Statistical evidences of unusual Southern Oscillation Index patterns suggests possible climate change impacts on ENSO behaviour

The Southern Oscillation Index (SOI) is an important large-scale predictor for seasonal climate forecasts in agriculture and other climate sensitive sectors worldwide. Questions are now being asked if changes in monthly SOI might be affected by climate change and if there is evidence for such changed SOI behavior. We quantified evidences for such changes by using a Bayesian time series model, proposed by Huerta and West (HW). We fitted the HW model to the monthly SOI time series from 1876 up to a range of candidate change points (CP) for the period between 1930 and 1989. Using Monte Carlo Markov Chain we generated ten thousand 20-year long time series of SOI from the fitted HW model following each candidate CP and investigated three metrics of change: (i) the number of negative SOI values (NNEG), (ii) the median SOI (MEDSOI) and (iii) the 5th SOI quantile (Q5SOI). Posterior predictive distributions (PPD) provided evidence for CP detection by calculating the frequency of metric values more extreme than observed in the respective historical SOI record. When considering all measures conjointly, our analysis revealed that the lowest subset of p (0.02 to 0.08) occurred between 1974 and 1978, providing strong empirical evidence for a changed SOI behavior following this period. This is consistent with mechanistic approaches to CP detection and provides additional evidence for secular or climate change impacts on SOI dynamics: the observed SOI patterns since the late 1970s are unprecedented in the instrumental SOI record.