I'm really confused

[Rieyeuxs]

I've been following 97 with everybody else here but I have two main questions I hope somebody here could answer.

1) Why aren't the models picking this wave up very much? It's been the best shot we've seen this season, yet we've seen the models (GFS, which I know has it's problems) go gung ho over other systems that haven't done anything. Remember our June Alex wannabe? I guess the question is more what "triggers" the models and why 97 hasn't done so.

2) Since I'm not really seeing the models pick up on 97's track, I've seen posts that take it into Mexico, into GA/SC (which is why I'm somewhat interested. THAT forescast perked me up, since it put this on my doorstep) and just about everywhere in between. Are the models, such as they are, that unsure where this is going? I haven't even seen a consensus on where this is going 2-3 days out. I've got family flying in Sunday morning, and I'm mildly concerned on what the possibilities are.

[Anonymous]

The disturbance is still relatively weak and the globals have been trying to force some of the moisture northward, which does not seem all that likely at this point. This disturbance in the Caribbean no longer poses a threat to SC.

[Derecho]

I've been following 97 with everybody else here but I have two main questions I hope somebody here could answer.

1) Why aren't the models picking this wave up very much? It's been the best shot we've seen this season, yet we've seen the models (GFS, which I know has it's problems) go gung ho over other systems that haven't done anything. Remember our June Alex wannabe? I guess the question is more what "triggers" the models and why 97 hasn't done so.

Model initialization is a subject of IMMENSE complexity. Computer models are among the most complex computer programs ever written, I think only the Department of Energy nuclear explosion models are probably more complex.

The main problem with models aren't the models themselves but the data that goes into them. The wave has never been initialized that well....but the main reason for it is that there really has never been that much "there."...very little at the surface.

2) Since I'm not really seeing the models pick up on 97's track, I've seen posts that take it into Mexico, into GA/SC (which is why I'm somewhat interested. THAT forescast perked me up, since it put this on my doorstep) and just about everywhere in between. Are the models, such as they are, that unsure where this is going? I haven't even seen a consensus on where this is going 2-3 days out. I've got family flying in Sunday morning, and I'm mildly concerned on what the possibilities are.

The East Coast stuff has largely been from people not paying attention or that have very little idea of what they're talking about. There's a low/vort area that forms off the East coast in a few days, but it doesn't have much of anything to do with this wave.

[Anonymous]

The East Coast stuff has largely been from people not paying attention or that have very little idea of what they're talking about. There's a low/vort area that forms off the East coast in a few days, but it doesn't have much of anything to do with this wave.

The GFS did have the wave progged north of the Greater Antilles in the medium range, but that was when it was still in the central Atlantic. This is in addition to the vort you mentioned.

[hurricanetrack]

This is exactly what I have been asking myself- and colleagues for some time now. Why aren't the global models "doing" anything with 97L? I guess the reason is, it's not going to develop much, if at all. The globals are there for a reason and they are not making much out of this system.

Of interest though for the future- is this from the 8:05 TWD:

TROPICAL ATLC...
THE MID/UPPER LEVEL RIDGE REMAINS SPLIT OVER THE CNTRL ATLC WITH
TWO UPPER LEVEL ANTICYCLONES CENTERED NEAR 20N36W AND 18N56W.
DEEP ELY FLOW EXTENDS S OF THE RIDGE TO 10N...AND THEN SLIGHTLY
MORE NELY S OF 10N BETWEEN 20W-35W. TROPICAL WAVE COMBINED WITH
SOME UPPER-LEVEL DIVERGENCE IS ENHANCING TSTMS ALONG THE ITCZ
BETWEEN 35W-45W. RELATIVELY DRY AIR IN MID/UPPER LEVELS IS
OVER THE ATLC E OF 30W. WIND SHEAR IS LOWER THAN AVERAGE IN
MUCH OF THE DEEP TROPICAL ATLC WITH PRESSURES BELOW AVERAGE E OF
45W AND NEAR AVERAGE W OF 45W. WLY MONSOON FLOW IS QUITE
STRONG/DEEP INTO W AFRICA AS NOTED BY THE DAKAR SOUNDING SHOWING
WLY WINDS UP TO 850 MB.


That last part sounds really encouraging if you like long track hurricanes- well, the potential for long track hurricanes anyway.

It seems that all of the parameters are in place now for a whopper of a peak to the season. Give it a month for that dust to settle down- however it does so- and we'll be tracking them left and right. I just don't think so right now- that certain spark is just not there yet.

[Wthrman13]

1) Why aren't the models picking this wave up very much? It's been the best shot we've seen this season, yet we've seen the models (GFS, which I know has it's problems) go gung ho over other systems that haven't done anything. Remember our June Alex wannabe? I guess the question is more what "triggers" the models and why 97 hasn't done so.

