ECMWF vs GFS - over the top gust predictions?

franksingleton

Well-known member
Joined
27 Oct 2002
Messages
3,383
Location
UK when not sailing
weather.mailasail.com
I just downloaded this >
send GFS:48N,47N,20W,19W|0.01,0.01|0,12,24..72|=
WIND,PRESS,APCP,HTSGW,WVPER,WVDIR,GUST

So should be 0.01Deg resolution?

Had to shrink the area otherwise it exceeds the email file size.

about 0.6Nm spacing in Lat between wind barbs in Opencpn. So doesn't look like the 0.11deg is actual model run data - could be interpolation by saildocs?
As I said at #38, I think this is a quirk of Saildocs. I will try to come back on this one. I am sure that NOAA only issues GRIB at 0.25 degree res.
 

franksingleton

Well-known member
Joined
27 Oct 2002
Messages
3,383
Location
UK when not sailing
weather.mailasail.com
As I said at #38, I think this is a quirk of Saildocs. I will try to come back on this one. I am sure that NOAA only issues GRIB at 0.25 degree res.
It appears that saildocs remaps GRIBs according to the request. If no resolution is stated it chooses a resolution depending on the rectangle requested. try “send gfs:48N,50N,2W,1E“. You will get 1.5 deg resolution.

I requested 0.01 degree resolution and got it (more or less) - try it.
It is only repackaging and mashing that is going on, and not an output of GFS at a finer resolution than 0.25 deg.
 
  • Like
Reactions: GHA

franksingleton

Well-known member
Joined
27 Oct 2002
Messages
3,383
Location
UK when not sailing
weather.mailasail.com
So other than a busier display there's no point downloading at a higher resolution than 0.25Deg?
No point whatsoever. I use XyGrib with values only at the published grid points. I can ask it to interpolate but prefer not to do so. Using PocketGrib, it automatically interpolates as do Ventusky and Windy. Also, you have bigger downloads for the same information. A no brainer.
 
  • Like
Reactions: GHA

GHA

Well-known member
Joined
26 Jun 2013
Messages
12,104
Location
Hopefully somewhere warm
Visit site
No point whatsoever. I use XyGrib with values only at the published grid points. I can ask it to interpolate but prefer not to do so. Using PocketGrib, it automatically interpolates as do Ventusky and Windy. Also, you have bigger downloads for the same information. A no brainer.
I've a little bit of computer code which downloads an area to dropbox 4 times a day, just checked, it's set to 0.5deg, plenty good enough :cool:
 

Roberto

Well-known member
Joined
20 Jul 2001
Messages
5,031
Location
Lorient/Paris
sybrancaleone.blogspot.com
I just downloaded this >
send GFS:48N,47N,20W,19W|0.01,0.01|0,12,24..72|=
WIND,PRESS,APCP,HTSGW,WVPER,WVDIR,GUST

So should be 0.01Deg resolution?

Had to shrink the area otherwise it exceeds the email file size.

about 0.6Nm spacing in Lat between wind barbs in Opencpn. So doesn't look like the 0.11deg is actual model run data - could be interpolation by saildocs?

That's interesting :)
Using these things mostly from the sea (and very limited bandwidth), I never delved too much in such high resolutions from GFS, file sizes are definitely too big. It might indeed be some further interpolation from saildocs or the viewers. I tried looking at GFS 0.01° grids and values are often interpolated all along the intervals. Saildocs queries with the ECMWF, RTOFS etc are also interpolated.
Oddily, the same Saildocs query with 0.01 grid spacing but asking data from COAMPS model for example only gives data to 0.2°, with no interpolation.

