The Southern Ocean is a pivotal component of the global climate system yet it is poorly represented in climate models, with significant biases in upper-ocean temperatures, clouds and winds. Combining Atmospheric and Coupled Model Inter-comparison Project (AMIP5/CMIP5) simulations, with observations and equilibrium heat budget theory, we show that across the CMIP5 ensemble variations in sea surface temperature biases in the 40–60°S Southern Ocean are primarily caused by AMIP5 atmospheric model net surface flux bias variations, linked to cloud-related short-wave errors. Equilibration of the biases involves local coupled sea surface temperature bias feedbacks onto the surface heat flux components. In combination with wind feedbacks, these biases adversely modify upper-ocean thermal structure. Most AMIP5 atmospheric models that exhibit small net heat flux biases appear to achieve this through compensating errors. We demonstrate that targeted developments to cloud-related parameterisations provide a route to better represent the Southern Ocean in climate models and projections.

We present an observation‐based heat budget analysis for seasonal and interannual variations of ocean heat content (H) in the mixed layer (H_mld) and full‐depth ocean (H_tot). Surface heat flux and ocean heat content estimates are combined using a novel Kalman smoother‐based method. Regional contributions from ocean heat transport convergences are inferred as a residual and the dominant drivers of H_mld and H_tot are quantified for seasonal and interannual time scales. We find that non‐Ekman ocean heat transport processes dominate H_mld variations in the equatorial oceans and regions of strong ocean currents and substantial eddy activity. In these locations, surface temperature anomalies generated by ocean dynamics result in turbulent flux anomalies that drive the overlying atmosphere. In addition, we find large regions of the Atlantic and Pacific oceans where heat transports combine with local air‐sea fluxes to generate mixed layer temperature anomalies. In all locations, except regions of deep convection and water mass transformation, interannual variations in H_tot are dominated by the internal rearrangement of heat by ocean dynamics rather than the loss or addition of heat at the surface. Our analysis suggests that, even in extratropical latitudes, initialization of ocean dynamical processes could be an important source of skill for interannual predictability of H_mld and H_tot. Furthermore, we expect variations in H_tot (and thus thermosteric sea level) to be more predictable than near surface temperature anomalies due to the increased importance of ocean heat transport processes for full‐depth heat budgets.

The net surface energy flux is central to the climate system yet observational limitations lead to substantial uncertainty. A combination of satellite‐derived radiative fluxes at the top of atmosphere adjusted using the latest estimation of the net heat uptake of the Earth system, and the atmospheric energy tendencies and transports from the ERA‐Interim reanalysis are used to estimate surface energy flux globally. To consider snowmelt and improve regional realism, land surface fluxes are adjusted through a simple energy balance approach at each grid point. This energy adjustment is redistributed over the oceans to ensure energy conservation and maintain realistic global ocean heat uptake, using a weighting function to avoid meridional discontinuities. Calculated surface energy fluxes are evaluated through comparison to ocean reanalyses. Derived turbulent energy flux variability is compared with the Objectively Analyzed air‐sea Fluxes (OAFLUX) product, and inferred meridional energy transports in the global ocean and the Atlantic are also evaluated using observations. Uncertainties in surface fluxes are investigated using a variety of approaches including comparison with a range of atmospheric reanalysis products. Decadal changes in the global mean and the interhemispheric energy imbalances are quantified, and present day cross‐equator heat transports are reevaluated at 0.22 ± 0.15 PW (petawatts) southward by the atmosphere and 0.32 ± 0.16 PW northward by the ocean considering the observed ocean heat sinks.

The study of energy flows in the Earth system is essential for understanding current climate change. To understand how energy is accumulating and being distributed within the climate system, an updated reconstruction of energy fluxes at the top of atmosphere, surface and within the atmosphere derived from observations is presented. New satellite and ocean data are combined with an improved methodology to quantify recent variability in meridional and ocean to land heat transports since 1985. A global top of atmosphere net imbalance is found to increase from 0.10 ± 0.61 W m⁻² over 1985–1999 to 0.62 ± 0.1 W m⁻² over 2000–2016, and the uncertainty of ± 0.61 W m⁻² is related to the Argo ocean heat content changes (± 0.1 W m⁻²) and an additional uncertainty applying prior to 2000 relating to homogeneity adjustments. The net top of atmosphere radiative flux imbalance is dominated by the southern hemisphere (0.36 ± 0.04 PW, about 1.41 ± 0.16 W m⁻²) with an even larger surface net flux into the southern hemisphere ocean (0.79 ± 0.16 PW, about 3.1 ± 0.6 W m⁻²) over 2006–2013. In the northern hemisphere the surface net flux is of opposite sign and directed from the ocean toward the atmosphere (0.44 ± 0.16 PW, about 1.7 ± 0.6 W m⁻²). The sea ice melting and freezing are accounted for in the estimation of surface heat flux into the ocean. The northward oceanic heat transports are inferred from the derived surface fluxes and estimates of ocean heat accumulation. The derived cross-equatorial oceanic heat transport of 0.50 PW is higher than most previous studies, and the derived mean meridional transport of 1.23 PW at 26° N is very close to 1.22 PW from RAPID observation. The surface flux contribution dominates the magnitude of the oceanic transport, but the integrated ocean heat storage controls the interannual variability. Poleward heat transport by the atmosphere at 30° N is found to increase after 2000 (0.17 PW decade⁻¹). The multiannual mean (2006–2013) transport of energy by the atmosphere from ocean to land is estimated as 2.65 PW, and is closely related to the ENSO variability.