Stratosphere-Troposphere Exchange
STRATOSPHERE-TROPOSPHERE EXCHANGE
Jame . HoltonsR 1
2
Peter H. Haynes 2 and Michael E. Mclntyre
3 Ann . DouglasseR Richard an 3 . RooddB Leonhard Pfister4
Abstract paste th ,n I studie. f stratosphere-troposo - strongest wave-induced forces e northeroccuth n i r n sphere exchange of mass and chemical species have hemisphere winter season, the exchange rate is also a mainly emphasized the synoptic- and small-scale maximum at that season. The global exchange rate is mechanisms of exchange. This review, however, in- not determined by details of near-tropopause phenom- cludes als global-scale oth e aspect exchangef so , such suca penetrativen s ha e cumulus convectio r smallno - transpore th s a t acros isentropin sa c surface (potential scale mixing associated with upper level fronts and temperature about 380 K) that in the tropics lies just cyclones. These smaller-scale processes mus cone tb - abov tropopausee eth , nea 100-hPe rth a pressure level. sidered, however n ordei , o understant r e fineth d r Such a surface divides the stratosphere into an "over- details of exchange. Moist convection appears to play n extratropicaa world d an " l "lowermost strato- an importan tropict rolthe ein accountin in s the gfor sphere" that for transport purposes need to be sharply extreme dehydratio r enterinai f n o stratospheree gth . distinguished. This approach places stratosphere-tro- Stratospheri bacy r wa findk ai cs int it e tropos oth - posphere exchange in the framework of the general sphere through a vast variety of irreversible eddy circulatio helpd n an clarif o st differen e rolee th y th f so t exchange phenomena, including tropopause folding mechanisms involved and the interplay between large anformatioe dth f so-calleno d tropical upper tropo- and small scales. The role of waves and eddies in the spheric troughs and consequent irreversible exchange. extratropical overworld is emphasized. There, wave- General circulation models are able to simulate the induced forces drive a kind of global-scale extratropi- mean global-scale mass exchang seasonas it d - ean cy l cal "fluid-dynamical suction pump," which withdraws cle but are not able to properly resolve the tropical air upward and poleward from the tropical lower dehydration process. Two-dimensional (height-lati- stratospher pushed an e t polewari s d downwaran d d tude) models commonly use assessmenr dfo humaf o t n intextratropicae oth l troposphere resultine Th . g glo- impacte ozonth n eo layer include representatiof o n bal-scale circulation drive stratosphere sth e away from stratosphere-troposphere exchange that is adequate to radiative equilibrium conditions. Wave-induced forces allow reasonable simulation of photochemical processes may be considered to exert a nonlocal control, mainly occurring in the overworld. However, for assessing downward in the extratropics but reaching laterally change lowermose th n si t stratosphere strone th , g longi- into the tropics, over the transport of mass across tudinal asymmetrie n stratosphere-tropospheri s - ex e lower stratospheric isentropic surfaces. This mass change render current two-dimensional models inade- transport is for many purposes a useful measure of quate. Either current transport parameterizations must global-scale stratosphere-troposphere exchange, espe- be improved elser o , , more likely, such changee b n sca cially on seasonal or longer timescales. Because the adequately assessed onl three-dimensionay yb l models.
1. INTRODUCTION predictio f globano l change f speciaO . l significancs ei the transport of trace chemical species, natural and Dynamical, chemical, and radiative coupling be- anthropogenic, between the stratosphere and the tro- tween the stratosphere and troposphere are among the posphere. For instance, anthropogenic species trans- many important processes that must be understood for ported from the troposphere into the stratosphere ini- tiat ee chemistrth muc f o h y responsiblr fo e stratospheric ozone depletion, according to a large Departmen f Atmospherio t c Sciences, Universitf o y body of evidence [e.g., World Meteorological Organi- Washington, Seattle. zation (WMO), 1995]. Conversely, downward trans- 2Centr r Atmospheriefo c Scienc e Departmenth t a e f o t Applied Mathematic Theoreticad san l Physics, Universitf yo port fro e stratosphermth t onlno ey constitutee th s Cambridge, Cambridge, England. main removal mechanis r manmfo y stratospheric spe- laboratory for Atmospheres, NASA Goddard Space cies, including those involve ozonn di e depletiont bu , Flight Center, Greenbelt, Maryland. also represent significansa t inpu f ozono t othed ean r 4NASA Ames Research Center, Moffett Field, California. reactive species into the tropospheric chemical system
Copyright 1995 by the American Geophysical Union. Review f Geophysicsso / November 4 , 33 , 7995 pages 403-439 8755-1209/95/95RG-02097S75.00 Paper number 95RG02097 • 403 • 404 • Holton et al.: STRATOSPHERE-TROPOSPHERE EXCHANGE 33, 4 / REVIEWS OF GEOPHYSICS
[e.g., Levy et al., 1980]. Chemical effects from strato- terest. It is necessary therefore to take into account sphere-troposphere exchange (STE) can turnn i ,- in , e speciesth ' photochemical sensitivitie t differena s t fluence the radiative flux balance in the troposphere altitudes and latitudes, and the global-scale circulation and lower stratosphere and can do so in more than one including the spatiotemporal structure of transport way [e.g., Ramaswamy t al.,e 1992; Toumi t al.,e within the stratosphere. Such considerations will be 1994]. STE can therefore have a significant role in the given emphasi thin si s review, which willo t als y otr radiative forcin globaf go l climate change. elucidat relatioe eth n betwee global-scale nth e circu- It is now widely appreciated that the dynamics of lation and the processes, some of them on small troposphere th stratosphere th d ean e are principlen i , , scales, effecy tha r affecma to t t transport acrose th s inseparable [e.g., Hoskins et al., 1985]. Indeed, one tropopause itself. might pos questione eth , Doe makt si sensl y al ean t ea What, in any case, do we mean by the tropopause? to distinguish between the troposphere and the strato- By convention of the World Meteorological Organiza- sphere? If not, then STE would just be part of the tion (WMO) tropopause th , defines ei lowese th s a dt
overall pattern of transport, with no particular signif- leve t whica l temperature hth e lapse rate decreaseo st icance on its own. However, there are strong reasons 2 K km" o1 r less and the lapse rate averaged between why the distinction between troposphere and strato- this level and any level within the next 2 km does not sphere will remain useful. One of them is that vertical excee km"K d2 1. Thu tropopause sth characters ei - chemicatranspord an r ai f lto species throug depte hth h ized by an increase in the static stability in moving troposphere ofth occun e ca timescale n ro shors sa s a t from the troposphere to the stratosphere; this shows a few hours via moist convection and on timescales of eve monthla n i y mea t fixena d altitude latituded san s days via baroclinic eddy motions in middle latitudes. (Figure 1). However, this conventional "thermal" def- By contrast, vertical transport throug hsimilaa r alti- inition of the tropopause obscures the fact that the tude range in the stratosphere takes months, indeed a tropopause often behaves, chemically speaking, as if it year or more in the lower stratosphere, and this ver- wer emateriaa l surfac varyino et g degree approxf so - tical transport must be accompanied by radiative heat- imation quasi-materiae Th . l behavio rationale b n ca r - ing or cooling. ize notingy db , first, thatropicae th t l tropopause cor- There are important implications therefore both for responds roughl n isentropia o t y r stratificatioo c n chemistr r radiativfo d yan e flux balances e differTh . - surface (whose potential temperature th n i K 0 ~e0 38 ence betwee verticae nth l transport timescalee th n si annual mean) and, second, thae extratropicath t l stratosphere and troposphere is part of what lies be- tropopause corresponds roughly to a surface of con- hind, for instance, the rapid increase in ozone mixing stant potential vorticity (PV), provided that the PV is e rapith ratidd decreasoan waten i e r vapor mixing defined in the conventional way (see glossary follow- ratio with altitude observed just abov tropopausee eth . ing section 9). Wit conventionae hth l definitio PVf no , n Wofteca e n identify "tropospheric d "stratoan " - extratropicae th l tropopaus founs ei d from observation spheric r parcel"ai theiy sb r differing chemical com- to be remarkably close to the 2-PVU potential vortic- positions. If only for this reason, it seems likely that ity surface, where PVU denotes the standard potential we shall continu distinguiso et h between troposphere vorticity unit (1 PVU = 1(T6 m2 s"1 K kg"1). The and stratosphere and to regard quantitative measures tropopause slopes downwar polewardd dan , intersect- transporofthe t between interestinthean mas g aspect ing isentropic surface s suggestea s e heavth y yb d of transport. curv Figurn ei usualls i ed 1an ,y distinguishe eacn do h The stratosphere and troposphere are sometimes such intersecting isentropic surfac bana y sharf eb do p regarde isolates da dexchange boxeth d san e problem gradients of PV, with values typically near 2 PVU, as merely one of estimating a single mass transport whose instantaneous configuratio y wele b ma nl rate between the two boxes. However, the relatively strongly deformed, indeed irreversibly deformedy b , long timescale verticar sfo l transport withi stratoe nth - synoptic-scale baroclinic eddies and other extratropi- e inhomogeneitspherth d an e e stratospherith f o y c l weather-relateca d disturbances. photochemical environment, as regards both actinic Figure 2 and Plate 1 (also Plate 2 below) illustrate radiative fluxe d chemicaan s l tracer distributions, how large these deformation e instantaneouth f o s s compe tako t s a leseu l s simplified viewpointE ST . typicaa tropopaus n o e l b intersectin n eca g isentrope, cannot, in reality, be thought of in terms of slow thi320in s K case . Figur showe2a coarse-graisa n view transport between two well-mixed boxes; single-num- isentropie oth f c distributio , estimatePV f no d froma r measurebe f exchango s e thereforear e y themb , - state-of-the-art operational weather analysis. Figure selves limitef o , d utility. More useful measureE ST f so 2b show correspondine sth g satellite imag5.7e th -f eo must concern not only the transport across the to 7.1-fxm water vapor absorption band, which is sen- tropopaus t alse rat bu et oth whic a e h tropospheric sitive to total water vapor above about the 500-hPa material is supplied to and removed from the regions in surface. Darker areas correspond to regions of low the stratosphere in which there are chemical sources water vapor and provide a finer-scale view of and sinks, for whichever chemical species are of in- tropopause deformations that bring dry stratospheric 33/ REVIEW 4 , GEOPHYSICF SO S Holton et al.: STRATOSPHERE-TROPOSPHERE EXCHANGE • 405
30 E=
1000 -90 60 90 South Pole A/ortA? Pole Latitude
Figure 1. Latitude-altitude cross section for January 1993 showing longitudinally averaged potential temperature (solid contours) and temperature (dashed contours). The heavy solid contour (cut off at the 380-K isentrope) denotes the 2-PVU potential vorticity contour, which approximates the tropopause outsid tropicse eth . Shaded areas denot "lowermose eth t stratosphere," whose isentropi r potentiaco l temperature surfaces span the tropopause. Data are from UKMO analyses [Swinbank and O'Neill, 1994]. In these dat tropicae ath l tropopause occur , whicsK nea 0 somewhas hi ~0 38 r t higher tha meae nth n tropical tropopause potential temperature (6 ~ 370 K) determined from radiosonde data by Reid and Gage [1981]. Polar tropopauses are far more variable and less well defined than might be suggested here. Figure courtes . AppenzellerC f yo .
air equatorwar downward dan d tha availabls ni e from by mixing "PV substance" (see glossary) along the the coarse-grain PV analysis alone. Plate 1 shows an tropospheric portions of isentropic surfaces. Such even finer scale vie f tropopauswo e deformations a s mixing does f courseo , , involve irreversible deforma- derived from an accurate simulation of an advected tion of PV contours on isentropic surfaces, as passive chemical tracer [Appenzeller et al., 1995]. The glimpsed in Figure 2 and Plate 1, and erodes or en- tracer follows the instantaneous tropopause, defined trains stratospheric material into the troposphere. 2-PVe herth s ea U contour, fairly closely ove daysr4 . Lindzen [1994] argues fro notioe mth f "baroclinino c The coloring marks air that had PV values >1 PVU adjustment" that such PV mixing on isentropes span- initially; most of this air, with initial PV values ^2 ning the troposphere should have a self-limiting char- PVU, stratospheree stayth n si . Fine details aparts i t i , acter, tendin t onl gno sharpeo y t tropopause nth t ebu clear that the instantaneous tropopause on the 320-K also to push it to higher altitudes and hence to exert a isentrop makins ei g north-south excursion 300f so m 0k negative or stable feedback control over its position in or more; such excursions are commonly observed, middle and high latitudes. The idea that tropospheric especially near the ends of storm tracks, as in this baroclinic eddies influence extratropical tropopause example, and especially in what meteorologists call height was first suggested by Held [1982]. However, a "blocking , moror " e generally, "low zonal index" quantitative overall picture of how different processes situations. combine to maintain the observed climatological-mean It is generally thought that the climatological-mean tropopaus stils ei l lacking. structure and position of the tropopause, indicated Give quasi-materiae nth l naturtropopausee th f eo , approximately in Figure 1, is determined both by ra- it follows that reversible deformations of its shape, diative-convective adjustment (especiall trope th n -yi however large and rapid, are of little significance to ics where deep convection clamps tropospheric tem- transpor themselvesn i t s alreadwa s yA . suggested, peratures close to a moist adiabat [e.g., Emanuel et what is significant is the irreversible transport across al., 1994]) and also by the action of synoptic-scale the tropopause, more precisely, across some represen- baroclinic eddieextratropicse th n i s e latteTh . r idea tative mean positio e tropopauseth f no e b , y thama t fits well wit e suggestiohth f Mclntyreno Palmerd an associated, for instance, with the synoptic-scale ba- [1984] that such eddies can sharpen the isentropic roclinic eddie r witso h smaller-scale processes exem- gradients of PV marking the extratropical tropopause, plifie Figurn d i Platr dwit o an 1 eh2 e global-scale Figur . Isentropi2a e 320-e c, 1992th contourn K14 o t 120 , y calculatea ,surfac V 0P UT Ma f r so efo d from European Centr r Medium-Rangefo e Weather Forecasts (ECMWF) operational analyses instantae Th . - neous tropopause appear bane th dottef ds o s a d contours firs(0.5e PVU5 th t 1. ,d soli 1.0d )an dan , contour highed ( an 2als e PVU)U ar ro showPV 4 ; contourd nan solid3 r sfo . (From Appenzeller t al.e [1995].)
Figure 2b. Meteosat 5.7- to 7. l-jjim water vapor image for the same time as Figure 2a. (From Appenzeller . [1995].eal t ) 33, 4 / REVIEWS OF GEOPHYSICS Holto t al.ne : STRATOSPHERE-TROPOSPHERE EXCHANG7 40 E•
14/5/92 12GMT
Plate 1. Passive tracer on 320-K isentrope for the same time as Figure 2, May 14, 1992, 1200 UT, estimated from ECMWF-analyzed winds using a high-precision Lagrangian tracer-advection algorithm ("contour advection") [see Norton, 1994; Waugh and Plumb, 1994]. The tracer contours were initialized 4 days earlie coincideo rt that a , t time 2-, , ,1- wit 3-e ,5-PVh d th 4- an , U contour smootheda n so isentropic showinV maP f po small-scalo gn e structure. Values correspondin higher o colorede U rar PV 1 o g.t (From Appenzeller . [1995].al t e )
diabatic ascent or descent. But how do events near the inFigurn i g show3 e e regioth s n withi e loweth n r tropopause fit into a global-scale picture of general stratosphere most directly affected by these eddy circulatio transportd nan nexe ?Th t section sketches transport effects. This region will be referred to as the what is involved and gives the plan of the rest of the "lowermost stratosphere" or, when clarity demands review. "lowermose th its a , t extratropical stratosphere."s i t I also the stratospheric part of what Ho skins [1991] calls the "middleworld." GLOBAE TH D L AN CIRCULATIONE ST . 2 , r chemicaFo l purposes especially e lowermostth , GENERAL APPROACH REVIEE PLAD TH F AN ,W NO stratosphere mus e sharplb t y distinguished froe mth tropospheri e middlc th par f o te world, where moist Figurattempn a s i e3 summarizo t e par whaf o t s i t convection strongly transports material across isen- involved usefus i t I .distinguis o t l hbetweenl firsal f o t , tropic surfaces. Equally lowermose th , t stratosphere hande on e ,o ntransporth t along isentropic surfaces, mus e distinguisheb t de stratofro th e res f mth o -t of the kind illustrated in Plate 1, which may occur sphere t beini , onle gth ye stratospher parth f o t - eac adiabatically (wave yth arrown o d Figurn i san , e3) cessible fro troposphere mth transpora evi t along isen- other hand transport across isentropic surfaces, which tropic surfaces tere mTh . "lowermost stratosphere" may require diabatic processes, including small-scale will help keep both distinctions clear. three-dimensionally turbulent processes. Since the Now air in the region where isentropic surfaces lie tropopause intersects the isentropes, transport can entirele stratosphereth n i y , referre y Napieb o t dr occur in either way and is likely to occur in both ways. Shaw and Hoskins as the "overwork!," cannot reach e e upperegioIth th n f ro n tropospher d lowean e r the troposphere without first slowly descending across stratosphere consistin f isentropigo c surfaces tha- in t isentropic surfaces, a process that must be accompa- tersect the tropopause, air and chemical constituents nie diabatiy db c cooling. Conversely e "unth -n i r ai , ca irreversible nb y transporte adiabatis da c edd- ymo derworld," where isentropic surfaces lie entirely in the tions lead to large latitudinal displacements of the troposphere, cannot reach the stratosphere without tropopause Figurn i Platd s an a ,e e2 1, followey b d first rising across isentropic surfaces isentropie Th . c irreversible mixin smaln go l scales darkee Th . r shad- surface bounding the overworld and the lowermost 408 • Holton et al.: STRATOSPHERE-TROPOSPHERE EXCHANGE 33, 4 / REVIEWS OF GEOPHYSICS
10
30
.a
3 100 2 0_ Some cumulonimbus clouds penetrate 300 stratosphere
Two-way exchange blocking anticyclones cut-off cyclones tropopause folds 1000 Pole Equator Latitude Figur . 3 eDynamica l aspect f stratosphere-tropospherso e exchange tropopause Th . s showe ei th y nb thick line. Thin line e isentropiar s r constanco t potential temperature surfaces labele kelvinsn di e Th . heavily shaded region is the ''lowermost stratosphere," where isentropic surfaces span the tropopause and isentropic exchang tropopausby e e folding occurs regioThe . n abov 380-the e K surfacthe is e "overworld," in which isentropes lie entirely in the stratosphere. Light shading in the overworld denotes wave-induced forcing (the extratropical "pump") wave Th . y double-headed arrows denote meridional transport by eddy motions, which include tropical upper tropospheric troughs and their cut off cyclones, wels a theis a l r midlatitude counterparts including folds [see Hoskins al.,t e 1985, Figur eddl eal t 2a]y No . transports are shown, and the wavy arrows are not meant to imply any two-way symmetry. The broad arrows show transpor global-scale th y b t e circulation, whic extratropicae drives hi th y nb l pump (see sectio Thi. n3) s global-scale circulatio primare th s ni y contributio exchango nt e across isentropic surfaces (e.g., the 400-K surface) that are entirely in the overworld.
