Impact of Sea Level Assimilation on ENSO Initialization and Prediction: The Role of the Sea Level Zonal Tilt and Zonal Mean

Abstract

t present, most models forecasting the El Niño?Southern Oscillation (ENSO) use data assimilation, which constrains models physics using available observations. In this article, an ENSO model of intermediate complexity is constrained by sea level observations: sea level from the Simple Ocean Data Assimilation (SODA) reanalysis is assimilated in the model forced by SODA winds, using an ensemble Kalman filter. In addition, retrospective ENSO forecasts over the period 1958?2007 are computed. The assimilation of sea level observations slightly improves the model?s predictive skill, which is due to the correction of the recharge?discharge process simulated by the model. To assess this, two indices relevant to the ENSO recharge?discharge theory are considered: the zonal tilt and zonal mean of sea level in the equatorial Pacific. The assimilation of those two observed indices alone leads to results that are qualitatively similar to the assimilation of full maps of sea level observations. This partly results from the fact that the leading statistical modes of the model errors on sea level have a zonal tilt and zonal mean structure. The data assimilation corrects in particular a too weak amplitude of the zonal mean sea level and its associated subsurface variability in the model. The authors suggest that insight on the role of the recharge?discharge process in other models could be gained by comparing the assimilation of full maps of sea level observations with the assimilation of the two indices of sea level.