On the origin of polar vortex air

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 106, NO. D24, PAGES 33,485-33,497, DECEMBER 27, 2001 On the origin of polar vortex air Joan E. Rosenfield General Sciences Corporation, Beltsville, Maryland, USA Mark R. Schoeberl NASA Goddard Space Flight Center, Greenbelt, Maryland, USA Abstract. Forward and backward three-dimensional stratospheric polar vortex trajectory computations between fall and spring for each of the years through in the Northern Hemisphere and the years in the Southern Hemisphere have been carried out. We find that the forward and backward trajectories give very different pictures of polar vortex descent. The backward trajectories show a complex distribution of parcels in which one population originates in the upper stratosphere and mesosphere and experiences considerable descent in the polar regions, while the remaining parcels originate at lower altitudes of the middle and lower stratosphere and are mixed into the polar regions during vortex formation. The forward trajectory calculations do not show this second population since the parcels in the forward calculation originate within the protovortex. The amount of descent experienced by the first population shows little variability from year to year, while the computed descent and mixing of the remaining parcels show considerable interannual variability due to the varying polar meteorology. Spring methane values reconstructed from Halogen Occultation Experiment (HALOE) fall data and the back trajectories compare fairly well with typical HALOE spring values. These results imply that using a comparison between prewinter and postwinter tracer profiles to estimate the amount of descent over the fall-to-spring time period might result in unrealistically large values of descent, since the fall profiles are not representative of the same air as the spring profiles. Additional back trajectory calculations of descent in the lower stratosphere over 3-month time periods from spring to midwinter show that very little mixing occurs at these times and altitudes in the Northern Hemisphere during winters when the vortex is undisturbed, and in the Southern Hemisphere for all years studied. Thus, for this shorter time period, tracking constant mixing ratios of tracers within the vortex should give reasonable estimates of descent amounts, except for years when there are meteorological disturbances in the Northern Hemisphere. 1. Introduction spring methane mixing ratios in the Southern Hemisphere (SH) stratospheric polar vortex were as low as those found in The discovery of polar ozone depletion has generated a great deal of interest in understanding dynamical processes in the mesosphere. Lahoz et al. [1993] also presented evidence using N20 and H20 from the Microwave Limb Sounder the polar regions, in particular the quantification of subsidence (MLS) on UARS of the descent of mesospheric and upper and the degree of isolation of polar from midlatitude air. Observed data are often analyzed by using estimates of the stratospheric air to the middle and lower stratosphere. These results provided the impetus for a number of theoretical studdescent of polar air over the winter months. This necessitates ies of polar vortex descent. Fisher et al. [1993], using a forward an accurate measure of the diabatic descent occurring during trajectory calculation based upon the wind fields generated by this time. In addition, there is also a need for an understanding a stratosphere-mesosphere general circulation model, found of the transport of air into and out of the winter polar vortices, that the air in the SH stratospheric polar vortex could descend since the amount of midlatitude air which has been trans- from the mesosphere during the fall. Nielsen et al. [1994] used ported into the polar vortex will impact estimates of polar ozone loss. Russell et al.'s [1993] observations of methane from the a global spectral mechanistic model with a full radiation scheme to study tracer evolution and also found rapid descent of polar air from the mesosphere into the stratospheric polar Halogen Occultation Experiment (HALOE) instrument on the vortex from fall to winter, similar to the results of Fisher et al. Upper Atmosphere Research Satellite (UARS) showed that [1993]. Eluszkiewicz et al. [1995], in forward trajectory calculations using winds from the Geophysical Fluid Dynamics Laboratory SKYHI general circulation model, found that parcels initialized in the polar mesosphere and upper stratosphere Now at University of Maryland Baltimore County, Baltimore, rapidly descended to the middle stratosphere. Rosenfield et al. Maryland, USA. [1994] used a radiative transfer model and a one-dimensional Copyright 2001 by the American Geophysical Union. Paper number 2001JD /01/2001JD $ vortex interior descent model to compute diabatic descent in the stratospheric polar vortex for the and Northern Hemisphere (NH) and the 1987 and 1992 SH 33,485

