Zezo forum: Any news about GFS 0.25� ?

Any news about GFS 0.25� ?

Post by Inicio | 2023-01-24 | 16:24:23

Zezo can adopt this future change?

45 Comments

Post by Inicio | 2023-01-24 | 16:24:58

0.25�

Post by zezo | 2023-02-14 | 16:47:25

The new race announced today (Gone with the wind) comes with 0.25� data, and it's in native GRIB format. And 1 knot minimum wind.

Post by Hardtack | 2023-02-14 | 19:12:49

That's actually the most reasonable thing to do, the compression provided by the GRIB format is hard to beat. And they might get rid of some discrepencies between server and frontend.

Post by zezo | 2023-02-16 | 08:29:13

0.5 bytes/sample indeed. But they are jpeg compressed and therefore slightly different from the original NOAA data.

Post by Inicio | 2023-02-14 | 18:56:44

Zezo will change to this format?

Post by zezo | 2023-02-15 | 15:38:53

I'm looking into it.

Post by zezo | 2023-02-16 | 14:05:12

Experimental support added at

http://zezo.org/gone/chart.pl

Wind display is not updated so it will be way off. DO NOT USE the track TWA.

Also no the gate will not work very well initially, but you know that ;-)

Post by zezo | 2023-02-16 | 19:01:12

OK, I've fixed the wind display to some extent, but it uses the old rounding/encoding scheme, so will be slightly off.

BTW don't trust the moving wind display in the game either. It seems to be showing the 1� winds. The 1-knot areas are not visible.

Post by Inicio | 2023-02-16 | 22:23:53

Thanks!

Post by Inicio | 2023-02-18 | 07:42:26

It seens work well...
many thanks!

Post by Paspox | 2023-02-18 | 08:42:14

The http://zezo.org/grib/gribv1/all.grib file is 1deg (I imagine it is the file used for the other races that still use the 1deg gribs).
Is there a link to see the 0.25deg all.grib file ?
Thanks again for all you do.

Post by zezo | 2023-02-18 | 11:03:12

You can get them directly from VR. Use the browser dev tools to see the URLs.

Post by Reves_de_gosse | 2023-06-28 | 16:21:19

Hi Cvetan
It's easy to get the .wnd files, but I have no idea on how to read / decode them ?
Some idea ?
Thanks.

Post by guayacan | 2023-07-29 | 18:07:58

Hello Cvetan,
Could you please explain more in detail where to find the VR grib URL in the browser dev tool: is it in the "sources" tab (which file?), other tab?
Many thanks.

Post by Flying Phil | 2023-02-18 | 14:36:46

Hi Cvetan,
I don't know if you found the way VR handles the 0.25� winds, but if it helps:
They disregarded the +3h and +6h gribs and the interpolation with 4 gribs at the time of the weather update is over. They integrate the +9h later so the disturbances related to the new weather also occur later 😉
Ex. example for the weather of 9h40UTC
- until 12h00 UTC, we use only the gribs of 00Z+9h and 00Z+12h (while in GFS 1� we added the 06Z+3h and 06Z+6h)
- from 12h00 UTC, beginning of the integration of the new weather, we use the gribs of 00Z+12h and 06Z+9h
In addition to the advantages in terms of calculation, there is also the fact that NOAA can be late by an hour or two, no impact.
Interpolation by 10 minutes is also finished, now it's continuous. And the calculation of bilinear interpolation is also standard. no more averages/coef/ratios/conversion to 8 bits , in your get_interp_wind function, speed_uv is the right result (minimum 1knt).
Good luck for the routeur update and thanks a lot for all you do for the VR players.

Post by Flying Phil | 2023-02-18 | 14:39:09

And better to use the VR winds (differents frome NOAA's)

Post by Reves_de_gosse | 2023-07-04 | 13:22:24

Post by Hardtack | 2023-02-18 | 17:49:14

@Phil, this is very useful information!
Do you know the duration of the integration? Are the two cycles merged over the period of one hour, or longer?

Post by zezo | 2023-02-18 | 17:54:55

No merge, it seems. Now the references are not encrypted and you can see the actual grib stamps.

https://vro-api-client.devel.virtualregatta.com/dl?t=fwr

{"references":["20230218.06.009","20230218.06.012","20230218.12.009","20230218.12.012" ...

Winds are quite different from the original GRIB because they are JPEG compressed. That brings the files to about 450k each. The original data is about 2M

Post by Flying Phil | 2023-02-18 | 20:20:11

The JPEG 2000 compression used is lossless and do not modify the data. it's an option of the packing (simple, complex1, complex2, complex3, jpeg) .
This is an algorithm of compression very efficient for compression but very slow when you decompress the data (I confirm ;-) ).

IMHO, if there is a change/compression of the data, it's made by VR before packing or VR uses a source other than NOAA.

Post by zezo | 2023-02-18 | 17:58:55

@Phil Thanks, I already did exactly that, but it's nice to have a confirmation.

Post by Flying Phil | 2023-02-18 | 20:20:54

You're welcome, great we have the same result ;-)

Post by zezo | 2023-02-18 | 21:16:36

grib_dump shows no difference in metadata between NOAA and VR. Only the compression method and the actual data itself.

that includes the forecast source

# US National Weather Service - NCEP (WMC) (grib1/0.table)
originatingCentre = 7;

and cnvgrib -nv -p40 -g22 out.grib out2.grib (to jpeg grid) gives a 1.5M files

so there is probably some smoothing in between.

