Interannual Changes of the Stratospheric Circulation: Relationship to Ozone and Tropospheric Structure

Transcription

1 3673 Interannual Changes of the Stratospheric Circulation: Relationship to Ozone and Tropospheric Structure MURRY L. SALBY AND PATRICK F. CALLAGHAN University of Colorado, Boulder, Colorado (Manuscript received 21 June 2001, in final form 12 July 2002) ABSTRACT Interannual changes of stratospheric dynamical structure and ozone are explored in observed variations over the Northern Hemisphere during the 1980s and 1990s. Changes of dynamical structure are consistent with a strengthening and weakening of the residual mean circulation of the stratosphere. It varies with the Eliassen Palm (E P) flux transmitted upward from the troposphere and, to a lesser degree, with the quasi-biennial oscillation (QBO). These two influences alone account for almost all of the interannual variance of wintertime temperature over the two decades, even during unusually cold winters. Stratospheric changes operating coherently with anomalous forcing of the residual circulation are coupled to changes of tropospheric wave structure. Those changes of dynamical structure share major features with the Arctic Oscillation. Both involve an amplification of the ridge over the North Pacific and an expansion of the North Atlantic storm track. Changes of tropospheric wave structure lead to a temperature signature of anomalous downwelling in the Arctic stratosphere. Accompanying it at a lower latitude is a temperature signature of anomalous upwelling. That compensating change operates coherently but out of phase with the temperature change over the Arctic. However, it is an order of magnitude smaller, making it difficult to isolate in individual years or in small systematic changes that characterize trends. Interannual changes of dynamical structure are mirrored by changes of total ozone. Like temperature, ozone changes are large at high latitudes. They are accompanied at lower latitudes by coherent changes of opposite sign. Those compensating changes, however, are an order of magnitude smaller like temperature. Ozone changes operating coherently with anomalous forcing of the residual circulation track observed changes. They account for most of the interannual variance. What remains (about 20%) is largely accounted for by changes of the photochemical environment, associated with volcanic perturbations of aerosol and increasing chlorine. The close relationship between these changes and observed ozone is robust: It is obeyed even during years of unusually low ozone. Total ozone then deviates substantially from climatological-mean levels. However, it remains broadly consistent with the relationship deduced from the overall population of years. 1. Introduction The stratospheric circulation plays a key role in regulating polar temperature and ozone. Temperature over the Arctic during polar darkness is maintained significantly warmer than radiative equilibrium by adiabatic warming, which accompanies downwelling in the residual mean circulation of the stratosphere. Driven by the absorption of planetary waves, the residual circulation simultaneously transfers ozone into the winter hemisphere. Such transport is responsible for the large wintertime increase of ozone each year. These central roles make the stratospheric circulation essential to understanding observed changes of temperature and ozone. The polar-night vortex undergoes large variations between years. Some of this variability is correlated to the equatorial quasi-biennial oscillation (QBO) on which Corresponding author address: Dr. Murry L. Salby, University of Colorado, Campus Box 311, Boulder, CO prior work has concentrated (e.g., Holton and Tan 1980; Labitzke 1982; Hasebe 1983; Garcia and Solomon 1987; Gray and Pyle 1989; Lait et al. 1989; Tung and Yang 1994a). The QBO favors a vortex that is warmer and weaker during winters of equatorial easterlies than during winters of equatorial westerlies. However, while some winters obey this pattern, others do not. Changes of the circulation are also implied by total ozone, which likewise undergoes large interannual variations. In addition to the QBO, ozone changes are correlated to changes of planetary wave activity, which drives the residual mean circulation (Kinnersley and Tung 1998; Fusco and Salby 1999). The interpretation of ozone changes and the identification of anthropogenic contributions rests on observed changes of temperature [World Meteorological Organization (WMO) 1999]. They are small in the Tropics, overshadowed by much larger temperature changes at high latitudes. Understanding interannual changes of both temperature and ozone relies on isolating which changes are introduced by changes of the circulation American Meteorological Society

2 3674 JOURNAL OF CLIMATE VOLUME 15 They are investigated here in observed changes of dynamical and chemical structure over the Northern Hemisphere during the 1980s and 1990s. The population of 20 winters affords the construction of a stable and selfconsistent picture of interannual changes: first for the circulation and then for contemporaneous changes of ozone. Both are shown to operate coherently with anomalous forcing of the residual circulation. The analysis isolates the major mechanisms responsible for observed changes, quantifies their respective contributions, and then establishes their corresponding structures. 2. Data Daily reanalyses from the European Centre for Medium-Range Weather Forecasts (ECMWF) provide global records of dynamical structure up to 10 mb from 1979 to Those records have been continued through 1998 in ECMWF operational analyses. They are supported by the National Centers for Environmental Prediction National Center for Atmospheric Research (NCEP NCAR) reanalyses, which likewise span the two decades The investigation of interannual changes concentrates on monthly mean data. However, quadratic quantities like the Eliassen Palm (E P) flux are first evaluated from daily data and then collected into their monthly means. The record of ozone column abundance, or total ozone, comes from the Nimbus-7 Total Ozone Mapping Spectrometer (TOMS) (V7), which operated from 1979 to The satellite record of ozone has been continued after 1993 in broken records from TOMS instruments on board the Russian satellite Meteor-3 (1994) and Earth Probe ( ). Collectively, these data provide nearly continuous records of the circulation and ozone spanning two decades. However, they are not without limitations. During winter, TOMS does not observe ozone in polar darkness. Left ambiguous are changes beyond the polar-night terminator and how they contribute to changes of overall ozone abundance. In addition, ozone measurements from Meteor-3 suffer from a drift in equatorial crossing time, which impacts TOMS data through the solar zenith angle. They also suffer from a long gap during the winter of In combination, these limitations of data from Meteor-3 introduce a mismatch relative to observations from the other satellites. 