Uncertainty evaluation of the AMOC transport calculation at 11°S

The Atlantic meridional overturning circulation (AMOC) is a key feature of the global ocean circulation and has a big impact on regional weather and global climate. To project future AMOC variability, it is vital to understand and properly assess the past variability. The AMOC transport and in particular its geostrophic component is measured at several latitudes by specific observing systems. However, the transport can be calculated using multiple combinations of instruments which complicates both the comparison of the observed AMOC at different latitudes and to model transports.

Here, we present results from a systematic comparison of methods to compute the upper branch of the AMOC at 11°S, utilizing data from the Tropical Atlantic Circulation and Overturning at 11°S (TRACOS) array. The TRACOS array comprises 10 years of observations, including data from Pressure Inverted Echo Sounders, tall moorings, and ship sections at the eastern and western boundaries as well as supplementary data from Argo floats and satellites. By subsampling the observational setup in two ocean models (INALT20 and VIKING20X, both based on NEMO), we quantify uncertainties in AMOC transport estimates on different time scales.

We find that bottom pressure measurements, despite being prone to sensor drifts, effectively capture the seasonal variability. The largest potential source of error lies in the choice of vertical structure between the measurement points. Longer-term variability assessments based on moored density measurements require particularly high vertical resolution in the upper 500m. For both methods, the error associated with replacing the eastern boundary data with a climatological seasonal cycle is small, indicating low uncertainty resulting from loss of instruments in that region. We also consider uncertainties in the Ekman transport, which has a magnitude of about one-third of the geostrophic transport at 11°S. Ekman transport estimates vary by a few Sverdrups depending on the choice of wind stress product or drag coefficient used. All in all, we find that the TRACOS array can assess AMOC signals but we also show potential improvements in the array design to reduce uncertainties regarding longer-term variability.