MikeBz
Well-Known Member
I watch it a lot and it generally is uncannily accurate even down to short pals and troughs. The only thing they got "wrong" on Saturday was the height of the highest peak but I can happily forgive that.
Surge prediction
- Thread starter MikeBz
- Start date
I watch it a lot and it generally is uncannily accurate even down to short pals and troughs. The only thing they got "wrong" on Saturday was the height of the highest peak but I can happily forgive that.
Bru
Well-Known Member
Meteogroup claim to be involved in the development of and investors in the ECMWF, UKMO, GFS, HIRLAM and WR weather models
Whether they are or not, from a purely personal perspective I consistently find their forecasts more useful and more accurate than the output from the Met Office
Whether they are or not, from a purely personal perspective I consistently find their forecasts more useful and more accurate than the output from the Met Office
Ali-alshira
Well-Known Member
It is valuable to understand what all this modelling involves. The various national met offices generally construct their own models and then which will be most accurate over their own territories and less so for the rest of the world. They will each define a 3-d grid of points covering (often) the whole world, but the grid points will be more closely spaced in their area of interest and more widely spaced further away. At each grid point, various properties will be calculated eg temperature, pressure... . Those will be filled in from observations for a few days in the past, then the big computers will calculate what should happen, step by step into the future using physical laws. Obviously, the finer the grid, the better the local predictions. Why am I telling you all this? Because if you look at a US Model, then their grid will not be as fine over the UK as a UK model. The French one may well be ok over the channel, but take it with a pinch of salt further away from French territory.
Then when you get to the weather sites, you need to look at which model they are using - often two nominally independent sites use the same model (GFS etc), so if they come out with the same prediction, it is not that surprising and does not mean that that prediction is accurate.
My policy is to use the national met office for the country I am in, and compare it with those from neighbouring countries (all using their own independently compiled models) plus maybe another GFS based one. If they all agree, you can be pretty sure they have it right, and if they disagree, you know the situation is unstable. The Norwegian one even tells you how reliable they rate their predictions. Given all that you still have to look out of the window!
Then when you get to the weather sites, you need to look at which model they are using - often two nominally independent sites use the same model (GFS etc), so if they come out with the same prediction, it is not that surprising and does not mean that that prediction is accurate.
My policy is to use the national met office for the country I am in, and compare it with those from neighbouring countries (all using their own independently compiled models) plus maybe another GFS based one. If they all agree, you can be pretty sure they have it right, and if they disagree, you know the situation is unstable. The Norwegian one even tells you how reliable they rate their predictions. Given all that you still have to look out of the window!
Boz
Well-Known Member
I have no beef with that Frank, as a weekend sailor all summer apart from the mid summer cruise, all I want is a reasonably accurate idea on the Wednesday or Thursday of any week, what I can expect for the coming weekend and frequently the predictions from the Met Office are either exaggerated or 12 - 24 hours out, where sites like Windfinder or Windyty seem to offer more accurate analysis.
I think this season I will try to keep a log of predicted versus actual to see if it's not just my prejudices coming to the fore
and that my tax £s are being well spent
franksingleton
Well-Known Member
I have no beef with that Frank, as a weekend sailor all summer apart from the mid summer cruise, all I want is a reasonably accurate idea on the Wednesday or Thursday of any week, what I can expect for the coming weekend and frequently the predictions from the Met Office are either exaggerated or 12 - 24 hours out, where sites like Windfinder or Windyty seem to offer more accurate analysis.
I think this season I will try to keep a log of predicted versus actual to see if it's not just my prejudices coming to the fore
To which I might say - "Read my book." My own way is to use the US GFS - zyGrib or one of the various apps. I look at the forecast every day when cruising for 10 days ahead. Not because I have any faith in a 10 day forecast but to look for consistency in successive forecasts. Quite often, certainly not always, we can plan about a week ahead. When there is inconsistency, then the situation is uncertain and we may be only planning a couple of days ahead.
Last year returning from St Peter Port to the Dart, we knew with some certainty when we would leave St PP.
For today, I rely heavily on the GMDSS texts and use the GFS to help interpretation. I use the GFS because it is available and as good as any current model over the next few days.
Last edited:
franksingleton
Well-Known Member
Perhaps I can add a little to your description of NWP.
I do not think that many modellers run a variable grid as you describe. The French do but the UK, US, the HIRLAM countries all use a global model with a grid of around 15 km (13 for the GFA and 17 for the UK). They then use the vales from these to provide boundary conditions for smaller areas. Take a look at http://www.metoffice.gov.uk/research/modelling-systems/unified-model/weather-forecasting. This shows how the UK runs a global model, and uses that to run a 4 km model from mid-Atlantic to the Urals. That is then used to run a smaller grid just for the UK.
Some notes of caution. Fine scale models require fine scale data. In situ observations only scratch the surface. You need satellite and radar data to get down to the very small scale of a km or so. I am not aware of any private company that can handle those data. Most simply interpolate from the GFS 0.25 d3gree (25 km) output. A grid can only represent shapes on a scale of around 4 grid lengths. So, they only start with a crude analysis. They can model l topography well nut results are always likely to be compromised by lack of weather detail. Not quite GIGO.
The other note of caution is that small weather details have short lifetimes. Something the size of a large thunderstorm cloud may only have a 6 hour life span. Something around a km or so will last less than an hour. This determines theoretical predictability. If we had a perfect weather model and knew exactly what was happening on a 1 km grid then the detail shown on that scale would have disappeared before a forecast model could be run.
Apologies for the length but predictability is something not well understood by many sailors, nor by some of those who run models and make claims about precise forecasts. Weather is not precise. Forecasts cannot be.
For anyone interested in looking at different models, Meteociel,fr is worth a few minutes’ study.
PS. To see how it all started and got to where we are now, see http://www.metoffice.gov.uk/researc...umerical-weather-prediction-at-the-met-office.
Last edited:
Ali-alshira
Well-Known Member
My comments on Frank's post have been mixed up with the quote - if you are interested, untangle them!
franksingleton
Well-Known Member
Sorry, I do not follow. In my post I was trying to expand on your description of NWP which was considerably better than many that I have seen.
Ali-alshira
Well-Known Member
I think we both agree with each other!
I was trying to explain why relying on just one model of weather, especially one not really designed for your area, can be problematic. Your point about the temporal resolution was nicely explained.
franksingleton
Well-Known Member
Basically, you are correct. I would expect the UK detailed model to out-perform, say, the WRF for the UK simply because there will have been some optimisation to maximise results. Further the UK Met Office will be using more detailed data. For that reason, I would always back the local National Met service against the rest. They will not always do better but should do so overall.
What I am always trying to argue against is when I hear people saying that they prefer XCWeather or Passageweather or whatever. Almost all GRIB data available to us come from the GFS. When it comes to genuinely different global models, GFS, NAVGEM, CES, JMA, UK etc, my gut feeling is that differences on specific occasions are more a question of different data analyses rather than model differences. There is no unique way of analysing the data, especially satellite data. Small differences in data analysis may give big differences in outcome. That is why they run ensembles of many forecasts using slightly different analyses of the same data.
Ali-alshira
Well-Known Member
yes, agreed!