Are climate models biased in the atmosphere-ocean partitioning of poleward energy transport?

Observational estimates of the partitioning of poleward energy transport between the ocean and atmosphere are compared to that in an ensemble of comprehensive coupled climate models. Due to the differing availability and reliability of observational data and model output, the techniques for partitioning energy transport in models and observations differ: a) in models ocean heat transport is diagnosed as the spatial integral of net surface energy flux whereas; b) in observational estimates ocean heat transport is diagnosed from the residual of satellite derived top of atmosphere radiation and atmospheric heat transport calculated from high frequency atmospheric reanalysis. We demonstrate that the differing approaches used to partition poleward energy transport yield nearly identical results in a perfect model setting in which high frequency atmospheric fields -- akin to the reanalysis data used in the observational estimate of atmospheric energy transportare output. The observed total (atmosphere plus ocean) poleward energy transport in each hemisphere is similar to the inter-model mean despite a substantial (20%) inter-model spread. However, models partition the transport between the ocean and atmosphere differently than the observations; the model-mean poleward ocean energy transport is biased low, compared to the observed, in the sub-tropics of both hemispheres and the mid-latitudes of the Southern Hemisphere. We speculate that this bias toward smaller ocean energy transport in models results from model overestimates of evaporation in the subtropical ocean: in models surface radiative heating is primarily balanced by evaporation whereas in observations a larger fraction of the surface radiative heating is exported by the oceanic circulation. We additionally partition atmospheric energy transport into contributions from the mean overturning circulation, stationary eddies and transient eddies. We find that, in the inter-model mean, the poleward energy transport by stationary eddies is biased low relative to observations in the subtropics of both hemispheres and in the Northern Hemisphere mid-latitudes whereas the modeled transient eddy poleward energy transport exceeds that in observations in both Hemispheres.