Anyone ever read Nate Silver's The Signal and the Noise? He has a chapter on weather forecasting which is quite interesting. But the most interesting part of it for me was that most local weather reports purposefully overstate the chance of inclement weather as people are more likely to remember times when the report was for good weather and they got bad (ruining their picnic or whatever) than the opposite. So in order to make people feel more confident in the weather predictions, they purposefully give worse predictions. Aren't humans great!
I really enjoyed this book. I'm glad that fivethirtyeight is starting up again this year.
Edit: It was more than that, jonts. The finding was the local weather reports overstated the chance of inclement weather
when the forecasted chance was low. If the models predicted a 70% chance, they tended to report a 70% chance. But when the model predicted a 10% chance, they inflated it. Viewers don't internalize probabilities well. If they report a 10% chance of rain, viewers will think "the weather guy says it won't rain". If they report a 70% chance of rain, viewers will think "the weather guy says it will rain". If the weather model forecasts a 10% chance of rain, then 1 out of 10 days it really should rain.