This step is as mysterious as how a cloud attains its rotation. There are only two possible mechanisms. Either the tornado concentrates pre-existing rotation in its environment, or some mechanism generates the rotation locally. The following hypotheses and models are variations on these two possibilities.
In the following paragraphs I will discuss a little of each model's merits and shortcomings. You'll note that each model shares some features with one or more of the other models.
Kuo's model has as one drawback that the tornadoes ought to form coterminously with the thunderstorm, and, I suppose, might even precede full development of the thunderstorm. This seems suspicious. On the other hand, Kuo's model leads to vortexes that are very much like tornadoes in size and intensity. His model provides an energy source for the tornado, but does not provide a source of vorticity. The vorticity is supposed to pre-exist in the general area of the parent thunderstorm. The region beneath the thunderstorm is symmetric and there is no mechanism that localizes the tornado. Instead the actual location of a tornado beneath the thunderstorm depends on some symmetry breaking feature or occurrence. Perhaps some feature on the ground.
The Maxworthy hypothesis does a great job of explaining dust-devils. It provides a source of vorticity in the wind shear near the surface and an energy source in the instability of heated air. Considering how thin the boundary layer is, Maxworthy's hypothesis seems very difficult to apply to tornadoes and even more difficult to apply to thunderstorm rotation. Once again the elements for tornado formation are in place before the thunderstorm is even formed, so a person has to wonder about the timing of events.
Eagleman's model has several appealing aspects. It provides a nearby, ready source of vorticity for the tornado, it provides a mechanism to concentrate circulation as the tornado vortex is drawn downstream in the anvil and stretched, and it provides a place to dump the core flow and vorticity. It ties the tornado to the jet stream directly; although, the jet stream may be linked to the generation of a tornado only indirectly through the convergence a jet-streak produces. It will work even if the layer of atmosphere below the thunderstorm is stable and has to be sucked into the storm. Unfortunately Eagleman's estimates of wind speeds and central pressure deficit are so badly out of agreement with observation that people may have dismissed the theory for this reason alone. Moreover, does a thunderstorm actually behave like this? The dipole flow has to develop from interaction with the ambient flow. How long does it take this to develop? Do surface pressures beneath a thunderstorm support this model?
This drawing, summarized from Eagleman's report shows the mature thunderstorm with both the thunderstorm mesocyclone and its embedded tornado curving downstream in the anvil. In fact Eagleman suggests the anvil in this case is like a wingtip vortex.
Starr's view leaves out most details. Apparently the eddy inflow must exist before the tornado proper, so what starts the eddy inflow? One interesting question in regard to Starr's idea is that the eddies could withdraw angular momentum from the region surrounding the tornado vortex, and leave this region with a deficit or even with negative angular momentum. It ought to be possible, and I suggest that the data exist already somewhere, to decide whether or not this is true. If we find that a tornado is a feature with angular momentum of one sign embedded within a region having angular momentum of opposite sign, then Starr's hypothesis almost certainly has something to do with tornado genesis. There is no other means of explaining the negative viscosity that this would imply.
The secondary circulation hypothesis, which I have refered to as Bluestein's hypothesis offers the advantage of getting the timing of tornado occurence right. The secondary circulation does not exist until the thunderstorm is mature. There are few detail's about what such secondary circulations might be like. In his book "Tornado Alley" Bluestein only refers to dynamic piping, which allows a tornado to descend from within the parent cloud. Eskridge once suggested that the downdraft induced by precipitation is the triggering mechanism for tornado formation, but, then, what of LP storms where there isn't even any virga? Starr's hypothesis may even fit into this category if the momentum-supplying eddies are found to be part of a secondary circulation.
Project Suggestion: Could doppler radar demonstrate whether or not the thunderstorm anvil has vorticity of the same sign as the updraft in the parent storm? This might help verify some of Eagleman's hypothesis.
One topic back or one topic forward