Read on Twitter
After some confusion with my initial tweet, here's another attempt to show the EPS/GEFS split a different (hopefully clearer) way.
If you got up this morning and saw GEFSv12 data, you'd see that by Sunday AM #97L most likely is TD/TS but there's an outside chance it's stronger.
If you got up this morning and saw GEFSv12 data, you'd see that by Sunday AM #97L most likely is TD/TS but there's an outside chance it's stronger.
If #97L never really gets going, or moves into Central America as a TD or weak TS, or moves into the Gulf as a TS, you won't be surprised. Each of these outcomes was included in your ensemble.
For the GEFS, the same can be said about a GOM hurricane! Likely? No. Possible? Yes.
For the GEFS, the same can be said about a GOM hurricane! Likely? No. Possible? Yes.
If you got up this morning and saw EPS data, you would only see the TD/TS into Central America/GOM scenario.
If that happens, great! You saw it coming.
But IF #97L becomes a hurricane, you'd be totally blindsided. Your ensemble didn't even show you that was a *possibility*!
If that happens, great! You saw it coming.
But IF #97L becomes a hurricane, you'd be totally blindsided. Your ensemble didn't even show you that was a *possibility*!
So my tweets about "what's going on w/EPS?" are arguing that the odds of #97L becoming a hurricane are probably >2% (1/50).
How much higher? GEFS has 6/31 members developing (19%). I think that's about right. Still a 4-in-5 chance of no hurricane but WAY different than 99-in-100
How much higher? GEFS has 6/31 members developing (19%). I think that's about right. Still a 4-in-5 chance of no hurricane but WAY different than 99-in-100
Probabilistic thinking is hard but one lesson I've learned reading @NateSilver538's book and posts about elections is that odds that feel similar (80% chance no hurricane is basically the same as 99% chance right!) are really quite different.
