Model guidance - lessons so far from Dennis

[x-y-no]

As most here will recall, we had a decidedly mixed performance from the models last year.


The GFS, still the model I look at first, had big problems with the steering of systems moving westward under a ridge, tending to break down that ridge prematurely allowing them to turn north. We saw this with Ivan and Frances in particular.

With Dennis, the GFS performed very well at first, until he began to intensify. Then the old problem turned up again. The conclusion, unfortunately, is that while this winter's tweaks to the model may have helped with relatively weaker systems, we cannot trust how it handles very strong systems moving south of a ridge. Still to be seen is whether it does better once the storm begins moving around the end of the ridge - we'll get a good test of that as Dennis reintensifies and moves through the Gulf. As this was something the GFS handled fairly well last year, I expect it still will, and I think the Pensacola-area landfall it's predicting looks reasonable.


The Canadian had a shining moment with Ivan last year, but otherwise performed pretty badly. With Dennis, it blew it yet again, insisting on a non-credible Florida peninsula landfall for days, and then eventually flipping too far west. Worth looking at for synoptic ideas, but useless for track guidance.


The UKMet seems to have the opposite problem from the GFS, overestimating the effect of the ridge resulting in a fairly substantial leftward error. Interestingly, this error didn't seem as much influenced by the strength of the storm as the GFS error is. This model does a great job on the global synoptic situation, though, and it has the virtue of not flip-flopping around so much. This is the one I always go to for a sanity-check on the GFS, and it definitely still rates that role given its perforance so far with Dennis.


Maybe someone else can comment on the European - I tend to give it a glance, but they release so little of their data publicly that I find I can't make much practical use of it.


Bottom line, none of the globals can be relied on as a guide in isolation. They all offer useful input, though, which one hopefully can combine with looking at real-time satellite and surface data to gain an understanding of what will happen. In that realm, I clearly need to sharpen my skills at understanding what's happening with the steering layers, since my reading of the satellite wind analysis was apparently not too accurate.

Jan

[Agua]

Good post. Thanks Jan. I suppose I remember the failings of the GFS from last year, vis a' vis ridge breakdown, simply because it seemed the NHC appeared to give a great deal of weight to that model.