To the best of my knowledge, few places have tracked temperature longer than the last 150 years and beyond that we have estimates and calculations that have variances/tolerances built in. Is 150 data points for a given location enough to KNOW both the "correct temperature" and that it does/doesn't move?
Most people never understand this aspect of statistics. Living in Houston, and having survived 3 out of 4 years with an above-100-year floods, I understand firsthand the clumping of events. But the "typical" guy on the street (oops, now I just violated my own rule!) has no clue what a distribution is.
Not sure how to solve this problem. Teaching statistics to everyone in not a solution (see the Monty Hall Problem [0] ).
1) Scientists predicting the probability of an earthquake in a region that just had one. "If it doesn't occur, they'll say it was only a probability, and if it does, they'll claim they predicted it! Stop wasting taxpayer's money on this nonsense!"
2) "My mother chain smoked all her life and she's in her 90's. So smoking doesn't cause cancer."
So I don't see what the problem is with charting, say, the 98% confidence interval, calling it normal, and stating that values outside that range are 1-in-50 year events.