This wouldn't really be effective with the general public, but I think it would be beneficial to start looking at above-below average temperatures in terms of standard deviation as opposed to just the boring ole departure.
Hell, I think it would be great to be able to do something based on a historical basis. For example, say the average high temperature at BDL is 52. It would be interesting to see each high temperatures for each March 23 in the database. Judging a 60-65F reading against that average may seem quite impressive, however, if the data is skewed towards that range is it as impressive as it really seems?
also, outliers should be removed from the calculations (I wonder if they are or not).