2 33,486 ROSENFIELD AND SCHOEBERL: ORIGIN OF POLAR VORTEX AIR late fall and winters. Their computed upper stratospheric descent of 14 km during the month of July agreed well with the descent of 12 km computed by Fisher et al. [1993]. Computations of parcel descent rates using a forward trajectory model and a radiation model in the NH and SH for the years 1992 and 1993 by Manney et al. [1994] were in general agreement with those of Rosenfield et al. [1994]. Schoeberl et al. [1995] used CH 4 observations from HALOE to compute more precise vertical descent rates in the 1992 late winter-spring Antarctic vortex by tracking the altitude of constant CH 4 mixing ratios with time. Schoeberl et al.'s computations were also in agreement with the radiative transfer estimates of heating and descent rates of Rosenfield et al. [1994] and Manney et al. [1994] in the SH. In the trajectory model [Schoeberl and Sparling, 1995] the parcels are advected using horizontal winds, interpolated to the parcel positions, from the U.K. Met Office (UKMO) data assimilation system [Swinbank and O'Neill, 1994]. The model vertical extent is hpa. The parcel potential temperatures are allowed to change using diabatic heating rates. The numerical time scheme is fourth-order Runge-Kutta, and the time step is 0.05 days. The diabatic heating rates are computed off-line using a radiative transfer model with the UKMO temperature data. The radiation model is as described by Rosenfield et al. [1994], except that the heating due to carbon dioxide in the thermal infrared uses the transmission function parameterization of Chou and Kouvaris The long survival of UARS and the HALOE instrument on [1991]. Troposphericlouds are as specified by Rosenfield et al. board has provided a multiyear data set of HALOE CH 4 ob- [1997]. servations, covering the period from 1992 to the present. Re- The trajectory model was run forward for 7 months from fall cently, Kawamoto and Shiotani [2000] estimated the middle to spring, with parcels starting in the fall at 1200 K (-4 hpa). stratospheric descent during the period by compar- In the NH the starting date was September 1, while in the SH ing Antarctic fall and spring HALOE methane profiles. They it was March 1. For each hemisphere, parcels were initialized found a biennial oscillation and a decreasing trend for the poleward of 60 ø latitude on a 2.5 ø x 3.75 ø latitude-longitude period studied. grid. The starting and ending points of this integration coincide The existence of the multiyear HALOE CH 4 data set, towith the highest-latitude HALOE observations. gether with some comparisons of forward with backward tra- Plate 1 (top) shows the latitude-potential temperature locajectory calculations which we have carried out, has motivated tions of the parcels on various days of the year for the us to reexamine the question of polar vortex descent. Different 1996 NH and 1997 SH fall-to-spring forward trajectories. measures of descent have been used in the literature, as illus- These years were chosen as typical years for the purpose of trated by the above studies. Fisher et al. [1993] and Manney et illustrating the features we discuss below. The parcels descend al. [1994] refer to the mean descent of parcels as computed in fastest in the fall months, with more than 50% falling to 800 K forward trajectories, Rosenfield et al. [1994] refer to the mean (8-10 hpa) or lower by December 1 in the NH and June 1 in change of potential temperature calculated using diabatic coolthe SH. Since cooling rates increase with altitude, the parcels ing rates, while Nielsen et al. [1994], Schoeberl et al. [1995], and descend faster in fall when they are at higher altitudes. In Kawamoto and Shiotani [2000] refer to the mean change of addition, a number of the parcels have moved to more equatracer mixing ratio profiles with time. The results of our study torward latitudes by this time as well. The expulsion of parcels will show that these measures of descent can be quite different. from the vortex is more evident in the NH than in the SH. We have used adiabatic trajectory model to investigate the During springtime, when the parcels in the polar regions have year-to-year variability and the nterhemispheric differences in net winter polar descent for the years 1992 through We reached altitudes of K, there is very little descent. will show that the forward calculations as used by Fisher et al. Although descent is greatest near the poles in the fall, by spring [1993] and Manney et al. [1994] can give misleading informathe net amount of descent is roughly the same at all latitudes tion on the amount of descent, as well as on the degree of poleward of 60 ø. isolation of polar from midlatitude air, occurring over the For comparison, a series of test parcels were fixed at the fall-to-spring time period. Our back trajectory computations same latitude-longitude locations and were allowed to descend indicate that for some years a significant fraction of the air in from the starting points using only the local heating rates. the springtime polar vortex has been mixed in from the low- These fixed parcels were not allowed to circulate. The plate and middle-latitude stratosphere during vortex formation. In shows the vertical spread in the final parcel locations due to the order to check the computations, spring methane values have variation of the heating rates with longitude. The ensemble been reconstructed from the back trajectories using fall HA- mean fixed parcel descent is greater than the forward trajec- LOE CH 4 data, and these reconstructed values have been tory descent at all latitudes. This is because the horizontal compared with observed spring HALOE CH 4 data. We first present the results of computations of net winter 7-month forward and backward trajectories for the years parcel motions and the variations in heating rates with latitude and longitude reduce the time-integrated cooling. The horizontal winds move the parcels to lower latitudes where cooling We then show the methane reconstructions and their rates are much less. For example, in the NH winter, comparison with observations. This is followed by a discussion 50% of the parcels which ended at latitudes greater than 76øN of computed descent rates over the shorter midwinter to spring time period. Finally, the summary and conclusions are presented. in the spring were at latitudes less than 45øN at some time during the 7-month period. The characteristic features discussed above were seen for all the years studied. 2. Trajectory Calculations Although the forward calculations estimate the maximum amount of descent that can occur, they are not necessarily a good indication of the actual origin of the springtime polar Three-dimensional (3-D) trajectory calculations have been vortex air. This is because low- and middle-latitude air can be carried out for the 7-month fall-to-spring period in both the NH and SH polar stratosphere. The winters included were through in the NH and 1992 through entrained within the vortex during its formation. To examine the origin of the springtime stratospheric air in the hpa altitude range, the trajectory model was run backward for

