Research

Speaker: Ioana Colfescu, NASC

A machine learning-based approach is used to quantify how much the predictability skill of the El Niño-Southern Oscillation (ENSO) is attributable to oceanic predictors and their interplay, and to isolate the geographical regions and time scales contributing to the skill. While tropical SST represents the main source of ENSO seasonal predictability skill at scales of up to 1 year, 10mu plays an important role from late fall up to late spring and for timescales of 11 to 21 months in advance. The long-lead signal is generated by a coupled wind-SST mechanism originating in the Indian Ocean and later propagates across the Pacific via the atmospheric bridge mechanism. A linear correlations analysis supports this mechanism, and further suggests the precursor link between SST anomalies in the western Indian Ocean (IO) and anomalies westerlies in the eastern basin originating nearly 2 years ahead of an ENSO event.