Besides the material exchanges across the interface between the stratosphere and the troposphere(orthe tropopause), the main stratosphere-troposphereinteractions lie in the dynamical coupling between thetwo layers. Typical dynamical interactions in winterhemisphere include the leading stratospheric oscillation mode and the associated changes in the meridional circulation, which are driven primarily by breaking planetary waves in the extratropical stratosphere(Matsuno, 1970). Anomalous circulation signals in thestratosphere propagate downward during this oscillation, thereby affecting both the circulation and theclimate in the troposphere(Baldwin and Dunkerton, 2001; Thompson et al., 2002; Cai, 2003; Hu, 2006;Deng et al., 2008; Li et al., 2008; Xiang et al., 2009;Chen et al., 2013; Lu and Ding, 2013b). Because of the longer timescale of stratospheric anomalies and their temporal lead(2-3 weeks)relative to that inthe troposphere, stratospheric signals used to be regarded as another useful predictor besides ENSO forthe climate variability in the troposphere(Baldwin and Dunkerton, 2001). However, this general pictureof stratosphere-troposphere interactions is insufficientfor the practical application of stratospheric signals inpredictions because changes in the tropospheric circulation are also affected by short-timescale variations. The dynamical coupling processes are also much morecomplex than our current underst and ing.

As the largest source of planetary waves in theNorthern Hemisphere, the Tibetan Plateau(TP)canlargely dominate the location and strength of the EastAsian westerly jet and the formation of the East Asiantrough(Zou et al., 1992a, b). A number of studies byChinese scientists have shown that the thermodynamicforcing provided by the plateau is a key factor in theinitiation and maintenance of the Asian summer monsoon, the formation of climate patterns in Asia, and the global atmospheric circulation(Ye et al., 1957; Wu et al., 2002, 2005; Liu et al., 2007, 2012; Wu et al., 2007, 2012a, b; Duan et al., 2011; Liu et al., 2013). Due to the powerful thermal forcing of TP in summer, the vertical transport and the exchange of mass and constituents by deep convections near the TP arealso an important pathway for global stratosphere-troposphere exchange(STE)(Zhou and Luo, 1994;Zhou et al., 1995, 2006; Fu et al., 2006;Park et al., 2007; Mao et al., 2008; Xu et al., 2008; Zhan et al., 2008; Bian et al., 2011, 2013; Chen et al., 2012). Theeffects of this transport can in turn substantially influence the regional climate and environment(Chen et al., 2006; Hu et al., 2009; Lü et al., 2009). However, the strongest irreversible STE events are related tostratospheric polar vortex oscillation(PVO)processesaccompanied by an anomalous meridional circulation(Brewer-Dobson circulation)(Brewer, 1949; Dobson, 1956)or other large-scale dynamical processes suchas irreversible mixing and tropopause folding(Yang and Lü, 2003; Chen et al., 2006; Guo et al., 2007;Lü et al., 2008; Liu et al., 2009). The dynamics ofstratosphere-troposphere interactions around the TPtherefore forms a basis for underst and ing regional and global STE processes. However, a definitive consensus on the dynamical impact of the TP impact onstratospheric PVO and stratosphere-troposphere interactions is still lacking due to limitations in dataavailability and research techniques.

According to the isentropic potential vorticity(IPV)theory(Hoskins et al., 1985; Hoskins, 1991), the atmosphere can be divided into three layers fromthe troposphere to the stratosphere: the "underworld"where isentropic surfaces are entirely within the troposphere and intersect with the ground; the "middle world" where isentropic surfaces intersect with thetropopause but not with the ground; and the "overworld" where isentropic surfaces are entirely withinthe stratosphere.

Potential temperature and potential vorticity areconserved under adiabatic frictionless conditions. Potential vorticity(PV)surfaces can be approximatelyregarded as material surfaces, and are therefore often used to define the tropopause in studies ofstratosphere-troposphere interactions. Moreover, thepolar vortex edge that separates high PV cold polarair from low PV warm air outside is characterized bya high IPV gradient(i. e., the baroclinic zone). Largescale irreversible mixing of cold and warm air acrossthis boundary occurs during stratospheric oscillationevents when the polar vortex collapses. These eventsare typically associated with anomalous meridionalcirculations, including anomalous tropical "upwelling" and extratropical "downwelling"(Holton et al., 1995). Dynamical coupling of the stratosphere and troposphere in the midlatitude "middle world"(where isentropic surfaces and the tropopause intersect)is characterized by large-scale vortices(e. g., blocking highs and cut-off lows) and tropopause folding events. Furthermore, the "underworld"(where isentropic surfacesintersect the ground)represents the main source and sink of PV for the entire atmosphere because the sumof mass-weighted IPV within a layer surrounded byclosed isentropic surfaces does not change even in thepresence of diabatic heating or friction(Haynes and Mcintyre, 1987, 1990; Hoskins, 1991). Studies of theatmospheric circulation in the TP region have suggested that one of the most significant impacts of the TPon the atmospheric circulation is its role as a sourceor sink of IPV(Liu et al., 2001; Wu et al., 2002).