stratosphere generall a potentia s ha y l temperature essentia e misleadingl b (indeed y ma t i ,measuro t ) e around 380 K, depending on cloud top heights (see STE by the transport across the tropopause. For many Figure 3). In this review the 380-K isentrope will be purposes, transport across another control surface used to designate the lower boundary of the over- ma more yb e relevan mord an t e effectively evaluated. world, but the reader should keep in mind that the r exampleFo , conside case chemicaa th rf eo l species actual boundary is somewhat variable. (e.g., methane (CH4), nitrous oxide (N2O), or one of It is useful then to distinguish transport in the over- the chlorofluorocarbons (CFCs)) that has a tropo- world from transport in the lowermost stratosphere. spheric source and a stratospheric sink, with the sink Transport in the lowermost stratosphere requires con- overworlde beinth n i g , abovtranse so r Th . o e - 1m 8k sideratio detaile th f f synoptic-scalno o s smalld ean - port across the 380-K isentrope (see Figure 3), which, scale processes and how these link to the overworld. s wil a arguee largeln b l ca , sectione n 4 di yb d an 3 s Horizontal mixinespecialle b n gca y significane th n i t understoo global-scale pars th d a f to e circulatioe th f no lowermost stratosphere and, most of all, as was al- overworld, is then a perfectly acceptable measure of ready mentioned, during blocking events, when me- exchange, indeed often more relevant because of the ridional motions are enhanced. Thus exchange be- higher location of the photochemical sink. tween the troposphere and the lowermost stratosphere The same applie a specie o t s sa strato thas ha t- can be significantly faster than exchange between the spheric source and a largely tropospheric sink. In this overworld and the lowermost stratosphere. context, detail transpore th f so t acros tropopausee sth , The distinction between the overworld and the low- involving processes suc s tropopausa h e foldind an g ermost stratosphere is therefore of prime importance. overshooting tropical cumulus anvils, are largely irrel- t shoulI cleae db r that owin thio gt s distinctiot no s i t ni othee evantth n r O hand. , such detail importane sar t 33, 4 / REVIEWS OF GEOPHYSICS Holton et al.: STRATOSPHERE-TROPOSPHERE EXCHANGE • 409
for determining the locations and intermittency of STE other timescales of interest. For example, to under- and for calculating measures of STE relevant to spe- stand why air moves toward the inlet of any suction cies that have appreciable sources in the part of the pump, one needs to recognize that the travel times of stratosphere belo 380-e wth K isentrope (and therefore acoustic waves are short in this sense and that there is lowermose inth t stratosphere) exampler fo , , from air- a corresponding nonlocal effect. Mass conservation craft emissions. has a key role and, for practical purposes, acts instan- Indeed, recent development atmospherin si c chem- taneously, as a nonlocal constraint. The time delay istry suggest thalowermose th t t stratospher- up d ean between turning on the pump and the establishment of per troposphere are more important for global chemi- the flow toward the suction tube is of the order of an l processeca s s thaformerlwa n y thoughtr Fo . acoustic propagation time, usually negligible in com- example , e heterogeneousomth f o e s chemistr- re y parison wit l othehal r timescale interestf so . This non- sponsible for observed ozone depletion occurs in the local picture is to be contrasted with statements like lowermost stratosphere, and much of the tropospheric "air moves toward the suction tube because the pres- nonurban photochemical ozone production occurn i s sure-gradient force pushes it," which mis poine th s t the upper troposphere [WMO, 1995]. Understanding thapressure th t e gradient adjusts nonlocallyn i t fi o t , and modeling the chemistry of these regions requires with the mass-conservation constraint. The pressure- tha knoe w ttemporae wth spatiad an l l distributiof no gradient force, whether written on the right or the left trace chemical transport in the upper troposphere and of the momentum equation, cannot usefully be re- lowermost stratosphere, both alon acrosd gan s isen- garde causins da flow e nonlocae gth Th . l charactef ro tropic surfaces. The wavy arrows in Figure 3 are not, adjustmene th e expresseb n ca t d mathematicalla yvi of course, meant to suggest the complete picture, even an elliptic (see glossary) partial differential equation for transport along isentropes. The "exchange" need for the pressure, formed by taking the divergence of nosymmetrice b t ; therwely strongee eb ma l r trans- momentue th m equation. ports along isentropes within the lowermost strato- There is a fundamentally similar nonlocal effect sphere, as well as outside it. having direc tproblem E relevancST e th , o namelyet , framewore Outh rn i thefirsplac o m t E s ai nti k eST the effect of the extratropical stratosphere and mesos- of the general circulation and to clarify the roles of the phere upon the tropical stratosphere. This depends on different mechanisms involve varioun do s scales. Sec- face th t thaglobal-scale th t e travel time f acoustiso c begi4 d tionreviewiny nb an s3 r understandingou f go waves and large-scale gravity waves are short in com- global-scale transport, considering firs mais it t n mech- parison with other timescales of interest. The effect anisa relevan a mvi t idealized mode d thee an lth n s beeha n demonstrated, indee s beedha n intensively application to the real atmosphere. Section 5 considers studied dynamicae th n i , l literature beginning wite hth quantitative estimates of global transport made on the pioneering work of Eliassen [1951] and Die kins on r theoreticabasiou f o s l understandin f globalo d gan - [1968]. These and many other studies have shown that scale dynamica chemicad an l l observations. Section6 the extratropical stratospher mesospherd ean pert eac - reviews synoptic-scal d smaller-scalan e e aspectf o s sistently upon the tropical lower stratosphere as a kind exchang extratropicse th n ei . Sectio considern7 e sth of global-scale fluid-dynamical suction pump, driven tropics, where certain aspect exchangef so , principally by certain eddy motions. The distinction between concerned with water vapor, clearly do depend cru- tropics and extratropics arises from the Earth's rota- cially on small-scale details. The representation of tion whicn o , e extratropicahth l pumping actio- nde both global-scale and small-scale aspects of exchange pends. in model consideres si section di . Finallyn8 suma , - confidene b n Onca ee reasot w abouy sige nwh th t n mar conclusiond yan givee sar section ni , followen9 d implied by the word "suction" is that statistically by a glossary of technical terms. speaking mose th , t importan edde th yf o tmotion n si question, so-called "breaking Rossby waves- re d "an lated potential-vorticity-transporting motions, have a . 3 GLOBAL-SCALE DYNAMICAL ASPECTF SO one-signed or ratchet-like character related to the EXCHANGE: THE EXTRATROPICAL PUMP sense of the Earth's rotation. (For detailed examples and historical remarks, seer instancefo , , Mclntyre A prime requirement is to understand how the pro- [1993, and references therein]). These eddy effects cesses suggeste Figurn di t togethefi e3 r mechanisti- hav givestrono a t persised a gan tendencp u - d ad o yt cally. One prerequisite to such understanding is to tently one-way pumping action, whose strength varies recognize the existence of nonlocal dynamical effects seasonall interannualld yan whic n grads i i d r hai -yan atmospheree inth . ually withdrawn from the tropical stratosphere and Recognition of such nonlocal effects is crucial to pushed polewar ultimateld dan y downward. Wherr eai understanding any fluid-dynamical system that sup- parcels are being pulled upward (as in the tropical ports fast wavese appropriatth n i , e sense, tha, is t stratosphere), adiabatic cooling pulls temperatures be- waves whose travel times are short in comparison with low their radiative values; wher parcelr eai beine sar g 410 • Holton et al.: STRATOSPHERE-TROPOSPHERE EXCHANGE / REVIEW 4 , 33 GEOPHYSICF SO S
pushed downward (as happens most strongly in high nominae th dr s i y airlm k (constant =,H 1 ) density latitude n wintei s d spring)an r , adiabatic warming scale height used in the log-pressure coordinates, a is pushes temperatures above their radiative values. the radius of the Earth, and p0 is a nominal basic state This mechanically pumped global-scale circulation density proportional to exp (-z/H). In (3), N2 is the is often referred to by dynamicists as the "wave- squar buoyance th f eo y frequenc measura y( statif eo c driven circulation. s existencIt " t onlno ey explains stability), defines da why observed temperatures differ fro e radiativmth e values toward which they would otherwise tend to R relax [e.g., Dickinson, 1969; Pels, 1985] but is also an essential part of what controls the tropospheric life- times of long-lived trace chemicals such as CFCs, where referenc a J s 0i e temperature dependent onln yo nitrous oxide, and methane that have photochemica R/c= 5 K pl , d wheran z e specifieth cs i pt a c r heaai f o t sinks in the stratosphere. The main uncertainty today constant pressure. Finally , Q/(fd Qe shortan s ar ) - is not the physical reality of the extratropical pumping wave and long-wave radiative heating rates, respec- action, but its precise strength and its seasonal and tively e zonath , s ui l wind component teme th s -i T , interannual variability certaid an , n nonlinear aspects perature, and tJ* and w* are the latitudinal and vertical of the resulting circulation dynamics that bear on the velocity components, respectively. These quantities precise way the pumping action reaches into the trop- may wils a explaine,e b l d later interpretee b , cers da - ics. tain zonally averaged quantities in the real atmosphere e nonlocaTh l effect n questioi s n operate juss a t three-dimensionan i r o l atmospheric models (thas i t robustly in a zonally symmetric, that is, longitude- why the overbars are added). For the moment, how- independent, model atmospher reae th l n thes i e a o yd ever, u, J, tJ*, and w* are simply taken as the zonally atmosphere. We can gain more detailed insight into the symmetric thoughe fieldth f o s t experiment whicn i , h pumping action, including the way in which the im- we shall imagine, furthermore, thawave-inducee th t d plied adiabatic warming and cooling interacts with forc prescribeds i eG equatione Th . s will then describe radiatio n differeno n t timescales y considerinb , a g pumpina g actiokine questionth n di f no global-scal:a e thought experiment performed on a zonally symmetric manifestatio f fluid-dynamicano l nonlocalness. atmosphere in which the one-signed Rossby-wave ef sake - th mathematicaf eo r Fo l simplicity, these par- fects are parameterized simply as a retrograde or west- ticular model equation approximatione th e sus - so f so ward wave-induced force. This force plus the Earth's called quasi-geostrophic theory [e.g., Andrews et al., rotation is the essential cause of the pumping, which 1987] quasi-geostrophie Th . c approximation adee sar - thoughe "quasi-gyroscopicb a s n a ca f to " effece th n i t quate for gaining qualitative insight into the mecha- sense that pushing a ring of air westward has a ten- nism of the pumping action and the timescales in- dency to make it move poleward. The relevant theo- volved. Studie n whici s e approximationth h e ar s retical formulations wer- e first pa forwar e pu t th n di relaxed, usin morthe g e accurate "primitive equa- per Eliassey b s Dickinsod nan n already referre, to d tions," are known to give qualitatively the same re- and Mclntyre [1992] provides an elementary introduc- sults [e.g., Haynes t al.,e 1991]. e Equatioth s i ) n(1 tion highlightin e underlyinth g g physical principles zonal momentum equation t stateI . s thawavee th t - with some illustrative thought experiments. induced force is balanced by the zonal flow accelera- A suitabl modef o t ese l equation sucr sfo hzonalla y tion minus the Coriolis force due to the latitudinal symmetric atmosphere is as follows, with the wave- motion'tJ*. Equatiothermae th s i ) ln (2 win d equation. induced force per unit mass denoted by G: t stateI s thae verticath t l zonae sheath f lo r wins di proportional to the latitudinal gradient of the temper- du ature. This important coupling between the wind and — - 2fl sin <|>t7* = G (1) dt temperature fields follows from the assumption that e zonath l hydrostatin floi s wi geostrophid can c bal- dT R du ancee theorth . w yThi ho s includepari sf o t e th s sin <)> — + — — = o (2) dz aH d$ all-important nonlocal effects. Equatiothere th s -i ) n(3 modynamic energy equation t stateI . t sne thae th t ar HN' diabatic heating is balanced by the sum of adiabatic Hi' ~R~ (3) cooline verticath y gb l e temperaturmotioth d nan e tendency distinguise w ) (3 n I . h betweee pare th nf th to l la diabatic heating tha largels i t y independen temperf o t - cos = 0 (4) s c|>co dc|a > diabatic heating). Equation (4) is the mass-continuity control exerted by a localized forcing G upon the or mass-conservation equation. It relates the latitudi- vertical mass flux pw*. Whae spatiath s i t l structure nal and vertical components, respectively, of the ve- of the response to a localize0 d forcing? Inspection of (5) locity in the latitude-height plane. This velocity field is shows that all aspects of the response depend on the often referred to as the "meridional circulation." parameter a/a. As a/a -> <», the adiabatic limit studied Equations e regarde(l)-(4b y s ma predictiv)a d e by Eliassen [1951 extended an ] sphericaa o dt l geom- equations for the "unknowns" du/dt, dT/dt, t7*, and etry by Plumb [1982] is approached. A value for a of w*. Given the wave-induced force, which appears 1/(20 days) is often taken as representative of the lower solely as the G term on the right-hand ^ide of (1), and stratosphere. The adiabatic limit will therefore hold for the contributions to diabatic heating, Qs and Qh they timescale lessof sdays thafew na . This limi approis t - may be combined to give a single equation for one of priate for transient fluctuations in G, such as those the unknown quantities. In order to simplify the cal- induce y Rossbb d y wave n stratospherii s c sudden culation timescale e focuo t th d n so s an importancef so , warmings. In that case, /a/(/a + a) —> 1, and since it is convenient to follow Garcia [1987] and to assume away from the forcing region the two terms on the that the time dependence is harmonic, with constant left-hand side in (5) must balance, simple scaling then frequenc . (Sinca y problee eth mlinears i , more gen- implies tha verticae th t l pscaln i 0 w z relates *i eA o dt eral time dependencies can be_deduced by Laplace the latitudinal scale ak$y b i(Tt transformation.) (Gee Thu R Qwritd e s= sw an ) eG e (Qe R characterizd l(Jtan = ) e responsth ee th y eb 2ft sin 4ft2 sin2 <|> (aAc|>)2 Ax z ~ma single variable w such that the vertical velocity w* = N N2 H Re [vH4>, z)e]. As was emphasized earlier, it is important thatlat the relaxational dependence of the dia- where the left-hand part of the expression applies batic heating field on the temperature be taken into when Az < H and the right-hand part applies when account conveniens i t I . parameterizo t t e this depen- A H.z> Note that whe left-hane nth d part appliese th , dence by assuming Q to be given by Newtonian cool- rati f verticao l scal latitudinao et l scale equale th s l 1 ratio of the Coriolis frequency to the buoyancy fre- ing with constant timescale a" , that is, Qt = -aJ, temperature th w witno hT e anomaly. Wit above hth e quency, which in middle latitudes is of the order of assumptions, e combine(l)-(4b n ca ) d int a osingl e 10~exampln A . 2 f thieo s adiabatic limi s showi t n ni equation for w of the following form: e arroweFigurTh . 4a de e mass-flucurveth e ar s x streamlines of the response to a simple forcing, con- d_ I 1 d(p0w) sisting onlwestwara f yo dapplie0 e forc< th n eG di dz \p dz shaded region. Note that the mass-flux pattern extends 0 mainly downward and sideways and that the right- 2 N d /COS c|> dw hand part of the above expression for Az is relevant. /a + a / 4£l2a2 cos sin2 cj) dcj) The pumping action immediately reaches to a great depth proportional to the square of the latitudinal scale a (RIH) _d_ /cos (fr d& of the pattern, the density scale height H being far smaller than the vertical scale of the pattern. For /a + a/ 4£l2a2 cos \sin2 c|) practica l e botto purposey thinth f ma f o km o e w s 1 a cos aG\ Figure 4a as the bottom of the stratospheric over- (5) world, around 380 K or 15 km above the Earth's cos sn surface pumpine Th . g action reaches sideways across This equation provides a suitable model for studying the tropic int d opposite soan th e hemisphere agreen i , - nonlocal control of the meridional circulation by the ment with observation large-scalf so e transient events in the real atmosphere [e.g., Dunkerton et al., 1981; wave-induced force and by the short-wave heating Qs, which we also, of course, take to be prescribed. Randel, 1993]. Since the operator acting on w on the left-hand side If a > a, that is, for timescales of a month or more of (5) is elliptic (see glossary) when a/a ^ 0, it follows when a ~ 1/(20 days), the characteristic scales of the tha wave-inducea t d force a short-wav r o , e heating, response are related by localize a particula o t d r region, will give a ris o t e a\1/2 2fl sin response that extends away from that region. Con- Az ~ max versely givea n i nw , region wil affectee lb changey db s cr/ TV in the wave-induced force or the short-wave heating elsewhere. This is how the mathematics expresses the 4fl2 sn (6) nonlocalnessthereford thay an sa w t n eca e vve ar * W . H "under nonlocal control." For reasons to emerge (e.g., Figure 4), we shall be squar e latitudinae Th th f eo l scal agais ei n relevano t especially intereste e downwarth n di sidewayd dan s response th largese th givey t ea an t n r scalesfo d an , 41 2• Holto t al.ne : STRATOSPHERE-TROPOSPHERE EXCHANGE 33, 4 / REVIEWS OF GEOPHYSICS 601——i—————,——————r- ~i——:———r a 50 40 30 20 10 0 -80 -60 -40 -20 0 20 40 60 80 Figure 4. Idealized numerical experi- Latitude response mentth meae n th so f neo strato- 60 spheric zonally symmetric circulation in the latitude-height plane to a westward force 50 applie shadee th n di d region. Contoure sar streamlines, wit same hth e contour interval use eacn di h panel ) Adiabati(a , c response 40 for a/a :» 1. (b) Response for a/a = 0.34, correspondin annuao gt l frequenc- 20 d yan 30 day radiative damping timescale; the solid % and dashed contours show the response that s f phaso t eou wit° e phas90 n hth i d s i ean 20 forcing, respectively ) Stead(c , y stat- re e sponse (a/a <$c 1). Note that the magnitude 10 of the response in the mass stream function increases as a/a decreases and that it is 0 given quantitatively by (7) for Figure 4c. 0 6 0 4 0 2 -80 0 0 -2 -40 80 The model is unbounded below, so the or- Latitude dinat s arbitrar- re ha e e b yn origica t nbu garded, for instance, as showing approxi- 60 mate vertical distance above the lower boundary of the overworld. 50 40 30 20 10 0 0 6 0 4 0 2 0 0 -2 0 -4 0 -6 0 -8 80 Latitude latitudinal scale the response extends still deeper ver- response must therefore have a deeper vertical scale if tically than is the case in the adiabatic limit. This these two terms are to balance (as they must away occurs becaus forcine th s ea g timescale increases rel- fro forcine mth g region). Such deepenin s seegi n ni radiative ativth o et e damping timescale, thaa ta/ iss a , Figure 4b, where a/a = 0.34, corresponding to annual decreases e importancth , e coefficienth e f o th e f o t frequenc radiativa d an y e damping days 0 tim2 f eo . latitudinal derivativ eleft-hane terth n mo d) sid(5 f eo Also e sidewayth , s control reaches still farther into diminishes relativ verticae th o et l derivative terme Th . and across the tropics. 