Decadal variability of prediction skill of ENSO

Wednesday, 17 December 2014
Tamaki Yasuda, Meteorological Research Institute, Ibaraki, Japan
The variability of SST in the tropical Pacific associated with ENSO is well predicted in time scales from 6 months to 1 year advance. However, the prediction skills of ENSO in many ENSO prediction systems are lower in the 2000s than those in the 1980s and 1990s. Since late 1990s sea surface height and SST have increased in the western equatorial Pacific and easterly surface wind has increased in the decadal time scale. It has not been fully understood whether the decadal variability in the tropical Pacific could cause the lower prediction skill of ENSO during the recent decade. In this study, decadal variability of prediction skill of ENSO using ocean-atmosphere coupled general circulation model and its relationship with decadal variability of tropical Pacific are examined. It is found that prediction skill of ENSO fluctuates in the decadal time scale. Anomaly correlation coefficient of NINO3 SST for the 1979-2006 period is 0.76 at 6-month lead time, but 0.40 for the 2003-2011 period. NINO3 SST error at 6-month lead time changes in the decadal time scale and has increased since late 1990s. The period of large SST error corresponds to that of large decadal anomalies of observed NINO3 SST. It is shown that decadal variability of NINO3 SST error is consistent with that of magnitude of initial SST anomalies in the prediction. Decadal anomalies of zonal gradient of observed thermocline depth in the equatorial Pacific are large in the early 1990s and late 2000s. However, those anomalies decrease rapidly in the prediction. Positive temperature errors at the depth of thermocline at the beginning of the prediction develop with propagating eastward accompanied by westerly surface wind anomalies. It is important for successful ENSO prediction to reduce errors of thermocline depth at the beginning of prediction and to understand how decadal anomalies in the equatorial Pacific are maintained in the observation and prediction.