Post by Flying Phil | 2023-02-18 | 23:08:41

;-) i show the same thing for the metadata

But I found a difference in the scale and packing option (cf https://www.cpc.ncep.noaa.gov/products/wesley/wgrib2/scale.html )
VR scale=-2,6: val=(-4034.49+i*2^6)*10^-2, i=0..127 (#bits=7)
NOAA scale=-2,0: val=(-4034.49+i*2^0)*10^-2, ref=0..4095 (#bits=12)

for the NOAA, numbers ar stored using 12 bits (7 bits for VR)

here is the command to convert NOAA file to VR format with 7 bits:
I used wgrib2 tool (input file : NOAA30 , output FILE NOAA30-7)

wgrib2 NOAA30 -set_grib_max_bits 7 -set_grib_type jpeg -grib_out NOAA30-7

And, bingo ! NOAA30-7 is identical to VR file

Post by zezo | 2023-02-19 | 11:10:33

Nice.

There are only so much distinct values in each file indeed - between 50 and 90.

This can help doing lossless grib to png conversion AND save some file size. And my internal data structures can co back from short to char ;-)

EDIT: Yep, 7-bit losless webp is even smaller then the jpeg2000 data

Post by Reves_de_gosse | 2023-02-20 | 14:45:52

Hello Flying Phil
Thank you for your information, quite surprising.
To be sure to be sure,
If I understand correctly, the first new winds available, the 06Z+6, are never used, which reduces the calculations and allows NOAA to be a few hours behind without any visible inconvenience?
Is that right ?

Post by Flying Phil | 2023-02-20 | 20:02:35

Hello,
That's right. Gribs +3h and +6H are no longer used by VR in the gfs 0.25 races .

Post by YourMomSA | 2023-02-20 | 23:50:28

Great job in the Gone with the Wind Race, Cvetan (Big Bird)!

Post by zezo | 2023-02-21 | 11:34:21

Thanks! I mostly stayed up all night during the weekend to observe how the router behaves, but it helped the game too.

Or it's that I've never set foot on a real races but have sailed a lot of cruisers ;-)

BTW the new winds make the game much more interesting. Now boat a mile to your side can see quite different wind.

Post by Flying Phil | 2023-03-29 | 16:19:06

Hi,
Great news, FYI, for the new GFS 0.25 race, VR added a flag in the race definition ("getboatinfos" frame) to indicate if we have to use the GFS 0.25 winds :
"fineWinds":true,
I don't think this flag exist for the previous 0.25 races

Post by zezo | 2023-03-29 | 17:28:20

It was present in the previous race too.

Post by Flying Phil | 2023-03-29 | 19:16:48

Oups, i missed it ...

Post by Keskyrest | 2023-05-12 | 16:41:08

Hello all,
On Pro Sailing Tour #1/3, the wind reported by the dashboard, and on zezo are very different: 16.5kt / 241� on dash, 15.5 kt / 242� on zezo.
Do you observe the same?
Boat: Bacemoi
Update time 18h48 (Paris time)

Post by Inicio | 2023-05-12 | 18:50:02

Yes

Not only at the moment, but also in the future predictions.

TWA-HDG are wrong, it seems.

Post by shockwave11 | 2023-05-13 | 01:02:34

This pro sailing race is the worst discrepancy's between the interface, zezo and the dashboard I've ever seen. Pretty much unplayable, it was telling me i was in 10 knots when i was in 4 and it has taken me straight into 1 knot of wind.

Post by Inicio | 2023-05-13 | 09:25:04

There is no problem with VRZEN.
I have put programming with VRZEN and the route has remained perfectly stable.
But Zezo maintains a difference of about 17� with respect to the wind, both the wind of the interface and the server.

Post by Inicio | 2023-05-31 | 08:19:56

The last version has "improved" the GFS 0.25 winds.

Have you any idea how it works this "improvement" ?

Post by Inicio | 2023-06-01 | 17:32:35

📢 VRO update 📢

After some complications with version 5.1.5, we've finally been able to correct the errors and bring you version 5.1.6.
Web link: https://www.virtualregatta.com/fr/offshore-jeu?versionRelease=20230601-691
Available tomorrow at the latest on IOS and Android
We apologize for the inconvenience ⛵

Post by jphe | 2023-06-03 | 08:35:46

Thanks to Flying phil and Cvetan for all this great work.

Only one question remains for me on the duration of the transition. In the example of phil
"from 12h00 UTC, beginning of the integration of the new weather, we use the gribs of 00Z+12h and 06Z+9h,"

But at what time are we on 100% 06Z ?
As advised by Cvetan , i also looked at the list of vr references but for example there is a mix at the moment, {"references":["20230602.18.012","20230603.00.009","20230603.00.012",
maybe it's because the 0.25 is not in use at this moment.

Another question 0.25 gribs are every hour until 120h. I guess interpolation at 12:05 UTC must be between 00Z+12h and 06Z+7h

Post by Flying Phil | 2023-07-02 | 23:34:50

As for the GFS 1�, VR uses the files every 3 hours.
Too bad, we would have had even more finesse in the winds, but VR would have had to load 3 times the amount of data for the first 5 days....