Support Sailonline

If you haven't already - join the SAILONLINE YACHT CLUB!

Please also consider making a donation - all amounts are greatly appreciated!

Board » Technical Support » NOAA vs SOL weather

Hi all,
I have noticed a difference between TWD/TWS as calculated from NOAA grib data and the ones shown in SOL. Would it be possible to know the way SOL interpolates in space and time wind speed and direction?
Thanks in advance and regards
would love to hear that!

--- Last Edited by Ghibli at 2010-12-03 08:50:51 ---

So, about the interpolation of weather in time and space :-) Maybe some out there is longing for a short one-liner as explanation but I am afraid there is a little more to it… Here is in short the steps we take to get an "observation" (as we call it) at an arbitrary time (called t) and location based on forecasts.

1) Find the two forecast time steps (here called t1 and t2) surrounding the time t.
2) In space we simply use linear interpolation (in the 2 dimensions lat/long) of the forecast-data in the 4 closest forecast-nodes in the grib-file. Since the forecast has separate fields, one for northerly wind and one for easily wind we need to treat these entities separately, in our case by using complex notation. We also need to do this for both frames t1 and t2 from step 1 which thus gives us a set of two complex-valued space-interpolated wind speeds, one at t1 and one at t2.
3) Now we need to interpolate in time as smoothly as we can. We chose to do this using simple shape functions ensuring continuous wind speed and actually in our case continuous time derivatives.

Now, there are naturally many ways of doing this. One draw-back of this kind of interpolation is that we loose energy. Imagine the situation with constant wind speed but rotating wind direction so that at t1 the wind is due easily and at t2 it is due westly. In our interpolation we would (falsely) get that the wind speed at the time (t1+t2)/2 would be zero.

Also note that we do not always use the GFS model from NOAA. In particular around Sweden we have more detailed forecasts.

Hope this helped a little.

...one of the guys behind the game...
Hi Jakob,
to keep in one line, the model smooth and make continue the derivative of a complex 3 variable function to avoid that the function has some jump. [ actually more than one line ;-) ]
need to think about it deeper, may be come back later.
thanks for moment
Would it be possible to differentiate between those wind arrows (highlighted color perhaps) that represent the actual data nodes & those arrows that are derivative which comprise the field.
None so blind
Nice topic indeed.

So, in my own words: two frames of weather grib are interpolated in space using bilinear. This operation gives two different values for the wind vector which are separated in time by the gap between the two frames (3h with GFS or whatever).

Now, in order to have a smooth transition between frames a simple linear approximation is not good because it will give discontinous values of first derivative corresponding to the frames change. And so the need for a proper interpolation function that can smooth the derivative at the beginning and the end of the interval.

One can argue that the only possible value of the derivative at these point is zero. But another important aspect is that, given the shape function, we have steeper gradients in time in order to compensate for the nulling of the derivative at the extremes. Am I right? And this is extremely dependent by the shape of the interpolating function.

my 2 cents
Hi all,

You are basically all correct :-) I am the first to admit that there are many ways of doing this and that we did not spend weeks of thinking when choosing algorithm - but it works. Also (Hubert) the game winds will only correspond exactly with the GFS-forecasts when the time is exactly the time of a certain forecast frame and the point of the "observation" is exactly that of a forecast nodal point. Thus, colour coding would not be of any use at all...

Why this attention? Is it just curiosity or weather-rounting ?

...one of the guys behind the game...
it curiosity driven by weather routing :)

as you remember I was coding my own router and, although I did it for pure fun, I am now checking it against some of the market available competition like maxsea or deckman. All of them share the common feeature of giving different wind values at the same place, same time and same wind data!! So the interpolation is critical and although I can rely on more elaborate intepolation technique, I will definite give a look to this zero-derivative approach.
...well you should be careful with those softwares... especially when there is a Sailplanner available for free ;-)
...one of the guys behind the game...
well... I need to admit that the question from my side was originated because I have notice some differences... and I am using SOL to prepare my sailing licence exams from navigation point of view, but you pass me by far in the discussion...
I appreciate your effort to educate the SOL community and I believe that I must open the old book to follow you now...... thanks!
one of the guys enjoing the game ;-)

Please login to post a reply.


Next Race: 00d 00h 00m

Current Races:

SOL's Bermuda Triangle TIMED Race 2023
Inspired by the America’s Cup racing in Bermuda, this is a fun triangular course of 90nm, which we originally raced with psail and bonkhoot’s collaborative polar for the AC72 foiling catamaran. Having tried the course in the more pedestrian Open 60 in 2021, this year we return to the AC72, in fact to the MODIFIED AC72v2, modified to remove the inconsistencies that obtained from the original interpolation of the limited data issued by the AC 2003 Principal Race Officer in connection with penalties determination. This is a TIMED race, which means you get multiple chances to try out the ramifications of the revisions. Be careful though, the polar remains a handling challenge with major stall-out potential!
RE-REGISTER HERE to race again after finishing a run
Race #1704
INFO by brainaid.de
WX Updates:
0430 / 1030 / 1630 / 2230
15 December at 2300 UTC
Race starts: Dec 02nd 12:00 Registration Open!

▶ Flash

Christmas to Christmas Island 2023
Two waypoints, start and finish, and between them 5827 NM of two oceans. Not much time if we want to sit at the Christmas Eve table. You will need to maintain an average speed of over 10 knots, so prepare your Volvo70s well and choose the best of the many possible routes. Hands up everyone who knew there were TWO Christmas Islands?! OK, maybe you have sailed this Sailonline course before... but it's time to get ready for the 2023 challenge of racing between Christmas Island in the Pacific to Christmas Island in the Indian Ocean! It also marks the conclusion of our prestigious Ocean Race Championship 2023. Please have fun! Fair winds!
Race #1745
INFO by brainaid.de
WX Updates:
0430 / 1030 / 1630 / 2230
Ranking: OCQ4 - OCCH - SUPSOL - SYC
Race starts: Dec 01st 23:00 Registration Open!

▶ Flash

Go to race archive

SYC Ranking

  1. Sailonline Yacht Club Member WRmirekd
  2. Sailonline Yacht Club Member Patrick70119
  3. Sailonline Yacht Club Member Pit8008
  4. Sailonline Yacht Club Member FreyjaUSA
  5. Sailonline Yacht Club Member Kipper1258
  6. Sailonline Yacht Club Member Sax747
  7. Sailonline Yacht Club Member rafa
  8. Sailonline Yacht Club Member Vida_Maldita
  9. Sailonline Yacht Club Member SlideRule
  10. Sailonline Yacht Club Member HappyHour

View full list


Mobile Client

SYC members have the benefit of access to our mobile/lightweight web client!

The mobile client