Recently, we employed the IPV theory tostudy the stratosphere-troposphere interactions associated with PVO processes, and identified the systemic meridional and vertical propagation of circulation anomalies from the stratosphere to the troposphere during the oscillations, and proposed a newstratosphere-troposphere coupling mechanism underthe global mass circulation framework(Cai and Ren, 2006, 2007; Ren and Cai, 2006, 2007, 2008). Our recent studies have also shown the lagged effect of ElNiño-Southern Oscillation(ENSO)variability in thestratosphere(Ren and Xiang, 2010; Ren, 2012a, b;Ren et al., 2012).

In this paper, we provide a systematic review ofthe above studies of stratosphere-troposphere interactions and summarize the thermal and dynamic effectsof the TP as revealed by applications of IPV theory. We also discuss the climatological significance of theTP with respect to stratosphere-troposphere coupling. 2. Propagation of stratosphere-tropospherecirculation anomalies associated with PVO 2. 1 A PVO index

Stratospheric PVO is an oscillation in the extratropical stratosphere between zonally symmetric strongwesterly winds and zonally asymmetric weak westerlywinds. The zonally symmetric case corresponds to ananomalously strong stratospheric polar vortex, whilethe zonally asymmetric case may correspond to ananomalously weak stratospheric polar vortex, or a vortex that shifts away from the polar region. In thisregard, the traditional zonal averaging may not always properly reflect changes in the strength, location, and zonal structure of the polar vortex. By contrast, the distribution of IPV can accurately characterize thepattern, intensity, and location of the polar vortex, aswell as the westerly jet along the vortex edge(coincident with the strong IPV gradient)(Baldwin and Holton, 1988; Waugh and R and el, 1999). We therefore constructed a semi-Lagrangian coordinate systemwith potential temperature(θ)as the vertical coordinate and IPV as the latitudinal coordinate. To do this, we calculated the area enclosed bYeach selected IPVcontour, and assigned each IPV line a latitude value(named IPV latitude, or PVLAT)when the spherical area surrounded by the IPV line equals that surrounded by this latitude circle. Different from conventional Eulerian or spherical coordinates, the PVLATvaries following the temporal variation of IPV. In thisθ-PVLAT coordinate, the zonal means are the averaging along the PVLAT(or IPV lines), which are verylike the "Lagrangian" averaging and effectively represent the averages on the air mass with similar properties. The contrasts between the air mass within thepolar vortex and the air mass outside are thereforemore pronounced(Fig. 1), which helps us to effectively capture the key thermodynamic processes associated with PVOs(Ren and Cai, 2006; Cai and Ren, 2007).

Fig. 1. Zonal means of(a)PV(PVU; 1 PVU = 10-6m2s-1K kg-1), (b)zonal wind(m s-1), and (c)temperature(K)at θ = 650 K on 2 February 1996. The curveswith solid dots are obtained by averaging along PVLA T, while thin curves are obtained by averaging along regularlatitudes. [From Ren and Cai, 2006]

Figure options

EOF analysis of daily IPV anomalies in the θ-PVLAT coordinate system indicates that the firstEOF mode explains as much as 69% of the variancein the daily IPV anomalies over the entire NorthernHemisphere, reflecting the intensity oscillation of thestratospheric polar vortex, or the out-of-phase relationship of IPV and temperature anomalies betweeninside and outside of the polar vortex. The "semiLagrangian" coordinate system acts as a natural filter for excluding the effects of advection, so that thetime series of the leading EOF mode of PVO is quitesmooth with less synoptic-scale disturbances, clearlydemonstrating the contrasting features between quietsummers and disturbed winters with on average 1-2 PVO events in every winter season(Cai and Ren, 2007). The maximum lead/lag correlation(-0. 91)between the PVO and the NAM(Northern Hemisphereannular mode)is at 20 hPa, when PVO leads theNAM by around 12 days. The maximum correlationstend to propagate downward with time(Cai and Ren, 2007). Since the peak phase that the PVO capturesis the most salient features of the changes in stratospheric polar vortex intensity, the downward propagation of the maximum correlations between PVO and the NAM index may therefore reflect the evolution features of circulation anomalies during the PVO(see below). Scale analysis of the PVO index shows that theaverage period of one complete PVO cycle includingboth positive(strong polar vortex) and negative(weakpolar vortex or warming event)events is roughly 116days. 2. 2 Propagation of circulation anomalies inthe str atosphere and tr oposphere

