Extended Data Fig. 8: Statistical analyses of observation-based AMOC indices.
a Observation weights in the EN4 dataset for the eastern and southwestern South Atlantic. b 5-year running average of the SSESA and SSWSA indices. The grey lines indicate the period excluded from the current analysis due to nearly zero weights attributed to observations in the EN4 dataset. Solid black line indicates linear trend of the SSWSA index from 1995 to 2022 (−0.13 psu.dec−1; p = 2e-6). The dashed black line indicates the linear trend of the SSESA index from 1975 to 2022 (−0.013 psu.dec−1; p = 3e-6). c SST-based indices calculated from the HadISST dataset. Blue line indicates SSTSPG-noPA (as per ref. 23), which does not account for polar amplification, showing no significant trend (p = 0.06). Green line indicates the SSTSPG-PA index corrected by polar amplification7 (trend: 0.1 K.dec−1; p = 3e-5). Magenta line indicate the SSTDIPOLE* index (trend: 0.22 K.dec−1; p = 5e-11). Thin lines represent annual mean values and thicker lines their respective 5-year moving average. Dashed lines indicate linear trends from 1950 to 2023. Bandings indicate two standard deviations range (equivalent to the 95% CI) of the reference period (1900 to 1950) for each index, according to their respective line colours. Vertical magenta line indicates the time-of-emergence (ToE) of the AMOC weakening signal for the SSTDIPOLE* index, defined as the year at which the time series emerges from the two standard deviations envelop with no returning62,63 In b-c, statistical significance of the trends is evaluated through a two-sided Mann-Kendall test with no adjustments.
