Impact of Parameterized Isopycnal Diffusivity on Shelf‐Ocean Exchanges Under Upwelling‐Favorable Winds: Offline Tracer Simulations Augmented by Artificial Neural Network

Abstract

Mesoscale eddies with horizontal scales of 10-100 km are ubiquitous in the world ocean and dominate the transport and mixing of climatologically relevant tracers, such as heat, salt, and carbon (Busecke & Abernathey, 2019;Gnanadesikan et al., 2015Gnanadesikan et al., , 2017Jones & Abernathey, 2019). Owing to their relatively small size, these eddies are not adequately resolved by predictive ocean climate models, most of which still adopt a horizontal grid spacing of 1° in longitude/latitude to accommodate long integration periods between centuries and millennia (Farneti & Gent, 2011;Farneti et al., 2010). Consequently, tracer transports driven by mesoscale eddies must be parameterized in these coarse-resolution models.Typical approaches to parameterizing the mesoscale eddy-driven transports include the Gent and Mcwilliams (1990) scheme (or the GM scheme), which rearranges fluid parcels adiabatically by flattening the isopycnals and thus extracting the large-scale potential energy, and the Redi scheme (Redi, 1982), which drives down-gradient tracer fluxes along isopycnal surfaces. Numerical implementation of the GM and Redi schemes