UK Met Office to Release Data to Amatuers

Sounds odd doesn't it? I wonder what extra data they think people would need to improve their forecasts? There is already a huge array of free data on the internet, models, forecasts from just about everybody and his dog, you name it ... and I would even dare say competition has improved forecasts, if only because the national agencies now realize that they are not a monopoly.

As I've been saying for about five years on various forums, what we really need in the long-range forecasting part of the "business" (emphasis on busy) is an accepted and widely used if not universal validation system so we can get past this stage of rushing about madly from guru to guru without a scorecard. It is sort of like having the Open Championship without keeping score and the pundits saying "Tiger seems to be doing well" or "Padraig is in good form" then picking a winner by consensus of onlookers. JUST SCORE THE FORECASTS LIKE WE DO HERE IN OUR FORECAST CONTESTS ... and perhaps there will be some clarity but the problem is that we need to do this for about three to five years before "form" is established. This idea would also help us determine if various players were improving or not. Something that is not at all apparent despite all the talk is whether there is any actual improvement going on. Despite all the talk about teleconnections and indices, sometimes it strikes me (more appropriate to North America perhaps) that long-range forecasts are in the same general condition as they were thirty years ago. This is a form of heresy I suppose, but even back in the distant pre-internet past, people were making long-range forecasts and sometimes they weren't half bad, despite the almost total lack of awareness of oscillations and index values. It was more along the lines of "big ridge west big trough east" but dammit, if that's the case then you've got the basis for a long-range forecast whether there's sunspots or bubbles of warm air at 78,000 feet or whatever you like.

More data is not really the issue, we need better models. Some people seem to think that better models will flow from more and more and more data, but I remain convinced that you can only expect better models (past three-four days) if you have better theories that can interact with your massive data sets in real time. It's not a case of catching more fish but knowing how to cook fish, if you see what I mean.