There's this new site called cumul.us (cool pun in the name) that aims to do for weather forecasting what the election futures markets are doing for political forecasting (they've typically done better than individual journalists or networks at predicting election results).
Since a lot of weather information is available freely (it's collected or produced by the US government), the concept is not that far fetched. One of my friends does weather forecasting for energy trading and his constant refrain is that he doesn't have to be right -- he just needs to be better than the models. So, the real question is whether a bunch of amateurs and weather enthusiasts can outperform the models when they (by necessity) will rely on them. One thing that humans (especially crowds) are good at, and that computers are not, is at assessing the probability of different scenarios.
Bottom line: I would expect the site to do somewhat worse than a typical weather forecaster on "normal" days and somewhat better than a typical weather forecaster on harder forecasts. To the meteorologists: feel free to comment.
No comments:
Post a Comment