A comparison of two methods for detecting abrupt changes in the variance of climatic time series

摘要: Abstract. Two methods for detecting abrupt shifts in the variance – Integrated Cumulative Sum of Squares (ICSS) and Sequential Regime Shift Detector (SRSD) have been compared on both synthetic observed time series. In Monte Carlo experiments, SRSD outperformed ICSS overwhelming majority modeled scenarios with different sequences regimes. The advantage was particularly apparent case outliers On other hand, has more parameters to adjust than ICSS, which requires experience from user order select those properly. Therefore, can serve as a good starting point regime shift analysis. When tested climatic series, most cases detected same change points longer series (252–787 monthly values). only exception Arctic Ocean sea surface temperature (SST) when found one extra that appeared be spurious. As shorter (66–136 yearly values), failed detect any even doubled or tripled another. For these is recommended. Interestingly, all tested, tropics, had thing common: last each toward high-variance regime. This consistent findings increased climate variability recent decades.