Post by iconPost by SnMan | 2018-11-17 | 11:44:26


I dont understand why i have a difference of Wind prevision between Qt and Zezo.
I load this one at 20h10 and 8h10 for QT.
It's possible to have which Grif use Zezo ???

I try to compare the 2 roads prevision but with 2 Grids si different it's not possible

Thanks all et good job

Sorry for my english,;)

commenticon 5 Comments
Post by iconPost by zezo | 2018-11-17 | 12:34:13
You can check (and download) the GRIB files used at

all.grib is a bundle with the current weather forecast. There are also separate files and archives of the entire forecast.

Also in the /grib/original you can find the original GFS data (for the past 24 hours). You can also check the download schedule and progress there. Times are in UTC.

One possible source of differences is the interpolation algorithm. Data at integer coordinats should match pretty well (there is a small error introduced by the VR number representation) but the middle of the square around a front is a different case.
Post by iconPost by yaxa2vrai | 2018-11-17 | 18:31:07
I see that the files are downloaded and processed arround 3 hours after their time (ex. download of t12z files start at 15:35)

Any reason for this delay? (if my understanding is correct)
Post by iconPost by zezo | 2018-11-17 | 18:45:07
12Z is the _starting_ point of the model.

That is, at 12:something all available worldwide weather data from the 12Z point of time is loaded in the NOAA supercomputers. They start crunching the numbers and a bit later come up with the "analysis" - that's a model of the current (now slightly past) starting conditions, interpolated from the finite number of data points.

Then the computers start computing the model and about 3 hours later come up with the first "forecast" - for which is now the actual weather.

The process continues, with 3 hours worth of data becoming available every minute or two. So it takes another 2 hours to calculate all future snapshots.
Post by iconPost by yaxa2vrai | 2018-11-17 | 22:07:26
Waouh, that's long computation.

My conclusion, supercomputer they use are not enough super ;-)

Thanks for the reply.
Post by iconPost by zezo | 2018-11-17 | 22:16:11
They should only be fast enough to reliably finish the computation within the 6-hour cycle, so the resources are free for the next GRIB.

It's a medium-term forecast. There is no real benefit to get it ready in 2 hours instead of 3.
Topics list
Copyright 2009 by ZEZO.ORG. All Rights Reserved.