Clark
(Meteorologist)
Wed Oct 19 2005 06:52 PM
Re: Wilma Continued

It is very, very tough to do distributed computing with the current state of computer models. Due to the time continuity of a model, you can't just assign an individual computer one time stamp to do because it depends on other computers doing the prior time stamp(s); this is unlike the SETI@home project, where you can assign a particular region to search at a time to individual computers...spatial dependence is much easier.

Essentially, a computer model is going to be constrained speed-wise by the slowest machine on the cluster. We are in the process of upgrading all of the machines on our cluster to 2.8GHz processors and dedicating 16 CPUs to the effort; this should help speed things up somewhat. For distributed computing to work, you'd need a large amount of fast machines always connected to the Internet with a high speed connection, all able to work in concert with each other -- not as individual cells. That's tough to do even with the best distributed computing solutions. If you want the best model, one you'd need someone to write a model that could handle distributed computing -- currently no such model exists, with the current models being designed to run in parallel computing environments -- and add in all of the layers of complexity that lead to things like the GFS & NOGAPS model (e.g. data assimilation and such). Unfortunately, it's just not feasible computationally and writing such a model would probably take 5 years.

As for the track of the system, I'm still not buying the storm not being captured by the trough. I don't have any large reason to go away from model consensus, though the trends I am seeing suggest the turn may be harder to the right than initially anticipated. This would be good news for the west coast and bad news for the east coast of Florida were it to pan out, but I'm not sold on that yet either. Reintensification is possible over the next day, depending on how long it takes that inner eye to collapse and how long it takes for that outermost eye (the 45mi one) to contract inward to bring about a new cycle.



Note: This is NOT an official page. It is run by weather hobbyists and should not be used as a replacement for official sources. 
CFHC's main servers are currently located at Hostdime.com in Orlando, FL.
Image Server Network thanks to Mike Potts and Amazon Web Services. If you have static file hosting space that allows dns aliasing contact us to help out! Some Maps Provided by:
Great thanks to all who donated and everyone who uses the site as well. Site designed for 800x600+ resolution
When in doubt, take the word of the National Hurricane Center