GFS changeover and new version

NCEP is scheduled to change over to the new GFS model (supposedly somewhat more accurate!) on June 12, at 12Z. This change is more significant than most upgrades, partly because there are some changes to URL’s, at least for the web based downloads. Sam (Bohler) and I have been conferring on this and preparing for the changeover, and hopefully it will happen seamlessly on Wednesday. However, our testing on the “parallel” data revealed a problem with reading the sea level barometric pressure data: the 0 hour data was fine, but erroneous values were appearing starting with hour 3. This triggered an error, so we switched temporarily back to the old (still “official”) data, so we could work on the problem.

We’re actually working on that tonight, and are pretty confident we will have it fixed. However, as a precaution, I have just posted new versions of WXSIM (2019, Build 1.2) and WXSIM-Lite (2.5), which sidestep the problem by coming up with substitute pressure values. WXSIM’s is very accurate, based on the 1000 mb height (which is being read fine), temperature, and station elevation, and should not compromise the forecast in any significant way. WXSIM-Lite will simply default it to 1013 mb, BUT it’s not really used in what WXSIM-Lite does (I was grabbing it mainly for possible future use), so that should be fine. You shouldn’t even see any evidence of these substitutions, except for a brielf message in WXSIM, if your running it manually.

Again, I expect this will not even be needed, but I wanted to get out ahead of any problems!

As for Chris McMahnon’s GFS data source, I’ve communicated with him on this, and he’s using an FTP source that he thinks will not be changed. I do not know whether his script will encounter problems with that pressure value, but if it does, this WXSIM upgrade will take care of it anyway.

SO, please run www.wxsim.com/wxsimupg.exe, just in case!

Thanks, as always, for using my software! :slight_smile:

Tom

Hello Tom

Thought I would load and run your new update prior to the NWS change.

I ran wxsimate, wxsimlite, wret and wxsim and found no issues. Also the forecast SLP pressures look reasonable.

Thanks
James B
Tucson AZ
DW4536

Tom,

All my scheduled runs have successfully completed since I installed your update early this AM. The results look normal.

Steve

Same here as the afternoon and early evening runs were good with the new updates.

:smiley:

MikeyM

Update: Sam found the problem and fixed it. The problem was that NCEP is now putting two different accumulated precipitation fields in the data (the old one accumulated over the last 3 or 6 hours, and the new one since initialization, and BOTH are now in there), and they were labeled APCP(1) and APCP(2), and we’d been searching for the tag APCP. This garbled up a couple of the variables. In any case, it’s fixed, and we’re poised to switch over for the 12Z data tomorrow.

I don’t know the details of Chris McMahon’s script in this regard. It may or may not be affected by this redundancy, but if it is, I think my precautionary move will at least prevent errors (though it’s conceivable something else might be affected). In any case, he will be home in a few days to address anything that might come up. Meanwhile, the Bohler data should be sound.

Tom

@ Tom,

Yesterday, June 11, installed now June 12, 8:40 AM and after two runs today no problems here.

Kind regards,

Piet

https://www.meteosauwerd.nl
https://www.sauwerdmeteo.nl
https//www.weerstation-sauwerd.nl

Hi Tom,

It seems, that all works as it should <-----> without any Problems. :wink:

Stefan

Hi Tom,

Thank you as always for all you do.

All working here for many auto updates with new version, we will see what happens during auto run of 12Z data tomorrow.

Kind Regards Always,

If I remember correctly my script extracts the APCP value so it might well be affected by the same problem. I don’t remember seeing this change in the logs or if I did then the significance didn’t sink in. I’m just about to leave on a business trip and won’t get home until late tomorrow so if there’s a problem I won’t be able to look at it until Friday evening.

Thanks, Chris, for checking the forum! I think your script quit working with the changeover, because I’m still seeing 06Z as your current data, and 12Z has been out for a while. I think Sam’s script is all good now, though we did still have a problem with the APCP at 06Z (fixed as of 12Z). Those two fields are defined differently, and they are easily confused! I’ll email you with what we found, to make it easier to fix when you get back. Right now, though, Bohler is a working option, so there’s no “emergency”. Have a great trip! (“great” for a business trip, anyway) :slight_smile:

Tom

Update on GFS changeover. Bohler should be working fine now, though I think 06Z of June 12 had bad precipitation data at some hours, due to us not fully understanding the changes they’d made. That should be fixed now. It appears that Chris McMahon’s script may have stopped working after 06Z (though the data was still good at that point). He’s out of town until Friday, and should be able to fix it soon after that, but meanwhile, Bohler is working fine, for both WXSIMATE and WXSIM-Lite version of the data.

Thanks for your patience during all these changes! I’m hoping NCEP will just leave it alone for a while, now! We’ve had to respond to a lot of URL and data changes this year!

Tom

since this update, i get this screen after every forecast


Capture.PNG

I see that if wxsimate is set to use McMahon data, Bohler is fine. This will be fixed hopefully in the next few days when Chris McMahon returns.

Stuart

Switching wxsimate to Bohler is working fine for me also.

James B
DW4536

Bohler works for me.

Thanks

I’m looking at my scripts now. I’ve found and fixed one problem where there seems to have been an unannounced change to the directory sructure which meant my scripts weren’t seeing any data to process.

I now need to wait for a set of data to download to see if it will process. I couldn’t see any changes announced to the data structure but having now found the directory structure change I’m a little concerned that there are other changes that I don’t know about.

I found another problem which I think I’ve fixed. The scripts have run through and indicate a success but I don’t have time to check output now. I’ll look later but you might be able to tell me if the data looks OK before I get home again.

Chris I just changed to your data for my 13.00 forecast ans allwent fine, so looks like you nailed it.

Stuart

There may still be something wrong. There are a number of new variables, one being the second APCP variable that Tom mentioned. I think that only the APCP variable is used by WxSim.

Unfortunately the naming convention for the new variables means that my scripts don’t currently differentiate between the two variables. I think by accident my scripts are reading the correct variable in each case, but I’ve had to make some changes to ensure that the correct variable is always selected. I need the next run (12z) to become available to confirm that my mods have worked.

I’ll take a look later to see if things are OK.

My scripts are still broken. An unfortunate variable naming convention in the GFS data means that different hours sometimes have two different variables with exactly the same name. They probably contain the same data but my scripts need to know how how many variables to expect for each hour and when you have two with the same name some hours have more variables than I expect to see. I’ll find a way to work around this, but I’m going out with friends tonight so I’ve run out of time for today.