Understanding the interplay between ENSO and related tropical SST variability using linear inverse models.

The impacts of tropical interbasin interaction (TBI) on the characteristics and predictability of sea surface temperature (SST) in the tropics are assessed with a linear inverse modelling (LIM) framework that uses SST and sea surface height anomalies in the tropical Pacific (PO), Atlantic (AO), and Indian Ocean (IO). The TBI pathways are shown to be successfully isolated in stochastically-forced simulations that modify off-diagonal elements of the linear operators. The removal of TBI leads to a substantial increase in the amplitude of El Niño-Southern Oscillation (ENSO) and related variability. Partial decoupling experiments that eliminate specific coupling components reveal that PO-IO interaction is the dominant contributor, whereas PO-AO and AO-IO interactions play a minor role. A series of retrospective forecast experiments with different operators shows that decoupling leads to a substantial decrease in ENSO prediction skill especially at longer lead times. The relative contributions of individual pathways to forecast skill are generally consistent with the results from the stochastically-forced experiments. Qualitatively similar results are obtained from an additional set of forecast experiments that partially apply initial conditions over specific basins, but several important differences were also found due to differences in the representations of each TBI pathway. Finally, the cause of contrasting SST anomalies over the AO after the extreme 1982/83 and 1997/98 El Niño events is explored using LIM forecast experiments to demonstrate the strength and flexibility of our LIM-based approach. [ABSTRACT FROM AUTHOR]