Extended Data Fig. 2: Replicability of summer temperature reconstructions.
From: Millennial-scale climate variability over land overprinted by ocean temperature fluctuations
Timeseries of odd and even sub-groups show good replicability of the millennial scale variability for all continents (the correlation coefficient r is indicated in purple, see Methods). d,e,f, To test how the replicability compares to the expectation from time-uncertain irregular timeseries, we repeat the analysis shown in as a,b,c, but using a surrogate database (without time-uncertainty τε) created using the TraCE-21ka simulation (see Methods). g,h,i, Same as d,e,f, but for a realization of the surrogate database with time-uncertainty of τε = 500 years (see Methods). The 90% range of the correlation obtained for an ensemble of 100 realizations of the 500-year time-uncertainty case is given (purple). The single realization shown was selected because it has correlation close to the mean of the ensemble and is thus representative of the mean behaviour. Comparing d,e,g with g,h,i shows that the replicability of our pollen reconstructions is consistent with the expectation of time-uncertain temperature timeseries. However, the amplitude of the average timeseries should not be interpreted (see Extended Data Fig. 1).