Kodera et al. (1990)identified downward propagation of stratospheric signals in the extratropics using monthly data. They found that stronger zonalmean westerly winds in the extratropical upper stratosphere in December can sustain westerlies in the lowerstratosphere and are associated with stronger westerly winds in the extratropical troposphere during thefollowing February. Later studies noted that the positive and negative phases of the AO(Arctic Oscillation), the NAO(North Atlantic Oscillation), and theleading modes of variability in the extratropical troposphere happen to correspond to the positive and negative phase of the NAM in the stratosphere. TheAO and NAO may thus be considered as surface signals of NAM variability. Links between the stratospheric NAM and the surface AO have been attributedto the systematic downward propagation of anomaliesin zonal-mean zonal wind(Kodera and Kuroda, 2000;Kodera et al., 2000) and geopotential height(Baldwin and Dunkerton, 1998, 1999). The circulation anomalies in the stratosphere lead the circulation anomaliesin the troposphere by approximately 2-3 weeks. Several studies have also showed some signs of polewardpropagation of zonal-mean zonal wind anomalies during NAM events(Feldstein and Lee, 1998; Dunkerton, 2000; Kuroda, 2002). Kodera et al. (2000)further showed some signs of simultaneous poleward and downward propagation of zonal wind anomalies by analyzing tra jectories on a phase plane defined by thetwo leading EOF vectors of zonal wind anomalies.

We used the PVO index and the "Lagrangian"zonal means in θ-PVLAT coordinates described in Section 2. 1 to analyze the spatiotemporal evolution of thecirculation anomalies from the stratosphere to troposphere and from the tropics to high latitudes, associated with PVO events. We found that the circulationanomalies indeed exhibit a systematic downward propagation, which not onlYexists in the extratropics butalso in all latitudes from the tropics to the polar region, and for both positive and negative PVO events. Besides, co-existed with the downward propagationsare simultaneous poleward propagations of circulationanomalies in the stratosphere, which also prevail inboth positive and negative PVO events. Moreover, the simultaneous propagation of temperature anomalies always leads that of other circulation anomalies. The arrival of these anomalous signals in the polarregion corresponds to the occurrence of a positive ornegative PVO event. The average time required forthese anomalies to propagate from the tropics to thepolar region is about 50-60 days(half of the PVOevent cycle)(Fig. 2; see also Fig. 7 in Cai and Ren, 2007).

Fig. 2. A series of vertical-time cross-sections showinglead/lag regressions of isentropic temperature anomalies(shading; K) and zonal wind anomalies(contours; m s-1)against the PVO index. The abscissa shows the lead-timerelative to a PVO event and the ordinate of each panel isthe isentropic vertical coordinate(ranging from 400 to 650K). The anomalies are averaged over the PVLAT b and indicated by the text to the right of each panel. The dottedblack, solid black, and white contours show positive, zero, and negative values, respectively.

Figure options

More importantly, synchronized with the poleward propagation in the stratosphere(Figs. 2 and 3a)is an equatorward propagation of circulation anomalies in the lower troposphere(Fig. 3b). Despite of thedifficulties in capturing the propagation signals in thelower troposphere due to the intersections of the isentropes with the ground, the equatorward propagationof the tropospheric anomalies still can be identifiedfrom the composites for different latitude zones(Figs. 3 and 6 in Cai and Ren, 2007).

Fig. 3. 15-day running mean composite isentropic anomalies(temperature in K in shadings; zonal wind in m s-1incontours)as a function of PVLAT(abscissa) and time(ordinate)averaged between(a)400-650- and (b)270-290-Kisentropic surfaces poleward of 30°N and 280-300-K isentropic surfaces equatorward of 30°N. The time axis(ordinate)covers two PVO cycles. Zero contours are omitted for clarity. The solid black line in(b)indicates the position of 30°N. The contours in(b)are plotted at 0. 1, 0. 4, 0. 7 m s-1(solid white), and -0. 1, -0. 4, -0. 7 m s-1(dashed black). Thecomposites shown here are based on the relative intensity of the PVO index(see detailed descriptions in Cai and Ren, 2006, 2007).

Figure options

The success in capturing the systematic propagation features shown above has benefited from the useof the semi-Lagrangian PVLAT coordinate as well asthe PVO index we constructed to represent the leading oscillation of the stratospheric polar vortex. However, the θ-PVLAT coordinate system also has somedrawbacks due to the potential for isentropes to intersect the ground. To further confirm the existence ofthese propagating signals, and to avoid the difficultiesin zonal averaging in the θ-PVLAT coordinate due tointersections of isentropes with the ground, we performed lead/lag regressions of the circulation anomalies in pressure coordinates against the same PVO index. The results show that the temperature anomalies at pressure levels located in the stratosphere consistently propagate poleward, while the temperatureanomalies at pressure levels located in the troposphereconsistently propagate equatorward. We even identi-fied changes in the direction of propagation on pressure levels that intersect the tropopause. Moreover, the downward propagation of temperature anomaliesis typically confined in the stratosphere and tends todiminish around the tropopause. As a result, the vertical extent of downward propagation at higher latitudes is typically deeper relative to the vertical extentof downward propagation at lower latitudes, consistentwith changes in tropopause height with latitude(Ren and Cai, 2007). 2. 3 Out-of-phase temp er ature anomalies in thestr atosphere and tr oposphere

