Decadal and Multidecadal Variability in ERSSTv5 Global SST during 1879–2018

Decadal and multidecadal variability in the ERSSTv5 global SST dataset are studied in terms of implicit fast (noise) and slow (signal) processes that affect variability on decadal time scales. Using a new method that better estimates the fast, or noise, component of decadal variability, estimates of the modes of variability in the slow component are possible. The fast component of decadal variability has a leading fast mode, which explains 62% of the variance, and it is shown that this fast variability, or decadal climate noise, is well represented by any of the indices associated with intradecadal or interannual variability in the tropical Pacific Ocean. Three slow modes are identified, representing 69% of the slow multidecadal variance, after removing the radiative forcing trend. These modes are shown to be related to variability in the Atlantic multidecadal oscillation (AMO) and SST multidecadal variability in the central western Pacific and in the Indian Ocean gyre region, respectively. The first and third slow modes represent two phases of a propagating mode with a period of about 80 years. The second slow mode represents multidecadal variability of the western Pacific warm pool, which is less robust than the other two and shown to be weakly related to the AMO with a lag of about 30 years; fast variability in this region is related to the leading fast mode. Three regions of significant slow variability are identified south of Australia, south of Africa, and near the Drake Passage in association with the Antarctic Circumpolar Current.