Idealized models have demonstrated the chaotic behavior of the ocean variability at high Reynolds number, up to multi-decadal timescales. Unlike laminar ocean models used in most current climate projections, eddying OGCMs that will be used in future climate projections spontaneously generate a substantial intrinsic variability from eddy scales to multidecadal/basin scales, with a chaotic character, and a marked signature on SSH and SST where air-sea fluxes are maximum in Nature. Whether and how this ocean-driven low-frequency chaotic variability may ultimately impact biogeochemistry, the atmosphere and climate is an important but unsettled question.
Before addressing this question in fully-coupled simulations, it is necessary to explicitly simulate, characterize and quantify over a long period the stochastic character and scales of the low-frequency oceanic variability at high Reynolds number under full reanalyzed forcing, with a focus on climate-relevant indexes. In the framework of the OCCIPUT and PIRATE projects, we have performed and are currently analyzing a 50-member ensemble of 1/4° global ocean/sea-ice NEMO-based 1/4° hindcasts driven by the same reanalyzed 1960-2015 atmospheric forcing. After a common spinup, the spread of the ensemble is seeded by applying stochastic perturbations within each member for one year; eddy-eddy interactions then take control of the subsequent growth of the ensemble spread and of its cascade toward long space and time scales.
Along with reduced-size North Atlantic sensitivity experiments, this global ensemble simulation provides a probabilistic description of the global ocean/sea-ice evolution over the last 5 decades over a wide range of spatio-temporal scales, and direct estimates of the chaotic ocean variability (from the ensemble spread) and of the actual constraint exerted by the atmosphere (variability of the ensemble mean). We will present our strategy and the strong imprints of the atmospherically-modulated ocean stochastic variability on Ocean Heat Content, AMOC, SSH and other climate-relevant indices, with a focus on interannual and longer time scales. We will present complex structure of the ocean variability revealed by these ensemble statistics, and the non-gaussian metrics (based e.g. on the Information Theory) we are developping to more thoroughly characterize the features, scales and imprints of the oceanic chaos and of its atmospheric modulation.