In the first blog on this data
http://www.science20.com/virtual_worlds/blog/global_annual_daily_temperatures_19292010-81063
I included a graph that starting in the early 80's has a -10F imbalance, where 10 degrees more heat is lost over night than we received during the previous day.

To help verify this, I decided to review the data on a monthly basis, that way I should be able to detect seasonal changes. But to get the seasons, I had to split the averages up based on location. So I selected 23 degrees Latitude as my dividing line, 23N, 23S, and between 23N/S as the tropics.

Here's Todays and Tomorrows Annual low temp averages for N,S,T where you can see different trends based on Lat.

It's interesting to note that NH temps start rising in the 70's, yet Tropic temps hold flat, and SH temps decline.
Then the NST daily imbalance. Where you can see N/S Hemisphere temps both show the negative imbalance in the early 80's, and the Tropics go negative in ~76.

Something happened to the ISD data in the early 80's, and it looks like it's a real effect in the data.
Comments