Possum Comitatus has noticed a very interesting change in Roy Spencer’s presentation of his satelite temperature data.
This is the November version:
And this is the December version:
Spot the difference!
Update: Gavin reminds me that in April Spencer was using a ridiculous degree 4 fit to the data:
If he’d stuck with that, the current graph would look like this:
The polynomial is still decreasing at the end, but the divergence from the data is striking.