Do you trust GFS and RTOFS models?

[psousa]

Hi!

In my particular situation, I get them from Saildocs but I tried Windyty too and it's similar so, I guess, the source is not relevant for my question.

Anyway, today I was checking the gribs for Biscay and I found a big difference for the wave size when compared with a reading from a ship (from http://www.ndbc.noaa.gov/radial_search.php?lat1=45.230N&lon1=5.000W&uom=E&dist=250).

So, at 9am, the ship, located at 44.10,-4.30, has reported a WVHT reading of 9,8ft (~3mt).
For the same time and position and talking about prediction (models are all about predictions, I guess), Windyty using "ECMWF WAM" model shows 1,5mt, Windyty using 'Wavewatch 3' shows 1,3mt. Models from Saildocs shows 1.3mts too.

Anyone knows how do these ships reads this? I know the models it's a average of the higher 1/3 of the waves or something like that, right? Could this ship reading to be a pontual max reading instead of an average?

Thank you!

 

[lpdsn]

I tend to rely on the GFS 0.25 with all the usual caveats regarding using models, especially assessing it's consistency over several runs for a particular time and place.

I use the wave models less. I do tend to look only at the Irish HIRLAM output and the wave heights it gives seem to be reasonably reliable.

In your example I would expect the equipment on the ship and the officers to be fully aware of the concept of significant wave height, so sounds like it's a bit choppier than forecast.

 

[AntarcticPilot]

Wave height is very difficult a) to specify and b) to measure. The usual measure obtained from satellite observations on a worldwide basis is Significant Wave Height (SWH), which is these days defined as 4 times the standard deviation of the surface elevation of the water. It was chosen because it correlates pretty well with wave-heights as recorded by experienced ship observers, and is reasonably easy to extract from satellite radar altimeter data. However, the instantaneous height of the sea surface is obviously difficult to observe, and there will always be a substantial number of waves considerably larger than the SWH. Satellite altimeters do provide an instantaneous measure of the distribution of wave-heights within the footprint of the altimeter, so the values they measure are pretty "accurate". But the other thing is that the distribution of heights is not a normal distribution, and the shape of the distribution can change quite a lot from place to place and time to time. This is especially noticeable in coastal regions, where the water depth, diffraction round headlands and other similart effects can modify the wave pattern substantially. For example, the North Sea has generally shorter, steeper waves than the west coast, because the water depths are small enough to inhibit the very long wavelength oceanic swell.

 

[psousa]

lpdsn, can you pls post a link to Irish HIRLAM output? ty

AntarcticPilot, thank you very much for your valuable explanation!

 

[lpdsn]

It's just on the Met Eireann website.

http://www.met.ie/forecasts/5day-ireland.asp

The wave tab shows the forecast sea state around Irish coasts.

From what I understand from posts made by FrankSingleton, and there's no guarantee I've understood correctly, the HIRLAM model is used by a number of smaller European countries to produce enhanced local area forecasts based upon input from global models (ECMWF I presume).

 

[KellysEye]

Long distance sailing we used UGRIB http://62.148.188.51/ which uses GFS, no forecast is 100% accurate but it was good enough to use, having tried others. In Emsworth the most inaccurate is the Met Office, Wndfinder is the most accurate.

 

[franksingleton]

As far as I know -


Voluntary observing ship reports of wave height, period and direction are visual estimates by the officer of the watch. An experienced bridge officer can distinguish up to three wave trains.


Data buoys and light vessels use a frequency analysis of the movement of the observing platform. They cannot give directions.


For satellite measurements, see https://www.tandfonline.com/doi/abs/10.1080/01431161.2012.685975. I do not know how well satellites actually measure wave heights. I suspect that much depends on empirical relationships between the scattering of a radar signal and in situ observation. Satellites provide surface wind data by measuring scattering at the molecular level using two beams at right angles.


Computer predictions use data based on the energy imparted to the sea by the predicted winds. Swell should be pretty well forecast. I think that there will be problems in calculating the local wind sea and adding this to the swell. There will be errors near the coast because of bottom topography effects and in areas of strong tidal stream.

PS. See https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2003GL017743. This emphasises that satellite radar altimetry is not as accurate as you might suppose. Like the technique described above it also depends on relating the scattering of a radar beam to in situ data. Although the name might imply radar type accuracy it does, in fact, depend on ststisticsl relationships with “real” data.

 

[AntarcticPilot]

As far as I know -


Voluntary observing ship reports of wave height, period and direction are visual estimates by the officer of the watch. An experienced bridge officer can distinguish up to three wave trains.


Data buoys and light vessels use a frequency analysis of the movement of the observing platform. They cannot give directions.


For satellite measurements, see https://www.tandfonline.com/doi/abs/10.1080/01431161.2012.685975. I do not know how well satellites actually measure wave heights. I suspect that much depends on empirical relationships between the scattering of a radar signal and in situ observation. Satellites provide surface wind data by measuring scattering at the molecular level using two beams at right angles.