3. Changes of dynamical structure a. Annual and interannual variation Interannual changes reflect deviations of the annual cycle from one year to the next. For the stratospheric circulation, they are represented in the column-averaged temperature: FIG. 1. (a) Mean annual cycle of column-averaged temperature at mb (solid) and upward E P flux at 100 mb (dashed), each averaged over N, along with the std dev of their interannual variability (shaded). (b) As in (a), but for total ozone averaged over N (or polar-night terminator). Months of early winter and late winter indicated by shading at the bottom of (a) and (b), respectively. 100 mb 1 T T dp, (1) p 10 mb averaged over the northern extratropics T, where the overbar denotes horizontal average. Here T is, in effect, the mb thickness averaged over the extratropics of the Northern Hemisphere. Figure 1a shows, for the population of 20 years, the mean annual cycle of T, along with its interannual variance. Following the summer solstice, mean temperature (solid) decreases steadily, until reaching a minimum shortly after the winter solstice. This seasonality is mirrored in the standard deviation of T (shaded). It increases sharply after autumn, reaching a maximum in January. Superposed in Fig. 1a is the mean annual cycle of upward E P flux from the troposphere F z (dashed), averaged over the northern extratropics at 100 mb. Through the divergence theorem, F z reflects the collective E P flux convergence ( F) overhead and, hence, overall wave driving of the residual circulation (see, e.g., Andrews et al. 1987). Its annual cycle is out of phase with that of T : Fz increases during winter, when planetary waves propagate upward from the tropo-

3 3675 sphere, also reaching a maximum around January. Westward momentum transmitted upward from the troposphere by planetary waves is absorbed in the middle atmosphere. The E P flux convergence ( F 0) there drives a mean poleward drift * 0 and downwelling w* 0 over the winter hemisphere, which is accompanied by adiabatic warming. Consequently, F z represents a warming tendency in the annual cycle of T, which offsets radiative cooling inside polar darkness. The interannual variance of F z has much the same seasonality. Its standard deviation (shaded) increases after autumn, until reaching a maximum in January the same months when the interannual variance of T is large. It then decreases through spring, when stratospheric easterlies block vertical propagation. Figure 1b shows similar information for total ozone O 3, averaged over the northern extratropics. The mean annual cycle of O3 resembles that of T, but lagged 3 months, reflecting the additional time for ozone-rich air to reach the extratropics. Mean O 3 (solid) increases steadily during winter, until reaching a maximum in spring. Its standard deviation (shaded) is also large during the disturbed months of winter. About their mean annual cycles, temperature, ozone, and upward E P flux from the troposphere each undergo large changes from one year to the next. They describe a cluster of some 20 realizations of the annual cycle, as characterized by the rms variation in Fig. 1 (shaded). An individual realization then defines the interannual anomaly for that year: the deviation from the 20-yr mean annual cycle. The analysis below explores how, during an individual year, the anomaly in one of these properties is related to anomalies in the other properties. We address this in two intervals (indicated in Fig. 1): 1) early winter (October January), the interval preceding the minimum of Northern Hemisphere temperature, during which we consider interannual changes of its wintertime decrease (Fig. 1a); and 2) late winter (November March) the interval preceding the maximum of Northern Hemisphere ozone, during which we consider interannual changes of its wintertime increase (Fig. 1b). b. Interannual anomaly FIG. 2. Anomalous Jan temperature (50 90 N), as a function of year, compared against the anomalous tendency of temperature during the preceding months Oct Jan of early winter. The record of anomalous T (not shown) reveals interannual variance that is concentrated over the Arctic. Changes at lower latitudes are an order of magnitude smaller. However, they tend to occur with the sign opposite to changes at high latitudes. The interannual variance of T is also concentrated during winter (Fig. 1a). During an individual winter, anomalous temperature has two contributions: (i) The anomalous initial temperature at the start of that winter season. It reflects changes carried forward from preceding seasons. (ii) The anomalous wintertime decrease during that season. Figure 2 illustrates these contributions by comparing the anomalous minimum temperature at high latitudes T Jan (solid), as a function of year, against the anomalous wintertime decrease during the preceding months (dashed) Jan Oct Jan Oct T T T. (2) Here T Jan Oct represents the wintertime tendency of T. [Note: It is the tendency, not temperature directly, which is coupled to the residual circulation (e.g., through the thermodynamic equation and adiabatic warming and cooling).] The records in Fig. 2 are almost identical. The temperature minimum achieved during an individual January is thus controlled chiefly by the anomalous wintertime tendency during the preceding months. Little memory is carried forward from previous winters. Interannual changes of minimum temperature can thus be understood from changes in its wintertime tendency. It controls how much T decreases during an individual disturbed season. Figure 3 plots, as a function of latitude and height, the anomalous wintertime tendency of temperature (2) for Anomalous T Jan Oct is marked over the Arctic by a deep negative anomaly (temperature is colder than average), which intensifies upward to 8 K by 10 mb. At lower latitudes, T Jan Oct has the opposite sign (temperature is warmer than average), but it is comparatively weak. 1 During the same year, equatorial easterlies of the QBO span the entire lower stratosphere (not shown). They displace the critical line of planetary waves ( u 0) into the winter hemisphere, from 100 mb upward. This favors a polar-night vortex that is anomalously warm and weak. Yet, during this particular year, the vortex is actually anomalously cold and strong. Superposed at the bottom of Fig. 3 is the anomalous upward E P flux at 100 mb averaged over the same winter. Anomalous F z is negative at high latitudes, where upward E P flux from the troposphere is concentrated. Reduced F z implies diminished F above the tropopause, weakened residual motion, and a west- 1 Exceptional is a temperature anomaly at the equator, which is associated with the vertical shear of the QBO.

