Во вторник 19 ноября в 11:00 в конференц-зале ИФА состоится институтский семинар.
Pavel Berloff (Imperial College London, Department of Mathematics)
Some novel approaches for parameterizing mesoscale eddies
This talk focuses on some new approaches for parameterizing oceanic mesoscale eddy effects for use in non-eddy-resolving and eddy-permitting general circulation models. The context is provided in terms of discussing the existing ideas and problems with their realizations. Specific example of eddy-rich eastward jet extensions of western boundary currents and their adjacent recirculation zones is considered in the classical multi-layer quasigeostrophic model of the wind-driven midlatitude circulation.
First, the key dynamical mechanism operating in the eddy-resolving model and maintaining the eastward jet is identified as the ``eddy backscatter'', which is based on persistent and positive time-lag correlations between the transient part of the nonlinear eddy forcing and the large-scale flow response.
Second, this mechanism has to be ultimately parameterized, and discussing how this can be done is the main part of the talk. We will systematically (but not too technically) discuss 4 different, novel parameterization approaches, which are complimentary to the existing ones: (1) direct stochastic forcing (DSF); (2) implicit stochastic footprints (ISF); (3) data-driven eddy emulations (DEE); and (4) local eddy amplification (LEA). DSF approach explicitly adds statistically constrained stochastic forcing to the coarse model. ISF approach imposes statistically constrained stochastic forcing on an intermediate-complexity eddy-resolving model, obtains its nonlinear response in terms of the coarse-grained footprint, and then imposes local footprints on the coarse model. DEE approach emulates eddies via multi-layer nonlinear regression, then feeds them to the deterministic eddy forcing operator coupled to the large-scale flow fields, and adds the resulting forcing to the coarse model. LEA approach interactively identifies eddies and amplifies them locally and in a simple way - this is the simplest and also most practical approach for the present state of modelling. Relative strengths and weaknesses of these approaches, as well as some future developments will be also discussed.