Figure 2. A comparison of macrofossil- and pollen-based mean T_jul in northern Europe for the four early post-glacial time windows.

It should be noted that in Russia, the mean s.e. for pollen are somewhat higher at 1.05–1.15 °C, probably because the Russian calibration data set is smaller and the species-response estimates are thus less certain. (a) The spatio-temporal pattern of macrofossil- and pollen- based mean July temperatures (T_julP and T_julM, respectively) in early- and mid-Holocene time windows (Supplementary Table 3). The pollen-based T_julP for each site is calculated as the median of all reconstructed values from fossil samples dating to the time window in question. The macrofossil-based T_julM is the highest site-specific value for the time window in question (see Methods for details). In addition, essential information about the ice-margin position and post-glacial stages of the Baltic Sea are shown (adapted and modified from ref. 38). Modern T_jul isolines (in grey) are also presented. (b) Temperature anomalies between T_julM and T_julP reconstructions (Supplementary Table 3). Red colour indicates positive anomalies where macrofossil-based temperatures are higher than pollen-based temperatures, while blue colour indicates negative anomalies. Note that pale pink and pale blue spots can both indicate no difference (0 °C). However, it should also be noted that due to the nature of T_julM as a lowest estimate for mean T_jul, negative T_julM–T_julP anomalies are difficult to interpret, as it is impossible to verify whether the anomaly is for instance due to a false absence of macrofossils. In contrast, positive T_julM–T_julP anomalies can be more confidently taken to suggest a negative bias in the T_julP values (see Methods for more detailed methodological discussion and data interpretation). If pollen data were not available from the same lake as macrofossil data, the T_julM was compared with T_julP derived from an adjacent lake from the same climate zone.