Fig. 1: Simulated ENSO variability under past and future climate states.

a, ENSO variability simulated by CESM1.2 for climate intervals characterized by altered glacial, orbital and greenhouse gas conditions over the past 21 ka BP and under a doubling of atmospheric CO₂ concentrations (2×CO₂). The level of ENSO variability in each climatic interval is computed as the standard deviation of simulated monthly SST anomalies averaged over the Niño–3.4 region (170° W–120° W, 5° S–5° N). The box-and-whisker diagrams indicate the distribution of standard deviation values for randomly selected 100-year intervals from each simulation (in which the box depicts the interquartile range and the whiskers extend to show the full distribution). b, Frequency distribution of simulated Niño–3.4 SST anomalies averaged from November to January, when ENSO events peak. c, The contrast between the preindustrial (PI) and LGM (21 ka BP) simulations. d, Relationship between the mean amplitude of simulated ENSO events and the percentage of El Niño events that reach extreme amplitude. El Niño events are identified as those with peak amplitude, measured by the Niño–3.4 SST averaged from November to January, larger than 0.5 K. Extreme El Niño events are computed as those with peak amplitude exceeding 2 K.