33/ REVIEW 4 , GEOPHYSICF SO S Holton et al.: STRATOSPHERE-TROPOSPHERE EXCHANGE • 413 e timescal e limith I nth s a t e becomes very large, expressio thes ni n consistent wit notioe hth n thae th t only the vertical derivative term remains on the left- response is diffusive, with timescale proportional to hand side of (5). The equation then changes its char- (Az)2, if Az <$c H, and propagating, with timescale acte hyperbolico rt streamlinee th d an , s away froe mth ^z proportiona(se>H A f ei , Haynes. Az al t o e t l forcin linee g ar constanf so t latitude (Figuree 4c)th n .I [1991] for more details). It follows from the above that same limi heatine th t gright-hane terth n mo d sidf eo the timescal o react e a hstead y state decreases a s (5) vanishes. It follows that in this limit the mean eithe latitudinae th r l scale latitud e (ak$)th r o e itself circulatio controlles ni d entirel wave-inducee th y yb d increases. Conversely, in the tropics the timescale to force, eve imposen Qf nA i . ^0 d short-wave heating reach a steady state becomes very long. For this and cannot driv persistenea t meridional circulatios i t nbu other reasons to do with nonlinear effects in the trop- simply balance adjustmenn a y de b long-wav th n i t e ics, the steady state limit, and therefore the downward temperature-dependent part of the heating. Radiative control principle f direco s i , t practical interest onln yi equilibrium will hold everywhere except beloe wth the extratropics. region in which there is a wave-induced force. There, To gain some quantitative insight, we might take Az adiabatic heatin verticae th coolin d o gt an e l circugdu - equa e andb H o o t t l, again radiative th , e timescalo et lation drive the temperature away from radiative equi dayssteadinese 0 -2 th e f .I b s criterio satisfiee b o t ns i d librium. for variations with the annual frequency, in the sense This steady stat ealsy limitanae oma b a —, -0 a/ », thaleft-hane th tt leasa s i factoa t d ) 2 f (8 sid o rf eo lyzed by considering the momentum equation (1) di- greater tha right-hane nth d side latitud,° the45 t e na eth rectly. In the steady state the acceleration du/dt = 0, horizontal scale must be at least 1100 km, while at 20° and the wave-induced force is balanced by the Coriolis it increases to 2000 km. force t followI . s from combinin- gre (1d ) an wit ) h(4 When the zonally symmetric model considered here quiring that pwz — »*s a °°— >0 , that w s give*i y nb is extended to apply to a zonally averaged circulation 0 reae inth l atmosphere downware ,th d control princi- 1 cos c|> ple (equation (7)) provides a framework in which to Po' Gdz (7) COS 211 sin c|> r interpre wave tth e drivin meridionae th f go l circulation and als osimpla e quantitative recip calculatinr efo e gth Thus in the steady state limit the nonlocal control of strengt f thiho s circulation, given observational esti- mase th s flux expresses becomha ) (5 purelea y db y zonae mateth f lso mea nRossbo t forc e du ey waves downward control, as illustrated in Figure 4c. A spe- and to gravity waves. Since the climatological distri- cial case of this was noted by Dickinson [1968]. The bution zonaf so l mean Rossby-wav gravity-wavd ean e vertical velocit t somya e locatio controlles ni e th y db forces in the stratosphere and mesosphere appear to distribution of wave-induced forcing (and its latitudi- hemispherie b scaln ci e possibls (Figuri t i , e5) e that nal derivative) above that location, as is explicitly estimates of the seasonally and zonally averaged ex- show expressioy nb n (7). This "downward control" tratropical meridional circulation downe baseth n do - principl discusses ei Haynesy d b . [1991] al t e , together ward control principle coul fairle db y reliable t leasa , t wit s generalizatiohit n beyond quasi-geostrophic the- e solsticith n e seasons. This beeha s n verifiea o t d ory. remarkable degree with observational data and with downware Th d control principl valis ei de onlth n yi output from a general circulation model; see section 5. steady stat es Figur a limit d remind4 ean e , th , us s Wha especialls i t y remarkabl apparene th s ei t suc- approac t unifor limie no th s i to h latitudet mn i . Noting cess of such calculations even in the subtropics, be- e steadth tha n i ty state limi e seconth t de terth n mo yond their domai validityf nnotes o wa ds earlierA . s a , left-hand side of (5) vanishes, we deduce that the value the equator is approached, the timescale to achieve the of a (and hence the timescale) for which the response steady state n whici , e wave-induceth h d force G approaches steadines proportionas si radiative th o t l e comes into balance wit Coriolie hth s force associated time a"1 and is given by with the meridional flow, becomes longer and longer. e wave-induceThuth f i s d forc s nonzerha e o fre- N2 quency, there must always be a low-latitude region in sm Zonal Forcing, January 1993 These extra terms may be incorporated into the theory most conveniently if (1) is replaced by the angular momentum form: dm v* dm dm _ — + — — + w* — - (a cos c|>)G (10) dt a d$ dz e absolut s $(uth s c)>+ s ailco i ) co a wher e = em angular momentum. (Again e notatioth , s usei dnm because the equation also applies for zonally averaged quantities when the flow is zonally asymmetric.) Al- though m is dominated by the contribution from the 0 3 0 0 -3 0 -6 -90 60 90 Earth's rotation, the latitudinal gradient of m, which Latitude appears in (10), is influenced by the contribution from Figur . 5 eWave-induce d zonal forc r unipe et massn i , G , the relative motion u, particularly at low latitudes. The unit f metero s r seconpe s r dayd pe r Januar ,fo y 1993- de , extra term e incorporateb n (10i s n )ca o givt d a e duced indirectly fro momentue mth m balance (equation (1)) nonlinear generalizatio downware th f no d contro- ex l usin gradiativela y derived meridional circulatioe th d nan pression (equation (7)), which then becomes an inte- 1 1 observed zonal mean wind .s" m Contour 2 d"t a e sar gral along contours of constant absolute angular mo- intervals, with additional contours of ±1 and ±0.5 m s"1 1 mentum. Further details are given by Haynes et al. d" . The eastward force centered near 60°S and 40 hPa may [1991] fule lTh . expressio bees nha n use Rosenlofy db be an artifact caused by the ill-conditionedness in calculating radiative heating rates from observed temperatures, espe- and Holton [1993] in their estimate of vv* from obser- cially in the summertime lower stratosphere. (After Rosenlof vations. [1995]). An important aspect of the downward control prin- ciple that is weakened in the tropics is that the change changa o it n wave-inducee n wei *du d forcino n s gi l/4 ^2 1/4 longer given entirely by appropriate modification of min [Az2, (9) the G appearing in (7) or its nonlinear generalization. Rather, sinc esignificana changete parth f o t d forcing The right-hand side of (9) may be interpreted as giving may caus a ezona l flow acceleration e resultinth , g the minimum horizontal distance from the equator for change angulath n i e r momentum distribution must the steady state downward control regime to hold. For als takee ob n into account latitudesw lo t .A , thene th , takez A nd thearlieran e , a value , a , thif so s distance proble f determininmo e meridionagth l circulation is about 1900 km, though a more accurate theory may given the wave forcing is fully nonlinear; the meridi- well giv somewhaea t larger value through nonlinear onal circulation and the angular momentum distribu- effects e belowse ; . These considerations, take- nto tion have to be determined simultaneously. It can be gether with results like Figur , strongle4b y indicate argued, from consideration angulaf so r momentu- mad that the most effective location for wave-induced forc- vection, that the effective value of 11 in (9) is reduced controo t s i annuae ingt i th lf i , l variatio uf npo welling by such nonlinearities, extendin reace gth f extratho - at the equator (see Yulaeva et al. [1994] and section 5), ropical pumping action intd acros oan e tropic th s s e suth b tropicsn iannuae si th n O l. timescal- up e eth even in the steady state. welling response to forcing at high latitudes will decay Another possibility arising from nonlinearite th n yi substantially at low latitudes (evident from Figure 4b), tropics is when the zonal wind distribution may be whereas the response to a force at a latitude equator- such that the angular momentum contours become war thaf do t right-hane giveth y nb dappear) sid(9 f eo s closeinterio atmospherthe the d in rof e (tha theris, t e primaril n accelerationa s a y , duldt, rather than a n is an interior maximum in angular momentum). It is induced meridional circulation. This in turn suggests a then possibl havo et freeea , closed meridional circu- crucial role for the position of the subtropical edge of lation that is everywhere parallel to the angular mo- the "surf zone" in the wintertime middle stratosphere, mentum contours. Sinc advectioe eth angulaf no - mo r where Rossby-wave breakin leasignificann a go ca d t t mentu (10mn i ) then vanishes, suc hcirculatioa n nca contributio. G o nt absencexisthe a wave-induce in t of e d force. This A second limitation on the downward control prin- regime was investigated for the tropospheric Hadley ciple (equation (7)) at low latitudes is the neglect of circulatio y Schneiderb n [1977 d Heldd Houan ]an relative angular momentum gradients ,f lati o tha , - is t [1980]. Dunkerton [1989] investigate e extendth o t t tudina verticad an l l advection actin gradientn go uf so whic hsimilaa r regimrelevane b equatoe y th ema o t - e quasi-geostrophiinth c versio e zonallth f no y sym- rial middle atmosphere, particularly at the solstices, metric model equations given above. At low latitudes ,modeled an formatioe dth low-latituda f no e - freme e such advection becomes increasingly important. ridional circulation cell associated with the cross- / REVIEW 4 , 33 GEOPHYSICF SO S Holto t al.ne : STRATOSPHERE-TROPOSPHERE EXCHANG5 41 E• equatorial gradien f radiativto e heatin middle th n gi e andMdntyre [1987] in connection with the similar but and upper stratosphere. However e relevancth , f eo tighter eddy transport barriers found at the edges of this nearly inviscid free circulatio time-avere th o nt - the winter polar vortices, such barriers a wor a kvi aged upwellin e tropicath n gi l lower stratosphers ei combinatio f dynamicano l effects involving both hori- doubtful. There is little resemblance, even at low lat- zontal shear on the one hand and isentropic gradients itudes, between this free circulatio circulae th d nan - of PV or "Rossby-wave elasticity" on the other tion calculated diabatically from observations [e.g., [Hoskins et al., 1985, section 6]. Solomon t al.,e 1986; Rosenlof, 1995], particularly subtropicae Th l eddy transport pera barrie t -no s ri with regard to the latter's broad region of tropical fect barrier, especially in the layer between 380 and upwelling in the lower stratosphere. Dunkerton [1991] 480 K [e.g., Trepte et al., 1993; Boering et al., 1995], showed that the addition of a wintertime extratropical bu stils maii e t th l nmean-circulatioe reasoth y nwh n wave force to his model enhances the circulation, picture suggested in Figure 3 is relevant to global-scale giving systematic upwelling in low latitudes and ex- chemical transport. It is the strong correlation be- tendin downwelline gth g circulation well int wine oth - tween the global-scale vertical velocity and the chem- hemispherer te . Thus whil ecleaa r mathematical the- ical mixing ratio eithen o s e barrier th sid f o er that ory of the connection between the time-averaged make the mean circulation chemically significant. As extratropical force and the time-averaged tropical up- has been emphasized by R. A. Plumb (personal com- welling remain e completedb o t s e wave-induceth , d munication, 1994), the net transport by the mean me- forcing appears to make the dominant contribution to ridional y circulationcirculationan y y b an r f o o ,, driving the circulation, including exchange across the chemical into the overworld would vanish if the mixing base of the overworld. In other words, it seems clear ratio were uniform on, say, the 380-K surface. for practical purposes that the mass transport across usuae Th l approac bees heighth e ha focu o nt th n so - any quasi-horizontal surface, even the tropical up- latitude structur f dynamicao e chemicad an l l tracer welling s primarili , y controlle e extratropicath y db l fields by dividing all fields into a longitudinally aver- wave-driven pumping above that surface. aged part (the "zonal mean""eddyn a d an )" part. Thernumbea e ear f alternativro e way whicn si e hth zona e computelb meay ma n d (see glossary)e Th . 4. GLOBAL-SCALE DYNAMICAL ASPECTS separation between zonal mean and eddy contribu- EXCHANGEF O : APPLICATION tions to transport differs according to the type of av- REAE TH L ATMOSPHERO T E eraging employed. These differences are manifested, The previous section has explained how nonlocal for example e strikinth y b , g difference betweee nth contro meridionae th f o l l circulatio- ndy arisee th n si conventional Eulerian-mean circulation measured in namic a zonall f o s y symmetric model atmosphere. pressure or log-pressure coordinates and the Eulerian- Ho wthis i s relevan transporo tt tracf o t e chemicaln si mean circulation measure isentropin di c coordinates the real atmospher realistin i r eo c numerical models, (i.e., with potential temperature as a vertical coordi- where "waves and eddies," or departures from zonal nate [e.g., Tung, 1982]. Of the two, only the latter, symmetry, are often very strong? often referred to as the "zonal-mean diabatic circula- As was already suggested, the most direct rele- tion," or diabatic circulation for short, provides a vance is to exchange across the 380-K surface, and qualitatively correct approximation to the advective there are two basic points. The first is that the stron- transport of tracers. Indeed, in the high-latitude winter gest departures from zonal symmetry are usually stratosphere the conventional Eulerian-mean circula- found in the extratropics. The second is that in the tion and the diabatic circulation are in opposite senses. overworld, there is still a sharp distinction between This is one of the reasons why the use of isentropic tropics and extratropics, evident from the observation coordinates greatly facilitates conceptual modeling of that many chemical specie f interes o se over th n -i t transport. world have different tropical and extratropical mixing isentropie Th c coordinate formulatioe n th alss oha ratios (recent references among many being Trepte important conceptual advantage of clearly separating and Hitchman [1992] and Randel et al. [1993]). This transport along isentropic surfaces from transport suggest t onca s e that horizontal eddy exchang- be e across isentropic surfaces. One qualitative advantage twee e tropicath n d extratropicaan l l overworlds i s of such coordinates is that eddy dispersion, which is alss weakowa notes a , d lonfroo gag m classic obser- primarily along isentropic surfaces, appears nearly vation f nucleaso r bomb test debris. Recen eviw ne t- horizontal. Thi extene trus sth i o et t tha eddiee tth e sar denc s mentionei e n sectioi d . Fluid-dynamica7 n l quasi-adiabatic [e.g., Tung, accorn i 1982s i d ]witan h model studies [e.g., Norton, 1994; Polvani et al., 1995; direct evidence from model simulations [Plumbd an Chen al.,t e 1994] show that thi explicabls si termn ei s Mahlman, 1987] e zonallTh . y symmetric mean dia- oa subtropicaf l "eddy transport barrier" against batic circulation, on the other hand, generally provides transport along isentropes. As explained by Juckes usefua l approximatio cross-isentropie th o nt c trans- 416 • Holton et al.: STRATOSPHERE-TROPOSPHERE EXCHANGE 33, 4 / REVIEWS OF GEOPHYSICS thus i pord s particularlan t y relevan r questionfo t f so by implication upwellinge th n i , e primarils )i th o t e ydu STE. interhemispheric contras wintertimn i t e stratospheric On the other hand, pressure coordinates have some Rossby-wave activity and not to annual cycles in sea advantage r purposefo s f dato s a analysis r thiFo s. surface temperature tropicad san l convective activity. reaso nmodifiea d versio Eulerian-meae th f no n equa- Thi interestins sha g implication STEr sfo e , botth r hfo tions in pressure coordinates, the "transformed Eule- global-scale upwellin dehydrae g th ratals d r ean ofo - rian mean" (TEM), is often used in practice. The TEM tion mechanism, which requires sufficientlw lo y provide measursa global-scale th f eo e meridional cir- tropopause temperatures (and which will be examined culation tha similas ti diabati e th o rt c circulatioe th f no in more detail in section 7). isentropic coordinate formulation. The TEM equa- Refinements to this picture of nonlocal control of tions have exactly the same structure as the model the global-scale circulation by G must take account of equations (l)-(4), except that zonally averaged quan- the following: tities replace zonally symmetric quantities. Further- 1. The detailed latitude-height distribution of G more, eddy effect indeeo d s d appear gooo t , d approx- depend wavn so e refraction, wave breaking radid an , - imation s wave-inducea , d force thin i s s formulation. ative damping of waves in the middle atmosphere. Rossby-wave Th e par sucf to h forces involves "break- Although broad-scale tropical stratospheric upwelling, ing," as well as dissipation by infrared radiative relax- for instance largels i , y controlled_b extratropicae yth l ation hencd an , e involves eddy dispersion. Thue sw fieldG , ther feedbacke ear itselG n fo s through any- may apply insights gained fro zonalle mth y symmetric thing that affects the extratropical wave fields, includ- model to deduce that the global-scale meridional cir- ing tropospheric wave sources. culation, and therefore the zonal mean transport of 2. A limited role for transient responses to heating tracers across isentropic surfaces, is under nonlocal stratosphere inth tropospherd ean e neey b o edma t control by the zonally averaged wave-induced forces. recognized time-dependene Th . t linear theor symf yo - It follows therefore that on sufficiently long time- metric circulations reviewe previoue th n di s section scale transpore sth t acros lowesa r stratospheric con- does predict some penetratio e circulatioth f no - nin trol surface, including r examplefo , e 380-th , K isen- duce tropospheriy db c heating int lowese oth t parf o t trope largels i , y controlle wave-inducey db d forcen i s stratosphere th probabls i t ebu y more relevane th o t t extratropicae th l middle atmosphere, whic turn hi e nar seasonally varying signal [Dickinson, 1971b] than to believed to be mainly due to Rossby and gravity waves longtime th e average [Dickinson, 1971a]. Indeed, indi- originatin extratropicae th n gi l troposphere. Fro- mob cations fro e steadmth y state nonlinear theoriee ar s servations and modeling it appears that wave-induced that tropospheric thermal forcing cannot alone drive forces in both the stratosphere and mesosphere are any upwellin lowee th n gi r stratospher . PlumbA . e(R , important contributor e distributioth o t s f waveno - personal communication, 1994) verticae Th . l penetra- induced forcing (Figur wholee th e Rossbyn e 5)O th ., - tion distance, for a circulation forced from below by wave force makes the most important contribution to heating that varies on the timescale of the annual transport across lower stratospheric control surfaces, estimatee cycleb n ca , d fro requiremene mth t thae th t though there are important variations with latitude and two sides of (6) balance. If the latitudinal scale of the season, the mesospheric gravity wave force being par- tropospheric heatin followt takes gi 100e i , b o n0t km s ticularly importan southere th n i t n hemisphere winter thaverticae th t l penetration distancr o onls m ei k y2 [e.g., Garcia andBoville, 1994]. Furthermore, gravity so. All this suggests that, while there will certainly be wave forces in the summer hemisphere and middle and some influence of tropospheric heating on local up- upper stratosphere, though not well known, are almost welling rates in the very lowest part of the tropical certainl essentian ya risine th gw l brancparho f o t f ho stratosphere primare th , y control, determining whas ti the stratospheric circulation varies seasonally. take intp stratospherie nu oth c overworld exertes i , d Note wele implicatioth l f thio n s e picturth r fo e by extratropical stratospheric pumping. cause upwellinf so tropicae th n gi l lower stratosphere. morA 3. e accurate pictur role eddief th e o f eo n si The broad-scale upwelling mass flux averaged over the cross-isentropic transport e developedneedb o t s . tropic s largeli s y independen f locao t l processes. Even in the isentropic formulation the effect of the These include both the large-scale tropospheric heat- eddies on transport is not purely dispersive, nor is the ing field due to moist convection and the mass injec- cross-isentropic transport purely advective. This i s tio cumulonimbuy nb s turrets penetratin e lowegth r inevitable witkiny f hEuleriaan do n mean. Mahlman stratosphere e approximatTh . e independence th f o e et al. [1980], in their pioneering study, have noted that broad-scale upwelling from such heating and cumulo- theradvectivn a s ei eedde parth yf o ttranspor e th n ti nimbus effect s sometimei s s regarded with surprise, isentropic-coordinate formulation (not included in the inescapabln a s i t i t bu e consequenc fluid-dynamicaf eo l mean diabatic circulation) that plays an important role nonlocalness as manifested by extratropical pumping. in the steepening of tracer surfaces relative to isen- Thus Yulaeva . [1994al t e ] have argued thae th t tropic surfaces. Indeed, they sugges wintete thath n ti r annual cycle in lower stratospheric temperatures (and lower stratosphere this contribution, which is caused 33, 4 / REVIEWS OF GEOPHYSICS Holton et al.: STRATOSPHERE-TROPOSPHERE EXCHANGE • 417 by anomalous diabatic heating and cooling directly tracer whose sources were solely abov e uppeth e r associated wit eddiese hth numericalls i , y comparable control surface would be flushed out of, or build up in, to that from the mean circulation itself. Correspond- the region between the two control surfaces until the ingly, a therdispersivalsy e ob ma e ee effecth f o t ambient concentrations were such that the fluxes bal- eddies across, as well as along, isentropic surfaces anced. [Mahlman t al.,e 1984]. Evidence from model simula- onls i timescalen t yI o s long enoug systee th r mhfo tions suggests that such dispersio effectivn a s nha e statisticaa n i e b o lt steady state that fluxes acrose sth vertical diffusion coefficient Kz of no greater than a two control surface ssamee musth e ,b t assuming that few times 10"1 m2 s"1 [Plumb and Mahlman, 1987]. thersourco n intervenins e i sinr eth o n ki g regiont A . Local vertical o clear-air-turbulenct mixin e du g e currene th t accurac observationaf yo l estimation, that "pancakes" is thought to be typically of a similar is probably all that can be said. On the other hand, if order [Dewan, 1981]. Simple scaling shows that for a observational estimates were accurate for shorter ti- vertical velocit lengta yd scalan h scalW e e e D,th mescales, the e t surprisinwoulni b t dno - fino gt im d ratio of vertical diffusion to vertical advection is KJ balances betwee fluxeso tw e ,n th reflecting examr fo , - (WD). Since cross-isentropic advective transport is ple, seasonal variation of chemical tracer amounts in associated with vertical velocitie timew fe sa 10~f so 4 the lowermost stratospher seasonar eo l variatioe th n i m s"1, it is to be expected that the advective transport mean position of the extratropical tropopause itself will dominate (KZ/(WD) « 1) on vertical length [e.g., Robinson, 1980]. thin scaleI . s sens :justifiekm »s D i 1 t ei assumo dt e that the cross-isentropic transport is dominated by the advective transpor n sufficientlo t y large vertical 5. OBSERVATIONAL ESTIMATES OF GLOBAL- scales. However, the dispersive effects could be im- SCALE EXCHANGE portane evolutioth r fo t f thino n layer f volcanio s c aerosol or for aircraft emissions. We have seen in section 3 that the downward con- thin I s sectio s emphasie a nth E s beeST ha sn no trol principle (based on the zonal mean momentum measured by global-scale upwelling or downwelling budget; see (7)) can be used to provide a reasonable acros a suitabls e control surfac e380-e sucth s Kha estimat e extratropicath f o e l meridional mean mass isentropic surface. We have argued that such up- flow when the time constant of the motion is much welling or downwelling is best viewed as controlled in longer tha radiative nth e timescale alternativen a s A . , a nonlocal way by extratropical eddy forcing in the the zonally averaged thermodynamic energy equation for f wave-inducemo d forces wils e e seeA th .b l n ni e solvezonae b th r n w (3f d*i fo ca l) mean potential next section, this viewpoint not only is qualitatively temperature distributio diabatid nan c heating distribu- insightful, but allows useful quantitative estimates of tion are known. This calculation is, however, subject such transport, including many aspect seasonas it f so l to considerable uncertainty in the lower stratosphere, and latitudinal variation [e.g., Rosenlof, 1995]. t radiativwherene e th e heatin smala s gi l residuaf o l The foregoing picture does not imply that at every several large terms. The downward control calculation instant the flux of a constituent downward into the also suffers problems probably owing to unresolved lowermost stratosphere need equal the flux across the gravity wave forcin momentue th n gi m e budgeth d an t extratropical tropopause intuppee oth r troposphere, time-dependent departures in the tropics and subtrop- taking into account synoptic- and smaller-scale dy- illustrates ic Figury db . e4b namical events (cutoff cyclones, tropopause folds, Rosenlof Holtond an [1993] have utilize downe dth - etc.; see Figure 3). It is quite clear that these two ward control approximation to estimate the contribu- fluxes need not be equal. If we extend downward the tio f eacno h extratropical hemispher global-scalo et e zonal-mean, eddy formalism described earlie- in o e t solsticr th n i e E seasonsST . They use a d10-yea r clude the upper troposphere, then with realistic eddy recor dailf do y stratospheric geopotential height anal- statistic e transporth s t acros e tropopausth s e itself yses from the United Kingdom Meteorological Office appears primarily as an eddy dispersion along isen- (UKMO) as a data source and the 100-hPa pressure tropes. While the strength of the meridional circulation surface (about 16-km altitude) as a control surface for canno regardee tb wholls da y independen edde th yf o t calculationE ST e th , since this surfac s approxii e - dispersion [e.g., Holton, 1986], it is a highly nonlocal mately at the tropical tropopause (Figure 1). function of such eddy dispersion at many altitudes, as resulte Th Rosenloff so Holtond an e [1993th r ]fo a result of the pumping mechanism illustrated above. rat f maseo s transfer acros e 100-hPth s a surface ar e t followI s that ther esimplo neen e db e relationship shown in Table 1. They found that upwelling is limited betwee e flu f tracernth o x s acros a lowes r control equatorwaro t f ±15do ° latitude, excep southe th n i t- surface representative of the tropopause itself and the ern hemisphere winter, whe t extendni 30°So st e Th . flux across an upper quasi-horizontal control surface downward flow across the 100-hPa surface in the ex- at 380 K or so. Nonetheless, if the two were badly out tratropics is dominated by the strong meridional cir- of balance for a significant period of time, then any culatio e northerth f no n hemisphere winter season, 41 8• Holto t al.ne : STRATOSPHERE-TROPOSPHERE EXCHANGE 33, 4 / REVIEWS OF GEOPHYSICS AVGY aDA . 