Using CMIP5 model outputs to investigate the initial errors that cause the “spring predictability barrier” for El Niño events

Abstract

Most ocean-atmosphere coupled models have difficulty in predicting the El Niño-Southern Oscillation (ENSO) when starting from the boreal spring season. However, the cause of this spring predictability barrier (SPB) phenomenon remains elusive. We investigated the spatial characteristics of optimal initial errors that cause a significant SPB for El Niño events by using the monthly mean data of the pre-industrial (PI) control runs from several models in CMIP5 experiments. The results indicated that the SPB-related optimal initial errors often present an SST pattern with positive errors in the central-eastern equatorial Pacific, and a subsurface temperature pattern with positive errors in the upper layers of the eastern equatorial Pacific, and negative errors in the lower layers of the western equatorial Pacific. The SPB-related optimal initial errors exhibit a typical La Niña-like evolving mode, ultimately causing a large but negative prediction error of the Niño-3.4 SST anomalies for El Niño events. The negative prediction errors were found to originate from the lower layers of the western equatorial Pacific and then grow to be large in the eastern equatorial Pacific. It is therefore reasonable to suggest that the El Niño predictions may be most sensitive to the initial errors of temperature in the subsurface layers of the western equatorial Pacific and the Niño-3.4 region, thus possibly representing sensitive areas for adaptive observation. That is, if additional observations were to be preferentially deployed in these two regions, it might be possible to avoid large prediction errors for El Niño and generate a better forecast than one based on additional observations targeted elsewhere. Moreover, we also confirmed that the SPB-related optimal initial errors bear a strong resemblance to the optimal precursory disturbance for El Niño and La Niña events. This indicated that improvement of the observation network by additional observations in the identified sensitive areas would also be helpful in detecting the signals provided by the precursory disturbance, which may greatly improve the ENSO prediction skill.