An extension of LDEO5 model for ENSO ensemble predictions

Abstract

This study provided an extension to the latest version of Lamont-Doherty Earth Observation (LDEO5) prediction system. First, an ensemble coupled data assimilation (CDA) system, based on the Ensemble Kalman Filter, was established. Both the Kaplan sea surface temperature (SST) data from January 1856 to December 2018 and the ECMWF twentieth century reanalysis (ERA-20C) wind data from January 1900 to February 2010 were assimilated for prediction initialization. Second, an ensemble prediction (EP) system was established using stochastic optimal perturbation that represented the uncertainty in the physical process. The assimilation experiments showed that assimilating multi-source data yielded better results than assimilating single-source data. The analyses of Niño3.4 SST anomalies and zonal wind stress (ZWS) anomalies were in good agreement with the observed counterparts, respectively. The root mean square errors of both Niño3.4 SST anomalies and ZWS anomalies were found to be significantly reduced, compared to the values obtained before assimilation. The modeled upper layer depth anomalies along the equator, and subsurface temperature anomalies in the Niño3.4 region were also found to be similar to the observed counterparts. A long-term ensemble hindcast was conducted using the EP system for the past 163 years, from January 1856 to December 2018. Results showed that the predictions initialized by assimilating multi-source data yielded best deterministic skill, reaching the international advanced level. A comparative analysis revealed that the EP system predicted the warm events well, followed by cold and neutral events.