Estimates of Atlantic meridional heat transport from spatiotemporal fusion of Argo, altimetry, and gravimetry data

Abstract. Understanding how changes in Atlantic meridional heat transport (MHT) and the Earth's climate relate to one another is crucial to our ability to predict the future climate response to anthropogenic forcing. Attaining this understanding requires continuous and accurate records of MHT across the whole Atlantic. While such records can be obtained through direct ocean observing systems, these systems are expensive to install and maintain, and thus, in practice, records of MHT derived in this way are restricted to a few latitudes. An alternative approach, based on hydrographic and satellite components of the global ocean observing system, consists of inferring heat transport convergence as a residual from the difference between ocean heat content (OHC) changes and surface heat flux. In its simplest form, this approach derives the OHC from hydrographic observations alone; however, these observations are spatially sparse and unevenly distributed, which can introduce significant errors and biases into the MHT estimates. Here, we combine data from hydrography, satellite altimetry, and satellite gravimetry through joint spatiotemporal modelling to generate probabilistic estimates of MHT for the period 2004–2020 at 3-month resolution across 12 latitudinal sections of the Atlantic Ocean between 65° N and 35° S. Our approach leverages the higher spatial sampling of the satellite observations to compensate for the sparseness and irregular distribution of the hydrographic data, leading to significantly improved estimates of MHT compared to those derived from hydrographic data alone. The fusion of the various datasets is done using rigorous Bayesian statistical methods that account for the spatial resolution mismatch between datasets and ensure an adequate representation and propagation of uncertainty. Our estimates of MHT at 26° N agree remarkably well with estimates based on direct ocean observations from the RAPID array, both in magnitude and phase of the variability, with a correlation of 0.68 for quarterly (3-monthly) time series and 0.81 after applying a yearly running mean. For the period 2004–2017, the correlation increases to 0.78 and 0.92, respectively. The time-mean MHT at 26° N is also captured by our approach, with a value of 1.14 PW [1.01, 1.27] (5 %–95 % credible interval). Estimates of MHT at other latitudes are also consistent with what we expect based on earlier estimates as well as on our current understanding of MHT in the Atlantic Ocean.