Ocean reanalysis (ORA) products are widely used by the scientific community and are essential tools for monitoring and understanding tropical Atlantic interannual variability. Yet, large differences have been observed among these products in the tropical Atlantic Ocean. Here, an ensemble of eight ORAs is examined and compared over the period ranging from January 1993 to December 2017. The analysis reveals a considerable spread ( \(> 0.3\, ^{\circ }\text {C}\) ) in tropical Atlantic Ocean temperature anomalies in the upper 150 m. Along the equatorial Atlantic and southwest African coast, which are regions of high interannual variability, most of the discrepancies among the ORAs are located around the depth of the thermocline. This spread is linked to the amount of in-situ profiles assimilated by ORAs. The deployment of the PIRATA mooring array in early 1998 and the advent of Argo floats after 2005 improved ORA consistency along the equatorial Atlantic, as well as down the path of equatorial Kelvin waves and subsequent coastal trapped waves. Despite a decreasing trend of \(-\,0.067\, ^{\circ }\text {C}\) \(\hbox {decade}^{-1}\) of the ensemble spread averaged over the upper 150 m of the tropical Atlantic Ocean, a few regions with limited in-situ observations present smaller improvements, such as in the Benguela upwelling system. A comparison between ORAs, forced-ocean models, and an ensemble of ORA varying only in their atmospheric forcing suggests that a considerable amount of the spread can be attributed to discrepancies among the atmospheric fields used to force ORAs. The ensemble spread of the tropical Atlantic wind anomalies across four atmospheric reanalyses suggests that their consistency has not improved or even worsened.
