This seems to be an issue with undershoot. The NTDA uses only 0.5 degree elevation and at this range this corresponds to about 1600 ft AGL. That is likely well below cloud base. The upper right panel is mid-level rotation tracks and the panel immediately to the left of that is low-level rotation tracks. Note how rotation is more evident in the “mid” levels, which again underscores the disadvantage the NTDA has with just looking at 0.5. Side note: The NTDA and ProbTor (lower right) probabilities

Below is 2018 UTC:

The 2018 UTC screen shot shows a persistent and strong mid-level rotation track with good consistency on the ProbTor object. The NTDA has been jumping around a bit, again because it’s at a disadvantage because it’s still sampling below 3000 ft even at this range.

Below is another example of NTDA detections jumping around. In this case, perhaps it has more to do with the confusing velocity associated with merging cells.

-marfafront