Computer predictions use data based on the energy imparted to the sea by the predicted winds. Swell should be pretty well forecast. I think that there will be problems in calculating the local wind sea and adding this to the swell. There will be errors near the coast because of bottom topography effects and in areas of strong tidal stream.

Frank,
Satellite measurement of SWH is very well understood and quantified; I happened to be part of the team that developed the algorithms for use in processing data from ERS-1! It is a very good measure of the statistical distribution of the instantaneous surface elevation within the radar footprint, which of course is what SWH is. My particular work was with the measurement of ice surfaces, but I worked closely with oceanographers who defined the SWH algorithms. I can send you references to the massive tomes we produced, if you like. SWH was the statistic chosen because it correlates well with ship-borne observers.
Paul

 

[franksingleton]

It's just on the Met Eireann website.

http://www.met.ie/forecasts/5day-ireland.asp

The wave tab shows the forecast sea state around Irish coasts.

From what I understand from posts made by FrankSingleton, and there's no guarantee I've understood correctly, the HIRLAM model is used by a number of smaller European countries to produce enhanced local area forecasts based upon input from global models (ECMWF I presume).

Yes. The HIRLAM countries have developed a meso-scale model that uses ECMWF for boundary and initial conditions. The latter are merged with detailed data from satellites, perhaps radar, and such in situ data as are available. The analyses are 4-dimensional so that they use far more observational data than any “non-official” meso-scale models (LAMs) of which I am aware. Their use of detailed data more than counters the fact that ECMWF only run their model at 12 hour intervals and with later run/output times than the 6-hourly GFS upon which most of these other models depend.

I do not know to what extent the meso-scale model impacts on wave modelling. Presumably, it is more critical for the wind-wave calculations than the swell, the latter being generated over longer time periods and usually at greater distances than covered by a LAM.

 

[franksingleton]

Frank,
Satellite measurement of SWH is very well understood and quantified; I happened to be part of the team that developed the algorithms for use in processing data from ERS-1! It is a very good measure of the statistical distribution of the instantaneous surface elevation within the radar footprint, which of course is what SWH is. My particular work was with the measurement of ice surfaces, but I worked closely with oceanographers who defined the SWH algorithms. I can send you references to the massive tomes we produced, if you like. SWH was the statistic chosen because it correlates well with ship-borne observers.
Paul

Thanks for the clarification. I understand the principles but, fairly obviously, am not familiar with the detail. I was really trying to clarify the fact that, in general, satellites do not measure the atmosphere or the ocean directly. They measure effects (scattering, refraction, absorbtion) that depend on the state of the environment from which environmental data are deduced or inferred. The relationships may be very good (as you say for wave data) or not very precise (atmospheric temperature profiles) but still inferred.

 

[AntarcticPilot]

Thanks for the clarification. I understand the principles but, fairly obviously, am not familiar with the detail. I was really trying to clarify the fact that, in general, satellites do not measure the atmosphere or the ocean directly. They measure effects (scattering, refraction, absorbtion) that depend on the state of the environment from which environmental data are deduced or inferred. The relationships may be very good (as you say for wave data) or not very precise (atmospheric temperature profiles) but still inferred.

In fact, in this case, it is a very direct measure. Basically, the RA records the power returned to the radar in successive time (=range) bins. The power returned into a particular range bin is directly proportional to the proportion of the surface at that distance from the altimeter. There is a geometric effect - if the surface was perfectly smooth, the range bins would correspond to circular zones surrounding the sub-surface point - but this can be taken into account. The slope of the leading edge of the returned pulse turns out to be a direct measure of SWH, as defined as a statistic of the distribution of surface elevations. As the RA has a "footprint" of around 30km, it gets a good statistical sample.

The real problem is that the actual statistics of sea surfaces don't always correspond very well to the ideal ocean surface that is assumed! In particular, ISTR that there is recent(ish) research showing that extremely large waves have a higher probability than previously thought; i.e. the statistical "tail" is both higher and longer than depicted in the usual distributions.

These references include considerable reference to the problem of sea surfaces; the ice work is a development of the same techniques. Sea surfaces are much simpler!