4 3676 JOURNAL OF CLIMATE VOLUME 15 FIG. 4. Anomalous wintertime tendency of column-averaged temperature at mb (60 90 N), as a function of year, compared against the anomalous upward E P flux from the troposphere (50 90 N) integrated over the same season. Correlation equals 0.92 (99.999% significant). FIG. 3. Anomalous wintertime tendency of temperature during 1980, as a function of latitude and pressure. Superposed at the bottom is the anomalous upward E P flux at 100 mb. erly vortex that is anomalously cold and strong as is observed. c. Relationship to upward transmission of wave activity The correspondence during 1980 between the anomalous thermal structure and wave forcing by the troposphere, in fact, reflects a more general relationship. Figure 4 compares, for the two decades, the anomalous wintertime tendency of column-averaged polar temperature (2) against the anomalous F z, integrated over the same season. The relationship between T Jan Oct and F z is clear and it is strong: Interannual changes in the wintertime tendency of Arctic temperature closely track changes in planetary wave activity transmitted upward from the troposphere. With a correlation of 0.92, temperature changes operating coherently with anomalous F z are significant at the % level (F statistic). They account for almost all of the interannual variance of T Jan Oct. The tendency of polar temperature during early winter is also correlated to the QBO, but at a much lower level that is only marginally significant. Including changes of equatorial wind in a bivariate regression increases the explained variance of T Jan Oct only slightly because almost all of the interannual variance during early winter is already accounted for by changes of F z. A similar finding was obtained recently by Newman et al. (2001), from eddy heat flux averaged over latitude. It is noteworthy that, during early winter (October January), the records exhibit little evidence of systematic changes. A trend is not present in polar temperature. Nor is one present in F z, which represents anomalous forcing of the residual circulation. The latter agrees with the recent analysis of Hu and Tung (2002), who also considered E P flux during early winter. In contrast, they argue the presence of a trend in polar temperature. The trend they recover, however, depends almost entirely on their choice of starting year. After the mid-1970s, there is little evidence of a trend during early winter, consistent with the behavior in Figs. 2 and 4. This contrasts with behavior during late winter (as will be seen shortly), when polar temperature and anomalous forcing of the residual circulation both exhibit a trend. d. Anomalous structure of interannual changes The record of anomalous F z is now used as a reference time series, against which the structure of dynamical changes is composited. The composite structure follows from the local covariance between anomalous T Jan Oct and F z, which serves as the interannual clock, and then rescaling by the standard deviation. The procedure is equivalent to a univariate regression that projects the record of anomalous T Jan Oct onto the reference time series F z (see, e.g., Hendon and Salby 1994). The resulting structure then describes temperature changes that operate coherently with changes of upward E P flux from the troposphere and, hence, with anomalous forcing of the residual circulation. Figure 5 plots, as a function of latitude and height, the anomalous wintertime tendency of temperature introduced by a one standard deviation increase of F z. It represents the climate sensitivity of stratospheric temperature with respect to changes of tropospheric structure. Anomalous T Jan Oct is 95% significant (shaded) throughout. It therefore operates coherently with F z over the entire winter hemisphere even into the subtropics of the summer hemisphere. Over the Arctic T Jan Oct is large and positive. There, it opposes radiative cooling inside polar darkness and, hence, the wintertime decrease of temperature. Analogous structure characterizes the correlation between temperature and wave forcing of the residual circulation (see, e.g., Newman et al.

