GFS Predicts Category 5 Hurricane at 980 hours!

This is the general tropical discussion area. Anyone can take their shot at predicting a storms path.

Moderator: S2k Moderators

Forum rules

The posts in this forum are NOT official forecasts and should not be used as such. They are just the opinion of the poster and may or may not be backed by sound meteorological data. They are NOT endorsed by any professional institution or STORM2K. For official information, please refer to products from the National Hurricane Center and National Weather Service.

Help Support Storm2K
Message
Author
DoctorHurricane2003

GFS Predicts Category 5 Hurricane at 980 hours!

#1 Postby DoctorHurricane2003 » Sun May 14, 2006 1:38 am

Ok, so the purpose of this thread is to educated board members on the appropriate use of models, and generally how far out you should go before what they usually say is wrong.

Computer Models forecast based on a certain set of current conditions whenever the model runs. They forecast for 6 hours out, use the 6 hour forecast to forecast for 12 hours, etc. until the model reaches its particular time limit.

Now one problem with this lies in Chaos Theory's "Butterfly Effect." That is...one small error can lead to HUGE errors down the road. A simple example of the butterfly effect is...let's take the numeric value "a" as our initial point. From "a" we need to get the number "j", but to do this we have to go through numbers "b", "c", "d", "f", "g", and "h". If the value of "a" = 28.3, but we use "a" = 28 to make it simpler, and if "b" = 3.1"a", "c" = 3.1"b", etc. then our final estimated number "j" = 77035.3, differing from the actual value of "j" = 77860.7 by 825.4. So the error difference went from .3 to 825.4!

This is what happens in weather computer models, obvious the 6-hour forecast will be correct or very close to correct. The keywords here are "very close". This small error here, and in subsequent forecasts, the following errors, will eventually lead to errors so large that the forecasted weather pattern given by the models could be the exact opposite of what occurs! (It could forecast a trough, when a strong high actually becomes present!)

So now you ask yourselves "How far out should I go before I should ignore the model's output?" Generally anything within 72 hours is good, and anything within 120 hours is "okay." Beyond 120 hours you can pretty much consider bovine fecal matter.

I hope this clears up the misuse of models on these boards in the future and that all of you have learned something new about weather forecasting tools! :)
0 likes   

Brent
S2K Supporter
S2K Supporter
Posts: 38117
Age: 37
Joined: Sun May 16, 2004 10:30 pm
Location: Tulsa Oklahoma
Contact:

#2 Postby Brent » Sun May 14, 2006 1:39 am

:roflmao:
0 likes   
#neversummer

User avatar
senorpepr
Military Met/Moderator
Military Met/Moderator
Posts: 12542
Age: 43
Joined: Fri Aug 22, 2003 9:22 pm
Location: Mackenbach, Germany
Contact:

#3 Postby senorpepr » Sun May 14, 2006 1:41 am

Outstanding post... and you earn extra credit for using the phrase "bovine fecal matter."
0 likes   

User avatar
wxmann_91
Category 5
Category 5
Posts: 8013
Age: 34
Joined: Fri Jul 15, 2005 2:49 pm
Location: Southern California
Contact:

#4 Postby wxmann_91 » Sun May 14, 2006 1:42 am

Oh Thank goodness this was written.
0 likes   

User avatar
mempho
Tropical Storm
Tropical Storm
Posts: 227
Joined: Fri Feb 10, 2006 11:08 am
Location: Memphis, TN

Re: GFS Predicts Category 5 Hurricane at 980 hours!

#5 Postby mempho » Sun May 14, 2006 2:41 am

Nice Title :D
0 likes   

User avatar
stormtruth
Category 2
Category 2
Posts: 651
Joined: Thu Mar 16, 2006 4:15 pm

#6 Postby stormtruth » Sun May 14, 2006 2:50 am

Maybe in 1920 hours.
0 likes   

Coredesat

#7 Postby Coredesat » Sun May 14, 2006 5:31 am

Well, 980 hours is a little over 40 days, so hmm...early season monster? :think:



I'm kidding. :fools:
0 likes   

User avatar
bvigal
S2K Supporter
S2K Supporter
Posts: 2276
Joined: Sun Jul 24, 2005 8:49 am
Location: British Virgin Islands
Contact:

#8 Postby bvigal » Sun May 14, 2006 5:36 am

Bravo Dr. Hurricane!!! Great explanation of model reliability, and especially your illustration of butterfly effect. You deserve a round of applause! :clap:
0 likes   

User avatar
x-y-no
Category 5
Category 5
Posts: 8359
Age: 65
Joined: Wed Aug 11, 2004 12:14 pm
Location: Fort Lauderdale, FL

Re: GFS Predicts Category 5 Hurricane at 980 hours!