6-10JAN9215 H N 310Q K2Z [pvu] 90E It 3 180 : •: 0 90W NH 22JAN92 00 Z Q 310K [pvu] 90E 180 :-••-•-: 90W Plate 2. Potential vorticity distribution on the 310-K isentropic surface. The 2-PVU contour (shown in white) marks the approximate tropopause. (a) Five-day average of January 6-10, 1992, which shows the modulatio tropopause th f no e latitud low-frequence th y eb y Rossby samwavese Th e) fiel(b ,Januar n do y 22, 1992, showing strong meridional displacements of the tropopause associated with cutoff cyclone formation near the Greenwich meridian. / REVIEW 4 , 33 GEOPHYSICF SO S Holton et al.: STRATOSPHERE-TROPOSPHERE EXCHANGE • 419 TABL . E1 Mas s Flux Acros 100-hPe sth a Surface From only a rough approximation, however, since the atmo- Rosenlof and Holton [1993] sphere above the 100-hPa level cannot be regarded as a well-mixed layer. DJF JJA Annual Mean The annual-mean global-scale mass exchange can NH extratropics -81 -26 -53 also be estimated from the budget of a long-lived trace Tropics 114 56 85 constituen e tchlorofluorocarbo th suc s a h n CFC13, SH extratropics -33 -30 -32 whic tropospher e iners hi th n i t t photodissociateebu d Units *e . 10g ar Negativk s8 e sign means downward flux. e stratosphereth in r parcelAi . s passing upward Abbreviation e DJFar s , December, January Februaryd an , ; JJA, throug tropicae hth l tropopause will have CFC13 mix- June, July Augustd an , , northerNH ; n hemisphere , southSH d - an ; ing ratios characteristic of the well-mixed troposphere; ern hemisphere. r parcelai s passing downward fro e extratropicamth l overworld into the lowermost stratosphere will have smaller mixing ratios owin photodissociatioo gt n dur- which causes the compensating tropical upwelling to ing their passage throug overworlde hth t flune x e Th . be about twice as large during the northern hemisphere of CFC1 into the overworld, Fmust approximately C9 winter as during the southern hemisphere winter. equa difference th l 3 e betwee upware nth d flux inte oth If the tropical mass flux is assumed to be uniformly overworld in the tropics and the downward flux out of distributed betwee 15n± ° latitude verticae th , l velocity overworle th extratropicse th n di : at the 100-hPa level in northern hemisphere winter 1 shoul aboue db x 10~ s"m 3 t greate 4 r than during Fc = ~ Fad\s)(mclma) (11) southern hemisphere winter. Since upward motion Her upware th e downwar d Fe d Fadar dauan an r dai causes adiabatic cooling in a statically stable environ- mass fluxes, respectively, XT and Xs are the mean ment e annuath , l e extratropicacyclth n ei l pumping volume mixing ratio f CFC1o s 3 transported upward should produc annuan ea l cycl tropican ei l tempera- and downward across the control surface defining the ture near the 100-hPa level, with the lowest tempera- lower boundar e overworldth f yo md e can th lm, s i a tures in the northern winter and the highest in the ratio of the molecular mass of CFC1 to that of air. northern summer. Suc annuan ha l temperature cycle Conservatio masf no s requires tha averagen o t = , F 3 was found in conventional radiosonde data by Reed au Fad, and the subtropical eddy-transport barriers in the and Vlcek [1969] and has been confirmed in global overworld tend to keep the values XT and Xs weU satellite data by Yulaeva et al. [1994] using 12 years of apart. data from channel 4 of the microwave sounding unit Follows [1992] used observed values of XT and \s (MSU-4) flow NOAn o n A operational weather satel- an estimatn da Ff eo deduce d from rate datf th e o n ao lites. MSU-4 measures the mean temperature in a releas f CFC1eo c inte atmospheroth s ratit f eo d ean layer extending from abou e 40-hPe 150th th t o -t a increase in3 the troposphere to solve (11) for the air level, with a maximum response at about the 80-hPa mass flux Fau. To compare his results with those of level. Figur showe6 area-weightee sth d annual march Rosenlof Holtond an controe th [1993] t le l e surfacw , e of temperature in the lower stratosphere from monthly mean MSU-4 data. In the tropics the minimum is in northern hemispher ee maximu winteth d n i an r s mi 220- northern hemisphere summer, extratrope whilth n ei - extratropics e f cyclicsimilao th s s i e r amplitud t oppositbu e e 218- phase. Because of the strong compensation between 216- tropic extratropicsd san globae th , l mean temperature 214- in this layer is nearly constant. Strong compensation 212- whole globe between the tropical and extratropical signals is, of course, just what woul expectee db temperatura r dfo e 210' signal produced by the adiabatic heating and cooling of 208 an annually varying extratropically pumped meridi- 206 tropics onal mass circulation, and provides strong support for 204 the annual cycle in STE shown in Table 1 and for the 202 global-scale circulation suggested in Figure 3. J FMAMJJ ASOND The annual-mean tropical upwelling computed as an Figure 6. Mean annual march of lower stratospheric tem- solstico averagtw e eth f seasoneo equivalens si n a o t t perature based on MSU 4 data for the period 1979-1991, upward mass flow acros 100-hPe sth a x surfac 7 2. f eo averaged ove tropice th r s (30° 30°N)o St e extratropicth , s 10 kg yr" . Since the total mass above the 100-hPa (polewar f 30° entire d30°N)o d th S an d e an globe, , showing 17 1 notionaa surfac10x , 2 kg 5. s ei l turnovere timth f eo evidence of control by extratropical pumping. (After order of 2 years is implied for the atmosphere above Yulaeva . [1994]al t e ; copyright American Meteorological the 100-hPa level. Such a number must be viewed as Society.) 42 0• Holto t al.ne : STRATOSPHERE-TROPOSPHERE EXCHANGE 33, 4 / REVIEWS OF GEOPHYSICS in (11) be the 100-hPa level. From the vertical profile The resilience to transport along the isentropes is at the data ofGoldan [1980]. al t e , \s ~ 0.6xr 100-hPa provide e Rossby-wavth y db e restoring mechanism, level in the extratropics. If we use the estimate of which cause bane sth f stron do gradientV gP - be o st 7 Rosenlof and Holton [1993] that F = 2.7 x 10 k1 g have somewhat lik elastin ea c band. Horizontal shear 1 6 1 au yr" , then (11t )ne givee 10x yr" g th 0 k s 9 a F = c also plays an important role in limiting smaller-scale flu CFC1f xo 3 int overworlde oth . This agrees well with transports thin I . s sens tropopause eth mucs ei h like 6 1 Follows's estimate of Fc = 83 (±28) x 10 kg yr" stratospherie edge th th f eo c wintertime polar vortex thud an s provides independent suppordowne th r -fo t [e.g., Juckes and Mclntyre, 1987; Mclntyre [1992]. ward control calculation. However, there are a number of important ways in importance Th f carefulleo y choosin approprin ga - which the tropopause is rather different from the ate control surface when using conventional meteoro- stratospheric vortex edge r exampleFo . , displace- logical dat estimato at e global mass exchang illuss ei - e stratospheri e edgmentth th f o ef o s c polar vortex wortratee th ky d b ofHoerling . [1993]al t e . Following occur primarily through the upward propagation of Wei [1987], Hoerling . [1993al t e ] use isentropin da c Rossby waves fro tropospheree mth . Large displace- coordinate formulatio whicn i n e controhth l surface ments of the tropopause, on the other hand, are gen- specifies wa coincido dt e wit tropopause hth e (defined erally associated with growing upper tropospheric cy- approximately as in Figure 1), and the mass transport clones e distributioTh . f o horizontan l sheas i r across this control surface was directly estimated from e proportionadifferentth d an , l jump acrose th s wind fields interpolated to isentropic coordinates. Al- tropopaus larger s ei ordee th unityf f ro o , , whicr fa s hi though their estimated upward mass flux inte th o larger than the jump in PV across the edge of the polar stratosphere in the tropics in January was surprisingly vortex, suggesting that quasi-geostrophic theors i y Decembere closth o et , January Februard an , y (DJF) highly limited in its applicability and that ageostrophic estimat f Rosenlofeo Holtond an [1993], Hoerlingt e circulations have a significant role. al. [1993] inferred that substantial transport into the Exchange occurs whe tropopause nth s stronglei y stratosphere occurred across the tropopause at high distorte largy db e latitudinal displacements. Such dis- latitudes. Such transport f substantiali , , mus cone b t - tortion characterizee sar tonguey db f anomalouslso y fined to a layer very close to the tropopause; otherwise high PV air, formed through the action of the large- it consistenwoule b t dno t wit observee hth watew dlo r scale deformation field, extending equatorward- Un . vapor mixing ratios in the lower stratosphere at high der some circumstances, some parts of these tongues latitudes. This reinforces our theme that for most pur- ma e stretcheyb t intdou o long filaments unded an , r poses the tropopause itself is not the most suitable other o forst y rolthemp u lma y isolate d coherent control surface for study of mass exchange. structures containing high-PV air, generally referred to as "cutoff" cyclones (Figure 2 and Plates 1 and 2b). The ubiquity at midlatitudes of filamentary structures 6. SMALLER-SCALE PROCESSES NEAR nea tropopause th r e characterize high-Py db V anom- THE TROPOPAUSE: THE EXTRATROPICS alies and likely to be associated with exchange pro- SUBTROPICD AN S cesses is nicely shown by Meteosat water vapor im- ages. An example was shown in Figure 2b (see also lowee Ath t rlowermose edgth f eo t stratosphere, Appenzeller Daviesd an [1992]). wher tropopause eth e cuts across isentropic surfaces, Cutoff cyclones have been argue importane b o dt t n occu ca y isentropi b rE ST c transport, leadino t g mechanism [e.g.E ST ,f o Bambers . 1984]al - t e Ex . displacemen e tropopausth f o t e fro s equilibriumit m chang cutofn ei f cyclone occun sca convectivy rb r eo position illustrates a , Figurn d i Plat d an e2 1, followed radiative erosio anomalousle th f no tropopausw ylo e by nonconservative processes such as diabatic heating that is characteristic of cutoff cyclones, by turbulent or cooling, layerwise two-dimensional turbulent mix- mixing near the jet stream associated with the cutoff ing, and small-scale turbulent mixing. The boundary system, or as a result of tropopause folding along the between stratospheric and tropospheric air along isen- flank of the system [Hoskins et al., 1985; Price and tropes that span the tropopause is often marked by Vaughan, 1993; Wirth, 1995]. However, Price and an undulating band of strong PV gradients (see Plate Vaughan conclude that convective exchange caused 2a) that coincides with the axis of an upper tropo- by cutoff cyclones is small compared to exchange spheri t streamcje existence .Th thif eo s ban stronf do g caused by tropopause folding associated with upper PV gradients and indeed similarly strong gradients in level cyclogenesis. Cutoff cyclones, sube whicb n - hca mixing ratios of species such as ozone and water stantial in size, have been successfully simulated and vapor, itself suggests that there mus rathee b t r strong wil discussee lb d furthe section i r n 8. Their subtropical dynamical resilience to transport along the isentropes, counterparts, called "upper air cold vortices" by trop- since otherwise vigorous mixin f stratospherigo d can ical meteorologists, mus presumee b t d als contribo t - tropospheric air would destroy the band suggestes oa f E wavstrone th ST y o dg b t arro e ut w third from gradients. the bottom in Figure 3. 33, 4 / REVIEWS OF GEOPHYSICS Holto t al.ne : STRATOSPHERE-TROPOSPHERE EXCHANG1 42 E• Figure 7. Cross section through strong three-dimensional tropopause-folding event of March 13, 1978, showing re- gion of tropospheric air (stippled), po- tential temperature (thin solid con- tours), wind speed (in meters per second; dashed contours), research air- craft flight track (thin dashed lines), and potential vorticity tropopause (heavy solid line). (From Shapiro [1980]; copyright American Meteoro- logical Society.) The process of filamentation and the intrusion of with upper level baroclinic wave development [Bush stratospheric air deep into the troposphere, referred to andPeltier, 1994; Rotunno et al., 1994; Lamarque and as "tropopause folding, alsy somo ot ma " e extene b t Hess, 1994]. The first two of these papers stressed the regarde e systematiresule th th f s o dta c e effecth f o t fact tha mose tth t extensive tropopause folding occurs large-scale deformation field. Ther soms ei e evidence in three-dimensional baroclinic developments associ- supporo t t this from contour-advection simulationf so ated with very strong jet streams. As the baroclinic filamentary tropopause structure in reasonable agree- wave development proceeds near the tropopause, ment wit Meteosae hth t water vapor images (Plat. e1) strong subsidence develop regioe th n northerlf si no y In this sense again, the tropopause is similar to the flow to the west of the upper level trough, together stratospherie edgth f eo c polar vortex, where vertical with downwar isentropee dth tilf o t intrusiod san f no variation e quasi-horizontath n i s l velocity fiele ar d air of stratospheric origin into the troposphere as expected to give filamentary, or sloping, sheet-like show Figurn ni . e7 structures with small vertical and horizontal scales. The quasi-isentropic transport of air out of the low- However, for the midlatitude tropopause the effects of ermost stratosphere initiated by tropopause folding large-scalthe e deformatio enhancebe ncan ageodby - could in principle occur in the absence of the global- strophic frontogenetic circulations; indeed, in simple scale diabatic circulation than i t t case.Bu , there would two-dimensional model uppesof r tropospheric fronto- necessarily be on average an equal quasi-isentropic genesis [e.g., Hoskins, 1971] ageostrophie th , c circu- reverse transport of air from the troposphere into the latio s entireli n y responsibl e appearancth r fo e f o e lowermost stratosphere n ordei , o maintait r n mass small vertical scales, since the applied large-scale de- balance. If this were the case, careful observations formation flo heighwis t independent. Stretchinof g would be required to monitor the mesoscale correla- stratospheric intrusions to ever finer scales leads to tions between trace constituent mixing ratios and air irreversible transport, often speeded up by turbulence motion orden i s estimato t t rexchang ne e a eth f o e resulting from shear instabilities. Much of the ozone given constituent. Although some studies [e.g., Reiter transport from the lowermost stratosphere into the et al., 1969; Ebel et al., 1991; Hoerling et al., 1993] troposphere is believed to occur in connection with suggest that irreversible transfe f tropospherio r r ai c such tropopause fold events. Ther lona s ei g historf yo into the stratosphere can occur in association with case studies of midlatitude tropopause folding, which tropopause folding (and indeed, some two-way ex- is reviewed by WMO [1986]. More recent examples chang suggestes ei visualizationy db s like d Platan e1 e givear Browetly b n. [1987 . al d Kritzal t e t an ] e other Lagrangian tracer studies suc s thah a f Yango t [1991]. These studies have confirmed that large epi- and Pierrehumbert [1993]) e extremelth , watew ylo r occun ca associatiosodin i rE cST n with tropopause vapor mixing ratios observed throughout the strato- folding. sphere [e.g., Oltmans and Hofmann, 1995] indicate structure Th dynamicd ean foldine th f so g process that the quasi-isentropic exchange in middle latitudes associated with upper level frontogenesis was re- is mostl one-waya y transport inte troposphereoth , viewe Keysery db Shapirod an [1986]. More recently, except perhap e lowesth n e i s t th kilometef o o s r o r high-resolution mesoscale numerical models have stratosphere. been used to simulate tropopause folding associated The global-scale diabatic circulation is, of course, 422 • Holton et al.: STRATOSPHERE-TROPOSPHERE EXCHANGE 33, 4 / REVIEWS OF GEOPHYSICS require transporo dt t stratospheric constituents such plai observee nth d seasonal-cycle behavior illustrated as ozone downward from the overworld to the lower- in Figure 6. most stratosphere average Th . e rat t whicea h sucha The existence of the global-scale circulation itself species can be transported into the troposphere is thus is confirme chemicay db l tracer evidence includw no , - ultimately determine rate t whicth e a y ddyname b hth - ind vergan y som w precisne e e evidence froe mth ically controlled global-scale circulation transports Halogen Occultation Experiment (HALOE) to be de- mass from the overworld intlowermose oth t strato- scribed shortly. Older, well-known evidence begins sphere. For this reason the details of mesoscale with the fact that long-lived chemical species with tropopause-folding events may not be important for mixing ratios nearly uniform throughout the tropo- determining the global flux of ozone and other tracers sphere, suc methanes ha , nitrou sCFCs e oxideth d ,an , froe stratospheremth , although they will certainly (1) are observed to have their largest stratospheric strongly influence the time and space distribution of mixing ratios immediately above the tropical such transport. tropopause, matching tropospheric) values(2 d an , Recent developments suggest, however, that the show upward extending plume relativelf so y large mix- timd spacan e e distribution e importanb y ma r s fo t ing ratios throughout the depth of the tropical strato- some chemical purposes and that some caution is sphere, consistent with upward transpor thin ti s region required r instancefo ; , e relationshimucth f ho - pbe an ddegrea f isolatioeo n fro extratropicse mth . Such tween global-scale dynamics and photochemistry was isolatio turnn i , n,is understandable fro existence mth e developed s knowbeforwa t i en that heterogeneous of the subtropical eddy transport barriers mentioned in chemical processing occurrin n loweo g r strato- section 4. spheric aerosol s importantwa s .d Withiol e th n Also well factknowthe s thaare n t water vapor framework of gas phase chemistry, most photo- shows a similar upward extending plume, but that in chemical activit s confine uppee ywa th o rt d strato- contrast wit e aforementioneth h d tracers, verw lo y sphere, so that the global transport in the overworld valuewatee th f rso vapor mixing ratio extend upward provided a good average representation of transport fro mminimua m withi kilometerw fe n a tropica e th f so l ann adequata d e constrain n STEo t . However, tropopause. That minimum mixing rati w onls oi fe ya stratospheric reaction aeroson so l - surfaceoc y ma s part r milliospe volumy nb e (ppmv) an lower dfa r than cur in the lowermost stratosphere. Furthermore, typical tropospheric values. It is generally accepted, new initiative evaluato st impace eth f aircrafo t n o t with good reason, that the basic mechanism responsi- the environment require the consideratio nvaluew olo f e freeze-dryinlongis si th r -fo e bl g [Brewer, 1949], tudinally asymmetric emissions directly in the low- a proces whicn i s r passinhai g throug e tropicahth l ermost stratosphere. For these problems, it is nec- tropopause has its water vapor mixing ratio reduced to essary to at least evaluate the sensitivity to the more the ice saturation value at or near the tropopause. detailed synoptic-scale approac conclusiony an f ho s Satellite data since 1979 have broadly confirmed this thadrawe ar t chemican ni l assessment studies. classical global-scale pictur r enterinai f eo tropie gth - cal stratosphere from below, being dehydrated as it enters, and then being gradually moistened, by meth- 7. SMALLER-SCALE PROCESSES NEAR oxidatioe an perhapd nan weay sb k mixing froe mth THE TROPOPAUSE TROPICE TH : S extratropics, as it is drawn upward in the tropics be- fore being pushed polewar eventualld dan y downward As discussed in sections 3-5, the theory of the in middle and high latitudes [e.g., Remsberg et al., wave-driven circulation, illustrate Figurn d i an 3 e 1984; McCormick and Veiga, 1992; Can et al., 1995]. confirmed and calibrated by observational knowledge The satellite data show the lowest zonal-mean water relevane oth f t eddy effects, tell thas extratroe su th t - vapor mixing tropicaratiothe sin l lower stratosphere pical stratospher mesospherd ean t nonlocalleac n yo during northern hemisphere winter (about 2-3 ppmv), the tropics as a global-scale fluid-dynamical suction with values increasing with altitude and latitude to 4-6 pump. The pumping action is slow but inexorable and ppmv. It should be noted that most satellite measure- causes large-scale upward transfe f maso r s froe mth ments in the lower tropical stratosphere are highly tropical troposphere into the tropical stratosphere, at a uncertain because of cloud particle and aerosol con- rate tha largels ti y independen locaf to l conditions near tamination year3 r ,o s especiall 2 followin e th r yfo g tropicae th l tropopause. Indeed contrarye th n o , s a , major volcanic eruptions. Also, the sharpness of trop- pointes Yulaevay b wa Rosen- . [1994y t b al d ou t d e ]an ical temperature and water vapor minima are difficult lof [1995], the local conditions must respond to the to determine with satellite measurements that resolve pumping rate, insofa stronges a r r extratropical pump- only vertical scales greate. r km tha 3 n2- ing must ten pulo dt l tropical upper tropospherid can likela w y No implicatio e freeze-dryinth f no - hy g lower stratospheric temperatures below radiative equi- pothesis is that water vapor mixing ratios of air enter- librium, encouraging deep cumulonimbus and produc- ing the tropical stratosphere should vary seasonally in ing higher and colder tropopauses. This helps to ex- phase with the annual cycle of tropical tropopause 33, 4 / REVIEWS OF GEOPHYSICS Holto t al.ne : STRATOSPHERE-TROPOSPHERE EXCHANG3 42 E• temperature. Thus, just as a signal is recorded on a more detail. Hers importani t ei noteo t t , first, that magnetic tape as it passes the head of a tape re- eddy transport barriers cannot be described quantita- corder e minimuth , m saturation mixing ratio near tively in terms of eddy diffusivities [e.g., Mclntyre, the tropical tropopause shoul e "recordeddb n o " 1992 . 