Now looking at ILN radar, Here we see several detections that are associated with a complex velocity pattern. The alternating inbound/outbound velocity pattern suggests gravity waves are propagating northeastward. This is in the evening, so perhaps a nocturnal inversion is helping to duct these waves. Not surprisingly the NTDA appears to have some false detections with the 2D shear.

-marfafront

We see a case where the TDA is inconsistent in time, likely due to noisy velocity, KIND 28May2019 at 0116Z Wayne County, likely due to vertical velocity data smearing.

Here (below) we see the detection return and it’s impressive that it is assigned the same ID as before. This will help for creating time series of the features attributes. That’s good news. What’s less good news is we see a great setup for vertical data smearing as broad rotation with a 60+ dBZ core aloft is being aliased downward. There is no easy solution for this I’m afraid. Spectrum width is very high (>20 kt) and CC is ~ 0.8 or less so there is a possible flagl for degraded data quality. However, these values also could legitimately occur with a strong TDS.

-marfafront

NTDA performs well in mature supercell over ern IN. First time of detection to current time of detection over 1 hour indicative of long track of cell. Probs were over 90% for quite a few scans. Overlaid with prob severe, it is nice to see meso environment that the cell is embedded in.

Would be good to see some environmental data embedded in the algorithm to keep the screen less cluttered. Also, a total time of detection will be a good addition to the readout table.

– Bacon Wrapped Hail

1904Z

A tornadic QLCS is moving through South Carolina. KGSP is running in SAILS2, producing three cuts at 0.2° for every volume scan. The NTDA algorithm is running off of the 0.5° slice, which is only running once per volume scan.

When NTDA is loaded on top of the 0.5° slice, it matches perfectly. However. When NTDA is loaded on top of the 0.2° slice, there are problems that cause the NTDA to drop out (NO DETECTIONS) at certain time steps, likely due to some kind of error with time matching.

1936Z

A bowing segment in a QLCS is moving through western Newberry County, SC. This storm has a history of producing tornadoes, but it has not obviously done so in at least a half hour now. The NTDA algorithm is displaying relatively high values (>50%) from two different radars — KGSP (about 4000 feet AGL) and KCAE (about 2500 feet AGL). This is absolutely the type of area of the storm where you would look for tornadogenesis (north of the apex of the bow, following the three ingredients method). However, it is interesting to note that these probabilities are quite high, despite the fact that neither radar is truly depicting a tight gate-to-gate circulation. In a real time event I would likely be issuing a tornado warning for this, as tornadogenesis seems quite favored in this region of the storm, but the probabilities on the NTDA are a little higher than I would expect given that I think it’s more likely than not that the storm is not currently producing a tornado.

1944Z

This is an update on the area of interest mentioned in the above post. It was noted that the NTDA feature on the KCAE radar (ID: 3) has shifted somewhat southward, to a different part of the bowing segment (closer to the apex) while updating the probability to >80%. The overall shear axis remains somewhat not-tight, which at first confused me as to the high probability. However, upon a closer look, there was a very small/tight couplet that developed just on the west edge of the shear axis, tight enough that I would suggest it could indeed be tornadic. It appears that the NTDA feature was keying off of this, and not on the overall shear axis, and I think this is excellent because it brought my attention to a small, hard-to-see feature.

2000Z

A semi-discrete cell is approaching the KCAE radar site at close range. This cell has produced an intense downburst signature, characterized by inbound winds as high as 65kts at less than 1000 feet AGL. Qualitatively, I can state fairly conclusively that this signature is a severe downburst, and is thus non-tornadic. However, the NTDA algorithm is keying in on the north end of this downburst signature as a potential tornado, with probabilities as high as 49%. This seems quite high for a weather situation that appears to have almost no chance to actually be tornadic. It is possible that the algorithm is noting that you have an area where strong inbound winds are located just south of weaker inbound winds, but to my eyes it is a soft transition, not indicative of rotation.

–Insolation

We noticed several instances this morning in the PIH forecast area where the NTDA algorithm would detect possible TOR’s over mountain and or hill peaks.  This is common with the TVS algorithm and I wonder if there was some sort of rate of change filter that could be passed along to the RF model. It’s something that as a forecaster if you know your CWA well is easy to ignore but if there’s a way to filter those out it would be helpful.  In the two examples shown below, the ridgetops in eastern/central Bingham county are easily shown to be the source of the NTDA at least in the 0.5 degree slice.

– Spaceman Spiff