The butterfly-hurricane thing is based on a fundamental property of "chaotic" systems. Their end state is determined by their starting state. But a tiny variation in their starting state can cause huge variation in the end result. If you could start the weather systems in exactly the same state this evening as they were yesterday evening, then the weather would be identical, even down to that irritating shower at 3.37 that got your washing wet. But you can never start from exactly the same place, and the tiniest discrepancy makes a huge difference. This tiny displacement of a few cubic microns of air here causes a breeze to blow subtly differently there, which means that this delicately balanced air mass over there does, or does not, form a tiny cloud which is the nucleus of a minor storm there which deflects the jet stream just a few meters way, way over there ... and so on until the hurricane hits Bermuda.
But, consider this. If the tiniest variation in the starting conditions causes such huge differences in the result, then huge differences in the result must have been caused by tiny differences in starting conditions. (Yes, I know it seems a trifle obvious, but I prefer obvious arguments.) Cosmologists have used the chaos argument this way round to great effect. Apparently, tiny differences in the starting conditions of the universe have substantial effects on the universe now - like, whether it is still there. They can deduce the curvature of space-time in the first millisecond of the universe to one part in 1018 , simply because the Universe is not 10 feet across or filled with strawberry Jello.
I suggest we do the same thing with the weather. Running the weather prediction software forwards in time produces ever more inaccurate predictions, as the uncertainties in our measurement of what the world is today have ever greater and greater effects on our predictions for tomorrow. But running it backwards would be much easier. We can measure temperature, pressure, rainfall etc. today with moderate accuracy. Running the software backwards will allow us to sort through the number of possible yesterdays that gave rise to today’s weather. Almost all of them will give rise to Hurricanes in Bermondsey, never mind Bermuda, so we can rule them out. The few that are left, only differing from the others in the last decimal point, must be what the weather was like yesterday.
Using this extraordinarily accurate model (checked against what the weather actually was like yesterday) we can predict the day before, and so on. As we predict backwards chaos theory proves that our model will get ever more accurate. We can predict to 100s of decimal places what the wind speed at the top of the Eiffel tower at 11.37 on November 17th 1972 was, because if it was even the slightest fraction different then the next ice age would probably have arrived my now, or falls of frogs would have covered Bermuda instead of hurricanes.
But this is no idle exercise. We could, for example, predict the air flow and pressure patterns over downtown Dallas on November 22nd 1963. We could say precisely how the natural patterns were disturbed by the passage on a certain number of high velocity bullets, how many there were and where they were fired from. We could say how many people were standing on a certain grassy knoll, and if our model had a unit grid size small enough, take a good bash at mapping their faces. If their nose had been less than the size it was, after all, then, pace Professor Henry Higgins, hurricanes would be happening in Hazlemere and Hartlepool every alternate Tuesday.
For a large fee and a guarantee of protection, I am willing to refrain from expanding this theory any further.