Consistent with the downward propagation ofzonal wind anomalies indicated bYearlier studies, Fig. 3 also shows that zonal wind anomalies in the extratropics are consistently in phase between the stratosphere and troposphere during PVOs(contours in Fig. 3). By contrast, in the polar region, the temperature anomalies in the stratosphere tend to be outof phase with those in the troposphere(shadings inFig. 3). In other words, following the arrival of warm(cold)anomalies into the polar region in the stratosphere(shadings in Fig. 3), which leads to a polarwarming event, temperature anomalies in the troposphere are anomalously cold(warm) and begin topropagate equatorward(Fig. 3b). Linear regressionof temperature anomalies on pressure levels also con-firms this out-of-phase relationship between the stratosphere and troposphere(Ren and Cai, 2007). Thisimplies that temperature anomalies of the same signsdo not propagate directly downward into the troposphere. The question is how to underst and these outof-phase temperature anomalies in the context of thedownward propagation of zonal wind and geopotentialheight anomalies from the stratosphere to troposphere.

Further analysis in Cai et al. (2007)by applying the IPV theory(Hoskins et al., 1985)indicated that a positive/negative IPV anomaly in theupper atmosphere will induce a positive/negative vorticity anomaly and a negative/positive temperatureanomaly in the troposphere(i. e., a cold low or a warmhigh). While in a relatively shallower surface layer, thepolarity of the IPV anomaly is determined by the stability of the layer rather than the vorticity field, so thattemperature anomaly of one sign is accompanied withvorticity anomaly of the opposite sign(a warm low ora cold high). Therefore, the out-of-phase coupling of IPV anomalies(Fig. 4d)between the stratosphere and troposphere naturally leads to an out-of-phase coupling of the temperature anomalies(Fig. 4a), as wellas an in-phase coupling of geopotential height(Fig. 4e) and zonal wind anomalies(Fig. 4c). However, itshould be noted the in-phase relationship of geopotential height and zonal wind anomalies does not implya direct downward propagation from the stratosphereto the troposphere, but rather reflects the different responses of the atmosphere to IPV forcing in the upper and lower layers(as indicated in Fig. 4c by the minimum in geopotential height anomalies near 300 K).

Fig. 4. Vertical-time cross-sections of 15-day runningmean composite anomalies of(a)isentropic temperature(K), (b)static stability(s-2), (c)zonal wind(m s-1), (d)isentropic PV(PVU), and (e)Montgomery potential(m2s-2)averaged between 60° and 90°N. Positive anomaliesare shaded. The abscissa is the timeline of a compositePVO event in days. The ordinate is the isentropic verticalcoordinate in K. The vertical scale is enlarged slightly between 270 and 300 K as indicated by the heavy horizontalline. [From Cai and Ren, 2007]

Figure options

According to the theoretical result by Matsuno(1971)about the Stratospheric Sudden Warming(SSW), the critical altitude at which the planetary wave can reach will successively decrease following the weakening of westerlies due to wave breaking, seemingly related to the downward propagation ofstratospheric circulation anomalies in the extratropics. However, the systematic downward propagation during positive PVO events(stronger polar vortices)cannot be explained by using the Matsuno theory. Evenif the recovery of circumpolar westerlies after an SSWevent is assumed to progress from upper to lower layers and thus may imply a downward propagation(Lu and Ding, 2013a), it is unclear why downward propagation still exists during positive PVO events that arenot preceded by SSWs. Further, why does downwardpropagation consistentlYexist in all latitudes? What isresponsible for the opposite equatorward propagationin the troposphere? Is there any relation of this equatorward propagation with the poleward propagationin the stratosphere? Cai and Ren(2006, 2007) and Ren and Cai(2008)applied the global mass circulationtheory(Johnson, 1989)to investigate the stratospheretroposphere dynamic coupling and proposed a physicalexplanation for the observed vertical and meridionalpropagations. 3. 1 The glob al mass circulation in isentr opiccoordinates

