Evaluating Transport Observations of the Atlantic Meridional Overturning Circulation at 11°S Using an Ocean Model

Abstract. The impact of the Atlantic Meridional Overturning Circulation (AMOC) on weather and climate, both regionally and globally, motivated the installation of several observational arrays. It is vital to not only derive transport time series from these arrays but to also quantify the associated uncertainties. Here, an observing system simulation experiment is performed to assess the uncertainty and potential of the TRACOS (Tropical Atlantic Circulation and Overturning at 11°S) array to calculate the geostrophic AMOC transport (AMOCg) from it. Accordingly, the observational setup is subsampled in a high-resolution ocean model and various approaches to derive AMOCg are tested. We find that the currently used approach based on bottom pressure recorders (BPRs) can explain 56 % of the short-term (seasonal to interannual) AMOCg variability, though overestimating the seasonal amplitude. Observations of longer-term variability are limited due to the pressure sensor drift. Currently, long-term (decadal to multi-decadal) variability is only captured by boundary current measurements which explain 62 % of the basin-wide AMOCg long-term variability, though with high root mean squared errors. Regarding potential improvements of the current approaches, we find: 1) The nominal drift rates of the reference sensors currently installed in self-calibrating BPRs are still too high to reliably detect a linear AMOCg trend of the magnitude presently considered realistic, namely about 1 Sv per decade. 2) Acoustic round-trip travel times are limited in use for AMOCg computation at 11°S. 3) Combining BPRs with moored temperature and salinity measurements is a promising approach that can improve AMOCg estimates of both short-term variability (to 79 % explained) and long-term variability (to 61 % explained). Overall, we find that, despite its relatively sparse instrumentation, the TRACOS array is capable of capturing AMOC signals, while we also highlight areas where uncertainties could be reduced.