would prompt a warning. POD measures that. Since we don't want any missed events, warned events will always equal total events. This yields the optimal score of 1. **These statistics, ranging from July of 1994 through 2001 have been calculated using a slightly different method. National Weather Service Headquarters uses a method which allows tornado warnings to be verified by wind and hail reports(severe thunderstorm criteria is met, but no tornado was verified). The method used here eliminates that factor and keeps the two types of warnings separate.** | |
County-by-county verfication data can be obtained through a clickable mapof our county warning area. | |
| TABLES | |
| Severe Thunderstorms by County and Year | Tornadoes by County and Year |
| 2001 | 2001 |
| 2000 | 2000 |
| 1999 | 1999 |
| 1998 | 1998 |
| 1997 | 1997 |
| 1996 | 1996 |
| 1995 | 1995 |
| 1994 | 1994 |
| Severe Thunderstorms and Tornadoes by Month | |
| 2001 | |
| 2000 | |
| 1999 | |
| 1998 | |
| 1997 | |
| 1996 | |
| 1995 | |
| 1994 | |
| GRAPHS | |
| Severe Thunderstorms | Tornadoes |
| 2001 | 2001 |
| 2000 | 2000 |
| 1999 | 1999 |
| 1998 | 1998 |
| 1997 | 1997 |
| 1996 | 1996 |
| 1995 | 1995 |
| 1994 | 1994 |
Last Updated: May 13, 2002
