Figure 4: Composite Taiwan tree-ring δ¹⁸O-based reconstruction of NIÑO4 sea-surface temperature (SST) anomaly from 1190 to 2007 AD.

(a) Plot of composite tree δ¹⁸O-based NIÑO4 SST anomalies (SSTAs) averaged over March–May for each year, calculated with respect to the mean of observed SSTs during the 1950–2007 AD period (black horizontal line), plotted with 31-year low-passed version of the data (red line), and the blue horizontal line indicates the highest 31-year low-passed SST value of the time series. The grey area denotes ±2σ error bars, based on the statistical reconstruction across overlapping tree δ¹⁸O series⁵³. An anomalously high reconstructed SST value in 1651 AD is indicated by an orange arrow. (b) Thirty-one-year low-passed value of the composite Taiwan tree δ¹⁸O series shown in a. The grey area denotes ±2σ smoothed error bars⁵⁴. The blue horizontal line indicates the highest 31-year low-passed SST value of the time series⁴¹ (Methods). (c) Time series of 31-year running variance of internnual-scale variability (isolated with a 2–7-year band-pass filter) corresponding to each of the 16 individual raw tree-ring δ¹⁸O time series (thin coloured lines) and the average of these running variance time series (thick red line, for periods where they overlap). (d) The red line is the same red line in c and the pink area denotes ±1σ of the mean. The horizontal blue lines both in c,d indicate the highest 31-year averaged interannual variance value of the time series, centred on 1992 AD.