Simulating conditional deterministic predictability within ocean frontogenesis

Ocean mesoscale eddies are non-deterministic in that small errors grow in time so that accurate prediction is not possible without continual correction from observations. Ocean frontogenesis can be forced by mesoscale eddies through straining of buoyancy gradients, which produces filaments of surface divergence related to ageostrophic upwelling. The upwelling can result in thinning of the mixed layer. The frontogenesis predictability is tested through a series of Observation System Experiments (OSEs), the results of which indicate that if the strength and location of the mesoscale eddies are accurately predicted, then the associated frontogenesis features can be predicted. The frontogenesis features have a ‘conditional deterministic predictability’. The OSEs are started with perturbed initial conditions, and the OSEs assimilate an increasing number of satellite altimeter data streams. One experiment uses all available data to provide the most accurate analysis, which is labeled as the nature run. Relative to the nature run, ocean steric height correlations increases from about 0.87 with one altimeter and asymptotically reaches 0.99 with four altimeters, showing increasing skill in mesoscale prediction. Satellite data provide no information to dynamically correct frontogenesis processes in the numerical models. Even though not corrected by data, as the number of satellite altimeters increases from 1 to 4, the spatial correlation to the nature run of the frontogenesis forcing increases linearly from 0.27 to 0.59, the surface divergence correlation increases linearly from 0.27 to 0.57 and mixed layer depth correlation increases linearly from 0.67 to 0.89. The conclusion is that within the simulations the frontogenesis filaments are deterministically predictable conditioned on accurate prediction of the mesoscale.