Improved Climate Simulations through a Stochastic Representation of Ocean Eddies

Subgrid-scale phenomena in ocean models include eddies, internal waves, and turbulent mixing. The impacts of these phenomena are conventionally parameterized using deterministic closure schemes, such as the widely used Gent and McWilliams eddy parameterization. A major limitation of deterministic parameterization schemes is their assumption that the subgrid-scale processes (and their impacts on the resolved flow) are uniquely determined by the resolved flow. In reality, this slaving relationship is just a first approximation, because many possible subgrid-scale configurations will be consistent with any given resolved flow. Stochastic parameterization schemes are an attempt to improve on this approximation, by sampling the subgrid-scale variability in a computationally inexpensive manner.

Here we investigate whether climate simulations may be improved by implementing a stochastic representation of ocean eddies in a coupled atmosphere–ocean general circulation model. Our approach is to use a high-resolution simulation from HiGEM, which is an eddy-permitting version of the Hadley Centre climate model HadCM3, to calculate the eddy statistics needed to inject realistic stochastic noise into the ocean of FAMOUS, which is its low-resolution counterpart. We have run a suite of four stochastic experiments to test the sensitivity of the simulated climate to the noise definition, by varying the noise amplitude and decorrelation time within reasonable limits. We find that the addition of zero-mean noise to the ocean temperature tendency has a non-zero effect on the mean state of the ocean, significantly improving the temperature field both at the surface and within the deep ocean. The simulated Atlantic Meridional Overturning Circulation is also improved. We conclude that the insertion of stochastic noise into ocean models has the potential to reduce model error and thereby significantly improve climate simulations.