Changing temperature scales?

[ET-NCMP]

Not sure how relevant this is but they cite the original paper.

Pavlasek, P., Merlone, A., Musacchio, C., Olsen, Å. A. F., Bergerud, R. A. and Knazovicka, L. (2016), Effect of changes in temperature scales on historical temperature data. Int. J. Climatol., 36: 1005–1010. doi:10.1002/joc.4404
http://onlinelibrary.wiley.com/doi/10.1002/joc.4404/full

[ET-NCMP]

OK, seems relevant.

This paper looks at changes in the reference scales of temperature and how temperatures are precisely defined. The typical systematic errors in the usual SST range of -2 to 32 degC are of the order 10s of mK i.e 0.01K. As the temperatures are largely positive, the effect on globally aggregated SST would likely be systematic and amount to around 0.01K (one could easily apply an adjustment based on the method given in the paper) with larger errors (0.02K) at higher SSTs.

The necessary correction would have the effect of cooling temperatures from 1927-1968 and 1969-1989 relative to the period after 1990. It's not clear what to do prior to 1927 though.