The global mass circulation which was proposedby Johnson(1989) and his collaborators(Gallimore et al., 1981; Townsend et al., 1985)by using the climatological mean isentropic mass stream functions, is aHadley-like, hemispheric scale, single cell meridional circulation. It consists of a warm branch that upwells in the tropics and then goes poleward in theupper layer, and a cold branch that sinks in highlatitudes and then backs equatorward in the lowerlayer. They indicated that the global mass circulationis forced by the meridional heating/cooling gradient, as well as the existing wave effect. The strength ofthe mass circulation is therefore stronger in the winter hemisphere than in the summer hemisphere, and is generally stronger in the Northern Hemisphere thanin the Southern Hemisphere. Benefited from the increased resolution of data and the development of research methods, the global mass circulation has recently been further developed(Pauluis et al., 2008;Cai and Shin, 2014), particularly in the realization ofdiagnosis on daily distribution and variability of themass circulation. Based on the equation of isentropicmass changes, daily changes of isentropic mass consist of adiabatic mass convergence/divergence alongisentropic surfaces and diabatic mass transport acrossisentropes(Cai and Ren, 2007; Cai and Shin, 2014). In contrast to the conventional representations of themeridional overturning circulation(the residual circulation and the Brewer-Dobson circulations), the masscirculation is able to explicitly resolve the zonal variations of mass transport on a daily basis. 3. 2 Intr aseasonal variability of the glob al masscirculation and str atosphere-tropospherec oupling associated with PVO

Cai and Ren(2006, 2007)used the spatiotemporal evolution of isentropic mass anomalies to demonstrate the variation of the mass circulation duringPVO events. They found that, accompanying thePVO, the isentropic mass anomalies also exhibit asystematic and simultaneous poleward and downwardpropagation; particularly, the mass anomalies in thetroposphere also propagate oppositelYequatorward, in agreement with the poleward warm branch and the equatorward cold branch of the mass circulation. The positive/negative tropospheric mass anomalies always begin to propagate equatorward following thearrival of positive/negative stratospheric mass anomalies in the polar region. This manifests the successivestrengthening(weakening)of the upper warm branch and the lower cold branch of the mass circulationon seasonal timescales. Namely, the mass circulationstrengthens during negative PVO events and weakensduring positive PVO events(Figs. 5a and 5b). Moreover, the poleward propagation always seems to appear earlier in the upper stratosphere and later in thelower stratosphere, or the poleward propagating signals in the upper layer always lead that in the lowerlayer. This then explains the downward propagationof circulation anomalies existing in all latitudes.

Fig. 5. Latitude-time cross-sections of 15-day runningmean composite mass anomalies(kg)(a)above 550 K, (b)below the lowest isentropic surface(270 K), and (c)integrated over the total column. Positive anomalies areshaded. The abscissa in(a-c)is the timeline of a composite PVO event in days, and the ordinate in(a) and (b)isPVLA T. [Adapted from Cai and Ren, 2007]

Figure options

More specifically, because the peak phase of negative PVO events(weaker polar vortex)(around day75 of the abscissa in Fig. 5)just corresponds to thetime when warm(positive mass)anomalies arrive inthe polar region and just prior to the strengthening of cold branch, the out-of-phase relationship of temperature anomalies between the stratosphere and the troposphere is therefore inevitable. Meanwhile, the enhanced mass transport into the polar region by thestrengthened warm branch can lead to increases in surface pressure(because it is before the strengthening ofthe cold branch or before the cold air mass moves outof the polar region)(Fig. 5c). At this time, a coldsurface high exists in the surface layer below a warmhigh in the upper layer(stratosphere or upper troposphere). Similarly, at the peak phase of a positivePVO(stronger polar vortex)event, the warm branchis anomalously weak with negative polar mass anomalies. In this case, a warm surface low exists belowa cold upper-level low. These facts explain why thegeopotential height and the zonal wind anomalies inwinter polar region always exhibit an in-phase relationship between the upper and the lower layers. On the other hand, the fact that surface cold high(warmlow)must weaken with height while upper warm high(cold low)must strengthen with height(determinedby static stability relation), also explains the minimumvalues of geopotential height anomalies in the middlelayer when there is an in-phase coupling between theupper and the lower layers(ex. the minimum of geopotential anomalies around 300 K in Fig. 4e). Therefore, the downward propagation as well as the opposite meridional propagation of circulation anomaliesfrom the stratosphere to troposphere, is primarily determined by the coupling and successive changes ofthe upper and lower branches of the mass circulationduring the PVO.

