Why you shouldn’t get excited about long range forecasts …

I ran across a weather blog yesterday (sadly, by someone with an AMS seal) playing up a possible hurricane hitting New Orleans next week, based on the GFS model forecast.  Here’s the forecast map in question, the 225 hour forecast from the 18z GFS/FV3 run on Tuesday, August 13th (the Tuesday Afternoon run).  Looks bad for the sweltering, drunken masses on Bourbon Street …

But, as a reader of this blog, I know that while the new models are better, global weather models still have a tendency to spin up vortexes, so I waited to see what the next run on Tuesday night showed before freaking out …

Oh, now it’s just a tropical depression or storm, and not a problem for the weekly NOLA Friday Night Bacchanalia.   But, what does that mean for my oil and cattle futures?  Lets look at the early morning (06z, Wed. 14 August) for the same time (03z Friday, August 23rd, now “only” 213 hours away) …

Hey, where’s my storm?  Now it’s just a blob of rain on the Alabama/Louisiana border?

Forecast tracks from the global models have improved a lot in recent years, once a storm spins up, and are better at forecasting formation in the short term (0-5 days).  But this example is why these long range models are not that useful for long range hurricane forecasts, and why the National Hurricane Center only does outlooks out five days.  These models just aren’t designed for it, and this far away, the uncertainty is so great you can get exhausted tracking every little vortex that, over the life of the model, will spin up, but don’t even exist in the “real” world (not to mention the systems that spin up in the real world).  So anybody talking about storms forming a week away is probably just phishing for clickbait.  Now, talking about those who talk about them, that’s a public service and perfectly ok …

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.