5 3677 FIG. 5. Anomalous wintertime tendency of temperature as a function of latitude and pressure, introduced by a 1 std dev increase of F z. Follows three passes of a bivariate convolution. The 95% significance level is shaded. 2001). The T Jan Oct increases upward through the roof of the analyses. Coherent changes are, in fact, visible in lidar measurements even in the upper stratosphere and mesosphere, along with contemporaneous changes of ozone (Salby et al. 2002). In Fig. 5, T Jan Oct approaches 8 K by 10 mb. This is close to the overall standard deviation of stratospheric temperature, as changes operating coherently with F z account for nearly all of the interannual variance (Fig. 4). Anomalously warm temperature over the Arctic reflects enhanced adiabatic warming and downwelling, driven by increased F. South of 50 N, anomalous T Jan Oct has the opposite sign. It reflects enhanced adiabatic cooling and upwelling, which compensates the changes at high latitudes. It too is strongly coherent with changes of F z. However, it is an order of magnitude smaller. The disparate magnitude of temperature changes over the Arctic and at lower latitudes is noteworthy. Anomalous temperature at subpolar latitudes is small enough to go undetected in individual years, as well as in small systematic changes that characterize trends. The smallness of temperature changes in the Tropics has, in fact, been argued as evidence for no change of the residual circulation, leaving the much larger temperature changes at high latitudes to be ascribed to ozone depletion (WMO 1999). The very different character of temperature changes in these regions reflects the much wider area over which upwelling occurs. It mirrors the distribution of vertical motion w *, which varies from sev- eral mm s 1 at high latitudes to only a few tenths of amms 1 in the Tropics. Figure 6 presents the anomalous wintertime tendency FIG. 6. As in Fig. 5, but for the anomalous wintertime tendency of zonal-mean wind. of zonal wind u Jan Oct introduced by a one standard deviation increase of F z. The u Jan Oct is strong and negative near 65 N, where it is 95% significant. Anomalous easterlies there are consistent with the warm temperature anomaly over the Arctic and thermal wind balance. They reflect a weakening of the polar-night jet, which accompanies increased F and intensified adiabatic warming. The anomaly intensifies upward to 12 ms 1 by 10 mb. South of 50 N, the wind anomaly has the opposite sign. However, it is comparatively small. The interannual changes in Figs. 5 and 6 are substantial. They reflect changes of the polar-night jet between neighboring winters (e.g., between 1 2 standard deviations of F z) as large as m s 1. The out-of-phase structure between high and low latitudes is an intrinsic feature of dynamical variability. Compositing interannual changes against the record of anomalous polar temperature recovers nearly identical structure. 4. Relationship to tropospheric structure The anomalous dynamical structure in Figs. 5 and 6 shares major features with the Arctic Oscillation (AO), discussed by Wallace and Thompson (2000a,b) and Baldwin and Dunkerton (1999). The AO refers to the leading EOF of sea level pressure (SLP) and variability operating coherently with it. Structure in the preceding figures, on the other hand, refers to changes operating coherently with F z. Temperature changes associated with the AO maximize over the Arctic in the lowermost stratosphere. They fall off sharply at subpolar latitudes, where anomalous temperature reverses sign and remains comparatively small. As in Figs. 5 and 6, those changes

6 3678 JOURNAL OF CLIMATE VOLUME 15 FIG. 7. Jan Oct mean 100-mb height during winters characterized by (a) a 1 std dev increase of Fz and (b) a 1 std dev decrease of Fz. correspond to an intensification and weakening of the polar-night vortex. Despite their structural similarity, temperature changes operating coherently with anomalous forcing of the residual circulation (Fig. 5) are deeper than those associated with the AO. They amplify upward through the roof of the analyses. More importantly, they are 2 3 times stronger than temperature changes operating coherently with SLP, which are of the order of 3 K and smaller (Thompson and Wallace 2000a). This distinction reflects the very high correlation to F z, which exceeds 0.90 (in places 0.95). It accounts for almost all of the interannual variance of T Jan Oct. In contrast, temperature correlations to the AO decrease above 100 mb to less than As both correspond to the similar structure in the stratosphere, the distinction suggests that tropospheric changes associated with the leading EOF of SLP represent but one of several contributions to changes of F z. Other tropospheric changes, which are embodied in higher EOFs or are uncorrelated with SLP, can also contribute to F z, which is strongly coupled to changes of stratospheric temperature. a. Origin of changes in F z It is instructive to identify what changes of tropospheric structure give rise to anomalous E P flux. They have been composited in a similar fashion, but from the seasonal-mean structure. Figure 7 shows 100-mb height corresponding to a one standard deviation increase of Fz and a one standard deviation decrease of Fz, each averaged over early winter. During winters when F z is strengthened (Fig. 7a), the circumpolar jet at high latitude is displaced out of zonal symmetry and into the Eastern Hemisphere by an amplified ridge over the eastern Pacific. Although most evident at high latitudes, that ridge can be tracked equatorward to 40 N. Associated with positive anomalous F z, the structure in Fig. 7a corresponds to the AO in its negative index. Each involves anomalously warm temperature over the Pole and an exaggeration of zonal asymmetry. In contrast are winters when F z is weakened (Fig. 7b). The flow at high latitude is then close to zonal symmetry. Associated with negative anomalous F z, this structure corresponds to the AO in its positive index. Anomalous height then weakens the ridge over the eastern Pacific. Simultaneously, anomalous height attains a deep minimum over the Arctic, which reinforces the polar low. Jointly, the opposing phases in Figs. 7a and 7b reflect an amplification and weakening of wavenumber 1, which efficiently propagates upward. Anomalous structure in Fig. 7 is evident well into the troposphere. Figure 8 shows the corresponding structure at 500 mb. During winters when F z is strengthened (Fig. 8a), an amplified ridge extends northward along western Canada, clear across the polar cap and into Siberia. 2 Corresponding to the AO in its negative index, this enhancement of wave structure distorts the polar low from zonal symmetry. It now involves two separate minima, which reflect an amplification of wavenumbers 1 and 2. In contrast, during winters when F z is weakened (Fig. 8b), the ridge is confined chiefly to Canada. This leaves the high-latitude flow centered over the Pole and close to 2 Positional eastward of the ridge at 100 mb (Fig. 7a), it reflects a westward tilt with height and upward propagation of planetary waves.

