Семинар

Опубликовано: 13 ноября 2019 г.
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
Abst­ract
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.