GFS model

Post Reply
jmagyar
Regular User
Posts: 13
Joined: Wed Mar 18, 2009 8:50 am
Location: Derby, KS

GFS model

Post by jmagyar » Thu Jan 06, 2011 12:16 pm

GFSrun aroun 86-110 is all jumbled up and some are duplicated with different values. Looks like the 0Z and 12Z runs are intertwined.

Just an FYI.

James
User avatar
Andrew Revering
Site Admin
Posts: 879
Joined: Mon Jul 04, 2005 9:37 pm
Location: Cambridge, MN
Contact:

Post by Andrew Revering » Thu Jan 06, 2011 5:38 pm

I can't see a problem. Which map? You should attach a screen shot.
Andrew Revering
F5Weather
http://www.f5wx.com/
jmagyar
Regular User
Posts: 13
Joined: Wed Mar 18, 2009 8:50 am
Location: Derby, KS

Post by jmagyar » Thu Jan 06, 2011 6:12 pm

I think I figured it out. My guess is that your server starts posting as soon as the new run has completed for a certain forecast hour. So the begining of a run might be on the newer and the end might still be on the older??

Right now, my GFS 96 hour...and looking at APRWX Total Accum Snow....anyways, 96 Hour is 12 noon Monday afternoon...and is the 18Z run. Now if I go to to 105...looking at the same APRWX Total Accum Snow...its showing 3pm Monday and the 12z run. So, when I animated the run from 86-120...It appeared all messed up as it would go to 9pm monday on the 18Z and then back up to 3pm on the 12z....My guess is either they do not update that far out on the 6 and 18Z? Or your releasing each hour as your backside processes it and not as a complete set?

Now that I understand whats going on...not as big of deal. Actually kind of neat to be able to see the differences in the two runs for 3pm :)

I wish, there was a way to download an entire run all at once for all the maps. That way we would quit pinging your server so much. And would speed things up on the end user side....and well again, could use it as comparision between the different runs.

What are we taking about...maybe a 15-30 meg file for all the maps?? Maybe an archive sytem? If we download a map...maybe it can archive it. If I Try to pull the same map...it checks to see if there is a newer map...if there is not...then it uses the cache??? Possible??

James

(EDIT) YEP...Now that I wait a few...more of the 18z are now shoing...So...its the server dumping the earlier 18z as it processes
User avatar
Andrew Revering
Site Admin
Posts: 879
Joined: Mon Jul 04, 2005 9:37 pm
Location: Cambridge, MN
Contact:

Post by Andrew Revering » Thu Jan 06, 2011 11:11 pm

jmagyar wrote:My guess is that your server starts posting as soon as the new run has completed for a certain forecast hour. So the begining of a run might be on the newer and the end might still be on the older??
That's exactly right. Also during the updating process its possible for a forecast hour to skip an hour temporarily due to a transfer error, but it should get back to updating that one after a few minutes.
jmagyar wrote:My guess is either they do not update that far out on the 6 and 18Z? Or your releasing each hour as your backside processes it and not as a complete set?
The 06z and 18z runs do forecast out to 384 hours, its just occasionally there are transfer problems and some of those 'frames' don't get updated. That's not supposed to happen, its a bug and I'm working on fixes for that issue.
jmagyar wrote: What are we taking about...maybe a 15-30 meg file for all the maps?? Maybe an archive sytem? If we download a map...maybe it can archive it. If I Try to pull the same map...it checks to see if there is a newer map...if there is not...then it uses the cache??? Possible??
A caching system was intended in the initial release of the software, but that never happened due to some complications. The problem with Caching is that you always want the latest data, and in order to know if you should cache or pull from the web is going to involve knowing if the web data is new or not. Well the data files are quite small, despite how long it takes to draw an image. That's because it is doing the rendering on your machine. So the download itself is pretty quick. So there isn't much difference between checking the server if the data is new and just downloading it. The only advantage I could see is if we had a status page that told the software (and perhaps the user) about everything's time stamp. That would be a much faster download for the software and it told the software about all products in one quick shot. ...

that status system is in the works. It is actually running to a certain extent but it's not 100% ready to go just yet. Once that is ready, we can use that to implement a caching system, as well as show the users the current status on each forecast hour for each model.
Andrew Revering
F5Weather
http://www.f5wx.com/
Post Reply