3 ROSENFIELD AND SCHOEBERL: ORIGIN OF POLAR VORTEX AIR 33,487 0 C 0 C 0 ojnlejodmo lebuolod C'Xl C i,r'-,r-. ojnlmodmoœ lebuolod ejnlmod[uoœ lebuolod ejnlejodtuo. L lebuolod

4 33,488 ROSENFIELD AND SCHOEBERL: ORIGIN OF POLAR VORTEX AIR UO!lOl J..I g o o o o 8 uo!loej-i o o o o o o uo!loejd I Ii'llllll Illl 1IIi I"! 1 illel!lille I Ii Ilili i o o g g g... _m,t... I h,,, c cl c o c c uo!loej..q UOl 0 J:! uotloej.q uo toejd

5 ROSENFIELD AND SCHOEBERL: ORIGIN OF POLAR VORTEX AIR 33, I... I... I... t...!...! I...,.,,,.,....,h,.,..h,,,...,.., o. o o 2 o o g o o o c o o c c o o o c; o c o o o o o c o o o o o c c. - uo!10.j:l uoll0. d u 10.J:l uo!10 J4 iiii '1111 [lllliiil iiit11111 ilillllll iii I!1' 1111TUlJlJllmtljilllll[f UOllOe;:l

6 33,490 ROSENFIELD AND SCHOEBERL: ORIGIN OF POLAR VORTEX AIR E._ 500 o Ensemble Mean Parcel Descent NH Forward Trajectory SH Forward Trajectory NH Fixed Parcel SH Fixed Parcel Figure 1. Interannual variability of the ensemble mean parcel descent for the NH and SH forward calculations. Bold (thin) solid lines refer to the NH (SH) forward trajectories, and the bold (thin) dashed lines refer to the NH (SH) fixed parcel computations. 7 months from spring to fall, with parcels starting in the spring at 650 K (20-25 hpa) in the NH and 700 K (13-20 hpa) in the SH. These potential temperatures approximate the average altitudes reached by the forward calculations in springtime. The starting dates were April 1 in the NH and October 1 in the SH, and ending dates were September 1 in the NH and March 1 in the SH. For the backward calculations, parcels were initialized poleward of 45 ø latitude on the same 2.5 ø x 3.75 ø latitude-longitude grid as the forward calculations. In the analyses described below, only those parcels which initially lay within the vortex boundary were selected. The vortex boundary was defined as the maximum gradient of Ertel's potential vorticity on equivalent latitudes [Nash et al., 1996]. Plate 1 (bottom) shows the back trajectory parcel locations at various times for the NH and 1997 SH springto-fall periods. The backward calculationshow that some par- cels have originated in the lower mesosphere, reaching potential temperatures greater than 2000 K. A limitation of the model is that parcels are not allowed to rise above 2300 K ( hpa), because there is no assimilated meteorological data above this altitude. Parcels are constrained to the top level of the model once they reach the top. Another grouping of parcels originates at much lower altitudes in the low- and middle-latitude regions. This second population appears to have mixed into the polar latitudes without experiencing as much vertical transport. Table 1. Fraction of Air Originating Above 1700 K Northern Hemisphere Southern Hemisphere Year Fraction Year Fraction Normalized probability distribution functions (pdf's) of each year's starting and final potential temperatures for the forward and backward trajectories are shown in Plates 2a and 2b for the NH and the SH, respectively. (For a discussion of pdf's, see Sparling [2000].) The final positions of the forward calculations (solid green) generally show a fairly sharply peaked distribution with comparatively little interannual variability. There is a slightly higher fraction of parcels above 700 K in the NH than in the SH, as a result of the weaker NH vortex. In contrast, the back trajectory final positions show a great deal of structure and interannual variability. Focusing first on the forward calculations, Figure 1 shows the year-to-year variability of the ensemble mean parcel descent from the forward trajectory and the fixed parcel calculations. The fixed parcel computationshow a greater amount of descent in the NH than the SH, averaging 728 K in the NH versus 641 K in the SH for the 1992 through 1999 period. This is consistent with the higher temperatures in the NH, which lead to more cooling. As seen in Plate 1, the forward trajectory ensemble mean descent is less than the fixed parcel descent, with the difference being greater in the NH than in the SH due to the greater variability of heating rates and the greater amount of wave activity in the NH. The NH also shows more interannual