ENSO predictability

The overarching goal of this work is to explore seasonal El Nino -- Southern Oscillation (ENSO) predictability. More specifically, this work investigates how intrinsic variability affects ENSO predictability using a state-of-the-art climate model. Topics related to the effects of systematic model errors and external forcing are not included in this study. Intrinsic variability encompasses a hierarchy of temporal and spatial scales, from high frequency small-scale noise-driven processes including coupled instabilities to low frequency large-scale deterministic climate modes. The former exemplifies what can be considered intrinsic "noise" in the climate system that hinders predictability by promoting rapid error growth whereas the latter often provides the slow thermal ocean inertia that supplies the coupled ENSO system with predictability. These two ends of the spectrum essentially provide the lower and upper bounds of ENSO predictability that can be attributed to internal variability.;The effects of noise-driven coupled instabilities on sea surface temperature (SST) predictability in the ENSO region is quantified by utilizing a novel coupled model methodology paired with an ensemble approach. The experimental design allows for rapid growth of intrinsic perturbations that are not prescribed. Several cases exhibit sufficiently rapid growth to produce ENSO-like final states that do not require a previous ENSO event, large-scale wind trigger, or subsurface heat content precursor. Results challenge conventional ENSO theory that considers the subsurface precursor as a necessary condition for ENSO. Noise-driven SST error growth exhibits strong seasonality and dependence on the initialization month. A dynamical analysis reveals that much of the error growth behavior is linked to the seasonal strength of the Bjerknes feedback in the model, indicating that the noise-induced perturbations grow via an ENSO-like mechanism. The daily error fields reveal that persistent stochastic zonal wind stress perturbations near the equatorial dateline activate the coupled instability, first driving local SST and anomalous zonal current velocity changes that in turn induce upwelling and a clear thermocline response. Since the experimental design also isolates daily stochastic wind stress, analysis reveals that during spring when the ENSO signal is smallest and the signal-to-noise ratio is lowest, stochastic winds have the largest impact on perturbation growth that ultimately affect the development of ENSO. Results show that the spring predictability barrier is likely, in part, caused by stochastic springtime winds instigating coupled instabilities in climate models, a hypothesis supporting the notion that noise-driven errors provide an intrinsic limit to predictability.;On the other end of the spectrum, ENSO precursors, including the Pacific Meridional Mode (PMM), are thought to enhance predictability of ENSO and thus may provide potential usefulness in real-time prediction. An empirical orthogonal function (EOF) analysis identifies the dominant ENSO precursor in a coupled model as the low frequency coupled variability associated with the PMM. The robustness of the PMM/ENSO precursor relationship is verified as similar to observations and captured well by short lead-time dynamical seasonal climate predictions. The implied usefulness of the PMM precursor as a supplemental tool for ENSO prediction is then tested by using the March PMM SST state as an independent predictor of December ENSO. In forecast mode, PMM events predict eastern Pacific El Nino events in both observations and model forecasts with somewhat useful skill, yet with much less skill for central Pacific El Nino or La Nina events.;The competition between the PMM precursor and noise-driven perturbation growth in March-initialized forecasts is tested in the context of the 2014 so-called El Nino forecast "bust." Overall, the error growth ensemble approach implemented in the model experiment provides a more complete range of possibilities for longer lead-time forecasts and provides a useful measure of uncertainty to supplement the typical forecast spread calculation. Moreover, applying the error ensemble approach to the 2014 El Nino forecast reveals a key insight into why PMM is a reliable precursor to ENSO, but not a reliable predictor. Since the coupled system is especially sensitive to noise-driven perturbation growth beginning in spring, much of the predictive potential of, say the PMM precursor, which peaks in terms of SST in spring, can be overshadowed by the large forecast uncertainty associated with noise-driven errors instigated by coupled instabilities. As a result, attempting to "break through" the spring predictability barrier in the coupled model may prove difficult given that noise-driven errors provide a clear intrinsic limit to ENSO predictability at longer lead-times, despite the presence of ENSO precursors.