The Bjerknes feedback in the tropical Atlantic in CMIP5 models

Coupled state-of-the-art general circulation models still perform relatively poorly in simulating tropical Atlantic (TA) climate. To investigate whether lack of air–sea interaction might be responsible for their biases, we investigate the Bjerknes feedback (BF) in the TA, the driver of the dominant interannual variability in that region. First, we analyse this mechanism from reanalysis data. Then, we compare our findings to model output from the Coupled Model Intercomparison Project Phase 5. The feedback is subdivided into three components. The first one consists of the influence of eastern equatorial sea surface temperature anomalies (SST’) on zonal wind stress anomalies (τu\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tau _u$$\end{document}’) in the western basin. The second component is the influence of wind stress anomalies in the western TA on eastern equatorial oceanic heat content anomalies (HC’). The third component is the local response of overlying SST’ to HC’ in the eastern TA. All three components are shown to be present in ERA-Interim and ORAS4 reanalysis by correlating the two variables of each component with each other. The obtained patterns are compared to the ones from model output via pattern correlation per component. While the models display errors in the annual cycles of SST, τu\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tau _u$$\end{document}, and HC, as well as in the seasonality of the feedback, the impact of SST’ on wind stress and the impact of wind stress on HC’ are simulated relatively well by most of the models. This is especially the case when correcting for the error in seasonality. The third component of the BF, the impact of HC’ on SST’ in the eastern part of the basin, deviates from what we find in reanalysis. We find an influence of HC anomalies on overlying SSTs in the eastern equatorial TA, but it is weaker than in the reanalysis and it is not strongly confined to the equator. Longitude–depth cross sections of equatorial temperature variance and correlation between subsurface temperature anomalies and SST’ in the cold tongue region show that flawed simulation and slow adjustment of the subsurface ocean are responsible for this.