A new approach to El Niño Southern Oscillation origin and forecastingimplications for predictability = Un nuevo enfoque para el origen y la predicción de El Niño Oscilación Sur: implicaciones para la predictibilidad

Resumen

El Niño Southern Oscillation arises in the tropical Pacific due to coupled atmosphere-ocean interactions. The nature of the dynamical system of the atmosphere is chaotic and its predictability is sensitive to initial conditions, which constraints our ability to foresee the evolution of ENSO for very long time in advance. This dissertation is dedicated to extending the state-of-the-art prediction of the phenomenon. It focuses on the identification of precursory signals in the ocean and atmosphere that improve the understanding and long-lead forecasts of ENSO. A new statistical modelling technique based on dynamic components and state-space methods is developed. Very early premonitory signals that are a result of an in-depth analysis of the processes accompanying the origin and evolution of El Niño, especially in the subsurface ocean, which is less impacted by initial conditions, are established. These tracers are defined in the far western and central tropical Pacific and are shown to anticipate El Niño two and a half years before its peak. Initial intensification of the easterly winds at this time is associated with convergence of mass, downwelling and warming of the subsurface ocean layers in the far west. Thus the South Equatorial Current and the Equatorial Undercurrent are strengthened, which leads to the propagation of warm subsurface anomalies eastward. These anomalous patterns later lead to changes in the circulation and warming of the surface of the ocean in the central tropical Pacific, which leads to the suppression of the easterly winds. The area of tropical convection shifts to the east, which weakens the Walker circulation and triggers the Bjerknes feedback. This allows the further propagation of the subsurface warm anomalies, which reach the eastern Pacific and are upwelled to the surface, which marks the onset of an El Niño . Warm anomalies in the subsurface equatorial ocean have been previously used as precursors in statistical ENSO models via the integration of the upper ocean heat content or the incorporation of anomalies of the 20°C isotherm. In this way the propagation feature of the anomalies is not considered, and no direct connection is made between the first anomalous patterns and the occurrence of a warm event. Hence, the predictive potential of the incipient warming in the western Pacific is not harnessed. Thus, the work presented in the dissertation provides implications for the possibility to improve the long-lead capabilities of other models. The definition of ENSO predictors at specific depths and regions in the ocean and atmosphere requires the reliable surface and subsurface measurements of various climate variables. Regular measurements have only begun with the satellite era in the 1980s, and with the placement of an observation system after the Tropical Ocean Global Atmosphere Program (1985-1994). Forecasts with the model developed here substantially improve after 1994, and the change is distinct for the long-lead forecasts that rely on good-quality subsurface information about the ocean thermal structure. Therefore, the higher temporal and spatial resolution data sets of key variables are now long enough for statistical forecasting models to make better use of. The practical utility of multi-year forecasts is also explored. A well established link exists between ENSO and local climate in the coastal areas of Ecuador. Following El Niño is a warmer surface temperature and enhanced precipitation. These two variables control the dynamics of mosquito population, and in this way affect the incidence of dengue. An experiment is performed where long-lead forecasts of El Niño are used within a dengue model and the prospects for developing an early warning system is investigated.