ENSO prediction and predictability

In this work we examine the role that observed intraseasonal atmospheric variability may play in controlling and maintaining El Nino-Southern Oscillation (ENSO) variability and its implications for prediction and predictability. To this end, we study the response of two ENSO models to different estimates of observed intraseasonal variability. This variability is estimated as the residual of a multivariate linear regression between the NCEP/NCAR reanalysis of heat-flux and surface winds, and Reynolds Sea Surface Temperatures in the Tropical Pacific. It was found that this variability exhibits many spatial structures with decorrelation times of just a few days and therefore it represents Stochastic Forcing (SF) acting on the low-frequency ENSO models. The first model used was an ENSO model of intermediate complexity. The simplicity of the model allowed us to perform a very exhaustive study of the system response to SF and to elucidate the physical processes at work. Our results support the hypothesis that a significant fraction of ENSO variability may be due to SF acting on an otherwise state of rest. Using the ideas of generalized stability theory we have been able to isolate the dynamically important contributions of SF from the vast universe of the uncoupled atmospheric variability. The model response is largely described by what would be expected from a nonnormal linear system, where the low-frequency tail of SF produces thermocline perturbations that grow through constructive interference of the low-frequency leading eigenvectors of the coupled system. These optimal structures are excited in the model primarily by the SF in the western and central Pacific. In these regions the resulting SST perturbations can trigger anomalous deep penetrative convection and energetic zonal wind anomalies that feed back on the ocean. The insight gained about the ocean response to SF was used to better understand how and why strong-constrained data assimilation forecasts are degraded in the presence of SF. All of the assimilation experiments have shown that information about the thermal structure of the upper ocean (represented as thermocline depth in the model) is a key variable in determining the future evolution of the forecast. The presence of stochastic forcing during the assimilation stage decreases the forecast skill inherent in the model. This is due to the inability of the model to produce intraseasonal variability and therefore the stochastic signal appears as noise in the assimilated observations. (Abstract shortened by UMI.)...