Exploring 6-month lead predictability of the Atlantic zonal mode in CMIP

Abstract

Skillful prediction of the equatorial Atlantic zonal mode (AZM) remains challenging, with many prediction systems dropping below an anomaly correlation coefficient (ACC) of 0.5 beyond a lead time of 3 months. Since the El Niño-Southern Oscillation (ENSO) is well known to have global impacts, it could be expect to be a useful predictor of the AZM but its influence on the adjacent equatorial Atlantic basin is inconsistent. This is perhaps best exemplified by the fact that the extreme 1982 and 1997 El Niño events were followed by Atlantic zonal mode (AZM) events of the opposite sign. Here we re-examine the potential role of ENSO in the predictability of the AZM using pre-industrial control simulations (piControl) from the Coupled Model Intercomparison Phase 6 (CMIP6). The observed correlation between boreal winter (DJF) sea-surface temperature (SST) in the Niño 3.4 region and the following summer (JJA) SSTs in the ATL3 region is close to zero, indicative of the inconsistent relation between the two. Individual models, however, exhibit a wide range of behaviors with correlations ranging from about -0.5 to +0.5. While the influence of ENSO on equatorial Atlantic SST is inconsistent, the influence of ENSO on surface winds over the equatorial Atlantic is rather robust. All models show a negative correlation between DJF Niño 3.4 SST and boreal spring (MAM) surface winds over the western equatorial Atlantic. In addition, we find that SSTs in the South Atlantic act as a precursor to AZM events. Based on these relations, we construct a multi-linear regression model to predict AZM events in JJA based on Pacific and Atlantic SST in DJF. In most climate models, this simple scheme can predict AZM events with an ACC above 0.5 during ENSO years. We will discuss to what extent these insights may help in the prediction of real-world AZM events.