There was a public report of a small tornado just east of I-25 in CYS’s forecast area at 20:54Z. Base velocity data from the KCYS radar was pretty unimpressed showing some weak rotation at 20:51Z. The NTDA didn’t have a detection with it but the NMDA was detecting a meso with a medium azShear Max and low azShear Average. The stronger rotation was above the 0.5° with the NMDA triggering on the 0.9 to 2.4° tilts. I’m unsure that would have provided enough confidence to issue a tornado warning. But the NMDA could’ve given some clue to the potential for a tornado. There was no SAILS data for this event, so maybe if we had additional 0.5° scans then we would have seen a stronger rotation signal in the base velocity.

– Gerry Bertier

I became bogged down with WarnGen and the GUI with this event and it took away from some of my situational awareness.  I issued a few Tornado Warnings with the KOUN WarnGen GUI but it was customized differently from the previous several cases and real time events. I liked the Tornado Warning (QLCS) option, but the re-arrangement of the product types and the attributes threw me off.  I did toggle on the TLX or Tulsa option in the WFO knob to get back to reality a bit (Fig. 1).  I also had issues with the Track and Box.  I got a speed of 605 mph.  A new record.  I did not issue a warning until I got the speed back to 35-50 mph.

I switched primarily to SVR mode with TOR tags after I issued a few TORs as the warnings were not verifying.  I lowered the NTDA probs down to 20% thinking more TORs may be needed with maybe 40-60% probs.  I forecasted incorrectly with this.  I likely should have decreased the NMDA to the bottom threshold to reduce the numbers of MESOs detected. See the bottom two figures (Fig. 2 and Fig. 3) .  I likely should have issued a few TORs here late in the event for the south side of the line.  This would have been between 0430-0500 UTC.

– Yoda777

KINX

A QLCS system developed across Tulsa CWA and moved ENE to NE. Using the three ingredients method, along with the two algorithms, allowed me to get about 10 min lead time on a QLCS tornado in northern Cherokee Co.

Tornado warning was issued at 0432Z.

Over the course of the week, I’ve found that co-located detections (NMDA and NTDA) along with steady or increasing intensity trends in both, have given me increased confidence to issue or continue downstream warnings. While the NTDA probabilities can be somewhat low, at times (esp. QLCS systems), the monitoring the trends in the intensity output has been very valuable to me. Some sort of visualization of the trends (similar to ProbSevere) would be extremely helpful.

Earlier, the combination of the NMDA, NTDA, and three ingredients method gave me confidence to continue a tornado warning downstream of a current warning.

New Tornado Warning was issued between 0400Z to 0405Z and there was a tornado LSR

There was another QLCS tornado earlier in the event in which the NTDA showed quick and substantial increase in probabilities – 39% @ 410Z, 43% @ 412Z, 67% @ 413Z, 59% @ 415Z. This didn’t really provide much lead time as I was a little late on the Tornado Warning, but the probabilities behaved as they should have.

– Beltzer

This was a great example where both the NTDA and NMDA algorithms acted as a confidence nudger for a warning decision – ultimately resulting in positive warning lead time ahead of a tornado. Interrogating a local surge in a line of severe thunderstorms southeast of TLX at 0250Z, this area was a prime focus for potential tornadogenesis given line-normal 0-3km Bulk Shear of ~40kts according to RAP13KM data (using the Three Ingredients Method). Scanning aloft quickly, a tightening mesocyclone was in progress which led to a tornado warning decision. By this time, I did notice quickly increasing trends in NTDA probabilities (83.59% at 0250Z, 86.36% at 0252Z) which by this time, I was already beginning to draw the TOR in Warngen. Seeing this display update while drawing the warning gave enough confidence to not second doubt anything, and to “pull the trigger”.

The warning came across at 0253Z, providing a 8 minute lead time to the first tornado (past survey analysis performed looking at the Data Assessment Toolkit shows a small EF-1 tornado touched down at 0301Z).

Did the NTDA/NMDA algorithms make my decision? No, and honestly I am not sure when or where it would. What this does show is how these tools provided the confidence needed to make the decision, which in a stressful situation is vital. Confidence is hard to come by when making warning decisions, as there are so many tools (sometimes too many). I would safely say the NTDA/NMDA algorithms provided extra confidence to my warning decision which ultimately led to a positive lead time from my tornado warning. Great tool!

– Dusty Davis

The TSA/OUN case was a mixed bag of supercell structures and QLCS storm types. Overall the performance of both the NMDA and NTDA did well. They showed higher values where you would expect to see stronger rotation. I thought both did particularly well with identifying areas along the QLCS where you would expect tornadic development based on the 3 ingredients method. Most of the detections, especially moderate or strong detections, were in favorable regions.

There were two very noticeable issues during the event. The first was caused by a bad radial close to the northern bookend vorticy of the QLCS. The NTDA picked up on the shear near the bad radial and identified it with a higher prob value. But it failed to identify a real couplet slightly further east. This may have been due to the distance requirement between two detections. The NTDA may have filtered out the nearby real detection favoring the artificially higher shear a bit further west. The NMDA didn’t detect either area.

Later on in the event, the NTDA and NMDA detected an area of rotation closer to the KINX radar. The NTDA had a prob of 82% at 04:17Z but then at the next scan (04:19Z) was no longer detecting the circulation, even though it looked tighter than the previous scan. The threshold was set to >= 10% for the NTDA in the image below but I believe even adjusting it to 0% showed no detection. Looking at the velocity data it definitely seemed like there should be some detection on it. The MDA was problematic too. It seemed to detect the two areas of rotation at 04:19Z but they were located well left of the couplet. This could have been because of the known issue with having 3 sails scans in place.

– Gerry Bertier