[UPDATE, 9:05 PM: The GFDL has shifted back south again, and now shows a 165 mph landfall between Corpus Christi and Houston, rather than a 155 mph hit on central Louisiana. The GFS has also shifted south, and is now predicting landfall in Belize (not the Yucatan) with no significant re-emergence over water until reaching the Pacific — in other words, no threat to the United States. The NOGAPS and UKMET remain essentially unchanged, both taking the storm into Mexico and keeping it away from the U.S. You can view all of the “big four” models here, and even more model tracks here.]

The debate over Hurricane Dean’s future track is boiling down to a simple question: Does the GFDL computer model know something that all the others don’t? Over the last several model runs, the GFDL has been trending northward, while the other models have been trending southward. Take a look at Weather Underground’s model map as of 2:00 PM… the blue line is the GFDL, while the white, purple and red lines represent the other three of the “big four” dynamical models (UKMET, GFS and NOGAPS):

As you can see, the GFDL is stubbornly taking Dean into Louisiana, while the other models are leaning increasingly toward a Mexico landfall. So, what are we to make of this stark disagreement? Dr. Jeff Masters has an excellent post asking the question, “Which model do you trust?” After posting and analyzing a couple of charts of model accuracy, he concludes that “the official NHC forecast outperforms all the individual models, particularly at long ranges. Looking at the individual model plots can be helpful to determine the uncertainty in the forecast, but it’s tough to beat the NHC. In the case of Dean, where one model is an outlier from the rest, it is usually better to believe the consensus of the other models.”

Usually. But what if the GFDL is seeing something the other models are missing? Recall Alan Sullivan’s post from this morning, in which he wrote, “In the real world, as opposed to the world of computer models, I see something happening” — namely, a “mighty sluggish” upper low off Florida defying forecasts that it will quickly get out of the way — that could nudge Dean’s track to the right. Perhaps the GFDL is seeing the same thing Sullivan saw. (It should be noted here that Sullivan, like me, isn’t a meteorologist; he’s a layperson with an intense interest in, and a good deal of lay knowledge about, hurricanes. He’s also a self-described “contrarian.” None of which necessarily means he’s wrong.)

Not everyone thinks the GFDL is on to something. Earlier this afternoon, Eric Berger wrote that the GFDL “continues to jump all over the map, this time from a Galveston landfall [which it was predicting at 8:00 AM] to central Louisiana. This indicates one of two things: [either] 1) the model has not properly initialized Dean and is rendering its path incorrectly, or 2) the model is detecting a trend in the upper atmosphere the other models have missed. Given its inconsistency I’d lead toward the former answer, but there’s no way to know for sure.”

But one man’s “inconsistency” is another man’s “trend.” For the last 24 hours, the GFDL has been trending more to the right, or north, with each model run. At 8:00 PM yesterday, the GFDL showed Dean’s track pointing toward Corpus Christi; at 2:00 AM this morning, an area just south of Houston/Galveston appeared to be the target; at 8:00 AM, Houston/Galveston was in the bull’s eye; and then at 2:00 PM today, Central Louisiana was favored. If that’s a genuine trend — and with only four data points, I admit it’s a somewhat shaky contention — it would seem consistent with Sullivan’s observation that the upper low off Florida is continually refusing to move as predicted. It will be interesting to see what the model comes up with at 8:00 PM.

Berger, though, is looking at a different “trend” — the trend toward a more southward track in the other model runs:

The hurricane center’s afternoon update appears to have nudged the five-day forecast only very slightly southward. As noted below, the models have made a more significant move south. What is happening here is that the hurricane center doesn’t want to have its forecasts, in the wise words of Jeff Masters, have a “windshield-wiper-like” effect.

Therefore if this evening’s model runs show a similiar southward movement, as this morning’s did, I would expect a southward shift in the 10 p.m. forecast track.

As for the GFDL, he writes, “The key feature here is an upper-level low pressure system…that’s moving to the west ahead of Dean. The GFDL model (the outlier at this point) handles the movement of this system differently than the other forecast models. I’ll have more on this tomorrow morning if the models have not agreed upon how to handle the upper-level low, which could steer Dean into Texas or even Louisiana.”

The bottom line continues to be: Stay tuned. There’s still a lot more ink (er, and pixels) to be spilled on this storm before we figure out precisely where it’s going.

UPDATE: In comments, Sullivan writes:

I see the bifurcation in models as part of the same trend. Dean could also turn left if the upper low refuses to move. In effect the hurricane would be trying to sneak underneath the opposing circulation.

In other words, I think all the models may be seeking ways to resolve a single fact: that the hurricane is overtaking the upper low.

I had a similar thought rolling around in my head, but wasn’t sure quite how to express it, or whether it made sense. But yeah. Thanks, Alan.

UPDATE 2: Now Alan has a new post on his blog fleshing out that same thought. Excerpt:

Some [of the computer models] stay with the Jamaica, Yucatan, upper Mexico run — almost a straight line. Others are shifting left toward Nicaragua or even further south. And some are shifting right, into the Gulf.

I have been thinking about the righthand shift all day. The leftward shift in some models surprised me — but only for a moment. Then I had an insight. What if all these diverging models are responding to the same phenomenon — namely the sluggish movement of the upper low near Florida, which the hurricane is steadily overtaking? For Dean could also avoid the encounter by turning left, in effect bouncing off the periphery of the other system’s circulation.

Two effects, one cause…but which way will Dean turn? Or will it simply blow through the opposing system and stay on the straight course?

P.S. Bryan Woods makes an important point about the intensity forecast for Dean, post-Jamaica: “If, as currently indicated, Dean tracks directly over Jamaica, we will have to reevaluate everything. Jamaica has quite large mountains in the center of the island and Hurricanes do very poorly when passing over high elevations. Jamaica could do quite a number on our storm and knock his strength down significantly with the right conditions. Not only will the elevation condense out a lot of water from Dean’s circulation, unstable flow and wave breaking over mountains can cause a redistribution of potential vorticity, which can create several types of lateral instability in the flow.”