7 3679 FIG. 8. As in Fig. 7, but at 500 mb. For clarity, winters plotted for a 2 std dev change of F z. zonal symmetry. It corresponds to the AO in its positive index. Under these conditions, streamlines are compressed over the North Atlantic and across the Greenwich meridian. Strong westerlies that mark the North Atlantic storm track then intensify and expand eastward: across Scandinavia and into the Baltic states. Simultaneously, the trough over Europe retreats eastward and deepens. These swings of the tropospheric circulation between winters of strong and weak F z are characteristic of the AO, as well as its cousin, the North Atlantic Oscillation (NAO). Intensification of the Atlantic storm track during winters of weak F z (Fig. 8b) is accompanied by changes to its west. The trough over North America expands southward, where it compresses height contours. In the anomalous structure (not shown), this corresponds to a low of 500-mb height over eastern Canada and the Arctic. It deepens the trough in total height. Immediately south of the low in anomalous height is a high in anomalous height, which extends across the Atlantic. Separating those oppositely signed anomalies is a steep meridional gradient. It reinforces the mean gradient of height, compressing streamlines and strengthening westerlies from 90 W to Europe. A similar structure characterizes the AO in its positive index. The changes of tropospheric structure in Figs. 7 and 8 lead to dramatic changes in the stratospheric circulation. Figure 9 compares the wintertime tendency of 50-mb height corresponding to a one standard deviation increase of F z and a one standard deviation decrease of Fz. During years when F z is strengthened (Fig. 9a), the wintertime tendency of height produces an amplified Jan Oct Aleutian high (maximum ) that invades the polar Z 50 cap. It displaces the vortex well out of zonal symmetry, achieving cross-polar flow. Ozone-rich air moving along streamlines then invades the polar cap in a matter of Jan Oct only days. Notice that the maximum of Z 50 can be traced upward and westward from the amplified ridge at tropospheric levels in Figs. 7a and 8a. In contrast, during years when F z is weakened (Fig. 9b), the wintertime tendency of height produces a midwinter circulation that is much less disturbed. The Aleutian high is then shallow, scarcely crossing the Arctic Circle. This leaves the vortex deeper, stronger, and close to zonal symmetry, with ozone-rich air sequestered at lower latitudes. b. Feedback onto the troposphere Anomalous E P flux from the troposphere modifies the stratospheric circulation, in particular, downwelling over the Arctic. [On intraseasonal timescales (not shown), which collect in the wintertime tendency, anomalous temperature over the Arctic appears at lower levels with increasing lag from F z : first at 30, then at 50 mb, and so forth.] It is of interest to determine how far downward those changes are felt. This is illustrated in the anomalous wintertime tendency of height (rather than the seasonal-mean structure used to interpret changes of F z). Like the temperature tendency, Z Jan Oct operates coherently over the 20 years with anomalous Fz. Figure 10 plots the anomalous wintertime tendency of 100-mb height introduced by a one standard deviation increase of F z. A strong positive anomaly prevails over Jan Oct the Arctic, where Z 100 approaches 200 gpm. It opposes the mean wintertime tendency of height (negative)

8 3680 JOURNAL OF CLIMATE VOLUME 15 FIG. 9. Wintertime tendency of 50-mb height during years characterized by (a) a 1 std dev increase of F z and (b) a 1 std dev decrease of F z. that intensifies the polar low during October January. Flanking the positive anomaly to its south is anomalous negative height (also 95% significant). The structure in Fig. 10, as in preceding figures, is characteristic of the AO. However, like the zonal-mean structure (Figs. 5 and 6), it is substantially stronger at least at stratospheric levels. At 50 mb, standard deviations of the AO are associated with height fluctuations of 170 gpm over the Arctic (Thompson and Wallace 1998). Changes of 50-mb height operating coherently with anomalous F z (not shown) assume very similar structure, but exceed 415 gpm. The height anomaly in Fig. 10 reflects anomalous 100-mb thickness and, therefore, anomalous temperature beneath 100 mb. The signature of anomalous downwelling ( T Jan Oct 0) is, in fact, visible well into the troposphere. Figure 11 plots the anomalous wintertime tendency of 500-mb height introduced by a one standard deviation increase of F z. Although punctuated by ad- FIG. 10. Anomalous wintertime tendency of 100-mb height introduced by a 1 std dev increase of F z. FIG. 11. As in Fig. 10, but at 500 mb.