364p , ] and, second, that the behavn yca e asym- each layer of air moving upward in the large-scale metrically s clearli s A .y demonstrate e morth ey db tropical stratospheric circulation. Since, as was familiar example of the polar vortex edge, rates of mentione section di , horizontan4 l eddy transpory b t entrainmen r mixino t comparable b neen gi t dno o et isentropic mixing betwee tropice nth extratrod san - rate erosiof so mixinr no g out. Further information no pic observes si relativele b o dt y weak, layerr ai f so these important questions is now emerging from recent passing upward through the tropical tropopause airborne measurement campaigns (K. A. Boering, R. should retain their water vapor signal for some time . Wofsy, C . Plumb. A S d , personaan , l communication, perhaps many months, as they are advected upward 1995). by the large-scale circulation. With sufficient verti- remainine Th g questions abou e tapth t e recorder cal resolution, the imprint of the seasonally varying analogy concern the amplitude of the observed tape saturation mixing ratio of the tropical tropopause signal in water vapor mixing ratios. Here our under- should the observable nb time-heigha n ei t sectios na standin fror fa ms gi complete . [1995bt Moteal bu ,t e ] alternatin higd an gh layerwatew lo f so r vapor mix- suggest thae subtropicath t l barrie s relativeli r y per- ing ratio anomalies that move upward at the speed of meable neaoverworle base th rth f eo d (consistent with the large-scale circulation. classical bomb-debris observations), is less permeable This "tape recorde beerw effectnno s strikingl"ha y kilometerw fe a s farthe , wherup r e mixin minis i n gi - confirmed by analysis of water vapor mixing ratio data mal, and is then more permeable again in the middle from the microwave limb sounder (MLS) and the stratosphere, especially during strong westerly quasi- Halogen Occultation Experiment (HALOE), botn ho biennial oscillation (QBO) episodes, as in January- e Uppeth r Atmosphere Research Satellite (UARS) February 1993. Suc hpattera consistens ni t with what [Mote et al., 1995a, b]. The data provide new and we know of horizontal eddy motion extending into the powerful evidence on two points. The first is the con -e overworlth bas f o e d froe subtropicamth l tropo- firmatio f large-scalo n e upwelling speed d theian s r sphere [e.g., Hoskins t al.,e 1985, equation (33d), seasonal variation, entirely independent of the three Figure, 15] thi10 n ;i , s sregard2a , Figur oversims i e3 - observational approaches to global exchange rates plified e mixinTh . g e overworlnea e basth th rf o e d summarize section di seconne remarkabl5e Th . th s di y greatly attenuates the signal relative to what a simple long persistenc e "tapth f eeo signal," showing only estimate from 100-hPa synoptic-scale temperature weak isentropic mixing into the tropics during the analyses especiallwoulw no s di implyt i y d cleaan , r perio whicn i d h adequate data were available from that dehydration arguments based on such data must HALOE, January 199 Octobe3o t r 1994 roughld an , y treatee b d with caution reasonr ,fo s classigivee th n i c altitude inth e rang annualle e 20-2Th . 5ykm varying "stratospheric fountain" work si New ell and Gould- mixing-ratio signal produced by the freeze-drying pro- Stewart r additiona [1981fo d an ] e lb reasono t w sno cess remains noticeable for as long as 2 years over a summarized. still larger altitude range of 15 km, well into the middle tape eTh recorder analogy, with seasonally varying stratosphere. The implied ascent rates just above the tape spee witd dan h allowanc r mixingefo , helpo t s tropical tropopause are approximately consistent with make better sense, retrospectively, of many past ob- the calculated mass fluxes of Table 1, varying from servational studies, including the in situ ER-2 aircraft about 0.2 mm s"1 in northern summer to about 0.4 mm measurements fro fiele mth d campaigns base t Panda - s" northern 1i n winter. ama, Central America and at Darwin, northern Aus- The HALOE "total hydrogen" (2[CH 4[H+ ] 2O]) tralia. This important pair of campaigns, to be dis- f speciao date ar al interest here r well-knowFo . n cussed in more detail below, took place during chemical reasons connected with methane-oxidation northern hemisphere summe d winteran r , respec- chemistry, total-hydrogen mixing ratios are to good tively, and showed the striking differences in the av- approximation uniform in the extratropical overworld, erage water vapor mixing ratio profiles summarizen di with values close to, or weakly fluctuating about, the Figure 8 [Kelly et al., 1993]. Figure 8 shows that the time-averaged tape signal t followI . s thae totalth t - average lower stratospheric water vapor minimum in hydrogen signal can be due only to the "recording the northern hemisphere winter series (2.5 ppmv) oc- head" nea tropopause rth overworld e thatd th ean n i , , curs near the tropopause (10 flights from Darwin). isentropic entrainmen mixinr to extratropicaf o n gi r lai However, the average minimum in the northern hemi- can hardly do other than attenuate the signal. Thus the sphere summer series (3.5 ppmv) occur, s nearkm 9 1 total-hydrogen result s d verprovidan y w direcne e t which is well above the tropopause (7 flights from confirmation of, and perhaps the first quantitative in- Panama). Thi "hygropausey sdr " layer canno- ex e tb formation about weaknese th , f isentropiso c entrain- plaine termn di freeze-dryinf so transpory gb t through ment from the extratropics; Mote et al. [1995b] give locae th l tropopaus same th t eea tim yearf eo , northern 424 • Holton et al.: STRATOSPHERE-TROPOSPHERE EXCHANGE 33/ REVIEW 4 , GEOPHYSICF SO S 550 I I I I I convective mass flux than can ultimately be taken up . Panama —— Darwin e extratropicallbth y y controlled large-scal- up e welling .. [1995bMoteal t e ] argue that such hyper- ,500 ventilation is most likely in the northern summer monsoon season, when surface conditions tend to- ward strengthenin deepenind gan e cumulonimgth - £450 s convectio bu e global-scalth t bu n e upwellins i g slowest. Such hyperventilation would have to intro- duce a statistical bias toward colder, dryer dehydra- ^400 tion conditions thasynoptic-scaly nan e temperature analysis would indicate. Thi becauss si tape th en ei recorder analogy the deepest penetrations act like 350 recording heads thalocatee ar t d farthest alone gth 012345678 Water Vapor (ppmv) direction of tape travel. The dehydration signals from such recording heads have the greatest chance Figure 8. Panama (northern hemisphere summer, 1980) of surviving, that is, of not being erased or displaced and Darwin (northern hemisphere winter, 1987) mean water back downward or sideways by subsequent higher vapor profiles measured by the NASA ER-2 research air- injection notionae th s a s l tape moves upward. Mote craft verticae Th . l axi markes si unitn di f potentiaso l tem- . poin al t tha t e ou tt hyperventilation could therefore perature. Bars indicat e minimuth e d maximuman - mob served values for individual flights. The northern summe contributinra e b g facto determininn i r amplitude gth e minimum at Panama is about 75 K (~3 km) above the local of the observed tape signal in total hydrogen. It tropopause, while the northern winter minimum at Darwin is would reinforc effecte eth isentropif o s c entrainment coincident wit tropopause hth e (From . [1993].Kellyal t e ) or mixing in during northern summer and work against them during northern winter. Paname Th Darwid aan n campaigns both usee dth summer [e.g., Holton, 1984]. However, the tape re- high-altitude ER-2 research aircraft and aimed at corder analogy provides a basis for interpreting this testing by in situ observation whether, on the scale elevated hygropause, to a first approximation, as a of cumulonimbus anvils and their immediate sur- consequence of upward advection, during the inter- roundings, dehydration took place in the way sug- vening months saturatiow lo f o , n mixing ratios that gested by Danielsen [1982]. The campaigns were were nea tropopause rth e durin northere gth n winter, joint efforts of the National Aeronautics and Space probably modified by mixing effects. Administration and the National Oceanic and Atmo- As explained in section 5, the extratropically con- spheric Administration Paname Th . a campaigne th , trolled upward mass transport into the tropical strato- Water Vapor Exchange Experiment (WVEE) of Sep- sphere is believed to be about twice as large in north- tember 1980, was the first aircraft investigation of ern hemisphere winter as in northern hemisphere tropical convective exchange e fundamentaTh . l summer, and regionally averaged tropopause temper- measurement thin si s experiment were temperature, atures ove northere th r n Australia region during Jan- pressure, water vapor Lyman-Alph(usinw ne a g a enougw uar lo appeao yd e dehydrato hb t o t r o t r eai hygrometer capable of measuring very low mixing prevailing minimum stratospheric values [Selkirk, 1993] uniforA . m rising motion would, howevere b , ratios of water vapor [Kley et al., 1982], and ice expecte causo dt e water vapor saturatio formad nan - particles [Knollenberg et al., 1982]. The experimen- tion of a widespread thick cirrus cloud layer near the tal plan involved sampling air both within and down- tropopause. Robinson [1980] and others suggested that stream of convectively produced anvils so that the such cloudines t observeno s i sd therefor an d e that natur evolutiod ean convectivele th f no y influenced upward flow across the tropical tropopause must be air coul ascertainede db . associated entirely wit e mashth s flow from tropical e WVETh E observations demonstrated that con- deep convection that penetrates into the lower strato- vective systems do penetrate into the stratosphere and sphere sucw hHo . mass flows could occu lead o ran dt do mix tropospheric and stratospheric air together stratospheric dehydratio discusses nwa Danielseny db [Danielsen, 1982] thao e resultins , th t g mixturn ca e [1982] termn i , f small-scalo s e turbulent entrainment have stratospheric value f potentiaso l temperature6 , into overshooting cumulonimbus turrets, followe y , despitdb K 0 maximue e38 th > m observed tropospheric particle sedimentation within anvils, leaving sheets of equivalent potential temperatur [Selkirk,K 7 36 f o e dry air behind. 1993]. Evidence for such penetration was shown by It is possible that such cumulonimbus penetrations observations in a convectively produced anvil where could "hyperventilate e loweth " r tropical strato- high ice crystal amounts (several hundred ppmv of spheree sensth n ei , that there coula greate e b d r total water substance) were found in an imperfectly / REVIEW 4 , 33 GEOPHYSICF SO S Holto t al.ne : STRATOSPHERE-TROPOSPHERE EXCHANG5 42 E• 20 a 19 18 17 16 180 190 200 210 1.5 10 40 Temperature, K Water, ppmv Figure 9. Vertical profiles from the flight of January 23, 1987, during STEP/Tropical. (a) Temperature at anvie edge th th f l eo (heav y solid lineoutsidd an ) anvie eth l (light solid line) Tota) (b , le wateic d an r saturation mixing ratio at the edge of the anvil (triangles and heavy solid line, respectively) and outside the anvil (square lighd san t solid line, respectively). mixed layer between 16 and 17 km. The air in this layer Tropical), base Darwinn di , Australia, during January was clearly stratospheric, with a potential temperature and February 1987. The aim was to deploy the ER-2 observatione oTh f. abouK 5 t38 s also showed thae tth aircraf e "stratospherith n i t c fountain" regioe th f no ice crystals in the mixed layer (16-17 km) appeared to Indonesian maritime continent [Newell d Gould-an fall out, leavin gmixtura f stratospherieo tropod can - Stewart, 1981], where seasonally averaged tempera- spheri r witai c a hwate r vapor conten f abouo t 6 t tures at 100 hPa were estimated to be as low as 188 K, ppmv, essentially the saturation mixing ratio for tem- implying saturation mixing ratios of 2.4 ppmv. This is convective peratureth f o p to es e founsysteth t a dm already at the low end of climatological-mean ob- (about 193 K at 17 km, or about 2.5° colder than the served water vapor mixing ratios, 2.5-3.5 ppmv, for seasonal mean at Panama). Finally, the particle mea- e tropicath l lower stratosphere. Furthermoree th f i , surements [Knollenberg t al.,e 1982] showed thae th t coldest cloud tops were 2.5° colder tha meae nth t na e sampleth n i e dparticln i ic anvi s e bulth wa le f ko 100 hPa, then moist convection could conceivably sizes of 100 fxm or more, implying that the ice would inject air with mixing ratios as low as 1.6 ppmv. lesn i s m thafalk 1 hourln 1 . Thu evolutioe sth n from That extreme conditiono f d thi o s d s an orde n ca r the anvil center (with substantial excess ice) to the occur in reality was verified by one of the flights from anvil edge (with no excess ice) is consistent with the the Darwin campaign. Figure 9 shows the evolution of particle data. an anvil fro mlarga e continental thunderstorm near No cumulonimbus penetration events were Gule ob th f Carpentaria o -ff o soute th d hen heave Th . y served, however, that reached above 450 K and pro- solid lines and triangles show the anvil penetration. duced the minimum Panama mixing ratios associated Here the presence of both measurable radon gas (not wit dottee hth d curv Figurn ei . Thie8 s promptee dth shown) [Kritz t al.,e 1993 totad an ] l water exceeding second campaign e tropicath , l experimene th f o t saturatee ic e th d mixing ratio [Kelly t al.,e 1993] indi- Stratosphere-Troposphere Exchange Project (STEP/ cates that tropospheric air has mixed upward to 17.9 42 6• Holto t al.ne : STRATOSPHERE-TROPOSPHERE EXCHANGE 33, 4 / REVIEWS OF GEOPHYSICS km with clearly stratospheric potential temperatures moist convection in the Australian region [Russell et . AtmospheriK 0 o39 f cs source radoit th s t a enha al., 1993] contrasty B . ,mois e mosth f to t convection Earth's surfaces atmospheriit d an , c lifetim s govi e - during STEP/Tropical occurred near the northern erned by its 3.8-day radioactive half-life. The lowest coast during the active periods of the monsoon. More- temperatures, with a potential for dehydration to 1.7 over, averaged ove monte th r f Januarho y 1987e ,th ppmv, were found between 17 and 17.6 km (potential lowest saturation mixing ratios from radiosonde mea- f thio sl airAl . ,K) 3 38 d temperaturean K 2 37 f o s surements occurre t ove continente dno th r t ovebu , r e quasi-adiabatith n i includin r ai e gth c mixed layer the northern Australian coast [Selkirk, 1993]. An im- stratospheris ha , belokm 7 w1 c characteristic wells sa , portant featur thif eo s coastal moist convectios it s nwa s demonstratea e elevateth y db d potential tempera- lesser penetration into the stratosphere, as evidenced tures. It also has ozone mixing ratios of 100 parts per e mucbth y h lower (10-1 ) tropopaus2K e potential billio volumy nb e (ppbv r higheo ) r [Danielsen, 1993], temperatures. most likely showing isentropic entrainment froe mth Figures lOa to lOc show three additional examples extratropic suggestind san actuay gwh l mixing ratios of sounding t cloua s d tops taken durin e STEPgth / (squares in Figure 9b) are not quite as low as 1.7 ppmv. Tropical experiment. Along with Figure 9, they show ThuWVEEn i s sa , tropospheri observes wa r cai o dt diversite th moisf yo t convection durin Australiae gth n have penetrate mixed dan d int stratospheree oth e Th . monsoon (see Danielsen [1993] for additional exam- major differenc thas cloui e th t d top coldee sar d ran ples). Figure lOa is from a convective system over somewhat higher, with minimum saturated mixing ra- Darwin observed soon after monsoon onset, when tios of 1.7 ppmv (light solid curve in Figure 9b) instead tropopause and near-tropopause temperatures (as of around 6 ppmv. The light solid lines and squares in measured by radiosondes) were at their lowest points Figure 9 show a sounding outside the anvil, northwest e winteoth f f 198o r 7 [Selkirk, 1993] e minimuTh . m and roughly (though not exactly) downstream of the saturated mixing ratios are comparable to or lower heavy solid lines. Som 18-ko t e- warminm16 e th n gi than those in Figure 9 (1.6 ppmv), but they occur in layer is evident (mostly attributable to adiabatic de- what is essentially pure tropospheric air (ozone mixing scent [see Kelly et al., 1993], but the excess ice has ratios of 10 ppbv and potential temperatures of 360 K). fallen out, leaving air with minimum water mixing If these cloudcontributo t e ar s e substantialle th o yt ratios of 2.2 ppmv (squares in Figure 9b) at a potential stratospheric mass budget, the air in these anvils must temperatur , againK 2 ,37 clearlf o e y stratospheric. attain stratospheric potential temperatures, eithey b r Measurable rado presens ni t thia t s locatio s wellna , turbulent diffusio anvie [Danielsen,th p t lto na 1993r ]o clearly demonstrating the following: (1) Tropospheric by radiative heating and subsequent rise [Danielsen, s penetrateaiha r d inte stratospherth o d beean en 1982; Ackerman al.,t e 1988]. Though ther substans ei - mixed irreversibly with stratospheric air, (2) excess ice tial theoretical evidenc enhancer efo d radiative heating falleresultins e wateth s ha ) nha (3 outr rgd ai vapo ,an r of tropical anvils, the observational evidence is still mixing ratios similar to or below the climatological- quite inconclusive [Russell t al.,e 1993] f course(O . , mean value 2.5-3.f so 5 ppmv. radiative heating is indeed what would occur if the The STEP/Tropical experiment thus clearly showed stratosphere were "underventilated," with insufficient that convective exchange in the tropics can transport mass injection, in which case the extratropical pump tropospheric air upward and dehydrate it at the same would simpl inexorabld yan y pul upwarr lai larga n do e time. However t cleano s ri fro t i dat,e mth a gathered s alreadscalewa s yA . pointed out absence ,th a f eo on the remaining flights whether penetration events large-scale cirrus veil can be taken as evidence against like that of Figure 9 are frequent enough to dominate such underventilation.) the tropical stratospheric mas wated san r vapor bud- Figure lOb is from a flight north of Darwin about 2 get, or whether they are "too frequent" in the sense weeks later, nea e beginninth r a secon f go d active that hyperventilation migh e takinb t g- re place e Th . period durin 198e gth 7 Australian monsoon. Tropopause mainde f thio r s section discusses e quessomth f -eo radiosonde temperature e northerth n o s n Australian tions thus raised. coas aboue highetar K t4 r than during monsoon onset, Befor budgeea t calculatio r hyperventilationo - nes anaircrafe dth t measurement e clearlar s y consistent timate coul e undertakedb nr instance basefo , don , with this. The moist convection clearly penetrates into global satellite coverage e woulon , d nee o knot d w the stratosphere (as is shown by the excess ice), but whether the mechanism outlined by Daniels en [1982] only to potential temperatures of about 370 K. More- and illustrate Figurn di s representativi e9 f deeeo p over, this system is clearly hydrating the stratosphere moist convection that reaches and overshoots the below 16.3 km, with saturated mixing ratios of 3.2-4 tropopause. In fact, the cloud top characteristics of the ppmv. bulk of the moist convection observed during STEP/ FigurGule fros i northere th f c mo th f elO o d nen Tropical differed substantially from Figure 9. The Carpentaria over a tropical cyclone. A notable feature storm corresponding to Figure 9 occurred over the e deeith s p regio f negativno e (but clearly subadia- Australian continent durin a gperio f suppresseo d d batic) vertical temperature gradients below 17., 9km Flight 4, 1/16/87 Fligh , 1/31/89 t 7 Fligh , 2/12/814 t 7 m 19 19- a n————i————\————r 19 0 -n n 18- 18- 18- o 1 B 17 17- 16- 16- 16- o. 15. 