Seasonal Impact on Baseline Selection

I have taken the ERSST3 regional data, Tropics, 20N-90N and 20S-90S and used a short 2000 to 2012 baseline just to show how baseline dependent the temperature data can be.  The 1950-1980 "normal" baseline period is circled.  Just prior to 1950, SST data took a 0.2-0.4 C nose dive and these three regions shifted phase relationships.  Except for the period from ~1955 to 1998, the regions are close to be in the same phase.  Is the drop in the 1940s real?  Dunno, but the data is what it is, but using this baseline, the 20N-90N SST caught up to the rest of the oceans amplifying the most recent warming.  Using the southern ocean data 20S-90S though, there is a more consistent temperature rise from ~1910 with a hint of a peak value being approached, but is 2000-2012 the right baseline?

The Reynolds optimally interpolated SST data may have some clues.  Above it is converted to anomaly using the 2000 to 2012 baseline and seasonal cycle removed.  The variability in the NH data 20N-90N is much greater outside of the baseline range.

If you use the early part of the Reynolds SST data the variability shifts to the end of the record.  The most obvious cause is the volcanic activity in the early part had a greater impact on the northern oceans which have a smaller area and transfer energy to a much larger percentage of land mass.  Volcanic sulfate have a greater impact on radiant forcing than atmospheric forcing so the impact is greater in warming season.  The data has a seasonal baseline dependence that impacts "natural" variability.

Anyone working with the data knows this and knows that baseline selection impacts attribution, but not very many explain why they select a baseline only that others are cherry picking if they select a different baseline.

Odd that?

UPDATE:  The seasonal sensitivity can impact which year is the warmest EVAH.  That is always great fun.