9 3681 FIG. 13. Anomalous Mar total ozone at N (solid), as a function of year, compared against its anomalous tendency during the preceding months Nov Mar of late winter (dashed). In the latter, anomalous structure has been extended beyond 65 N by extrapolating ozone during Nov, when interannual variance is small (Fig. 1). FIG. 12. As in Fig. 10, but at 1000 mb. Jan Oct ditional eddy structure, Z 500 retains the basic pattern of anomalous positive height over the Arctic. Flanking it to the south is negative anomalous height (also 95% significant). The pattern is similar to that associated with the AO in its negative index: It involves a positive anomaly at high latitudes that is displaced toward Greenland, a negative anomaly that extends from Europe across the North Atlantic, and another one that extends from Siberia across the North Pacific (see also Baldwin et al. 1994). Additional eddy structure present in Fig. 11, but not in the AO, reflects variance that operates coherently with changes of F z, but is not represented in the leading EOF of SLP. 3 During the opposite phase, when F z is weakened (positive AO index), the anomaly in Fig. 11 reverses sign. Structure over the North Atlantic is then transformed into a positive height anomaly, which flanks negative anomalous height over the Arctic. Separating them is a steep meridional gradient. It compresses streamlines from 90 W to across the Greenwich meridian (cf. Fig. 8b). Strong westerlies marking the North Atlantic storm track then intensify and expand eastward. The signature of stratospheric changes over the Arctic is evident even at the surface. Figure 12 plots the anomalous wintertime tendency of 1000-mb height. It is a surrogate for SLP. Although punctuated by additional eddy structure, SLP still evidences the basic pattern of positive height anomaly over the Arctic. Displaced 3 By ordering EOFs according to their covariance, principal component analysis emphasizes the largest scales in the leading EOFs. Analogous to spatial filtering, this limits variations in the leading EOF (from which the AO is defined) to the gravest horizontal dimensions. along the Greenwich meridian, it weakens the Icelandic low in climatological-mean SLP. Like changes above, Jan Oct Z 1000 over the Arctic operates coherently with anomalous F z. Flanking it to the south is a negative height anomaly that extends from Europe to North America. Along with its companion over the North Pacific, this feature mirrors analogous structure in the AO. Both anomalies are evident at 500 mb (Fig. 11). During winters of opposite phase (positive AO index), anomalous height reverses sign. A negative anomaly then prevails over the Arctic, reinforcing the Icelandic low. It combines with a positive anomaly to its south to compress streamlines. This reinforces strong westerlies across the North Atlantic. Thompson and Wallace (1998) identify such behavior with intensified temperature advection from maritime regions, which favors warmer surface air temperature over Eurasia (see also Shindel et al. 1999). 5. Relationship to ozone Total ozone behaves much like temperature. Interannual changes in the spring maximum of O 3 follow chiefly from changes in its wintertime increase during the preceding months of late winter (Fig. 1b). Figure 13 compares, as a function of year, anomalous total ozone during March against its anomalous increase during November March. The O Mar Nov 3 represents the wintertime tendency of ozone. As for temperature, it is the tendency, not ozone directly, which is coupled to the residual circulation (e.g., through the continuity equation and transport and temperature-dependent chemical production). Over the two decades, O Mar 3 tracks O Mar Nov 3. Little memory is thus carried forward from previous winters (see also Hadjinicolaou et al. 1997; Fusco and Salby 1999). Exceptional are winters before 1983, preceding El Chichón, when anomalous March ozone is greater than the anomalous wintertime tendency. 4 A positive ozone residual is then 4 This discrepancy is visible for wintertime intervals extending backward from March as far as August, which recover behavior identical to that in Fig. 13. It is also visible in anomalous temperature during late winter (not shown).

10 3682 JOURNAL OF CLIMATE VOLUME 15 passed forward from preceding winters. It disappears in 1983, following El Chichón, and remains insignificant over the following decade, until In winters immediately following Pinatubo, notably in 1993 and 1994, anomalous March ozone is less than the anomalous wintertime tendency. A negative ozone residual is then passed forward from preceding winters, reflecting enhanced ozone depletion during those volcanically perturbed years. However, like the positive residual during the early 1980s, it disappears within a couple of years. By the end of the record, anomalous March ozone again coincides with the anomalous wintertime tendency. The anomalous ozone increase during late winter accounts for most of the interannual anomaly in spring ozone. Interannual changes in spring ozone can therefore be understood from changes in its wintertime tendency. The record of anomalous O 3 Mar Nov (not shown) reveals interannual variance that is large at high latitudes, like temperature. Changes at lower latitudes are much smaller. However, they too tend to occur with the sign opposite to the larger changes at high latitudes. During late winter, when ozone increases, the residual circulation is also influenced by the QBO. Equatorial easterlies displace the critical line of planetary waves into the winter hemisphere. This shifts strong wave absorption, and downwelling to its north, toward the polarnight vortex. Conversely, equatorial westerlies remove the critical line into the summer hemisphere. Strong absorption then exerts a weaker influence on the polarnight vortex. By controlling where wave activity is absorbed, the QBO modulates the residual circulation (see Tung and Yang 1994b). During some years, interannual changes of F z and the QBO reinforce one another. However, during others, these influences interfere (cf. Fig. 3). This accounts for the relationship between the vortex and the QBO being obeyed in some years, but not in others (section 3b). For most years, F z carries the day, as is suggested by Fig. 4 and its strong coherence with dynamical changes. Generalizing the univariate analysis in section 3 to a multivariate regression accounts for changes of both F z and the QBO. Included is the phase of the QBO. It varies with height and time, so it is tantamount to selecting equatorial wind at a particular level and time lag. The QBO s phase is determined, jointly with the multivariate regression, through a variational