[OFFICIAL] Hurricane Season 2007

The GFDL is one of the most reliable models, but it varies from season to season, even storm to storm. It, the GFS and the NOGAPS tend to be the most reliable, but they all vary a bit.

Right now, the trouble they all appear to be having is what to do with the upper level trough of low pressure sitting over the Bahamas and moving westward. Low pressure tends to pull other low pressure systems along with them. The models are trying to figure out how fast that upper level low is going to move (very important) and where it will wrap around the high pressure system to its north (even more important).

The problem is that it is hard to predict how it will behave because its movement is dependent upon the movement and behavior of other upper level conditions.

I looked over the initial runs for the 00z that just came out and they are concerning. The NOGAPS is about the same, maybe slightly north, but nothing surprising. However, the GFS has made a fairly sizable move north and is now passing through the northernmost part of the Yucatan and making landfall around the Texas/Mexico border. This is not a DRASTIC change given that it was at this point yesterday, but it doesn't continue the southward trend we've been seeing the past few runs. It looks to be a similar track to the last run of the UKMET.

Of course, this is the problem you have following late model forecast tracks on every pass of every model. If you watch the 72 hour forecasts, they've been remarkably stable, but that obviously isn't our concern here in Texas - not until Sunday anyway.

I'm waiting on the latest NWRF to come up on the NOAA site, but it is a little behind. Unfortunately, I can't find a good resource for the GFDL anywhere on the web that posts sooner than Wunderground or the South Florida site.

 

Well, there are a couple of ways.

http://www.meteo.psu.edu/~gadomski/ewall.html - this has all the relevant models, but it's a little tricky to navigate

http://www.nco.ncep.noaa.gov/pmb/nwprod/analysis/carib/gfs/ - GFS (you can see the time stamps)

http://www.nco.ncep.noaa.gov/pmb/nwprod/analysis/hwrf_nested/storm_1/ (this is for Dean and hasn't run yet - you can judge by the time stamps)

It is hard to see on the first link because the maps only show the continental US, but it appears the HWRF (listed as WRF on the first link above) is slightly south of the previous run...not by a lot, but still.

The GFS looks like it has nearly an identical track to the NHC forecast track.

When the third link above is updated (hopefully, soon), you'll be able to see a clearer picture of the HWRF, which was the second northernmost outlier to the GFDL in previous runs.

Unfortunately, I just found a UKMET run that is also further northward landing around Corpus Christi.

Granted, none of these appear to have a landfall above the south Texas coast, but still...

 

I should point out at least SOME good news. It would appear (with the GFDL being the one exception since I haven't seen the 00z run for it yet) that all the models are in good agreement of at least SOME landfall on the Yucatan. None of the runs I've seen has Dean missing land.

While not good for the Yucatan, any interaction with land means Dean will not come into the Gulf as a massive category 5 storm.

 

Ok, this shows the UKMET running at 18 00z, which is the latest run, and it is pretty much the same as the last run.



I tend to trust this more than my ability to read animated maps, particularly since this seems to have a lot of the 00z model runs on it. Still missing the HWRF, GFDL and GFS (represented on this map as the AVNO) 00z runs. This thing updates quicker than all the others.


It should be noted that because I'm grabbing this image from the page that hosts it, it may have updated (perhaps several times) by the time you read it.

 

Actually, I was just reading about their historical accuracy today. From the Chron SciGuy blog:

Here, according to National Hurricane Center data, are the average five-day errors for the models discussed above in 2006:

1. GFDL (234 miles)
2. GFS (250 miles)
3. NOGAPS (250 miles)
4. UKMET (294 miles)

And for the extremely active 2005 season, which had a lot more storms like Dean, the average five-day errors were:

1. UKMET (253 miles)
2. NOGAPS (275 miles)
3. GFDL (296 miles)
4. GFS (314 miles)

So this doesn't suggest the GFDL has been any better than the other models recently.

Here's the NHC's historical averages page:

http://www.nhc.noaa.gov/verification/verify3.shtml

Dr. Masters has some excellent info as well here:

http://www.wunderground.com/blog/JeffMasters/comment.html?entrynum=753&tstamp=200708

Ultimately, the NHC consensus forecast is, BY FAR, the most accurate predictor of forecasting, particularly in the later periods.

 

Ok, just got the latest model runs for both the GFDL and the HWRF. The GFDL has shifted south VERY slightly - about 50 miles - although in the near term it is WELL north of all of the other models pushing it north of Jamaica and even nearing Cuba.

The HWRF run shows it just grazing the northernmost edge of Jamaica (well within the margin of error of the other storms) and then moving across the Yucatan and into northern Mexico just south of the NHC track. This is very similar to its last run, just a little quicker.

The South Florida plot map doesn't have a 00z run for the GFS, NOGAPS or the HWRF, but I've seen all those already and posted the info previously.

Judging by this, I doubt the NHC will alter its track for the 5am update.

 

If you go by actual numbers, yes. But, the margin of error is so incredibly small - less than 20 miles for the last two seasons - that, statistically, there is no difference.

If you want to say the last two years that the GFDL was more accurate by an average of 17 miles, then, yes, it was more accurate.

 

I should note that I am not dismissing the GFDL at all, however, even Dr. Masters has suggested that it is more LIKELY that the dynamic model consensus is closer to accurate than one outlier, no matter which model it may be.

It's certainly possible that the GFDL is correct and the storm will make landfall along the central or upper Texas coast, but it is equally as likely that it will make landfall in Mexico at this point.

 

That's possible. There have only been a handful of cat 4 and 5 storms to come on shore in Texas. I think Beulah was the only cat 5 storm to ever hit Texas.