Advance and Prospect of t he st udies of El Ni no Predict abilit y by Nonli near opti mizati on Met hod

DUAN Wan-Suo,MU Mu
2013-01-01
Abstract:The st udi es of El Ni no Sout her n Oscill ati on ENSO predi ctabilit y by usi ng nonli near opti mizati on met hod are revi e wed i n t hi s paper . For t he key proble ms of ENSO predi ctabilit y9i ncl udi ng opti mal precursors 9 spri ng predi ctabilit y barri er 9how to Cuantif y ENSO predi ctabilit y9and ENSO asy mmetr y9t he aut hors use t he t heoreti cal model and t hen i nter medi ate model t o i nvesti gate t he dyna mics of ENSO predi ctabilit y . By t hese i nvesti gati ons9some nonli near characteri sti cs of ENSO behavi or are revealed t o a certai n extent . The mai n results are su mmarized as f oll owsC i Conditi onal nonli near opti mal pert ur bati on CNOP l ocal CNOP i s more potenti al t han t he correspondi ng li near si ngular vect or LSV i n evol vi ng i nt o El Ni no La Ni na event . CNOP and l ocal CNOP are t heref ore regar ded as t he opti mal precursor of El Ni no and La Ni na respecti vel y . These deri ved El Ni no and La Ni na events are asy mmetri c i n a mplit ude about t he cli matol ogi cal mean state i .e . Sea surf ace te mperat ure i s zero . Syste mati c t heoreti cal anal ysi s de monstrates t hat it i s t he nonli near te mperat ure advecti on process t hat enhances El Ni no si gnifi cantl y and suppresses La Ni na negli gi bl y t hen leadi ng t o t he asy mmetr y of ENSO. The NCEP reanal ysi s data duri ng 1980 Z00Z are used t o verif y t he above results . Then t hese t heoreti cal results are supported Cualitati vel y . ii The CNOP t ype i niti al error of ENSO i nduces t he pro mi nent seasonal-dependent error evol uti on . Furt hermore f or a l ar ge nu mber of i niti al error patter n dif f eri ng f ro m CNOP extensi ve nu meri cal experi ments suggested t hat CNOP-t ype i niti al error causes t he most obvi ous spri ng predi ctabilit y barri er SPB . The mechanis m of SPB is also i nvesti gated by anal yzi ng t he tendency eCuati ons of error evol uti on of ENSO. The results de monstrate t hat t he occurrence of SPB f or ENSO depends on not onl y t he cli matol ogi cal mean state but also t he ENSO event itself and t he confi gurati on of i niti al error . Theref ore SPB f or ENSOis caused by t he coll ecti ve ef f ect of t hese t hree f act ors . iii The l ower bound of t he maxi mu m predi cable ti me t he upper bound of maxi mu m predi cti on error and t he l ower bound of maxi mu m all owable i niti al error f or ENSO are establi shed wit h t hree dif f erent nonli near opti mizati on proble ms . To a certai n extent t hese t hree sub-proble ms of predi ctabilit y Cuantifi ed ENSO predi ctabilit y are based on a t heoreti cal coupled model . Then t hi s wor k tell us how t he nonli near opti mizati on met hod is used t o Cuantif y t he predi ctabilit y of ENSO. Besi des t he SPB f or ENSOis revealed f ro m t hese t hree dif f erent perspecti ves which also de monstrates t he seasonal-dependent evol uti on of error f or El Ni no . i v CNOP approach is also used t o st udy t he decadal vari abilit y of ENSO. It i s f ound t hat nonli near te mperat ure advecti on plays an i mportant role i n ENSOasy mmetr y which explai ns not onl y t he asy mmetr y of ENSO i n i nterannual scale but also t hat i n i nter decadal scale . Furt her more t he mechanis m of decadal asy mmetr y of ENSOasy mmetr y i s gi ven i .e . t he decadal change of ENSO asy mmetr y may be caused by t he coll ecti ve ef f ect of t he changes i n tropi cal background state and nonli nearit y associated wit h ENSO. At t he end of t hi s paper t he aut hors anal yze t he prospect of appli cati on of nonli near opti mizati on met hod i n ENSO st udy . It i s expected t hat t he nonli near opti mizati on met hod can be extended to t he st udi es of t he second predi ctabilit y proble ms f or ENSO.
What problem does this paper attempt to address?