The weather was silly yesterday. Weak fronts, a bunch of troughs, layers, convection on top of it all. Flip a coin.
Today was a lot clearer situation and the forecast was quite good.
Is the European forecasting service similar to the US forecasting system where it is publicly funded? I guess Im asking because it would be really expensive for any one private company to fund such a program.
PetitCessnaVoyageur wrote:
Do you consider the major update of last week, a significant improvement step in forecast accuracy ?
For all practical purposes it will have to run for at least 2-3 years before it becomes clear whether it has been an improvement or not. That goes for all the models.
Usually the circle is the same: people complain about a shortcoming and that part is tweaked or reworked. That on the other hand has influences over the whole behaviour of the model and may damage something which worked previously. A model which underestimates precipitation will be tweaked to show more precipitation and once it does, shows that situation correct but is worse in others previously correct. Model science is a real headache generator for those who work there and they are real “rocket scientists” so to speak, in most cases don’t even try to comprehend where they are coming from.
GFS does a fantastic job for what it is, but it has shortcomings, some of which are by nature of it’s grid witdh and data generation, others are because the developers have not found a reasonable way to fix it. I am not sure about the latest version but I remember reading that e.g. mountain chains are massively simplified to the point that if you were to plot them on a map, they would in some parts not even show. Again, this has to do with the grid width. GFS is a model which is designed to cover the whole globe at a manageable resolution. It is not designed for and not capable of calculating local annomalies.
Some of the highly specialized models such as Cosmo have partially ultra high orographic resolutions, which then accounts for other errors until they get ironed out. It was funny to see how e.g. the older generation Cosmo model with the wider grid constantly underestimated rainfall whereas the newer 1km grid one overestimates. The reason probably is that the 1km grid has much higher terrain resolution than the 7km one, amongst other things. It can do a lot of things GFS was never meant to, but it has other shortcomings which will cause a lot of head scratching in their development teams.
Doesn’t sound like a huge improvement however:
Still GFS in my experience is better that whatever lies behind Windy. Last weekend Windy showed no cloud tops above 8000 ft or so while actual tops were above FL200 with icing and all other stuff expected within occluded front. GFS was almost 100% accurate.
I think the best model Windy.com offers is the ECMWF model but it isn’t clear where its (Windy’s) cloud data comes from. It uses the same low/med/high cloud classification as GFS and that looks like it will always be a compromise, because clouds don’t form in three convenient layers. There are indeed the three broad processes AIUI but the reality is a fully 3D picture.
Windy is merely a visualization tool, by default it will give you ECMWF but has GFS raw output data as option
Still, it does bit of interpolation and smoothing of raw output data in space and time plus some weird substitutions/formulas (e.g. Peter’s example, there are no raw cloud/rainfall data out of ECMWF…)
If people are interested in the differences between US and European weather forecasts, they should get to see the fantastic play Pressure, written and acted by David Haig.
Timothy wrote:
should get to see the fantastic play Pressure, written and acted by David Haig.
Where?
