Extended Data Fig. 10: Local spectra of summer temperature for the non-infilled HadCRUT5 and the BEST interpolated products for varying spatial coverage.
From: Millennial-scale climate variability over land overprinted by ocean temperature fluctuations
The average spectrum of the non-infilled HadCRUT566 (dark purple) and the BEST34 station data (with at least 80 years of data, dark blue) overlapping with the pollen record locations (less than 300 km away in the case of the station data) are compared with the average spectrum of the BEST interpolated product for varying spatial coverage: at the pollen records locations (pale green), for all land extra tropic north of 20oN (dark green), and at the locations of overlap (legend uses the intersection symbol ∩ ) between the pollen records with the non-infilled HadCRUT5 dataset (pale purple), and with the station data (light blue); logarithmically spaced axes were used. This supports that in all cases the spatial coverage is broad enough to be representative of the expected behaviour for the northern hemisphere land extra tropics and that HadCRUT5 is not affected by a coverage bias in Fig. 1 The higher variability in both station data and HadCRUT5 suggests that the infilled BEST product is biased low.
