Horseshoes and Hurricanes

Several folk - half the world to be precise- have been going too and fro about the role that climate change has played in the latest weather catastrophes.  Eli had a say yesterday, and some of the Friends of the Bunny chimed in.  Perhaps Izen had the best tweet comment about cause and effect

To stretch the metaphor, many things can light a fire, but the size depends on how big the pile of fuel it can burn.

and Bernard J pointed out that the nature of the beast of chance is that when the statistics become strong enough so a single event can be unequivocally attributed to the changes we are making in the atmosphere, oceans and soil, then it is already too late and a considerable amount of pain has already been absorbed with more and worse to follow.  It may not be a good idea to be living on such a planet.

This, of course, raises the question of the skill in hurricane track forecasting  A difference of 70 km or so yesterday saved Miami and by an even smaller margin Tampa.  Others may debate to the value and duration of such a reprieve.  There are lots of folk who know more about this than Eli, and one of them pointed him to those who really care, the US National Hurricane Center page on Forecast Verification.  

Turns out there are multiple models from many places and a scorecard is for sure needed. NOAA provides one for the models it uses  Forecasts are done on a three hour cycle with the forecast due three hours after the start of the cycle.  Models are divided into early or late, depending on when in the cycle (or afterwards) they are available.  Early models which start to run at the beginning of the cycle and are available before the end are called early.  Those which take longer are called late.  The output of the late model is adjusted so that it feeds into the start of the next occurring cycle becoming a psuedo very early forecast for that cycle. 

Models can be statistical, depending on historical data, these tend to be early, or dynamical, at some (varied) level doing physics based climate modeling.  The most complex of these are pretty obviously late models.  Models can be dropped from the ensemble when they don't perform well or something better comes along.

That being said, how are the models doing.  Data goes back to the 1970s.

Things are improving, perhaps also because tracking is improving.  At early times the ensemble was heavily weighted to statistical models, dynamical models rose to the fore in the 1990s

There are multiple models and combining them into a single forecast has a bit of magic about it where the forecasters weigh the combination of various models.  This can be seen in the 48 hour forecasts from the "early" models and the official NHC (dark line) forecast,  The dotted line is a combination of two statistical forecasts (see links for details)

Anyhow the Florida peninsula is about 150 miles at the widest down to about 100 just north of Orlando.  You can get lucky at horseshoes.