North Tropical Atlantic Climate Variability and Model Biases

Remote forcing from El Niño-Southern Oscillation (ENSO) and local ocean-atmosphere feedback are important for climate variability over the North Tropical Atlantic. These two factors are extracted by the ensemble mean and inter-member difference of a 10-member Pacific Ocean-Global Atmosphere (POGA) experiment, in which sea surface temperatures (SSTs) are restored to the observed anomalies over the tropical Pacific but fully coupled to the atmosphere elsewhere. Both ENSO forced and local North Tropical Atlantic modes (NTAMs) develop with wind-evaporation-SST feedback. Local variability, represented by the preseason NAO and SST persistence, contributes considerably to NTAM variability. Including these two indicators, together with ENSO, improves the predictability of NTAM. The South Tropical Atlantic Mode can be forced by ENSO, and a cross-equatorial dipole is triggered by ENSO instead of local air-sea coupling within the tropical Atlantic.

Notable biases exist. The seasonality of the simulated NTAM is delayed by one month, due to the late development of the North Atlantic Oscillation (NAO) in the model. This suggests the importance of NAO in setting the seasonality of NTAM and of the extratropical-tropical teleconnection. Furthermore, a spurious band of enhanced SST variance (SBEV) is identified over the northern equatorial Atlantic. The SBEV is especially pronounced in boreal spring and due to the combined effect of both anomalous atmospheric thermal forcing and oceanic vertical upwelling. The SBEV is a common bias in Coupled Model Intercomparison Project 5 (CMIP5), found in 14 out of 23 models. While the tropical North Atlantic variability is only weakly correlated with the Atlantic Zonal Mode (AZM) in observations, the SBEV in CMIP5 produces conditions that drive and intensify the AZM variability via triggering the Bjerknes feedback. This partially explains why AZM is strong in some CMIP5 models even though the equatorial cold tongue and easterly trades are biased low.