Cai and Ren(2006, 2007)further indicated thatthe seasonal variability of mass circulation is accompanied by advancement of successive steepening(strengthening)or leveling(weakening)of a series ofisentropic surfaces(or baroclinic zones)from the equator to the polar regions and from the upper to thelower layer, and also the meridional exchange of cold and warm air masses(Fig. 6). Specifically, due to theearth's rotation, the westerly flow associated with thevertically sloped baroclinic zones that prevail from thetroposphere to stratosphere and from the subtropicsto the pole is a physical barrier for a direct meridional exchange of warm and cold air masses. Thedynamically-driven cross-frontal circulation associatedwith an intensification of meridional temperature gradient acts as a "pump" that pulls warm(cold)air poleward(equatorward)over(under)the westerly jet(Fig. 6a). At upper levels, the mixing of warm and coldair masses by the cross-frontal ageostrophic circulation leads to a poleward advancement of warm air, and thus a new development of frontogenesis in the cold airsector. At lower levels, the equatorward advancementof cold air results in a frontogenesis in the warm airsector. As a result, the cross-frontal circulation propagates poleward and downward as the baroclinic zonefurther leans towards the cold air sector accompaniedwith a poleward advancement of warm air mass above and an equatorward advancement of cold air below(Fig. 6b). A series of advancements of cross-frontalcirculations leads to simultaneous poleward and downward propagation of stratospheric anomalies of bothsigns from the tropics to the pole, whereas the tropospheric temperature anomalies adv ance towards thelow latitudes from high latitudes. A new round ofsuch a series of advancements of cross-frontal circulations would restart soon after the frontolysis process isover, as long as the system is continuously sub ject toa diabatic heating in low latitudes and cooling in highlatitudes. Obviously, accompanied with SSW event isthe most rapid leveling of the strongest baroclinic zoneat the edge of the polar vortex.

Fig. 6. Schematic diagrams showing the evolution ofthe circulation across a baroclinic zone based on semigeostrophic frontogenesis theory. (a)An earlier time and (b)a later time showing a more leveled baroclinic zone. The red lines are isentropic surfaces(θ3 > θ2 > θ1). Thecurved block arrows indicate the advance of warm air associated with the cross-frontal circulation(black arrows), and blue triangle wedges represent the advance of cold airnear the surface. Circles with a dot inside indicate a westerly anomaly and circles with a cross indicate an easterlyanomaly. The abscissa represents a latitude coordinatefrom south to north and the ordinate represents a pressure coordinate increasing downward. [From Cai and Ren, 2006]

Figure options

The discovery and confirmation of the Brewer-Dobson circulation serve to explain the distribution and transport of water vapor and ozone(Brewer, 1949;Dobson, 1956). The Brewer-Dobson circulation upwells in the tropics, extends poleward, and then sinksin polar region to complete the vertical and meridional transport of water vapor and ozone. Since itis inevitable that the Brewer-Dobson circulation alsotransports air mass, a back equatorward branch wasalso identified then(Brewer, 1949).

Based on the wave-mean flow interaction theory, Haynes et al. (1991)derived the "residual circulation" driven by an extratropical "air pump" due to waveforcing in the stratosphere. This theoretically provedexistence of the Brewer-Dobson circulation in thestratosphere, and on the other hand, indicated the significance of wave forcing to the Brewer-Dobson circulation. However, the Brewer-Dobson circulation wasdefined on long-term mean basis, and the "residualcirculation" was derived in a zonal-mean framework. In contrast, the mass circulation is based on the original total flow, and it can explicitly define the zonalstructure of mass distribution and transport. Therefore, the mass circulation can be defined accuratelyon a variety of timescales, and by definition, it manifests the effects of both the meridional gradient ofdiabatic heating and wave activity. Moreover, components of mass transport by diabatic(vertical) and adiabatic(horizontal)circulation can be distinguished. 4. Interannual variability of the stratosphericPVO

The major variability of the stratospheric circulation is dominated by internal variability on seasonaltimescales; however, significant interannual variability is also observed. The main factors responsiblefor interannual variability in the stratospheric circulation include ENSO, volcanic eruptions, and the quasibiennial oscillation(QBO). Volcanic eruptions are irregular and the impact of the QBO is modulatedby ENSO( Wei et al., 2007). ENSO, as one of themajor sources for the interannual variability of theclimate system, becomes a key impacting factor forthe interannual variability of the stratospheric circulation. EarlYevidence indicated that Northern Hemisphere stratospheric vortex tends to be anomalouslyweak/strong during warm/cold ENSO winters(vanLoon et al., 1982; Hamilton, 1995; Sassi et al., 2004;Taguchi and Hartmann, 2006; Manzini et al., 2006;Camp and Tung, 2007). Some evidence indicates thatthe strongest effects of ENSO on the stratosphere mayappear months after the mature phase of ENSO(Chen et al., 2003; Garćıa-Herrera et al., 2006). Our studies, which are based on multiple sea surface temperature(SST) and reanalysis datasets, indicated that theresponses of the extratropical stratosphere to ENSOforcing prevail in both the concurrent winter of mature ENSO and the next winter season after matureENSO, and the polar anomalies in the next winterare much stronger and with a deeper vertical structure than that in the concurrent winter. This lead/lagcoupling relationship between ENSO and the stratosphere is especially significant on the lead timescale ofENSO(3-5 yr), implying that the major impacts ofENSO on the extratropical stratosphere lie in its delayed effects(Ren and Xiang, 2010; Ren, 2012b; Ren et al., 2012). Actually, the delayed effects of ENSO onthe troposphere have been indicated in earlier studies. For example, strongest temperature anomalies in thetropical troposphere always lag the ENSO peaks by around one to two seasons(Newell and Weare, 1976;Angell, 1981; Yulaeva and Wallace, 1994; Huang and Huang, 2009). Although the tropical eastern PacificSST anomaly is weaker during the summer followingan ENSO peak, the response oFupper troposphericgeopotential height tends to be even stronger than thatin the previous summer(Kumar and Hoerling, 2003).

