About 14 months ago, Accuweather extended its long-range forecasting to 25 days. Forecasters at the Washington Post’s Capital Weather Gang and at the independent Phillyweather.net both expressed significant skepticism that any forecast could be accurate at that distance. Tom from Phillyweather.net ran a small sample, confirming his impressions.
About 6 months ago, I also ran a small sample test. I collected 25 days worth of Accuweather’s forecasts for a single future day, comparing them to each other and then to the final weather for that day. The forecasts weren’t that great, as I expected. But again, this wasn’t much of a sample size. I wanted to go bigger. Now, over 6 months after my last post, I now have data comprising all of (astronomical) winter and spring, and it’s time to see the results. Continue reading