#9 Postby x-y-no » Sun May 14, 2006 8:18 am

DoctorHurricane2003 wrote:So now you ask yourselves "How far out should I go before I should ignore the model's output?" Generally anything within 72 hours is good, and anything within 120 hours is "okay." Beyond 120 hours you can pretty much consider bovine fecal matter.


A pretty fair summary. I would only add that one can get some useful infomation further out by watching for developments which persist in approximately the same position and timepoint over several days worth of runs. That indicates a pattern supportive of that development is quite likely.

The details, however, aren't at all reliable beyond the three to five day range you suggest.
0 likes   

User avatar
benny
Category 2
Category 2
Posts: 593
Joined: Sun Mar 19, 2006 8:09 am
Location: Miami

Re: GFS Predicts Category 5 Hurricane at 980 hours!

#10 Postby benny » Sun May 14, 2006 8:45 am

DoctorHurricane2003 wrote:Ok, so the purpose of this thread is to educated board members on the appropriate use of models, and generally how far out you should go before what they usually say is wrong.

<uct>

So now you ask yourselves "How far out should I go before I should ignore the model's output?" Generally anything within 72 hours is good, and anything within 120 hours is "okay." Beyond 120 hours you can pretty much consider bovine fecal matter.

I hope this clears up the misuse of models on these boards in the future and that all of you have learned something new about weather forecasting tools! :)


I would have to disagree with the models being manure beyond 5 days. Sure they aren't as reliable beyond 5 days and they can certainly be wrong but there is certainly forecast value in them. Take a look at:

http://wwwt.emc.ncep.noaa.gov/gmb/STATS/html/acz6.html

Every one of the models has skill in the NH extratropics at Day 6. I'll just crudely define skill as a correlation > 0.6 (as many others have done). If you look at the die-off curves.. ie the amount of time it takes to get below 0.6...

http://wwwt.emc.ncep.noaa.gov/gmb/STATS/html/sjl2.html

You get to about Day 7 or 7.5 for the 0.6 marker.. and 5.5 or so for 0.8 correlation. That's pretty damn good beyond 5 days in my opinion and they are getting better. I don't have any stats for the ensemble means but they are better than the controls especially at longer range. The stats aren't as good in the summetime so the low end estimates are more realistic.

This is a fun graphic showing how the GFS has improved through the years:
http://wwwt.emc.ncep.noaa.gov/gmb/STATS ... imate.html

If you rollover the last year.. the edge of the "skill" of 0.6 has shifted from about 6 days to 7.5 days in the past 17 years. That pretty impressive in my book. You have to be more a generalist to use the models beyond a week... if more than model is indicating something, perhaps if the trends are on your side, if you have other longer-term climate features controlling things (ENSO/MJO)... it is a tricky business but there is certainly value there. Nothing like a short-range forecast though.

I will edit this and say that in terms of tropical cyclogenesis, there aren't that many studies, but if a bunch of models.. not just one.. are indicating something happening in the 5-8 day timeframe.. it is worth taking a look at... :P
0 likes   

HurricaneHunter914
Category 5
Category 5
Posts: 4439
Age: 31
Joined: Fri Mar 10, 2006 7:36 pm
Location: College Station, TX

#11 Postby HurricaneHunter914 » Sun May 14, 2006 9:33 am

For a minute there I thought the GFS was on crack. :lol: :lol:
0 likes   

User avatar
Hybridstorm_November2001
S2K Supporter
S2K Supporter
Posts: 2813
Joined: Sat Aug 21, 2004 2:50 pm
Location: SW New Brunswick, Canada
Contact:

#12 Postby Hybridstorm_November2001 » Sun May 14, 2006 9:57 am

Good one DoctorHurricane2003! I haven't had this good a laugh in ages!