algorithm that isolates the maximum projection of anomalous ozone onto the records of F z and the QBO. The collective procedure then has a parallel with EOF analysis in that it maximizes the covariance between observed changes. Covariance between anomalous O Mar Nov 3 and equatorial wind is maximized for peak easterlies near 10 mb with a lead of 1 month. This phase of the QBO positions maximum equatorial westerlies in the lowermost stratosphere and maximum equatorial easterlies FIG. 14. Anomalous wintertime tendency of total ozone at N (solid), as a function of year, compared against that operating coherently with anomalous forcing of the residual circulation (dashed). The latter is defined from F z at N during Dec Feb and the QBO, which is represented by equatorial wind at 10 mb with a lead of 1 month. Correlation equals 0.89 (99.999% significant). To eliminate spurious changes introduced by variations of satellite viewing geometry and missing data, 1994 has been omitted. near 10 mb. (It is equivalent to equatorial wind at intermediate levels, with a time lag of approximately 1 month km 1 below maximum easterlies.) The record of anomalous temperature happens to obey the same relationship to the QBO. In this phase, the QBO removes the critical line in the lowermost stratosphere into the summer hemisphere. Tropospheric planetary waves can then propagate upward freely (suffering little attenuation), until reaching the midstratosphere, where the critical line veers into the winter hemisphere. Strong absorption ( F) is then concentrated at higher latitudes (O Sullivan and Salby 1990; O Sullivan and Young 1992; Tung and Yang 1994b). There, it reinforces poleward drift, downwelling, and adiabatic warming at levels below 10 mb. Those are the same levels in which the column properties T and O 3 are concentrated. Interannual changes of F z and the QBO represent anomalous forcing of the residual circulation. Operating with it are coherent changes of temperature and ozone. Figure 14 plots the anomalous wintertime tendency of the extratropical ozone operating coherently with anomalous forcing of the residual circulation (dashed). Like temperature (Fig. 4), it closely tracks observed changes over the two decades (solid). Changes operating coherently with anomalous forcing of the residual circulation have a correlation to observed ozone changes of 0.89 (99.999% significant). They account for most of the interannual variance. Calculated in a similar fashion are coherent changes of temperature (not shown). They track anomalous forcing of the residual circulation with the same correlation as ozone. For each, the correlation of 0.89 during late winter is almost as high as the variance explained by anomalous forcing of the residual circulation during early winter. However, during late winter, ozone and temperature both exhibit a downward trend over the 1980s and early 1990s, before rebounding in the closing years of the record. As is apparent from Fig. 14, anomalous forcing of the residual circulation exhibits very similar behavior.

11 3683 The multivariate analysis accounts for interdependence of F z and the QBO, through their cross covariance. It is generally small, implying little influence of the QBO on upward E P flux at 100 mb. Feedback between those influences can be accounted for in a similar fashion, by expanding the multivariate analysis to include nonlinear terms. The resulting nonlinear regression increases the explained variance little. It reflects a second-order correction, as most of the interannual variance is already accounted for. Feedback of ozone changes onto dynamical structure can likewise be addressed through nonlinear regression. It leads to a similar conclusion. The close relationship between changes of ozone and anomalous forcing of the residual circulation (Fig. 14) is robust: It is obeyed even during years of unusually low spring ozone, like 1993 and 1997 (solid circles). It should be recognized, however, that changes of the residual circulation influence ozone through transport as well as photochemistry. The rate at which air drifts poleward across contours of photochemical lifetime affects chemical production and destruction. It too contributes to the wintertime tendency of extratropical ozone. That contribution becomes increasingly important at polar latitudes, where temperatures are cold. The decrease of temperature that accompanies a weakening of the residual circulation can then reinforce chemical destruction through heterogeneous processes that enhance chlorine activation (WMO 1999). At the subpolar latitudes observed continuously by TOMS, however, such processes are unlikely. The small variance in Fig. 14 that remains unexplained comes chiefly from those winters when anomalous ozone is carried forward from previous years (cf. Fig. 13). During winters preceding El Chichón, more ozone appears in the extratropics (solid) than is accounted for by anomalous forcing of the residual circulation (dashed). Conversely, during winters immediately following Pinatubo, less ozone appears in the extratropics than is accounted for by anomalous forcing of the residual circulation. However, this discrepancy also disappears within a couple of years. By closing years of the record, anomalous spring ozone in Fig. 13 reduces to its anomalous wintertime tendency. Simultaneously, the gap between observed ozone changes in Fig. 14 and those accounted for by the anomalous forcing of the residual circulation narrows to zero. Not accounted for in Fig. 14 are changes of photochemical environment. Ozone production and destruction during individual winters changes with chlorine and aerosol loading. Photochemical models that include heterogeneous processes indicate that ozone destruction depends on total chlorine loading [Cl y ] and logarithmically on aerosol surface area S (e.g., Solomon et al. 1996). Changes of photochemical environment are thus characterized by an ozone depletion factor FIG. 15. As in Fig. 14, but also accounting for changes of ozone depletion during individual winters, introduced by changes of chlorine and aerosol loading. Correlation equals 0.95 (99.999% significant). lns ODF [Cl ] 1, (3) y lns 1979 which varies over the two decades under consideration (Fusco and Salby 1999). The evolution of [Cl y ] follows from observed levels. They imply increases of column abundance from about to about molecules cm 2 (WMO 1999). The record of stratospheric aerosol is constructed from satellite measurements by the Stratospheric Aerosol Measurement II (SAM II) and the Stratospheric Aerosol and Gas Experiment I (SAGE I) and II (SAGE II) instruments. Integrating those data over the Northern Hemisphere yields a record of ODF