Cute, but... doesn't seem very significant?
Yes. In the 1700s The Earth was just beginning to emerge from 'the little ice age'. There is some controversy over whether the medieval warm period may have been as hot, or hotter than present.
Caldeira has apparently used the GISTEMP LOTI Global mean series, but this seems to have been smoothed to remove any peaks. I tried plotting this particular series at.
The pattern he describes in his tweet is not so clear with higher resolution data. In particular, there is a pronounced spike in the late 1930s, and another in the mid 1950s:
http://woodfortrees.org/plot/gistemp-dts/from:1880/to:2013
Furthermore, this does not seem to be replicable using other available temperature series. For instance HADCRUT4 global mean series from 1880 to 2013 seems to have a peak in the late 1940's that is not present in Caldeira's graph.
http://woodfortrees.org/plot/hadcrut4gl/from:1880/to:2013
It is even less visible in the earlier HADCRUT3 variance adj. global mean:
http://woodfortrees.org/plot/hadcrut3vgl/from:1880/to:2013
Or the CRUTEM4 variance-adj land global mean
http://woodfortrees.org/plot/crutem4vgl/from:1880/to:2013
Similar for BEST global land mean (preliminary)
http://woodfortrees.org/plot/best/from:1880/to:2013
In short, unless I am missing something, his claim is based on the one outlier among temperature series which can be made to support it (if you squint your eyes real hard, I suppose).
