The question today is: how does Hansen calculate quarterly and annual averages when there are missing values? It’s not obvious.
The issue arose in the previous puzzle in trying to calculate Hansen’s adjustment prior to combining scribal versions. Essentially what happens as reported in the prior post is that when there is a missing value, Hansen estimates the missing value (without doing the obvious thing and uses the available reading from a different version.) His estimates typically differ a lot from the actual value. He then says – aha, there is a bias between these versions and adjusts everything down.
As noted, Praha Dec 1986 is missing from one version and is 1.4 deg C in another version. Hansen then estimates the missing value at -3.6 deg C, a value of no less than 5 deg C lower than the actual value. There are 48 months in common, so he then concludes that the average bias is -0.1 deg C and adjusts all early Praha values down by 0.1 deg C. It sounds pretty absurd when written down. The methodology is not reported in the publications that NASA spokesman Gavin Schmidt has said to contain a comprehensive description of calculations.
Today’s puzzle: how did Hansen estimate Dec 1986 temperatures at Praha at -5 deg C (and other similar estimates)?