For swirling flows (e.g. cyclones), 2-equation turbulence models don't work very well, those flows are better predicted using Reynolds stress models, but I'm not sure your case is indeed a candidate for those (?)