:lol: :lol: :lol: :lol: :lol: :lol:
0 likes   

User avatar
george_r_1961
S2K Supporter
S2K Supporter
Posts: 3171
Age: 64
Joined: Sat Oct 12, 2002 9:14 pm
Location: Carbondale, Pennsylvania

#13 Postby george_r_1961 » Sun May 14, 2006 10:21 am

Doc I gotta agree with you. The GFS is famous for spinning up phantom storms and and overdoing the lows/storms that do form. The 120 hour runs are for entertainment purposes only.
Last edited by george_r_1961 on Sun May 14, 2006 11:23 am, edited 1 time in total.
0 likes   

User avatar
Tampa Bay Hurricane
Category 5
Category 5
Posts: 5597
Age: 37
Joined: Fri Jul 22, 2005 7:54 pm
Location: St. Petersburg, FL

#14 Postby Tampa Bay Hurricane » Sun May 14, 2006 11:20 am

excellent points doctor hurricane
0 likes   

jlauderdal
S2K Supporter
S2K Supporter
Posts: 7210
Joined: Wed May 19, 2004 5:46 am
Location: NE Fort Lauderdale
Contact:

Re: GFS Predicts Category 5 Hurricane at 980 hours!

#15 Postby jlauderdal » Sun May 14, 2006 11:37 am

DoctorHurricane2003 wrote:Ok, so the purpose of this thread is to educated board members on the appropriate use of models, and generally how far out you should go before what they usually say is wrong.

Computer Models forecast based on a certain set of current conditions whenever the model runs. They forecast for 6 hours out, use the 6 hour forecast to forecast for 12 hours, etc. until the model reaches its particular time limit.

Now one problem with this lies in Chaos Theory's "Butterfly Effect." That is...one small error can lead to HUGE errors down the road. A simple example of the butterfly effect is...let's take the numeric value "a" as our initial point. From "a" we need to get the number "j", but to do this we have to go through numbers "b", "c", "d", "f", "g", and "h". If the value of "a" = 28.3, but we use "a" = 28 to make it simpler, and if "b" = 3.1"a", "c" = 3.1"b", etc. then our final estimated number "j" = 77035.3, differing from the actual value of "j" = 77860.7 by 825.4. So the error difference went from .3 to 825.4!

This is what happens in weather computer models, obvious the 6-hour forecast will be correct or very close to correct. The keywords here are "very close". This small error here, and in subsequent forecasts, the following errors, will eventually lead to errors so large that the forecasted weather pattern given by the models could be the exact opposite of what occurs! (It could forecast a trough, when a strong high actually becomes present!)

So now you ask yourselves "How far out should I go before I should ignore the model's output?" Generally anything within 72 hours is good, and anything within 120 hours is "okay." Beyond 120 hours you can pretty much consider bovine fecal matter.

I hope this clears up the misuse of models on these boards in the future and that all of you have learned something new about weather forecasting tools! :)


I needed you a few days ago when there were members harping about gfs 16 days out developing a low, i quickly squashed that idea but of course received some hassle but i guess discussion is what it is all about.
0 likes   

User avatar
GeneratorPower
S2K Supporter
S2K Supporter
Posts: 1648
Age: 45
Joined: Sun Dec 18, 2005 11:48 pm
Location: Huntsville, AL

#16 Postby GeneratorPower » Sun May 14, 2006 1:35 pm

senorpepr wrote:Outstanding post... and you earn extra credit for using the phrase "bovine fecal matter."


New acronym... BFM.

TTYL
0 likes   

Dean4Storms
S2K Supporter
S2K Supporter
Posts: 6358
Age: 62
Joined: Sun Aug 31, 2003 1:01 pm
Location: Miramar Bch. FL

#17 Postby Dean4Storms » Sun May 14, 2006 2:08 pm

And all this time I thought the weather followed what the models told them to do!!!! :roll:
0 likes   

User avatar
Aslkahuna
Professional-Met
Professional-Met
Posts: 4550
Joined: Thu Feb 06, 2003 5:00 pm
Location: Tucson, AZ
Contact:

#18 Postby Aslkahuna » Sun May 14, 2006 3:43 pm

When it comes to monsoon rainfall in Arizona, any model beyond 48 hours falls into the BFM category because they just don't handle the pattern well at all.

Steve
0 likes   

User avatar
jusforsean
Category 1
Category 1
Posts: 395
Joined: Wed Nov 09, 2005 8:22 am
Location: South Florida

#19 Postby jusforsean » Sun May 14, 2006 4:31 pm

:fantastic:
0 likes   

DoctorHurricane2003

#20 Postby DoctorHurricane2003 » Mon May 22, 2006 11:36 am

bump because some people are still doing it
0 likes   


Return to “Talkin' Tropics”

Who is online

Users browsing this forum: cycloneye, Google Adsense [Bot], riapal and 41 guests