The delayed impact of ENSO on the stratosphereis associated with planetary wave activity and interannual anomalies in the global mass circulation thatare stimulated by ENSO. In specific, accompanying anENSO event on approximately 3-5-yr timescale(Fig. 7a), stratospheric mass anomalies in 430-700-K isentropic layer exhibit an interannual timescale poleward and downward propagation, which appears in low latitudes in the initiating stage of the ENSO event(Figs. 7b and 7c), and persists in the midlatitudes from theprevious summer before ENSO peak to the next summer after ENSO peak, though minor mass and temperature anomalies do appear in the polar region during the concurrent winter. Till the next winter seasonafter ENSO peak, the most significant and strongestmass and temperature anomalies arrive in the polarregion, completing a cycle of strengthening or weakening mass circulation successively from the tropics tothe polar region on the interannual timescale.

Fig. 7. (a)Auto-regression of the Niño3 index, and lead/lag regressions of(b)stratospheric mass anomalies in theisentropic layer between 430 and 700 K(shaded; kg m-2) and (c)stratospheric temperature anomalies averaged between30 and 100 hPa(shaded; K)against the winter filtered Niño3 index. The lead time of the index varies from -18 to +18months. Dashed line indicates the 90% confidence level. [Adapted from Ren et al., 2012]

Figure options

Nevertheless, when an ENSO event exhibits amuch shorter(less than 3 yr)or a much longer(morethan 5 yr)timescale, the peak of ENSO may not appear in winter, and the coupling relationship betweenENSO and the stratosphere may also be different fromthat described above. For example, an ENSO eventthat peaks in autumn will mainly affect the stratosphere in the following winter(Ren, 2012b). BothENSO and stratospheric variability are phase-lockedto the winter season, which defines the seasonality oftheir coupled relationship(Ren, 2012a). In general, due to the limitation of data length and numbers ofobserved ENSO cases, the statistical significance and stability of this coupling are not yet fully assured. Thislack of certainty is particularly for the coupling processes associated with different types of ENSO. Theeffect of warm pool ENSO on the extratropical stratosphere may largely be associated with the phase of theQBO(Xie et al., 2012). 5. Influences of the Tibetan Plateau on stratosphere-troposphere coupling

The source/sink of PV in the "underworld" is affected directly by diabatic heating and friction at thesurface. The Tibetan Plateau(TP)becomes an important source of sensible heating during spring and summer(Ye et al., 1957; Yu et al., 2011a, b), so that manyof the isentropic surfaces near the TP intersect withthe ground. The strong surface sensible heating overthe TP drives an "air pump" that forces deep convection, leading to strong convective latent heat releaseover the TP during summer(Wu et al., 1997). Budgetanalysis of PV fluxes over the TP in July shows thatthe diabatic heat source over the TP induces a cycloniccirculation near surface. The surface friction effects ofthe TP represent a source of negative PV in the upper troposphere. Surface friction and diabatic heatingtogether induce a strong anticyclonic circulation(theSouth Asian high or SAH)in the upper troposphere and lower stratosphere over the summer TP(Liu et al., 2001).

Numerical experiments with and without TP sensible heating show that the anticyclone at 200 hPa resulted from the source of negative PV over the TP, can stimulate a Rossby wave train that extends fromthe eastern coast of Eurasia to the eastern Pacific and North America. This wave train can pass the influenceof the TP to remote areas via energy dispersion and has significant effects on the local circulation(Fig. 8)(Liu et al., 2001; Wu et al., 2002). The 200-hPa pressure level is located in the stratosphere in high latitudes, implying that diabatic heating over the TP insummer may influence not only the tropospheric circulation over Asia, but also the stratospheric circulationin high latitudes.

Fig. 8. Differences in 200-hPa stream function between the control experiment(CON) and an experiment with nosurface sensible heating over the summer Tibetan Plateau. Red vectors mark the Rossby wave trains. [From Wu et al., 1997]

