Consider this experiment:
A pot of highly-flammable liquid burns with a small flame if placed on an extensive flat surface (observation A). Place to half-tubes around it which let in air in a (predominantly) circular fashion, let's say clockwise (seen from the flame). Now the flame is much higher (observation B) and it rotates like a tornado (observation C).
I would explain this in a turbulent vs laminar flow context. If a burning liquid is placed on an extensive flat surface, the flame is impeded by the turbulences it creates above itself. The air cannot rise fast enough to produce a decently long flame (explanation of A). Now, if you impose a directional air flow and this flow is laminar, then the flow velocity should be much larger and the flame is higher (observation C). The 'tornado' effect is just a side-effect of this type of rotating laminar flow (obs B).
Do you agree? What are your ideas?