15 15 a- 180 184 188 192 196 200 188 192 196 200 204 208 186 190 194 198 202 206 Si. Temp, K ______Temp, K TempK , 200 400 600 800 1000 0 200 400 600 800 1000 200 400 600 800 1000 Ozone, ppbv ______Ozone, ppbv______Ozone, ppbv 1 en 2468 8 6 10 4 -2 2 0 -20246 •D Ice Sat mr ppmv Ice Sat mr, TW ppmv Ice , SaTWmr t ppmv m 73 Figure 10. (a) Vertical profile through an Australian monsoon onset convective system from the flight of January 16, 1987. heave Th y solid line denotes temperature kelvinsn i , lighe th , t solid line denotes ozone, in part billior spe volumey nb d an ; dashee th d line denote saturatioe sic n mixing ratio partn i ,r millio spe volumey nb dottee Th . d361-e linth s Kei isentrope. (b) Vertical profile throug midmonsooha n convective system fro flighe m th Januarf o t , 1987y31 . Heavy solid line denotes CD I temperature in kelvins; light solid line denotes ozone, in parts per billion by volume; dashed line denotes ice saturation m 72 mixing ratio, in parts per million by volume; crosses denote total water mixing ratio, in parts per million by volume. The m dotte d367- linthe eKis isentrope Figurin As e lOb(c) . , excep tropicaa for t l cyclone fro flighm the Februar of t 1987y12, . m nX The dotted line is now the 375-K isentrope. i n 428 • Holton et al.: STRATOSPHERE-TROPOSPHERE EXCHANGE 33/ REVIEW 4 , GEOPHYSICF SO S which is capped by a strong inversion. The fact that several days rather than several hourse b ten o t d e entirth e regio s saturateni d withs i respec e ic o t t somewhat larger [Jensen and Toon, 1994], variations evidenc anvir efo l penetration factn I e 17. o s. t ic , 9km in tropopause temperatures on longer timescales may s observewa d ove e samth r e tropical cyclon daye4 s e b even more promisin r dehydratinfo g e th g earlier at 18.3 km [Kelly et al., 1993; Pfister et al., tropopause region. One possibility is the tropical 1993]. A succession of overshooting and detraining Kelvin wave havin gtimescala abouf eo weeka t . (See convective turrets have here produce a regiod f o n Andrews et al. [1987] for a review of Kelvin wave weak stability that is a mixture of stratospheric and properties.) . Tsuda[1994al t e ] have documentee dth tropospheric airremaine suce Th .on h f turreo s e ar t passage of such a Kelvin wave over Indonesia and apparen t 16. , a witt 5km h marke crystae dic l loading. noted a temperature variation from 185 to 193 K, These systems are clearly able to dehydrate the trop- implying a saturation mixing ratio variation from 1.8 to ical stratosphere, since the minimum saturated mixing 4.8 ppmv. If the particles that are formed are large rati abous oi ppmv5 2. t . enough to fall significantly in a few days, this is cer- e mostTh important implicatio f thino s varietf yo tainl potentiaya l water removal mechanism neare th convective cloud tops for tropical exchange and dehy- tropical tropopause. Similar slow cooling mechanisms dratio thas ni t complicatei t s attempt t globasa l bulk responsible ar dehydratior efo Antarctie th n i c winter calculations of water vapor and mass budgets. Though stratosphere [Toon et al., 1989]. e typealth l f observeo s d moist convection produce The WVEE and STEP/Tropical experiments upward transfe somf o r e sort continentae th , l systems showed that cumulonimbus penetration and dehydra- are probably most effective on an individual basis. tion event takn ca se plac t onleno y during northern Currently available cloudiness statistics are hardly ad- winter, but also even in northern summer when equate for estimating the spatial and temporal distri- tropopause temperature nearle s ar highe K y5 r than i butions of near tropopause cloud mass fluxes and northern winter, with correspondingly higher satu- henc degree eth whico et h convection hyperventilates rated mixing ratios. This argues agains e originath t l the lower tropical stratosphere. An additional problem suggestion of Newell and Gould-Stewart [1981] that involves ice crystal fallout. No very large ice crystals such penetratio s strictlni y confine e maritimth o t d e (MOO jxm in diameter) were observed during STEP/ continent in northern winter. Though, as is discussed Tropical. In fact, within the regions of high ice crystal in section 5, upward motio tropice th n i onl s si y about content (200 ppmv of ice or more), perhaps 2% of the hal s strona f northern gi n northere summeth n i s na r mass was in sizes of less than 10 jjim [Knollenberg et winter becaus weakef eo r extratropical pumping from al., 1993]. These crystals fall very slowly, taking over southere th n winter hemisphere [Rosenlof Holton,d an a day to fall 1 km. Thus given the possibility of signif- 1993], there is still considerable moistening at the icant anvil heating rates [Ackerman et al., 1988], there tropopause, as is shown in Figure 8. As was discussed is the real possibility of evaporation of these ice crys- earlier, this moistening, combined with the upward tals . a smalThoug s i l % percentage2 h , this could liftin f tropicao g r dehydrateai l d durin e previougth s account ppm4 r morr fo v o icef eo , certainly significant northern winter (the tape-recorder effect) explain ca , n stratospherifoe rth c water budget, particularl therf yi e the observed lifted hygropause shown for Panama in is substantial hyperventilation. Figurpoleware Th . e8 d sprea dehydratee th f do s i r dai e difficultTh f quantitativelo y y accountin- ob r gfo also the most likely explanation for the observed an- served tropical stratosphere dehydration e baseth n do nual minimu n watemi r vapor mixing e ratioth n i s penetrative cumulus convection model led Potter and midlatitude southern hemisphere lower stratosphere Holton [1995 o propost ] n alternativa e e schemn i e during March-July [McCormick et al. 1993]. In fact, which convectively generated buoyancy waves pro- significant southward spreading of dry air from the vide an important role in dehydration. Vertical parcel Australian monsoo s evidenni t from STEP/Tropical, displacements produce sucy db h wave producn ca s e wher a e100-hP a anticyclone clearly transportr ai s local saturation and promote the formation of thin ice from just abov tropopause eth northere th n ei n Aus- clouds in the lower tropical stratosphere (analogous to tralian monsoon regiomidlatitude th o nt t oveeje e rth wave clouds in the lee of mountains). Such clouds Southern Ocean [Danielsen, 1993; Kelly et al., 1993; would be expected to form in a quasi-random fashion Kritzet al, 1993]. in the vicinity of tropical moist convection. Even if typical ice particle sizes are very small so that particle sedimentation t rateresul e slowne sar e t th ,average d . 8 STRATOSPHERE-TROPOSPHERE EXCHANGE over many realizations of such clouds could be a IN MODELS significant downward flu f waterxo , which would tend dehydrato t lowee eth r stratospher thud ean s contrib- In this review we have emphasized that STE must ute to the maintenance of the observed water vapor be regarded as part of the general problem of global minimum. transport and that nonlocal dynamical forcing (extra- Since particles that form as a result of cooling over tropical pumping) plays a primary role in driving global / REVIEW 4 , 33 GEOPHYSICF SO S Holton et al.: STRATOSPHERE-TROPOSPHERE EXCHANGE • 429 transport. In section 5, observational evidence was ment of tracer transport, in a detailed study of model- presented in support of the extratropical-pumping simulated STE. mode f transporo l d STEan t . Because e mucth f o h The CCM2 qualitatively simulates the stratospheric extratropical pumping is generated by Rossby waves, transport drive extratropicay nb l pumping. However, general circulation models (GCMs) that correctly sim- the simulated mass transport across the 100-hPa sur- ulat generatione eth , propagation dissipatiod an , f no strongo facto s i e , whil transpore eth t highea t r alti- Rossby waves should correctly represent the global- tudes is too weak, consistent with the fact that the scale aspects of transport and STE. Thus GCMs rep- zonal forcing in the model is too strong in the lower resen valuablta e too testinr lfo conceptuae gth l model stratosphere and too weak in the upper stratosphere of transporpresenteE ST d an td above. and mesosphere. Nevertheless, the CCM2 captures It is also reasonable to ask whether the two-dimen- the correct seasonal variation in exchange; there is a sional (longitudinally averaged) models commonly maximum in transport across the 100-hPa surface in use chemican di l assessment studies properly repre- northern hemisphere winter and a minimum in north- sent global transport and STE. The nonlocally driven ern hemisphere summer. global-scale diabatic circulation (either externally The simulation of water vapor in the CCM2 is not as specified or determined internally) is used as the basis satisfactor simulatioe th s i masr s ai ya f sno exchange. for the advective transport in most two-dimensional Although ther s somei e evidenc tropicae th f eo l tape models. Thus such models should adequately repre- recorder effect in the model, latitudinal transport in the sent the global-scale mass exchange across the lower lower stratosphere is much too strong; the imprint of boundary of the overworld. Any other contribution to e seasonallth y varying tropopause saturation mixing STE in such models, such as by isentropic mixing in ratio decays with height much more rapidly than is the lowermost stratosphere onln includee ca ,y b a dvi observed. Furthermore, althoug e tropicath h l the eddy-flux parameterization. tropopause temperatures in the model are colder than those observed lowee th , r stratospher modee th n es i i l Three-dimensional models will include other contri- actually moister than that observed. The model does butions to STE through explicit simulation of plane- not resolve convective-scale processes, and the pa- tary-scal synoptic-scald ean e eddies t presena t bu , t i t rameterized representatio moisf no t convectioe th n i is only those with relatively high resolution that can mode apparentls i l y unabl mimio et dehydratine cth g resolve features such as tropopause folds. For exam- effects of overshooting cumulonimbus turrets. On the ple, Cox et al. [1995] report a recent simulation of a other hand, the model is rather successful in simulat- a globafol n i d l GCM, with horizontal resolutiof no ing isentropic exchange with the extratropical lower- about 100 km and vertical resolution of about 1.5 km most stratosphere cause wavy db e breaking associated near the tropopause. Coarser-resolution global models with synoptic-scale weather disturbances n otheI . r will represent exchange in such cases, largely through words modee th , soms ha l e representatio lowe th -f no parameterized mixing. Mesoscale (limited area) mod- ermost wavy arrows in Figure 3, including some of the els may not be so reliant on parameterization and may transport effects normally associated with tropopause be used for detailed simulation of local processes (see folding, even though the folds themselves are not re- resulte th difficuls i t t i spu sectio t froo t tbu m, n6) suc h solved realistically. models in a global context. This section will explore Two-dimensional chemistr transpord an y t models furthe two w three-dimensionad ho r-an l models actu- hav importann ea t rol playo et , t becausfeano -s i t ei ally represent STE. sible, even with current computin- g ex powern ru o t , Global constituent transport in general circulation tended multiyear simulations in three-dimensional models has been studied for more than a decade, GCMs wit hfula l representatio f chemistryno . Such particularly at the Geophysical Fluid Dynamics Lab- extended simulations are necessary to calculate, for oratory at Princeton University [e.g., Mahlman et al., example impace th , f increaseo t d fluorocarbond an s 1980, 1986; Plumb Mahlman,d an 1987; Strahand an aircraft exhaust pollutants on stratospheric ozone, in- Mahlman, 1994a, b]. However, there have been onlya cludin radiative rols th git n ei e balanc tropoe th f eo - few general circulation model studies that specifically sphere standars i t two-dimensionaI .e th n di l modelo st focuse STEn do earln .A y examplstude th f s yo ewa assume that transpore approximateb y ma t n a y b d Allam and Tuck [1984a, b], who attempted to simulate averaged circulation and a spatially varying, noniso- transpore th exchangd an t f wateeo r vapor. Thei- ef r tropic diffusio alsd mako an not e simplifying assump- forts were hampered both by the limited resolution of tions concernin chemistrye gth r examplefo , , neglect theis unrealisticallit ry b mode d an l y large ratf o e or approximation of diurnal variation. Together with diffusio subgrid-scaly b n e processes. More recently, smale th l numbe degreef ro freedomf so , such approx- . [1994Moteal t e ] have use Nationae dth l Center fo r imations greatly reduc e computationath e l require- Atmospheric Research community climate mode2 l ments. Such models have been used extensively for (CCM2) modea , l with much better resolutioe th n ni major stratospheric ozone assessments [e.g., WMO, regiotropopause th f no betted ean r numerical treat- 1986, 1995], and comparison with observation shows 430 • Holton et al.: STRATOSPHERE-TROPOSPHERE EXCHANGE 33/ REVIEW 4 , GEOPHYSICF SO S that the models can be useful for long-term simula- promises the robustness of the models' sensitivity to tions. However two-dimensionae th , l models hav- eob change. vious limitations particularn i d an , t seemi , s likely that For applications in which the sources of strato- for species thacontrollee ar t transpory db wels a t s a l spheric constituents have strong latitudinal, longitudi- by photochemistry, simulated temporal variability can temporad nalan , l variability inadequaciee th , e th f so be regarded with any confidence only for timescales of current two-dimensional formulations will have seri- a season or longer. ous consequences. However, two-dimensional models Two-dimensional model clearle sar t capablyno f eo are thought to be an appropriate tool for calculating representin detaile gth STEf so horizontale Th . , verti- impace th f chlorofluorocarboo t n increas straton eo - cal, and temporal resolutions are insufficient. How- spheric ozone. The chlorofluorocarbons CFC13 and ever, since they are based on the global-scale circula- CF2C12 are both photolyzed mainly above 25 km; these tion, they might be expected to represent the bulk molecules are released in industrialized areas but are magnitude of STE with some accuracy when the well mixed in the troposphere and thus represent a model exchange is integrated over a year. Also, since zonally symmetric sourc f inorganieo c chlorine th n ei e two-dimensionath l model circulatio s usuallni - yde upper stratosphere. The fraction of chlorine that par- rived from observations in the overworld, the annual ticipates directly in ozone destruction (Cl and CIO) mass exchange between the stratosphere and tropo e fractioth - d nan store n i reservoird s (HCd an 1 sphere calculated from a two-dimensional model [e.g., C1ONO2) are determined photochemically. The only Douglass et al., 1992] should be similar to that calcu- loss process for stratospheric chlorine is transport to late othen di r ways [Rosenlof Holton,d an 1993; Fol- the troposphere, where soluble specie e removear s d lows, 1992; Robinson, 1980]. by rainout. Although transport to the troposphere fro stratosphere mth t necessarilno s ei y zonally sym- two-dimensionae Th l model transport fro troe mth - metric, on an annual basis the net loss from the strato- pospher stratosphere th o et e takes place mainle th n yi spher s approximateei e ddiabati th well y b c circula- tropics. Transport from the stratosphere to the tropo- tion. advectivo t sphere du s ei e transpor t middla t higo et h hydrochlorofluorocarbone Th s (HCFCs- )in tha e tar latitudes, horizontal transport where the two-dimen- creasingly being use s substitutea d r CFC1d fo s an sional model tropopause height is changed by a grid CFC1 are shorter-lived and have significan3 t photo- 2 morer o verticad x an ,bo l diffusio troposphere th o t n e chemica2 l loss rates (and thus production rate reacf so - at all latitudes outside the tropics [e.g., Douglass et tive chlorine species lowee th n )ri stratosphere. Owing al., 1993; Shia et al., 1993]. Though observations in- to the difficulties of accurately representing transport dicate that most transport from the stratosphere to the d mixine lowean th n gi r stratospher two-dimenn i e - troposphere takes place at middle latitudes and is sional models, assessmen e relativth f o t e impacf o t associated with synoptic-scale storms (see sectio, n6) these specie n ozono s e using such model s mori s e in many two-dimensional models, stratosphere-to-tro- problematic. posphere transport tends to be placed poleward of the Effort evaluato t s e atmospherieth c effect f subo s - storm track t middlsa higo et h latitudes. Three-dimen- sonic and supersonic aircraft are motivating more re- sional model simulations, on the other hand, are able search into tracer transport and STE [e.g., Sparling et to simulate the observed distribution of stratosphere- al., 1995] y contrasB . t wite inorganith h c chlorine to-troposphere transport [e.g., Douglass et al., 1993; source, which the stratosphere sees as zonally sym- Rasch et al., 1994]. This difference in placement of the metric, supersonic aircraft emissions wil restrictee b l d downwar e mosth s ti d importan E brancST f ho t dif- to oceanic corridors and are thus a highly nonzonal ference between most two-dimensional and three-di- source of pollutants. Other aircraft emissions are also mensional tracer simulations when comparisons are nonzonal, reflecting the geography of the main air mad r seasonaefo r longeo l r timescales. traffic corridors. Pollutant transpor lowermose th n i t t Introduction of strong eddy diffusion in two-dimen- stratosphere is very much a three-dimensional pro- sional models to represent quasi-isentropic transport cess, involving rapid advectio shearind nan g alont gje acros e tropopausth s y havma ee unrealistic conse- stream wels sa edds a l y dispersion away from themn i quences. For instance, a pollutant that is introduced y illustratemucwa f these o hth l Figurn edi Al . 2b e into the northern hemisphere lowermost stratosphere processe generalle sar y much stronge winten i r r than may rapidly ente southere rth n hemisphere lowermost in summer. These three-dimensional aspects are not stratospher a rapivi e d diffusion along isentropes, dealt with adequatel parameterizationy yb currenn si t crossing the equator on the tropospheric portion of two-dimensiona e beyonb y l modeldma sucd an sh such isentropes. Such a case is discussed by Shia et al. models altogether, even with improved parameteriza- [1993]. Since, however, these processe basicalle sar y tion. specifie two-dimensionae th n di l formulation spece th , - issueMane th f yo s aircrafe raiseth y db t emissions ificatio changee b n morphologE nca mak o dST t e eth - problem require evaluatio three-dimensionaf no l trans- ically correct. Specification of such parameters com- port effects. This does not, however, mean that it is 33, 4 / REVIEWS OF GEOPHYSICS Holto t al.ne : STRATOSPHERE-TROPOSPHERE EXCHANG1 43 E• necessar fula e l generaus o yt l circulation modeln A . ward displacemen mixinf o t g ratio surfaces associated alternative approac three-dimensionao ht l modelinf go wit weakee hth r cutoff cyclone (an lese dth s intense constituent transport and exchange is to solve the ozone maximum t 35°a ) greates Ei r than that associ- continuity equations for the relevant trace constituents ated with the stronger cutoff cyclone at 315°E. This using wind data comthay tma e either fro mprevioua s suggests that the exchange is associated with more generaa ru f no l circulation model [Rasch t al.,e 1994; derive r sophisticatedd(o ) parameters that define th e Chipperfield et al., 1994] or from real data analyzed characteristic cutofe th f sfo cyclones. wit datha a assimilation system ,e STRA sucth s ha - Price and Vaughan [1993] have suggested that only stratospheriN TA c analysis system [Coy t al.,e 1994] a relatively small percentage of the STE is associated use Douglassy db . [1993]al r t eithee Fo . f theso r e with cutoff cyclones. Instead, they argue that strong approaches, the net mass exchange calculated diag- baroclinic folding events [e.g., Danielsen, 1968e ar ] nostically (e.g. downwary b , d controgenerals it d an l - more important. The basic features of the transport are izations) will be similar, provided that the general entirely consistent with the mechanisms proposed by circulation model produces a reasonable climatology Danielsen [1968]. Irreversibility is closely correlated of wind temperatured an s r botFo .h approachese th , with diabatic heating terms, both radiative and moist- meridional circulation (an annuae dth l mass exchange convective, consistent with the mesoscale studies of betwee e tropospherth n d stratospherean e ) wile b l Lamarque and Hess [1994]. Once again, however, similar to that calculated using a two-dimensional zeroth-order parameters such as surface pressure do model. not provide good predictors of the STE associated Global three-dimensional transport models address wit particulaha r event; quantities suc s statiha c sta- shortcomingsome th f eo two-dimensionae th f so l mod- bility, baroclinicity, and transience made irreversible els, in that the upper tropospheric synoptic scales are by Rossby-wave breakin importane gar determinn i t - explicitly represented in the model. For transport as- inamoune gth stratospherif o t whicr cai carries hi o dt sociated with synoptic scales store ,th me trackth d san altitudes low enough to lead to effective STE. There is upper tropospheric synoptic-scale activity are defined some modelin observationad gan l experienc sugo et - realistically when assimilated fields are used for the gest that rapidly growing events [see Ucellini t al.,e transport resulte Th . f simulationso verifiee b y dma s 1985] might pla yparticularla y important rol midn ei - with constituent observations; however, such compar- latitude STE. isons do not provide a strong test of the accuracy of Diagnosis of the representation of tropical STE in the net stratosphere-to-troposphere mass flux calcula- three-dimensional model receives sha d relatively little tion. Spatia temporad lan l resolutio sucf no h models si attention have w es argue.A d above overale ,th l ratf eo stilcoarso to l resolvo et actuae processeseE th ST l . upward mass transport acros tropicae sth l tropopause Small difference horizontan i s l behavio lean o ca rdt is determined nonlocall overale pars th y a f o t l general large difference vertican i s l transport; comparisonf so circulation and is not likely to be sensitive to the calculated and observed profile shapes are not a strong detail tropicaf so l moist convection. Tropicaf o E ST l test because irreversible transport is not clearly iden- water vapor and of short-lived tropospheric tracers is, tified e model Th adequately . ma s y represen- ef e th t however, likely to be strongly influenced by the man- processesE ST e fectth f ;o s judgmen thin o t s subject whicn i r hne tropical moist convectio parameters ni - requires careful analysis of three-dimensional model modee izeth s n discussedi a l d CCMe abovth r 2efo behavior, bot lonr hfo g (annual) timescale shorted san r simulation of Mote et al. [1994]. timescales associated with specific events. Consideration of specific events in global transport models illustrate heterogeneite sth theif yo r behavior. . 