On my eternal quest for problems to solve in order to make this world more awesome I realized one big one has been right in front of me for a while: weather forecasts. A commonly heard complaint these days is that weather forecasts are absolutely useless, never get it right and are a waste of time. Even so, people keep watching them.
I am no meteorologist, so I don’t mean to improve on the forecasts. I think it may be interesting to look into the actual accuracy. I don’t have a lot of experience with statistical analysis either, but collecting some data and graphing it should be doable.
The plan
For a year I will collect forecast data for 1, 3 and 7 days ahead. At the same time I will collect weather reports (that is, actual measurements). Multiple locations and weather services may be used, I’m not sure. At the end of that year all the data can be compared to eachother and conclusions may be drawn.
I imagine this has been done before, but I’d really like to see the data for myself.
The data
The only thing I’m pretty sure about is that I’ll draw report data from METAR. I don’t know where I’ll be getting forecast data, but preferably from two or three easily parseable sources.
Update 15-2-2012: Finding a source for forecasts that is easily parsable seemed impossible, until someone came up with the bright idea to look into RSS feeds. Yahoo’s weather service seems excellent, just have to find something that’s published by something that’s actually in the Netherlands. Might be a while until I get around to this, though.