• Rapley, C. G., H. D. Griffiths, V. A. Squire, M. Lefebvre, A. R. Birks, A. C. Brenner, C. Brossier, L. D. Clifford, A. P. R. Cooper, A. M. Cowan, D. J. Drewry, M. R. Gorman, H. E. Huckle, P. A. Lamb, T. V. Martin, N. F. McIntyre, K. Milne, E. Novotny, G. E. Peckham, C. Schgounn, R. F. Scott, R. H. Thomas and J. F. Vesecky, A study of satellite radar altimeter operations over ice-covered surfaces, ESA Report 5182/82/F/CG(SC), European Space Agency, Holmbury St. Mary, 1983.
• Drewry, D. J., N. F. McIntyre, A. P. R. Cooper and E. Novotny, Modelling the surface of the Antarctic ice sheet for satellite radar altimeter studies, Annals of Glaciology, 5, 202 (abstract), 1983.
• Novotny, E., A. P. R. Cooper and M. R. Gorman, Simulation of waveforms returned by realistic surfaces, in Proceedings of a Workshop on ERS-1 Radar Altimeter Data Products,Frascati, Italy, 8-11 May,, ESA SP-221, pp. 171-172, European Space Agency, Frascati, 1984.

• Drewry, D. J., N. F. McIntyre and A. P. R. Cooper, The Antarctic ice sheet: a surface model for satellite altimeter studies, in The Binghampton symposia in geomorphology: International series, 14: Models in Geomorphology, edited by M. J. Woldenburg, pp. 1-23, Allen and Unwin, London, 1985.
• Rapley, C. G., N. F. McIntyre, R. F. Scott, A. R. Birks, A. P. R. Cooper, E. Novotny and H. D. Griffiths, ERS-1 altimeter tracker performance simulation, in Proceedings of the EARSel conference, Strasbourg,,, 1985.
• Rapley, C. G., H. D. Griffiths, V. A. Squire, J. G. Olliver, A. R. Birks, A. P. R. Cooper, A. M. Cowan, D. J. Drewry, M. R. Gorman, M. Guzkowska, S. Laxon, I. M. Mason, N. F. McIntyre, E. Novotny, R. Paterson, R. F. Scott and F. A. Street-Perrot, Applications and scientific uses of ERS-1 radar altimeter data, ESA Report 5684/83/NL/BI, European Space Agency, Holmbury St. Mary, 1985.
• Novotny, E., and A. P. R. Cooper, Simulated waveforms in radar altimetry as an aid in determining terrain characteristics, in Proceedings of the poster sessions at the International conference of the Remote Sensing Society and the Centre for Earth Resource Management, 10-12 September 1985, pp. 199-207, Remote Sensing Society, 1985.

 

[franksingleton]

Paul, thank you for that. I think that it confirms what is well understood by those working in the field but far less well by the end users. Particularly, that it is extremely difficult/impossible to relate satellite areal observations to specific in situ data.

The altimetry data provide a distribution of wave height over an area of size 30 km. Ships and data buoys provide data at specific locations. Similarly, NWP model wave predictions are based on calculations of energy input. They also really give a distribution of wave height over an area rather than specific point values although the GRIB output is given as point values.

This is a general problem in weather prediction. A global model can only provide information on a 50-60 km resolution. Even meso-scale predictions can only represent detail with a 20-30 km resolution at best. Inevitably, sailors, and the general public at large, experience weather at specific locations. They know about and experience variations in space and time but still comment on or criticise differences between model grid point values or human interpretations of those data. Many of my posts have been aimed at- this particular dichotomy.

 

[AntarcticPilot]

Paul, thank you for that. I think that it confirms what is well understood by those working in the field but far less well by the end users. Particularly, that it is extremely difficult/impossible to relate satellite areal observations to specific in situ data.

The altimetry data provide a distribution of wave height over an area of size 30 km. Ships and data buoys provide data at specific locations. Similarly, NWP model wave predictions are based on calculations of energy input. They also really give a distribution of wave height over an area rather than specific point values although the GRIB output is given as point values.

This is a general problem in weather prediction. A global model can only provide information on a 50-60 km resolution. Even meso-scale predictions can only represent detail with a 20-30 km resolution at best. Inevitably, sailors, and the general public at large, experience weather at specific locations. They know about and experience variations in space and time but still comment on or criticise differences between model grid point values or human interpretations of those data. Many of my posts have been aimed at- this particular dichotomy.

As you say, the discrepancy between area based measurements and predictions and local experience will always be present, and is unavoidable. Indeed, it can be worse, as the SWH measurement itself shows. We take measurements that reduce a complex phenomenon to a single number, and that number may not express a value that is actually meaningful to a particular user. For example, the SWH will include long wavelength swells that may actually be imperceptible to a yacht, so the SWH gives a higher wave height than the yacht experiences. But short wavelength waves of a metre or so amplitude, which make a very uncomfortable chop for a yacht, will only appear as an SWH of a metre - not very much!

There is also the unfortunate fact that some measurements HAVE to be over large areas - passive microwave temperatures being one; the physics of the measurement means that the measurement has to be taken over large areas, as small areas simply emit too few photons at the relevant energy levels for meaningful observations to be made over the time-scale imposed by the passage of a satellite! And while sea surface elevation can be measured with smaller spot sizes by laser altimetry, the statistics underpinning the measurements become less secure.