9 SUMMAR CONCLUSIOND YAN S The variability of cutoff cyclone systems is illustrated by two examples which are found in the STRATAN In this revie havwwe e argued tha understanto t d 500-hPa geopotential height fields on January 21 and essentias STEi t i , includo t l t withiei nglobal-scala e 22, 1989 (Plates 3a and 3b). Relatively low values of dynamical perspectiv s usefud thai t an ei t l theo t n geopotential heigh e founar t t 35°a d315°d Ean n Eo distinguish between the part of the atmosphere whose January 21January b ; , bot y22 f thesh o e areae ar s isentropes lie entirely above the tropopause (the over- identifie closey b d d contours ozonn A . e simulation, world, 6 ^ 380 K) and the part of the atmosphere initialize n Decembeo d , 1988 28 rd tota an , l ozone whose isentropes span the troposphere and the strato- mapping spectrometer (TOMS) observations for Jan- sphere (whose stratospheric part we have called the uar , 1989y 22 givee ,ar area e 4bPlated n i Th .an sa s4 lowermost stratosphere). Further have ,w e argued that of the cutoff cyclones are characterized by relative chemical transport between the lowermost strato- maxim TOMe e simulationth th n ai Sn i datd an an I . sphere and the overworld is controlled largely by the both the data and the simulation, the maxima at 315°E same nonlocal dynamical processes that are responsi- largee ar r tha t 35°Ena . However, vertical cross sec- global-scale th r fo e bl e mean circulatio middle th n i e tion ozonf so sho) e5b w(Plated tha an downe a th ts5 - atmosphere, with upward mass flux in the tropics and 43 2• Holto t al.ne : STRATOSPHERE-TROPOSPHERE EXCHANGE 33/ REVIEW 4 , GEOPHYSICF SO S Plate 3. The 500-hPa geopotential height fields from the STRATAN data assimilation system for (a) January 21, 1989, and (b) January 22, 1989. Cutoff cyclones are found near 35°E and 315°E in Plate 3b. Plate 4. (a) Ozone simulation in a three-dimensional chemical transport model initialized on December 28, 1988, and (b) TOMS observations for January 22, 1989. Areas corresponding to the cutoff cyclones of Plate 3b show anomalously high total ozone in both data and simulation. 33, 4 / REVIEWS OF GEOPHYSICS Holto t al.ne : STRATOSPHERE-TROPOSPHERE EXCHANG3 43 E• 3DCTM OZONE [ppm-vj a) 35°EE LON ) 315"b 21JAN8E LO9 N 22JAN89 - 0.8 fl 0.6 H 0 0. I . , i . . . i . . i 1000. • . i , . i !. 2 17 31 46 61 75 90 17 31 46 61 75 90 Plat . 5 eVertica l cross section cutofo tw ozone f so cycloneth r efo s 315°E) nea (b 35°) d rPlat(a n Ei , an e 3b. downward mas extratropicsse fluth n xi . These upward force varyin seasonae th n go l timescale, muc f thiho s and downward mass fluxes largely govern global-scale mass flux is drawn up from the deep tropics. Therefore chemical transport, because of the weakness of eddy it is reasonable to regard the wave-induced force as transport along isentropes betwee tropice e nth th d san exertin a g"downward-sideway s controle th n o " extratropics, that is, because of the effectiveness, to a transport acros lowee sth r boundar overworlde th f yo , sufficient exten r thifo ts purpose e subtropicath f o , l even thoug troposphere hth turn ei n e appearth e b o st eddy transport barrier overworle th n si d explainen di ultimate source of the waves, and hence the wave- section 4 (and now directly evidenced by the "tape induced force, in the stratosphere and mesosphere. recorder effect" [Mote et al., 1995a, b]. A complete picture would nee o includt d e th e These considerations explain the remarkable fact pumping effects of gravity-wave drag in the upper that STE, viewed from this global perspective and on stratosphere and in the mesosphere, at present not seasonal and longer timescales, is controlled not by well quantified. Together with the transient response local events near the tropopause but by the generatio seasonae nth o t l cycl f solaeo r heating, these probably of large-scal d small-scalan e e waves mainle th n i y affec detaile tropicae th tth f so l upwellin uppee th n gi r troposphere, followed by their propagation into the stratosphere. But, more importantl purposee th r yfo s stratosphere and mesosphere and their dissipation to of this review, non f theseo e effects chang grose eth s produce a predominantly and persistently westward picture just summarized, particularl face yth t tha- up t wave-induced zonal force. In the extratropics, which welling mass fluxes in the lower tropical stratosphere strongle ar y influence Earth'e th y db s rotation, sucha are extratropically controlled. persistent westward force tends to come into balance It can hardly be overemphasized that this down- wit e Coriolihth s force associated wit hmeaa n pole- ward-sideways control vie f stratosphere-tropowo - ward mass flux n otheI . r words r persistentlai , y sphere exchange is a drastic departure from previous pushed westward has a persistent tendency to move conceptual models for exchange. In the extensive re- poleward, a kind of quasi-gyroscopic effect, producing literature vieth f wo stratosphere-troposphern eo - eex the pumping action discussed in section 3. By mass change give y WMOb n [1986] e focuth , s almosi s t continuit througd yan tendence hth y toward "down- entirel synoptic-scaln yo d small-scalan e e processes ward control" exhibited by the extratropical response occurring nea tropopausee th r . Althoug attempo hn t zonae th o lt force (the thought experimen Figurf to b e4 was made in that review to provide estimates of global being especially relevant) pusher ai , e mosth d f poleo t - mass exchange rates, it is clear from the context that ward mus o downwardg t . Mass continuit- re w no y authore th s felt tha t wouli t necessare db compilo yt e quire a correspondins g upward mass flu n lowei x r statistic climatologe th n so exchangf yo individuay eb l latitudes s Figura d suggestsb e4 an , , wit realistiha c tropopause-folding cutofeventy b d f san cyclones , cor- 434 • Holton et al.: STRATOSPHERE-TROPOSPHERE EXCHANGE / REVIEW 4 , 33 GEOPHYSICF SO S responding mostle lowermosth o yt t wavy arron wi or cutoff cyclone formation. They are, however, con- Figure 3, in order to derive seasonal estimates of sistent wit concepe hth f nonlocao t l control. Accord- exchang eacr efo h hemisphere. in thao gt t concept occurrence th , maximua f eo mn i Ther alss ei osensa e conveye discussioe th n di f no wave dissipation in the northern hemisphere winter tropical exchange by WMO [1986] that air is forced stratosphere lead maximua o st m downward mass flux into the stratosphere by tropical cumulus convective e extratropicinth hencd maximua an s o et m fluf o x turret thad an st estimate f tropicao s l mass exchange ozon othed ean r stratospheric tracers acros lowee sth r would thus require accurate measurement of the cu- boundary of the overworld into the lowermost strato- mulus mass flux nea tropopausee rth have .W e argued, sphere during the northern winter. This winter season however juss twa summarizeds a , t irre,ne tha-e th t e fluxepeath n i ks inte lowermosoth t stratosphere versible mass flux into the tropical stratosphere is inevitably leads to a buildup of tracers in the lower primarily determined nonlocall dynamicay yb l forcing stratosphere until the tracer mixing ratios become suf- forin the wave-drivem of n pumping fro extratmthe - ficiently large that the fluxes from the overworld can ropical stratosphere, which is the immediate cause of matchee b increasey db d fluxes inttropospheree oth . sloe th w diabatic ascentropicae th n i t l stratosphere. Thus a springtime maximum in flux of ozone and Because transport acros isentropie sth c surfaces near radioactive tracers into the troposphere would exist 380 K defining the lower boundary of the overworld is e rat evef th irreversiblo e f i n e mass transpory b t independent of the details of tropopause folding, of tropopause folding and other small-scale processes formatio f cutofno f e distributiocyclonesth f o d an ,n remained constant all year round. and intensity of tropical cumulus convection, it is Notwithstanding the attractive simplicity of the possible, as is discussed in section 5, to obtain sea- nonlocal-control concept and its implications for the sonal estimate totae th f l so upwar d transport inte oth global circulatio chemicad nan l transpor overe th n -i t tropical stratospher totae th ld downwarean d transport world, there are important temporal and spatial varia- int lowermose oth t stratospher eacn ei h hemisphery eb occurrencee tionth n si f tropopausso e foldinn i d gan estimatin extratropicathe g l wave-induced forcing e formatioth f cutofo n f cyclones. Problems sucs ha from general circulation statistics using methoe th f do aircraft emission impact assessments, remova volf o l - downward control. This substantiaha s l advantages canic debris seasonad an , l ozone depletion will almost over the strategy of adding up the exchange from certainly require the consideration of detailed mecha- individual events lattee Th . likel s vere i r b yo yt diffi- nisms. Detailed modelin geographicae th f go temd lan - cult to carry out, given the indications from modeling poral dependenc f transporeo t acros tropopause sth e studies (see sectio tha) n8 t ther largs ei e variabilitn yi requires careful consideration of the various processes associateE ST e th d with nominally similar synoptic that operate on the synoptic scale and on the meso- events. Furthermorethe , reasotherall no eis nwhy scale. Similarly, an understanding of the stratospheric exchange acros e meath s n tropopause should take water vapor budget and of the longitudinal variations place in a conspicuous way; there could be significant in transfer of mass and trace constituents across the global-scale subsidenc thar fo t, mattereor , ascentd an , tropical tropopause requires careful consideratiof no significant contributions fro smaller-scale mth e eddies physice th dynamicd san f tropicaso l convective sys- illustrated in Figure 2b. particulan i tem d e role th an sf overshootin o f so r g e recognitioTh f global-scalo n e nonlocal control, convective tower equatoriad san l gravity wavee th n si mainly in the downward-sideways sense explained possibly hyperventilated layer just abov tropicale eth above, also provides an explanation for the observed tropopause. These problems, importan s thea t y are, springtime maxima in the fluxes of ozone and radioac- shoul e clearldb y distinguished fro e problemth f mo tive tracers into the northern hemisphere troposphere. global mass exchange, e havwhicw es a hargue s i d s usuallha t I y been assumed [e.g., WMO, 1986] that determined as part of the general circulation of the these seasonal maxima are caused by a springtime global stratosphere. peak in upper level cyclogenesis and tropopause-fold- ing events. Mahlman [1969], however, showed that although surface radioactivity and upper level cyclo- genesis were strongly correlated on the timescale of GLOSSARY individual cyclones, the seasonal cycle in observed surface radioactive tracer debris concentration in the Baroclinic wave eddiesd an s : synoptic-scale dis- northern hemispher t significantlno s ewa y correlated turbances (horizontal length scales usually of the order with the seasonal cycle in an index of upper tropo- of 1000 km) that develop partly through a process spheric cyclone activity. know baroclinis na c instability whicn i , h potentia- en l The findings si Mahlman t consisten[1969no e ar ] t ergy stored in the background flow and associated with wit e viehth w thae seasonath t l e ratcyclf th eo n i e the horizontal temperature gradient is released. Ba- mass transfer int troposphere oth controlles ei e th y db roclinic waves and eddies are most prevalent in the frequenc amplitudd yan tropopausf eo e folding events storm track regions coincident with tropospherit cje 33/ REVIEW 4 , GEOPHYSICF SO S Holton et al.: STRATOSPHERE-TROPOSPHERE EXCHANGE • 435 stream d stronan s g low-level temperature gradients temperature e resultanth ; t diabatic circulation does associated for instance with land-sea contrasts. Such provid gooea d approximatio verticae th r nfo l advec- wave eddied san s (e.g., Figur Platd an lea) e1 e2 o d t tion of tracers. In the transformed Eulerian mean deformatio sharpenind nan tropopause th f go o t d ean (TEM) the average is taken at constant pressure, but irreversible exchange. e resultinth g mean meridional circulatio a goo s i nd Contour advection: an adaptive computational approximation to the zonal-mean diabatic circulation techniqu r accuratelfo e y followin a quasi-materiag l for seasonal and longer timescales. contour, for example, the isoline of a tracer, on an Frontogenesis: a meteorological term applied to isentropic surface by computing the isentropic trajec- certain cases of strain or deformation in the wind field torie larga f so e numbe f parcelo r s placed initiallt ya sharpening the horizontal gradient of an advected short intervals alon contoure gth . Resolutio mains ni - quantity. Such gradient sharpenin s commonplacgi e tained even in "surf-zone" conditions by systemati- and often produces streaky or filamentary features like cally redistributing and adding parcels as necessary. thos Plat n ethosi r o e1 e produce stirriny db g creamn o Diabatic circulation verticae th : l circulation asso- coffee. The classical cases called "frontogenesis" re- ciated with cross-isentropic flow. In the global-scale fer to the sharpening of large-scale horizontal gradi- stratospheric "overwork!" (see Figure 3), th maio e mostw nf o kinds te ar ent:d (1san ) thos whicn ei e hth persistent and chemically important contribution to advected quantity is potential temperature near the diabatie th c circulatio thas ni t causeextratroe th y db - Earth's surface, in extreme cases producing the sud- pical pumping discusse n sectioe resultini d Th . n3 g den temperature drops thaaccompann ca t y changen si vertical motion, upward in the tropics and downward surface weather thos) (2 whicn ed i advectee an , hth d in the extratropics, pushes temperatures away from isentropen quantito V P s yi s nea tropopausee th r n i , the radiatively determined values toward which they extreme cases producin highle gth y distorted streaky would otherwise relax, thus inducin requiree gth - ddi structures know s a "uppen r frontsai r r o " abatic heating or cooling. "tropopause folds" (see below), a section through one Downward control: the extratropical wave-driven of which is shown in Figure 7. pumping discusse long-time section di th n i n3 e limit- Geostrophic flow a hypothetica: l flo n whicwi h ing case of Figure 4c. In Figure 4c, a steady state has the horizontal component of the Coriolis force due to been reached in which the pumping action of a steady force extends purely downward. That is, the nonlocal the rotation of the Earth is exactly balanced by the control of the pumped circulation is felt, after a long horizontal pressure gradient force. Geostrophic flows i enough time, only directly below the extratropical a good approximation to the actual flow (with errors of locations wher e forcth es nonzeroi e a mor r eFo . the order of, say, 10% or better) for most large-scale realistic, seasonally varying global-scale forcee th , extratropical wind fields, the main exceptions being limiting configuration does not apply in the tropics. strong cutoff cyclones and strongly curved jet streams, This is illustrated by the thought experiment of Figure in which the relative centrifugal acceleration can be 4b. The control is then mainly downward-sideways; fairly important. that is, the extratropical pumping by such a seasonally Log-pressure coordinate (z)verticaa : l coordinate, varying force is felt mainly (as regards mass flux) approximately equal to geopotential height, which is below and to the side of the force, most powerfully to commonly used as an independent variable in dynam- the tropica llong-time sideth n I . e mea downwarde nth - ical meteorology H I- n defines (p/p= i t z I .s ),s a d sideway se tropic effecth n o st depend nonlinean o s r where H is a mean density scale height, p is pressure, aspect response th pumpinge f so th o et . and/?5 = 1000 hPa. Elliptic partial differential equations: these often Potential temperature: defined for a perfect gas as R/Cp aris n modeli e f physicao s l systems with nonlocal 6 = T(pQ/p) , where T is the temperature, p is effects approximated as acting instantaneously. Stan- pressures ga e th , /?e = c100d 0ar p an 0R hPa d an , dard textbook examples include the Laplace and Pois- constan d specifian t y airc dr ,hea r respectivelyfo t . son equations arising in electrostatics and in stati ce temperaturth Thus i 9 s r parcee ai tha n a lt would models of elastic membranes. Equation (5) is also attain if it were adiabatically compressed to pressure elliptic when a/a 7^ 0. Changing conditions at one PQ. Potential temperatur functioa s ei f specifino - en c location generally affect the solution everywhere else, tropy s thui alon d s an econserve y fluib d d parcels as when poking an elastic membrane. whe motioe nth adiabatics ni . Since diabatic processes Eulerian (zonal) mean: an instantaneous average operate on the timescale of weeks in the stratosphere, taken aroun a latitudd e e conventionacircleth n I . l on shorter timescales parcels move approximately Eulerian mea average nth takes ei t constanna t altitude along constant 6 surfaces. Isentropic coordinates use or constant pressure e resultanth ; t mean meridional potential temperatur r specifieo c entrop vertie th s -ya circulation doe t providno s a usefue l approximation l coordinatca havd advantage ean eth e that under adi- for the advective transport of tracers. In isentropic abatic conditions there verticao appearn - e b mo lo st coordinates the average is taken at constant potential tion. Thus adiabatic motion appear horizontalle b o st y 43 6• Holto t al.ne : STRATOSPHERE-TROPOSPHERE EXCHANGE 33, 4 / REVIEWS OF GEOPHYSICS two dimensional (albeit compressible) when viewed in Rossby-wave breaking concepa : t thebasie th -o ct isentropic coordinates. wave-inducef o y or d forces (e.g., Mclntyre [1993d ]an Potential vorticity (PV): somewhat analogouo t s many other references), as well as to the formation of spin angular momentum. For large-scale atmospheric the "surf zones" conspicuou stratospherin si d an V cP motion s conventionalli V P , y (£= define+ e P s a d tracer distributions. "Breaking" doe t measno n frac- f)(-gdti/dp), where £e is Rossby's "isentropic relative turing in any literal sense; rather, it refers to situations vorticity, a vorticity-lik" e quantity approximately in which PV contours deform irreversibly, or fold componenequae th o t l relativf o t e vorticity normao t l sideways n isentropio , c surfaces, instea f merelo d y an isentropic surface;/i Coriolie sth s parameter (twice undulating as in a simple Rossby wave. These irrevers- e locath l vertical componen e Earth'th f o t s angular ible deformation taky ema s placa complicated n i e , velocity); and g is the gravitational acceleration. Here layerwise two-dimensional turbulent manner illuss a , - replacee b monotoni y n an 6 ca y db c function F(6)t bu , trate Platn di e 1. the conventional choice, F(6) = 6, is empirically Tropopause folding: a process in which a thin useful for defining the extratropical tropopause. Since band of stratospheric air intrudes more or less deeply — dp/3Q may be regarded as a local measure of the inte tropospherth o e alon e stronglth g y tilted isen- dept pressurn (i h e e layeunitsth f ro )betwee o tw n tropes associated with an upper tropospheric frontal potential temperature surfaces increasn a , -dp/dtin ei zone Figurverticae a r Se . fo e7 l cross section through implies stretchin f vortego xincreasn a tube d san n ei a large midlatitude fold. It is usually found that part of absolute th e vorticity vicd an ,e versa, somewhat like the stratospheric air in the fold returns reversibly to the spin angular momentum of a ballerina or figure the stratosphere, and part is drawn irreversibly into skater. the troposphere by advection round an anticyclone; PV substance concepe th : t underlyin analoge gth y this is a special case of "Rossby-wave breaking" (see between PV and chemical mass mixing ratios. PV above). The distinction between reversible and irre- substance mean n imaginara s y quasi-chemical sub- versible behavio importans ri correca r fo t t estimation stance whose mixing ratio, or amount per unit mass of e resultinoth f g eddy fluxe f constituento s s froe mth air, is equal, by definition, to the PV. The term "PV lowermost stratosphere to the troposphere. substance" does not mean the amount of anything per unit volume. Haynes Mclntyred an [1990] show that the PV substance can be imagined to consist of mass- ACKNOWLEDGMENTS wise W thano ht . k Alan Plumb less, positively or negatively charged particles that, and Jerry Mahlman for their perceptive reviews, Julia Slingo despite appearance somn si e cases unable ar , croso et s for kindly providing unpublished insigh informatiod an t n no tropical convection, Warwick Norton for furnishing Plate 1, isentropic surfaces. So for inviscid, adiabatic motion Christod an f Appenzelle preparinr rfo s hi r g fo Figur d an e1 substancV P e th s simpli e y advected likn inera e t comment manuscripte th n so . This pape bases ri d partln yo chemical tracer, but it cannot, for instance, be mixed discussion e NATth t a sO Advanced Research Workshop, lik chemicalea . Stratosphere-Troposphere Exchange, hel n Cambridgei d , Quasi-geostrophic approximation e zonallth n i y: England Septemben i , r 1993 wise W than.o h t Scientifie kth c (longitudinally) averaged equations, an approximation and Environmental Affairs Division of NATO for its sup- assuming that the zonal wind component is in geo- port. Partial support was provided by the NASA Atmo- strophic balance wit latitudinae hth l pressure gradient spheric Chemistry Modeling and Analysis Office under and neglecting the effect of the mean meridional cir- NASA grants NAGW-379. U NAGW-66d e 4an th y b d 2an culation in the zonal momentum equation, except in . NaturaK l Environment Research Council throug Unie hth - e Coriolith e thermodynamith sn i term d an , c energy versities Global Atmospheric Modelling Programme. equation, excep statie th n ci t stability term. Rossby wave wava : e that owe s existencit s o et the spatial variation of potential vorticity along isen- REFERENCES tropic surfaces, for example, in the latitudinal direc- tion. Becaus variatioe th f eo potentiaf no l vorticitya , Ackerman, T. P., K.