Figure options

Figure 9 shows vertical cross-sections of the climatological mean July distributions of potential temperature(thinner lines), IPV(thicker lines), and meridional IPV advection(shadings)along the eastern flankof the TP(90°-115°E) and central Pacific(150°E-135°W), as well as the corresponding differences between these two regions. Diabatic heating over theTP raises the tropopause over the southern TP, whilethe isentropic surfaces are concave downward. Thetropopause break is more severe in the TP region thanin other regions, including the central Pacific. Isentropic surfaces that typically slope upward from southto north become much steeper and denser near the TP(thinner lines in Fig. 9a). Meanwhile, the powerfulnegative PV source in the upper troposphere over theTP yields strong negative PV anomalies in the uppertroposphere and results in a steepening of the slopesof PV contours over the TP(especially in the 350-150-hPa layer, where PV contours over the TP are almost vertical in pressure coordinates). These approximately vertical PV contours represent the dynamicaltropopause that separates the tropospheric low PV airto the south from the stratospheric high PV air to thenorth. The isentropic surfaces are approximately perpendicular to the PV contours over the TP. Northerlywinds prevail over the east flank of TP, indicating the eastern equatorward branch of the SAH. Thesenortherly winds tend to transport midlatitude tratospheric air with high PV into the subtropical Asiantroposphere along the steep isentropic surfaces. Themeridional gradients of isentropic and PV surfaces aremuch weaker over the central Pacific, so that meridional PV transport is mainly limited to the stratosphere(Fig. 9b). Comparison of Figs. 9a and 9bshows that the East Asian region is more strongly affected by the advection of high PV stratospheric airduring summer than the central Pacific region, primarily because of the existence of the TP. High PV aircan stimulate low-level cyclones and deep convection, which maYexplain the strength of monsoon precipitation over East Asia relative to other monsoon regions.

Fig. 9. Vertical cross-sections of potential temperature(thin lines; K), potential vorticity(thick lines; PVU), and meridional PV advection(shaded; PVU s-1)averaged over(a)90°-115°E and (b)150°E-135°W. The differences betweenthese two regions are shown in(c). The topography of the plateau over 75°-105°E is shown as grey shading, and theapproximate location of the tropopause is shown as green dashed line.

Figure options

The analysis detailed above strongly suggests thatthe TP is not only a main pathway for STE, but alsohas significant impacts on stratosphere-tropospheredynamic coupling in nearby regions(including EastAsia). Further investigation of the stratosphere-troposphere interactions over the TP region may become an important means oFunderst and ing how theTP affects the summer climate of East Asia.

As mentioned in Section 1, the TP is the largestsource of planetary waves in the Northern Hemisphereduring winter. Accordingly, the effects of the TPare fundamental factors in determining the location and strength of the East Asian westerly jet, as wellas the formation of the East Asian trough(Zou et al., 1992a, b). However, the role of the TP in coupled stratosphere-troposphere variability(includingthe PVO)remains poorly understood. It is well knownthat the TP is a relatively weak heat sink during winter, although it can also serve as a heat source insome areas(Yu et al., 2011a, b), while the springtime increase in surface sensible heating over the TP ismore rapid than that over other regions. Nevertheless, there is a general lack of studies focusing on the potential impacts of thermodynamic processes over theTP on stratosphere-troposphere coupling. In particular, further studies of the mechanisms by which theTP influences regional and the global stratosphere-troposphere interactions are needed. 6. Key issues in the studies of stratosphere-troposphere coupling

Studies of the stratosphere and STE have long been limited by a lack of reliable long-term observ ations in the stratosphere and ambiguity in key parameters, such as the definition of the tropopause. Theoretical underst and ing of wave-mean flow interactions explains the mechanism of stratospheric warmingevents, but cannot provide a reasonable explanationfor the meridional and vertical propagation of circulation anomalies during PVO processes. Althoughwe have provided a qualitative physical explanationwithin the framework of the global mass circulation, no theoretical proof has yet been produced. The lackoFunderst and ing regarding stratosphere-tropospherecoupling also limits our underst and ing of interannualvariability in the stratosphere itself. In particular, the scientific community has yet to achieve consensusregarding the impacts of ENSO, the QBO, and volcanic eruptions on the stratosphere. The inability ofcurrent numerical models to provide realistic simulations of stratospheric processes is another importantlimiting factor(Charlton et al., 2007; Ren et al., 2009, 2012; Liu Y. Z., 2012; Rao et al., 2014). Improvementof numerical models in turn requires reliable observ ational data and sound theoretical foundations. Thelack of data over the TP is particularly severe, so theimprovement of numerical model performance in thevicinity of the TP represents an even bigger challenge.

More and more satellite data have begun to coverthe stratospheric layer in recent years, and simulations of the stratosphere in numerical models havebeen greatly improved by increases in model resolution and improved representations of dynamical and chemical processes. These developments will substantially adv ance our underst and ing of the stratosphere. Stratospheric research has also been attracting increasing attention in China. A key project recentlyimplemented by the National Science F oundation ofChina(NSFC)titled "L and -air Coupling Processesover the Tibetan Plateau and Their Global ClimateEffects" has adopted "stratosphere-troposphere interactions and their effects" as one of its major scientifictopics. This project will provide an invaluable opportunity for studies on relationships between theTP and stratosphere-troposphere interactions. Wetherefore have reason to expect our underst and ingof the role of the TP in modulating regional and global stratosphere-troposphere interactions and climate change to develop rapidly in the near future.