A hypothesis, GFS has (better had, I just read 2017/8 they went into a different Fixed-Volume structure which I do not know at all) a T1534 spectral truncation, meaning the model physics are solved using 1534 spectral components, which is more or less equivalent to a 12km grid at the equator. The model output is then post processed into the variables we use, pressure Uwind Vwind etc etc: I wonder if the post processor does not create 0.12 gribs at all, but only the available 0.25°, in which case they would not be available in any provider, PreditcWind included. Just a thought, I do not know exactly what has been changed with the latest GFS.
 

franksingleton

Well-known member
Joined
27 Oct 2002
Messages
3,383
Location
UK when not sailing
weather.mailasail.com
That's interesting :)
Using these things mostly from the sea (and very limited bandwidth), I never delved too much in such high resolutions from GFS, file sizes are definitely too big. It might indeed be some further interpolation from saildocs or the viewers. I tried looking at GFS 0.01° grids and values are often interpolated all along the intervals. Saildocs queries with the ECMWF, RTOFS etc are also interpolated.
Oddily, the same Saildocs query with 0.01 grid spacing but asking data from COAMPS model for example only gives data to 0.2°, with no interpolation.

A hypothesis, GFS has (better had, I just read 2017/8 they went into a different Fixed-Volume structure which I do not know at all) a T1534 spectral truncation, meaning the model physics are solved using 1534 spectral components, which is more or less equivalent to a 12km grid at the equator. The model output is then post processed into the variables we use, pressure Uwind Vwind etc etc: I wonder if the post processor does not create 0.12 gribs at all, but only the available 0.25°, in which case they would not be available in any provider, PreditcWind included. Just a thought, I do not know exactly what has been changed with the latest GFS.
The GFS upgrade in June 2019 had several aspects including using some newer data sources from satellites. I am surprised that it has taken them so long to incorporate some of these.

I see no great point in us, users, getting too deep into the details, the finer aspects are well beyond me now. All we really need to know is what models provide. Although the GFS is a spectral model, https://www.weather.gov/media/notification/scn19-40gfs_v15_1.pdf says that “GFS Version 15.1 maintains a horizontal resolution of 13 km and has 64 levels in the vertical extending up to 0.2 hPa.” They issue GRIB output on a 25 km grid, ie 0.25 deg lat/long.

Whatever individual providers or viewer software do the fact remains that the data we receive can only be at a 25 km res. Attempts to derive higher res information are a total waste of time and unnecessary in terms of download sizes. No doubt some clever soul could hack into NOAA and get the 13km, 0.125 degree grid data but, until and unless that happens, the GFS provides 25 km grid data. I do not know what res ECMWF provides as used by Windy and PredictWind. Can a PW user please tell me at what res does PW provide ECMWF and GFS data.

PS. In the above, I assume that the GFS does archive GRIB dara on a 0.125 degree grid. That is how they store data products.
 
Last edited:

miniPiddy

New member
Joined
7 Oct 2020
Messages
16
Visit site
For ECMWF gusts, this is a known issue in under-predicting: Known IFS forecasting issues - Forecast User - ECMWF Confluence Wiki

"Although there is a helpful convective contribution in the computation of maximum gusts (as used in direct model output and the EFI), experience has shown that extreme gusts are generally under-represented, particularly when vigorous convection is involved, such as one might see with MCSs or squall lines - eg 60kt gusts or more might be observed when 30-40kt gusts are predicted. This relates to (i) an inability, at the current model resolution, to represent the 3-d circulation around convective systems, and (ii) the fact that it is impossible to design an adjustment in the gust computation that will work in all cases."
 

franksingleton

Well-known member
Joined
27 Oct 2002
Messages
3,383
Location
UK when not sailing
weather.mailasail.com
For ECMWF gusts, this is a known issue in under-predicting: Known IFS forecasting issues - Forecast User - ECMWF Confluence Wiki

"Although there is a helpful convective contribution in the computation of maximum gusts (as used in direct model output and the EFI), experience has shown that extreme gusts are generally under-represented, particularly when vigorous convection is involved, such as one might see with MCSs or squall lines - eg 60kt gusts or more might be observed when 30-40kt gusts are predicted. This relates to (i) an inability, at the current model resolution, to represent the 3-d circulation around convective systems, and (ii) the fact that it is impossible to design an adjustment in the gust computation that will work in all cases."
Agree totally. A 10 km grid is about the upper limit for use of non hydrostatic models. The surface - 10 m wind, itself, is derived using an algorithm from the bottom model level - 0.998 x surface pressure. Gust algorithms, likewise.
 
Top