Fig. 3: Reduced glacial temperature variability inferred from central and eastern equatorial Pacific IFA records.

a–d, Histograms and box-and-whisker plots (box represents the interquartile range and whiskers represent the full distribution minus outliers—depicted in b and d) of IFA anomalies for the late Holocene (LH; red) and LGM (blue) intervals from central (CEP; ML-1208: IFA-δ¹⁸O (a); this study—where δ¹⁸O is negatively correlated with temperature) and eastern equatorial Pacific sediments (EEP; ODP Site 849: IFA-Mg/Ca-temperature (c); ref. ¹²; see Extended Data Table 3 for details). e,f, Resampled histograms of simulated changes (1σ; percentage change from the 0 ka simulation) in site-based temperature variability detectable using the IFA technique^19,52 in two scenarios: (1) changes between the 21 ka and 0 ka preindustrial (PI) simulation at each site (teal) and (2) year-to-year variability from the PI simulation imposed on the local annual cycle from the LGM simulation (dark blue) to test whether changes in seasonal variability can explain the observed changes. Reconstructed change (1σ; %) in temperature variability inferred from IFA at the sites are also depicted (around 40% at ODP Site 849, EEP, and about 30% at ML1208, CEP). g,h, Quantile–quantile plots of late Holocene versus LGM IFA data (markers) and their regression lines (purple dashed lines), with uncertainty envelopes generated from bootstrap resampling (n = 10,000) of seasonal-only versus full-variability scenarios described above. All foraminiferal carbonate δ¹⁸O values are reported in permille relative to Vienna Pee Dee Belemnite (VPDB).

Source Data