-N. Liou, F. P. J. Valero, and L. Pfister, fluid parcel displaced alon potentiae gth l vorticity gra- Heating rates in tropical anvils, J. Atmos. Sci., 45, 1606- dient (and materially conserving its potential vorticity 1623, 1988. and potential temperature) will have potential vorticity Allam, R. J., and A. F. Tuck, Transport of water vapour in a stratosphere-troposphere general circulation model, I, different from that in the local environment and will . MeteorolFluxesR . J . Q , Soc., 110, 321-356, 1984a. induc perturbatioea n velocity disturbance. This will . TuckF . AllamA , d Transpor. J.R an , watef o t r vapoun ri cause parcel displacement e samth f e o sth sig o t n stratosphere-troposphera e general circulation model, II , west of the original displaced parcel and of the oppo- Trajectories, Q. J. R. Meteorol. Soc., 110, 357-392, sit eeast e resul e sigth tendenca V Th .o n t P s t i e th f yo 1984b. Andrews, D. G., J. R. Holton, and C. B. Leovy, Middle wave pattern in the potential vorticity field to undulate Atmospheric Dynamics, pp.9 48 , Academic Diegon Sa , , quasi-elastically, the undulations propagating west- Calif., 1987. ward relative to the mean flow. Appenzeller . DaviesC . H , ,d C.Structuran , f stratoo e - 33, 4 / REVIEWS OF GEOPHYSICS Holton et al.: STRATOSPHERE-TROPOSPHERE EXCHANGE • 437 spheric intrusions into the troposphere, Nature, 358, . RStolarskiS . . McCormick P . Fahey . W M . , D , d an , 570-572, 1992. Natural cycles, gases, in The Atmospheric Effects of . NortonAppenzellerA . . DaviesC W . ,d H Frag an , , C. ,- Stratospheric Aircraft: A First Program Report, NASA mentatio f stratospherino c intrusions . Geophys.J , Res., Ref.Publ, 1272, 33-61, 1992. in press, 1995. Douglass, A. R., R. B. Rood, C. J. Weaver, and M. C. Bamber, D. J., P. G. W. Healey, B. M. R. Jones, S. A. Cerniglia, Implication three-dimensionaf so l tracer stud- Penkett . Vaughan. TuckG F . d A , an , , Vertical profilef so r two-dimensionafo s ie l assessment e impacth f f o so t tropospheric gases: Chemical consequence f stratoo s - supersonic aircraft on stratospheric ozone, J. Geophys. spheric intrusions, Atmos. Environ., 18, 1759-1766, 1984. Res., 8949-8963, 98 , 1993. Boering . A. t al.K ,e , , Measurement f stratospheriso c car- Dunkerton . J.T ,, Nonlinear Hadley circulation drivey nb bon dioxid wated an e r vapo t northera r n midlatitudes: asymmetric differential heating, /. Atmos. Sci., 956, 46 - Implication r troposphere-to-stratospherfo s e transport, 974, 1989. Geophys. Res. Lett., pressn i , 1995. Dunkerton, T. J., Nonlinear propagation of zonal winds in Brewer, A. M., Evidence for a world circulation provided by an atmosphere with Newtonian coolin d equatoriagan l the measurements of helium and water vapor distribution wavedriving, J. Atmos. Sci., 48, 236-263, 1991. in the stratosphere, Q. J. R. Meteorol. Soc., 75, 351-363, Dunkerton . J. . T ,Mclntyre, E C.-P . . HsuM F . d ,an , Some 1949. Euleria Lagrangiad nan n diagnostic modea r sfo l strato- Browell, E. V., E. F. Danielsen, S. Ismail, G. L. Gregory, spheric warming, /. Atmos. Sci., 38, 819-843, 1981. and S. M. Bleck, Tropopause fold structure determined Ebel, A.. Hass. JakobsJ H , . H ,. Laure M , . MemmesheM , - from airborne lidar and in situ measurements, /. Geo- imer, A. Oberreuter, H. Geiss, and Y.-H. Kuo, Simula- phys. Res., 92, 2112-2120, 1987. tio f ozonno e intrusion cause tropopausa y db e fold dan . PeltierR . W ,. d BushG.TropopausB an ,. A , e foldd an s cut-off low, Atmos. Environ., , 2131-2144 25 Part , A , synoptic-scale baroclinic wave life cycles, J. Atmos. ScL, 1991. 51, 1581-1604, 1994. Eliassen, A., Slow thermally or frictionally controlled me- Carr . S. t al.,E ,e , Tropical stratospheric water vapor mea- ridional circulation in a circular vortex, Astrophys. sured by the microwave limb sounder (MLS), Geophys. Norv.,5(2), 19-60, 1951. Res. Lett., 22, 691-694, 1995. Emanuel . Neelin. S D . C A.. . K ,BrethertonJ d , an , n O , Chen, P., J. R. Holton, A. O'Neill, and R. Swinbank, Isen- large-scale circulations in convecting atmospheres, Q. J. tropic mass exchange betwee tropice nth extratropd san - R. Meteorol. Soc., 120, 1111-1143, 1994. stratospheree th icn i s , /. Atmos. Sci.,, 3006-301851 , Fels . B.S , , Radiative-dynamical interaction middle th n si e 1994. atmosphere, Adv. Geophys., 28A, 277-300, 1985. Chipperfield, M. P., D. Cariolle, and P. Simon, A three- Follows cross-tropopause th n J.. O ,M , e exchang airf eo , /. dimensional transport model study of polar stratospheric Atmos. Sci., 879-882, 49 , 1992. cloud processing during EASOE, Geophys. Res. Lett., Garcia, R. R., On the mean meridional circulation of the 21, 1463-1466, 1994. middle atmosphere, J. Atmos. Sci., 44, 3599-3609, 1987. Cox, B. D., M. Bithell, and L. J. Gray, A general circulation . BovilleA Garcia . B . d R., R ,"Downwaran , d controlf "o model study of a tropopause folding event at middle the mean meridional circulation and temperature distri- latitudes, Q. J. R. Meteorol. Soc., 121, 883-910, 1995. bution of the polar winter stratosphere, J. Atmos. Sci., Coy, L., R. B. Rood, and P. A. Newman, A comparison of 57,2238-2245, 1994. winds fro e STRATAmth N data assimilation systemo t Goldan . Kuster C . . Albritton. . L D.L P ,. . W , A D , d an , geostrophi balanced can d wind estimates . Atmos..J Sci., Schmeltekopf, Stratospheric CFC13, CF^ Nd 2Oan , 57,2309-2315, 1994. height profile measurements at several latitudes, J. Geo- Danielsen . F.E , , Stratospheric-tropospheric exchange phys. Res., 85, 413-423, 1980. based upon radioactivity, ozone potentiad an , l vorticity, . MclntyreHaynesE . conservatioe M th . d H.n P ,O an ,, n J. Atmos. Sci., 25, 502-518, 1968. and impermeability theorem r potentiafo s l vorticity, J. Danielsen . F. dehydratioE ,A , n mechanis stratoe th r m-fo Atmos. Sci., 47, 2021-2031, 1990. sphere, Geophys. Res. Lett., , 605-6089 , 1982. Haynes, P. H., C. J. Marks, M. E. Mclntyre, T. G. Shep- Danielsen . F. sitE ,n ,I u evidenc rapidf eo , vertical, irrevers- e "downwar. ShineP th . n K herdO , d an , d controlf "o ible transpor f loweo t r tropospheri r inte lowecai oth r extratropical diabatic circulations by eddy-induced mean tropical stratosphere by convective cloud turrets and by zonal forces . Atmos.J , Sci., , 651-67848 , 1991. larger-scale upwellin tropican gi l cyclones . Geophys.J , Held heightropopause e th th . M.I statif ,e n o tO th , cd ean Res., 98, 8665-8681, 1993. stability of the troposphere, /. Atmos. Sci., 39, 412-417, Dewan . M.E , , Turbulent vertical transpor o thit ne du t 1982. intermittent mixing layere stratospherth n si othed ean r Held, I. M., and A. Y. Hou, Nonlinear axially symmetric stable fluids, Science, 211, 1041-1042, 1981. circulation nearla n i s y inviscid atmosphere, /. Atmos. Dickinson, R. E., On the excitation and propagation of zonal Sci., 37, 515-533, 1980. winds in an atmosphere with Newtonian cooling, J. At- Hoerling, M. P., T. K. Schaack, and A. J. Lenzen, A global mos. Sci., , 269-27925 , 1968. analysi f Stratospheric-tropospherio s c exchange during Dickinson, R. E., Theory of planetary wave-zonal flow northern winter, Mon. Weather Rev., 121, 162-172, 1993. interaction, J. Atmos. Sci., 26, 73-81, 1969. Holton, J. R., Troposphere-stratosphere exchange of trace Dickinson, R. E., Analytic model for zonal winds in the constituents watee Th : r vapor puzzle Dynamicsn i , e oth f tropics, I, Details of the model and simulation of gross Middle Atmosphere, edited by J. R. Holton and T. Mat- feature e zonath f lo s mean troposphere, Mon. Weather suno, pp. 369-385, Terra Sci., Tokyo, 1984. Rev., 99, 501-510, 1971a. Holton . R.J , , Meridional distributio f stratospherino c trace Dickinson, R. E., Analytic model for zonal winds in the constituents, /. Atmos. Sci., 43, 1238-1242, 1986. tropics , VariatioII , tropospherie th f no c mean structure Hoskins, B. J., Atmospheric frontogenesis models: Some with seaso differenced nan s between hemispheres, Mon. solutions, Q. J. R. Meteorol. Soc., 97, 139-153, 1971. Weather Rev., 99, 511-523, 1971b. Hoskins, B. J., Towards a PV-theta view of the general Douglass . JackmanH . R.. . RoodA . ,AikinC , B C . . R ,A , , circulation, Tellus, Ser. A, 43, 27-35, 1991. 43 8• Holto t al.ne : STRATOSPHERE-TROPOSPHERE EXCHANGE 33/ REVIEW 4 , GEOPHYSICF SO S Hoskins, B. J., M. E. Mclntyre, and A. W. Robertson, On ment f earlo s y Pinatubo aerosols, Geophys. Res. Lett., the use and significance of isentropic potential vorticity 19, 155-158, 1992. maps, Q. J. R. Meteorol. Soc., Ill, 877-946, 1985. (Cor- McCormick, M. P., E. W. Chiou, L. R. McMaster, W. P. rection . Meteorol., R Q. . J Soc., 113, 402-404, 1987.) . LarsenC Chu. . OltmansJ S . Rind, D ,d an , , Annual Jensen, E. J., and O. B. Toon, Ice nucleation in the upper variation f wateso r vapoe stratospherth n i r upped ean r troposphere: Sensitivit aerosoo yt l number density, tem- troposphere observed by the Stratospheric Aerosol and perature, and cooling rate, Geophys. Res. Lett., 21, Gas Experimen . Geophys.J , II t Res., , 4867-487498 , 2019-2022, 1994. 1993. Juckes, M. N., and M. E. Mclntyre, A high resolution, Mclntyre, M. E., Atmospheric dynamics: Some fundamen- one-layer model of breaking planetary waves in the tals, with observational implications, in The Use of EOS stratosphere, Nature, 328, 590-596, 1987. for Studies of Atmospheric Physics, edited by J. C. Gille Kelly . Proffitt ,. ChanH K.. R . Loewenstein. M , K ,M , . R . ,J and G. Visconti, pp. 313-386, North-Holland, New York, Podolske, S. E. Strahan, J. C. Wilson, and D. Kley, 1992. Water vapor and cloud water measurements over Darwin Mclntyre role wavf th e o . E. n M ,O e propagatio wavd nan e durin STEe gth P 1987 tropical mission . Geophys.,J Res., breaking in atmosphere-ocean dynamics, in Theoretical , 8713-872398 , 1993. and Applied Mechanics 1992, . BodnerR edite . S y . dJ ,b Keyser . ShapiroA . M , revied D.structurA e , th an , f wo e Singer, A. Solan, and Z. Hashin, pp. 281-304, Elsevier and dynamics of upper level frontal zones, Mon. Weather Sci., New York, 1993. Rev., 118, 1914-1921, 1986. Mclntyre. PalmerN . "sure T d . E.Th , fM an , zonee th n "i Kley . SchmeltekopfL . A.. D , A , . Kelly. WinklerK ,H . R , , stratosphere . Atmos.J , Terr. Phys., , 825-84946 , 1984. . ThompsonL . . T McFarlandM d an , , Transpor watef o t r Mote, P. W., J. R. Holton, and B. A. Boville, Characteris- vapor through the tropical tropopause, Geophys. Res. tics of stratosphere-troposphere exchange in a general Lett., , 617-6209 , 1982. circulation model, /. Geophys. Res., , 16,815-16,82999 , Knollenberg, R. G., A. J. Dascher, and D. Huffman, Mea- 1994. surement e crystae aerosoic th f d o sl an l populationn i s Mote, P. W., K. H. Rosenlof, J. R. Holton, R. S. Harwood, tropical stratospheric cumulonimbus anvils, Geophys. and J. W. Waters, Seasonal variations of water vapor in Res. Lett., 9, 613-616, 1982. the tropical lower stratosphere, Geophys. Res. Lett., 22, Knollenberg, R. G., K. Kelly, and J. C. Wilson, Measure- 1093-1096, 1995a. ment higf so h number densitie crystale ic tope f th so sn si Mote . Rosenlof H . W. . . P ,Mclntyre E K , , . CarrM , S . E , of tropical cumulonimbus, J. Geophys. Res., 98, 8639- J. C. Gille, J. R. Holton, J. S. Kinnersley, H. C. Pum- 8664, 1993. phrey. Russel. WatersM W . J . ,atmo n J l A d ,III -an , Kritz, M., S. W. Rosner, E. F. Danielsen, and H. B. Selkirk, spheric tape recorder: The imprint of tropical tropopause Air mass origins and troposphere-to-stratosphere ex- temperatures on stratospheric water vapor, J. Geophys. change associated with midlatitude cyclogenesid an s Res., pressn i , 1995b. tropopause folding inferred from 7Be measurements. J , Newell, R. E., and S. Gould-Stewart, A stratospheric foun- Geophys. Res., 96, 17,405-17,414, 1991. tain? . Atmos.J , Sci., , 2789-279638 , 1981. Kritz, M., S. W. Rosner, K. K. Kelly, M. Loewenstein, and Norton . A.W ,, Breaking Rossby wave modea n si l strato- K. R. Chan, Radon measurements in the lower tropical sphere diagnose vortex-followina y db g coordinate sys- stratosphere: Evidence for rapid vertical transport and techniqua ted man advectinr efo g material contours, /. dehydratio f tropospherino c air . Geophys.J , Res.,, 98 Atmos. Sci., 51, 654-673, 1994. . HofmannJ Oltmans . D d . J.S an ,, Increas lowern i e - 8725-8736, 1993. stratospheric water vapour at a mid-latitude northern Lamarque . HessG . P , ,d Cross-tropopausJ.an , e mas- sex hemisphere site from 1981 to 1994, Nature, 374, 146-149, chang d potentiaan e l vorticity budge a simulate n i t d 1995. tropopause folding, J. Atmos. Sci., 51, 2246-2269, 1994. Pfister, L., K. R. Chan, T. P. Bui, S. Bowen, M. Legg, B. Levy, H., II, J. D. Mahlman, and W. J. Moxim, A strato- Gary, K. Kelly, M. Proffitt, and W. Starr, Gravity waves spheric source of reactive nitrogen in the unpolluted tro- generated by a tropical cyclone during the STEP tropical posphere, Geophys. Res. Lett., 7, 441-444, 1980. field program: A case study, J. Geophys. Res., 98, 8611- Lindzen, R. S., The Eady problem for a basic state with zero 8638, 1993. PV gradien . Atmos.pt J , -bu t0 Sci., 51, 3221-3226, Plumb, R. A., Zonally symmetric Hough modes and merid- 1994. ional circulation e middlth n i se atmosphere, J. Atmos. Mahlman, J. D., Long-term dependence of surface fallout Sci., 983-991, 39 , 1982. fluctuations upon tropopause-level cyclogenesis, Arch. Plumb, R. A., and J. D. Mahlman, The zonally averaged Meteorol. Geophys. Bioklimatol., , 299-311 18 Ser. , A , transport characteristics of the GFDL general circulation/ 1969. transport model . Atmos.J , Sci., , 298-32744 , 1987. Mahlman, J. D., H. Levy II, and W. J. Moxim, Three- Polvani, L. M., D. W. Waugh, and R. A. Plumb, On the dimensional tracer structure and behavior as simulated in subtropical edge of the stratospheric surf zone, J. Atmos. two ozone precursor experiments, J. Atmos. Sci., 37, Sci., 52, 1288-1309, 1995. 655-685, 1980. Potter, B., and J. R. Holton, The role of monsoon convec- Mahlman, J. D., D. G. Andrews, D. L. Hartmann, T. Mat- tion in the dehydration of the lower tropical stratosphere, suno, and R. G. Murgatroyd, Transport of trace constit- J. Atmos. Sci., 52, 1034-1050, 1995. uents in the stratosphere, in Dynamics of the Middle Price, J. D., and G. Vaughan, On the potential for strato- Atmosphere, edited by J. R. Holton and T. Matsuno, pp. sphere-troposphere exchang cut-ofn ei systemsw lo f . Q , 387-416 . ReidelD , , Norwell, Mass., 1984. J. R. Meteorol. Soc., 119, 343-365, 1993. . MahlmanMoximJ . W d ,. D.. J ThreeLev,an H ,, II y - Ramaswamy . SchwarzkopfD , . . ShineV.P M . , K , d an , dimensional simulations of stratospheric N2O: Predic- Radiative forcing of climate from halocarbon-induced tions for other trace constituents, J. Geophys. Res., 91, global stratospheric ozone loss, Nature, 355, 810-812, 2687-2707, 1986. 1992. McCormick, M. P., and R. E. Veiga, SAGE II measure- Randel, W. J., Global variations of zonal mean ozone during / REVIEW 4 , 33 GEOPHYSICF SO S Holton et al.: STRATOSPHERE-TROPOSPHERE EXCHANGE • 439 stratospheric warming events, /. Atmos. Sci., 50, 3308- measurements, 2, Tracer variability and diabatic meridi- 3321, 1993. onal circulation . Geophys.J , Res., , 10,319-10,33299 , Randel, W. J., J. C. Gille, A. E. Roche, J. B. Kumer, J. L. 1994b. Mergenthaler . Waters. W . FishbeinA F . . J ,. W E , d an , Swinbank . O'NeillA , d R. an stratosphere-tropospher, A , e Lahoz, Stratospheric transport fro tropice mth middlo st e data assimilation system, Mon. Weather Rev., 122, 686- latitude planetary b s y wave mixing, Nature, 365, 533- 702, 1994. 535, 1993. Toon . TurcoP . B.. O , R , . JordanJ , . GoodmanJ , . G d an , Rasch, P. J., X. Tie, B. A. Boville, and D. L. Williamson, A Ferry, Physical processe n polai s r stratospherie ic c three-dimensional transport middle modeth r efo l atmo- clouds, J. Geophys. Res., 94, 11,359-11,380, 1989. sphere, J. Geophys. Res., 99, 999-1017, 1994. Toumi Law,S. R. BekkiS. K. ,, Indirecand , t influencof e Reed, R. J., and C. L. Vlcek, The annual temperature ozone depletion on climate forcing by clouds, Nature, variation in the lower tropical stratosphere, J. Atmos. 372, 348-351, 1994. Sci.,26, 163-167, 1969. Trepte, C. R., and M. H. Hitchman, Tropical stratospheric Reid, G. C., and K. S. Gage, On the annual variation in circulation deduced from satellite aerosol data, Nature, height of the tropical tropopause, J. Atmos. Sci., 38, 355, 626-628, 1992. 1928-1938, 1981. Trepte, C. R., R. E. Veiga, and M. P. McCormick, The Reiter . Mahlman. GlasserD E . R.. . J E , M ,d role an ,f e,o Th poleward dispersal of Mount Pinatubo volcanic aerosol, e tropopausth n stratospheric-tropospherii e c exchange J. Geophys. Res.,98, 18,563-18,573, 1993. process, Pure Appl. Geophys., , 185-21875 , 1969. Tsuda, T.. MurayamaY , . WiryosumartoH , . HarB . -W . S , Remsberg, E. E., J. M. Russell III, L. L. Gordley, P. L. ijono, and S. Kato, Radiosonde observations of equato- Bailey, W. G. Planet, and J. E. Harries, Implications of rial atmosphere dynamics over Indonesia, 1, Equatorial the stratospheric water vapor distributio determines na d wave diurnad an s l tides, /. Geophys. Res., 99, 10,491- from the Nimbus 7 LIMS experiment, /. Atmos. Sci., 41, 10,506, 1994. 2934-2945, 1984. Tung, K. K., On the two-dimensional transport of strato- Robinson transpore . D.G ,Th , f minoo t r atmospheric con- spheric trace gases in isentropic coordinates, J. Atmos. stituents between troposphere and stratosphere, Q. J. R. Sci., 39, 2330-2355, 1982. Meteorol. Soc., 106, 227-253, 1980. Ucellini . Wash . H W. . L Keyser,. D ,C . Brill ,F d . an K ,, e seasonaRosenlofTh , H. l residuae . cyclK th , f eo l mean The Presidents' Day cyclone of 18-19 February 1979: meridional circulation in the stratosphere, J. Geophys. Influence of upstream trough amplification and associated Res., 100, 5173-5191, 1995. tropopause folding on rapid cyclogenesis, Mon. Weather Rosenlof . HoltonR . J . H.d ,K ,an ,Estimate stratoe th f so - Rev., 775,962-988, 1985. spheric residual circulation usin downware gth d control . PlumbWaughA . R . ,W.d D Contou, an , r advection with principle, /. Geophys. Res., 10,465-10,479, 98 , 1993. surgery: A technique for investigating fine-scale structure Rotunno . SkamarockC , R.. W ,. Snyder C d analysin an , A , s in tracer transport, J. Atmos. Sci., 51, 530-540, 1994. of frontogenesi numerican i s l simulation f baroclinio s c Wei, M.-Y., A new formulation of the exchange of mass and waves . Atmos.J , Sci., 51, 3373-3398, 1994. trace constituents betwee stratosphere nth troe th -d ean Russell, P. B., L. Pfister, and H. B. Selkirk, The tropical posphere, J. Atmos. Sci., 44, 3079-3086, 1987. experimen e stratosphere-tropospherth f o t e exchange Wirth, V., Diabatic heating in an axisymmetric cut-off cy- project (STEP): Science objectives, operations, and sum- clon related ean d stratosphere-troposphere exchange. Q . mary findings, J. Geophys. Res., 98, 8563-8589, 1993. J. R. Meteorol. Soc., 121, 127-147, 1995. Schneider, E. K., Axially symmetric steady-state models of World Meteorological Organization (WMO), Atmospheric e basith c stat r instabilitfo e d climatan y e studies, II , ozone 1985, WMO 16, Geneva, Switzerland, 1986. Nonlinear calculations, J. Atmos. Sci., 34, 280-297, 1977. WMO, Scientific Assessment of Ozone Depletion: 1994, Selkirk tropopause . B.H ,Th , e col dAustraliae trath n pi n WM , GenevaO37 , Switzerland, 1995. monsoon during STEP/AMEX 1987 . Geophys.J , Res., Yang, H., and R. T. Pierrehumbert, Production of dry air by 98, 8591-8610, 1993. isentropic mixing, /. Atmos. Sci., 51, 3437-3454, 1994. Shapiro . A.M ,, Turbulent mixing within tropopause folds sa Yulaeva, E., J. R. Holton, and J. M. Wallace, On the cause a mechanism for the exchange of chemical constituents of the annual cycle in the tropical lower stratospheric betwee stratosphere nth tropospheree th d ean . Atmos.J , temperature, J. Atmos. Sci., 51, 169-174, 1994. Sci., 37, 994-1004, 1980. Shia, R.-L., M. K. W. Ko, M. Zou, and V. R. Kotamarthi, Cross-tropopause transport of excess 14C in a two-dimen- sional model . Geophys.J , Res., , 18,607-18,61698 , 1993. Solomon, S., J. T. Kiehl, R. R. Garcia, and W. L. Grose, A. R. Douglass and R. B. Rood, Laboratory for Atmo- Tracer transpor e adiabatith y b t c circulation deduced spheres, Code 916, NASA Goddard Space Flight Center, from satellite observations, J. Atmos. Sci., 43, 1603- Greenbelt D 20771M , . (e-mail: [email protected]. 1617, 1986. nasa.gov; [email protected]) Sparling . SchoeberlR . . C. L . ,DouglassM ,R . . A J , . C , Mclntyre. E . . HaynePH .M d ,s an Departmen- Ap f o t Weaver, P. A. Newman, and L. R. Lait, Trajectory plied Mathematic Theoreticad san l Physics, Universitf yo modeling of emissions from lower stratospheric aircraft, Cambridge, Cambridge CB3 9EW, England, (e-mail: phh@ J. Geophys. Res., 100, 1427-1438, 1995. amtp.cam.ac.uk; [email protected]) Strahan, S. E., and J. D. Mahlman, Evaluation of the J. R. Holton, Department of Atmospheric Sciences, SKYHI general circulation model using aircraft NO2 measurements, 1, Polar winter stratospheric meteorology University of Washington, Box 351640, Seattle, WA 98195- and tracer morphology . Geophys.J , Res., , 10,30599 - 1640. (e-mail: [email protected]) 10,318, 1994a. L. Pfister, Space Science 245-3, NASA Ames Research Strahan, S. E., and J. D. Mahlman, Evaluation of the Center, Moffett Field, CA 94035. (e-mail: pfister@margie. SKYHI general circulation model using aircraft NO2 arc.nasa.gov)