The image below looks at 2 cells in central Oklahoma. The northern cell has a prob tor of 77%, but is falsely lightning up a shear zone with little chance of a tornado. The southern cell shows all the characteristics of a tornadic supercell, but has a prob tor of 37%. However, the southern cell is closer to producing a tornado (via live media).
Here’s another image using KVNX instead of KTLX, and shows only the prob tor.
The two cells to the north have 90 and 91 percent, respectively, while the southern supercell has only 33 percent.
Not to pick on the prob severe product, the meso detection algorithms a similarly struggling. Below is a 4-panel showing the legacy mesocyclone alogrithm (upper left), the digital mesocylone algorithm (upper right), the low level AzShear (lower left), and the experimental meso algorithm (lower right).
All three detection algorithms are flagging the circulations as equally important, but an examination of the base data shows otherwise.
Thorcaster
Just after 20z on the eastern fringe of the LUB CWA, the KFDR radar indicated an area of very high inbound velocity. However, this data is in question as the elevated velocity occurred in an area of low Z and high SW, and likely not representative of the actual storm. This may have been caused by a side lobe. This had cascading affects with algorithms being tested which could not filter out the bad data. Low level az shear spiked to over 0.01 in a group of stationary pixels. This caused algorithms that ingest the az shear product to spike including ProbTor which increased to over 90%, as well as CPTI which showed lower end probabilities of a violent tornado in progress.