Excellent question that drives right at the heart of one of the most difficult issues in the field of meteorology, and especially numerical weather prediction. Basically, it's as Derecho said, model initialization is by far the most likely culprit. And, as he said, the issue is very complicated and is the subject of entire scientific conferences (eg. the last AMS annual meeting). Essentially what the problem is is that we simply do not have enough information on what the atmosphere is doing in enough detail both in space and time to adequately initialize our models. Observations of the atmosphere, both remote (from satellites and radar) and in situ (surface weather stations, upper-air sounding balloons) simply do not give us a detailed enough picture, but weather prediction models are utterly useless without these observations to initialize them. Basically, what is done is that a "first guess" of the state of the atmosphere, which usually comes from a prior forecast of the same model is used to more-or-less fill in the gaps between observations. The observations are then fed into the model in a statistically robust manner to "update" the first guess field. The end result is an analysis of the atmosphere, which is subject to all kinds of errors from multiple sources, but nevertheless is our best guess of what the state of the atmosphere is at a given time. This analysis is then plopped into the model, which then uses a whole hoard of extremely complicated computations of the equations that govern fluid flow in the atmosphere, themselves subject to all kinds of questionable assumptions and constraints, to give us our forecast. With this in mind, it's absolutely amazing to me that the numerical forecast models give us any useful information at all about the future state of the atmosphere, and I've been working with one for the past 2 years! And yet, they are considered the backbone of modern meteorology, and deservedly so.

In the case of 97L, it's basically a crapshoot whether or not the models will pick up on it (note it could also be that the models aren't doing much with it because it really isn't going to actually develop). Generally observations over the ocean are of too low of a density to adequately resolve tropical cyclone-scale circulations unless you get really lucky. Sometimes, however, the atmospheric conditions that will go on to produce a developing tropical cyclone are present in the larger-scale flow patterns that the model CAN resolve, and this is why every once in a while you will see the models predict the development of a tropical cyclone 2 or 3 days before the wave or disturbance in question has even exited the African coast or developed to begin with. The problem is, other tropical cyclones form without such obvious pre-existant atmospheric "fanfare", at least on a scale resolvable by the global models. Tropical cyclones are among some of the most difficult to forecast events for this reason, in that they can develop in multiple distinct ways, some of which are more easily handled by the models.

HTH, and sorry for the long-winded reply!

Dan

[wxman57]

There's just so little data out there over the oceans for the Global models to work with. They're not designed with tropical systems in mind, so they don't "look" for them, necessarily. Oh, if a tropical system is big enough or strong enough then the Global models might lock onto it, though it's intensity probably won't be initialized well. That's why you may see the GFS take a Cat 5 hurricane and forecast it to be 1005mb in 3 days.

The tropical models are different. Special care is given to the initialization of data near weaker tropical systems. In fact, the center position and pressure may even be defined before the model is run (GFDL).

So it's no wonder the globlal models are generally clueless out in the tropics with weaker systems. One bit of good news is that the next generation of satellites will have advanced sounder capabilities that could provide high-resolution data out across the oceans for model initialization. I don't know if the GFS would use this data at first, but the new WRF model that will replace the GFDL in 2005/2006 will certainly be able to use the next-gen satellite data. Perhaps finally we'll do better at forecasting hurricane intensity. But don't expect much until maybe 2010 or so. NASA hasn't had much luck launching weather satellites lately.....

[Wthrman13]

Indeed, I think that improvements in remote sensing technology are our best hope for gathering more data for model initialization, especially over the oceans.

I realized I failed to mention in my previous post that in many models (i.e. the UKMET and GFS, to name just 2 off the top of my head) existing tropical cyclones are actually artificially inserted into the model initial fields, a process called "bogussing". A certain vortex structure based on idealized models of tropical cyclones is assumed and blended with the model initial fields in such a way as to try to make up for the lack of actual data at that spatial scale. The same is true of the GFDL, although they do try to "bend" the bogussed vortex towards actual observations as much as possible in their initialization routine. Needless to say, if the real system departs from this idealized bogussed vortex in any significant way, your forecast of said cyclone is going to suffer. I think this is part of the problem with the GFDL and it's often completely useless intensity forecasts.

[stormcloud]

I don't think that there will be much model improvement until the day arrives when sophisticated satellites will make weather balloons unnecessary. No waiting every 12 hours for data, as the satellites will provide world wide minute to minute atmospheric updates.

[Rieyeuxs]

Thank you all for the excellent answers. This board has been a wonderful learning experience.

I understand, sort of, if the dynamics aren't there that the models won't pick up on development. However, if I could get clarification on two other points.

1) If development potential isn't there, like what many people are speculating on 97, then what were the more specific triggers that set off the models for development for the June wave that most everybody agreed wasn't going to develop? And didn't? I'm trying to compare and contrast the two systems for my own learning curve as to the weakness in the modeling.

2) If the models have such inherent weaknesses, then why do we "trust" modeling concensus? I remember last year when there was the lively discussion as to where Isabel would end up and when it was going to curve northward despite a lot of -removed- for a Florida hit. The models had all fairly agreed on the NC location, and sure enough, that's where it ended up. Do the models run "better" on a fully developed system vs a system in the developmental stages? Is it they can't prog development very well, but once development has occurred, they can "steer" it more accurately? I ask this only because the path of 97 seems so much more uncertain than other, more developed storms.

And yes, I do see than even though there is uncertately on development for 97, the models are "agreeing" more on Mexico for where it's going to end up, even though that's still further out time wise.

Thank you all once again for your help!