Models

[flyingphish]

Do they initialize from ths current state or the anticipated state? For instance..is the model suite for a possible Ernesto based on a tropical wave or an official tropical depression ? Thanks

[SouthAlabamaWX]

It is intialized by points/grids spaces.

[flyingphish]

Thanks Sw Ala. I guess that is somewhat obvious. But,what is generally fed into the models ? The current statement or what is anticipated? What do they base criterion on ? I am not trying to challenge, as much as justify. Everybody seems to espouse models, and be the first to criticise them. So when do they initialize ?
Or if they are constantly initializing, when do you pay attention ?

[btangy]

There's a whole literature (Data Assimilation) on how to do this. It isn't as simple as plugging in observations and running the model forward. There are techniques such as 3DVAR, 4DVAR, Ensemble Kalman Filter (EnKF)... stuff that I don't really understand myself. It's a whole lot of linear algebra.

In particular for tropical cyclones, you sometimes see special missions flown around storms which then get fed into the 00Z models occasionally. I'm not sure if reconaissance data gets into the model... I suspect it doesn't. Satellite data around the storm is usually assimilated. What some models do (e.g. the GFDL) is that they put a 'bogus' storm into the model as initialization... i.e. a synthetic tropical cyclone with the minimum pressure and winds of the real one along with any size information. It's pretty tricky stuff.

[Wthrman13]

Indeed, atmospheric data assimilation is a whole discipline in itself. Basically though, all the techniques that btangy mentioned are ways in which various observations (surface station, upper air soundings, radar reflectivity and radial wind observations, and even satellite images and soundings) are all mixed together, using a previous model forecast as a "first guess" in such a way as to minimize the errors (differences between the analyzed initial field and the unknown "true" field) in the final analysis. Some techniques are better than others. Right now, most operational models use some form of 3DVAR, which is a way of creating an analysis at a single time, and then running the model forward from there. Some agencies use 4DVAR, which actually runs the model forward (and here's the wierd part: backward as well!) over a certain interval of time, assimilating observations in that interval, until the observations and the model agree as well as possible during that interval. The model fields at the end of that interval are then used for the subsequent forecast. This technique (4DVAR) has been slow to be adopted by many forecast centers due to its large demand on compuational resources, and the fact that a second model in effect has to be created, one that does all the processes of the original model, only backwards in time! The ECMWF model, however, has been using this technique for some time, and it's one of the reasons why the ECMWF consistently outperforms the GFS in the long range.

EnKF is another technique that is similar to 4DVAR, but only requires the model to be run forward in time, and assimilates data "on the fly". Some of my own research is delving into using EnKF to help assimilate radar data for simulating and forecasting individual thunderstorms.

[flyingphish]

THANKS. This is what makes this a great board !!

[flyingphish]

On the side .. A local principle was almost killed by lightning here last week. Keep up the good work. My uncle Jim graduated fom MIT in 66. Guess I have a bit of him in me somewhere.

[Wx_Warrior]

Just spoke with my brother (KFDM)...he said GFLD model has "ernesto" at 100mp and on the tip of Cuba on monday...GOM will be staring down the barrell of a loaded gun!

[tailgater]

Just spoke with my brother (KFDM)...he said GFLD model has "ernesto" at 100mp and on the tip of Cuba on monday...GOM will be staring down the barrell of a loaded gun!

Tuesday is the anniversary, hope there's nothing there to freak everyone out. Lots of Media I'm sure will be around and we don't need any more bad pub.

[jlauderdal]

Just spoke with my brother (KFDM)...he said GFLD model has "ernesto" at 100mp and on the tip of Cuba on monday...GOM will be staring down the barrell of a loaded gun!

20 or 12 gauge?

[NCWeatherChic]

Just spoke with my brother (KFDM)...he said GFLD model has "ernesto" at 100mp and on the tip of Cuba on monday...GOM will be staring down the barrell of a loaded gun!

20 or 12 gauge?

Double-barrel, sawed off

[P.K.]

The ECMWF model, however, has been using this technique for some time, and it's one of the reasons why the ECMWF consistently outperforms the GFS in the long range.

If I remember right the Met Office global model has also been using 4DVAR since Autumn 2004.

[KatDaddy]

GFDL has shifted well E to the Bahamas.

[Blown Away]

Just spoke with my brother (KFDM)...he said GFLD model has "ernesto" at 100mp and on the tip of Cuba on monday...GOM will be staring down the barrell of a loaded gun!

Tuesday is the anniversary, hope there's nothing there to freak everyone out. Lots of Media I'm sure will be around and we don't need any more bad pub.

Which tip, the East or West, looks like the GFDL is heading towards the Bahamas.

[Windtalker1]

Once this becomes a depression, all other models will follow the GFDL into the Bahamas and S FL....wait and see. The following post is NOT an official forecast and should not be used as such. It is just the opinion of the poster and may or may not be backed by sound meteorological data. It is NOT endorsed by any professional institution including storm2k.org For Official Information please refer to the NHC and NWS products.

[HURRICANELONNY]

I think many pro mets have said to take the models with a grain of salt. Especially when the wave hasn't been initialized. If recon heads out today and finds an LLC then just maybe some of the models will be looked at more cautiously. They have no meaning yet. I like looking at the negative factors that people post. Like the SAL and the shear maps to get an idea if it will even survive.

[Air Force Met]

I think many pro mets have said to take the models with a grain of salt. Especially when the wave hasn't been initialized. If recon heads out today and finds an LLC then just maybe some of the models will be looked at more cautiously. They have no meaning yet. I like looking at the negative factors that people post. Like the SAL and the shear maps to get an idea if it will even survive.

The models should be looked at and can give you a good idea of where it is going...do not disregard them...even with no LLC initialized.

The problem many (MANY) on this board have is they just look at a dot that a certain model puts out (like the 72 hour position for the GFDL...or the 120 hour position for the BAMD) instead of looking at the SYNOPTIC pattern.

You don't need a center to be formed to get a good idea for the synoptic pattern in the coming week. If a blocking ridge is going to be to the north...it going to be to the north...and it doesn't matter where the center finally forms...that blocking pattern will steer that system.

Now, every wobble and where the center forms does make a diff on the final landfall...but that is SOOO far down the road it's not even worth talking about....just like these tracks aren't worth talking about.

What is worth talking about is the synoptic pattern...because that is what influences the tracks. That will be there REGARDLESS of whethe a center forms or not.

Too many people totally ignore the pattern...they don't look at the actual maps of the 500mb, 700mb...etc...and they just look at the model spread.

Just like what the GFDL is doing now. Looking at the synoptic pattern...does it make sense? or has anyone even looked that far and just said "hey...that's a track that looks good" with no reasons?

It's all about the synoptic pattern.

BTW...a day or 2 ago...the GFDL had this going east of the PR.

[Air Force Met]

Once this becomes a depression, all other models will follow the GFDL into the Bahamas and S FL....wait and see. The following post is NOT an official forecast and should not be used as such. It is just the opinion of the poster and may or may not be backed by sound meteorological data. It is NOT endorsed by any professional institution including storm2k.org For Official Information please refer to the NHC and NWS products.

Why?

Not saying I agree or disagree...but tell us why.

It's real easy to make a dogmatic statement...but lets put some synoptic reasoning behind it.

Why?

[HURRICANELONNY]

Why. because he lives there. LOL. and he's probably a kid.

[Air Force Met]

Why. because he lives there. LOL. and he's probably a kid.

Well I wasn't going to go THAT far...

Although studies have shown that to be the leading factor in "Gut Call" forecasting.