over the two decades (not shown). It characterizes interannual changes of ozone destruction as air drifts poleward during winter. The record of ODF is punctuated by a sharp increase in 1983, following El Chichón, and an even larger one in 1992, following Pinatubo. Separating them is a plateau, during which ODF is almost flat. Increases of chlorine during those intervening years are offset by decreases of aerosol following El Chichón. Before 1983, ODF is distinctly smaller, whereas after 1993 it is distinctly larger. Those decadal changes have the same timing as the outstanding variance in Fig. 14. Incorporating the record of ODF into the multivariate analysis accounts for anomalous forcing of the residual circulation and anomalous photochemical environment. It recovers the anomalous wintertime increase of ozone in Fig. 15 (dashed). Eliminated is the minor discrepancy preceding El Chichón and the one following Pinatubo, which contributed most of the outstanding variance in Fig. 14. The correlation to observed changes has increased to 0.95 (99.999% significant). Collectively, ozone changes operating coherently with these two mechanisms account for virtually all of the interannual variation of ozone over the two decades. Included are years of unusually low ozone. Feedback between anomalous photochemical environment and forcing of the residual circulation can be addressed by expanding the multivariate analysis to include nonlinear terms. They account for changes in F z

12 3684 JOURNAL OF CLIMATE VOLUME 15 FIG. 16. Anomalous wintertime tendency of total ozone O Mar Nov 3, as a function of latitude, introduced by a 1 std dev increase in forcing of the residual circulation. Anomalous structure has been extended beyond 65 N by extrapolating ozone during Nov, when interannual variance is small. and the QBO introduced by changes of chlorine and aerosol. Including such interaction adds little to the ozone variance explained, at least at the subpolar latitudes observed continuously by TOMS. The interannual variance is largely accounted for by changes introduced directly through anomalous forcing of the residual circulation and anomalous photochemical environment. Calculations with a chemical transport model, in which photochemistry and observed dynamical changes are explicitly represented, lead to a similar conclusion (Chipperfield and Jones 1999; Hadjinicolaou et al. 2002). Anomalous structure of interannual changes The structure of ozone changes operating coherently with anomalous forcing of the residual circulation has been composited as before, now against collective changes of F z and the QBO. Figure 16 shows, as a function of latitude, the change of O 3 Mar Nov introduced by a one standard deviation increase in forcing of the residual circulation. It represents the climate sensitivity of ozone with respect to changes of the residual circulation. The anomalous wintertime tendency of ozone is positive in the extratropics. Approaching 30 Dobson units (DU) at high latitude, it reflects intensified downwelling that displaces (ozone rich) mixing ratio surfaces downward. This increases ozone number density at lower levels, where total ozone is concentrated. (The intermediate maximum near 40 N reflects the residual circulation of the QBO, which extends into the midlatitudes; the feature is absent in the structure that has been composited against F z alone.) In the Tropics, anomalous O 3 Mar Nov has reversed sign. It reflects intensified upwelling that displaces (ozone lean) mixing ratio surfaces upward, decreasing total ozone. However, like temperature, ozone changes in regions of upwelling are comparatively small. They, nevertheless, operate coherently with the larger ozone changes at high latitude. Both operate coherently with anomalous forcing of the residual circulation, The anomalous structure in Fig. 16 implies ozone changes between neighboring winters (e.g., between 1 2 standard deviations of the residual circulation) as large as DU. They are as large as observed changes that have been ascribed to ozone depletion (WMO 1999). It is noteworthy that the out-of-phase structure between high and low latitudes is an intrinsic feature of ozone variability as was found earlier for temperature. It is equally apparent in the latitude height structure of ozone changes, observed by the Microwave Limb Sounder (MLS) on board the Upper Atmosphere Research Satellite (UARS; Fusco and Salby 1999; Fig. 5). Compositing changes against TOMS ozone at the polar-night terminator recover a nearly identical structure. 6. Conclusions Interannual changes of stratospheric dynamical structure are consistent with anomalous forcing of the residual circulation. It varies between years with the E P flux transmitted upward from the troposphere and, to a lesser degree, with the QBO. These two influences alone account for nearly all of the interannual variance of wintertime temperature. This contrasts with the QBO alone, which tracks changes of the vortex during some years, but not in others. The close relationship between stratospheric temperature and anomalous forcing of the residual circulation is obeyed even during unusually cold winters, like 1989 and Changes of upward E P flux from the troposphere are coupled to changes of tropospheric wave structure, notably, to changes of the ridge over the North Pacific and changes of the North Atlantic storm track. They lead to changes of downwelling and adiabatic warming in the Arctic stratosphere. The signature of anomalous downwelling extends upward coherently through the roof of the analyses. It also extends downward into the Arctic troposphere, where coherent changes are clearly visible. Anomalous downwelling over the Arctic is attended by a signature of anomalous upwelling at lower latitudes. It is, however, much smaller, reflecting smaller vertical motion and the wider area over which upwelling occurs. Nonetheless, temperature changes at subpolar latitudes operate coherently with changes of opposite sign over the Arctic. Both vary coherently with anomalous forcing of the residual circulation. These changes of dynamical structure share major features with the AO. Each is manifested in pronounced changes of polar temperature. They penetrate well into the troposphere, where changes of thermal structure modify the circulation at high latitudes. Changes in the Arctic troposphere are accompanied by changes at subpolar latitudes: an amplification of the ridge over the North Pacific and an eastward expansion of the North Atlantic storm track. These features are also characteristic of the AO, as well as its cousin, the NAO. Changes of dynamical structure are mirrored by interannual changes of ozone. Like temperature, ozone