variability, with the forward trajectory descent ranging from 490 to 595 K, compared with 519 to 568 K in the SH. On average, the forward trajectories show a comparable amount of descent in the two hemispheres, with the greater amount of mixing tending to counter the larger amount of cooling in the NH. There is no obvious trend in the computed descent, in contrast to the findings of Kawamoto and Shiotani [2000], who compared fall and spring HALOE methane profiles in the Antarctic. Looking now at the back trajectory calculations, the finalposition pdf's shown in Plates 2a and 2b (solid red) indicate that in most years there is a population of parcels which have descended from altitudes above about 1700 K. In reality, this population has probably experienced more descenthan this, since the region of fall and winter high-latitude cooling extends to the mesosphere [Rosenfield et al., 1987], above the trajectory model top at 0.4 hpa. There is little year-to-year difference in the net amount of descent experienced by this population of parcels. However, there is a large interannual variability in the net fraction of air in this population. Using a cutoff of 1700 K ( hpa), the fraction of air which has descended from the upper stratosphere and mesosphere ranges from 5 to 90% in the NH and from 7 to 86% in the SH. The fractions for each year are shown in Table 1. The remaining parcels appear to have descended from a broad range of lower altitudes. The final positions of their back trajectories exhibit a large interannual variability in width and mean altitude as a result of the varying polar meteorology. The forward trajectory calculations do not show this second population, which has mixed into the vortex during vortex formation, since the parcels in the forward calculation originate within the protovortex. The NH winter is unusual in that only 5% of the spring vortex air appears to have originated in the mesosphere. That year exhibited the latest vortex formation in 18 years of the National Centers for Environmental Prediction (NCEP) record [Coy et al., 1997]. Late November to early December zonal wind speeds, averaged around the 900-K vortex edge, were m/s, compared with an average of m/s for the other years included in our study. This allowed a great deal of lower-latitude air to be entrained within the polar vortex.

7 ROSENFIELD AND SCHOEBERL: ORIGIN OF POLAR VORTEX AIR 33,491 6O HALOE Arctic Fall Methane O O lo o Latitude Plate 3. Halogen Occultation Experiment (HALOE) Arctic fall 1995 average methane. The black dots show the final positions of the spring-to-fall back trajectories for the winter. Another unusual NH winter was that of For this year, the back trajectory pdf shown in Plate 2a indicates that a large fraction of spring vortex air came from potential temperatures of K, with only 14% descending from the mesosphere. This was the result of a major stratospheric warming occurring in mid-december 1998, which led to a weak polar vortex through most of the winter [Manney et al., 1999]. The SH in 1994 was another case where the computations showed a relatively small percentage, 7%, of springtime vortex air originating in the mesosphere. The vortex that year also formed later than usual. Zonal winds averaged around the 900-K vortex edge in late May to early June were about 55 m/s, compared with an average of roughly 65 m/s for the other years. This allowed the mixing of more equatorward loweraltitude air into the polar region during vortex formation. 3. Methane Reconstructions In order to evaluate the computed descent, spring methane amounts were computed using the trajectory calculations and compared with observations. We computed spring methane mixing ratios by mapping version 19 HALOE fall observations onto the final latitude-altitude locations of the back trajectories. Chemical loss of methane during the 7 months of the trajectory was included using loss rates from the Goddard 2-D model [Fleming et al., 1999], interpolated to the parcel latitude and altitude positions. Plate 3 shows latitude-altitude mixing ratios of 1995 NH fall HALOE methane, averaged in longitude and time between September 15 and October 15. Regions of missing data were filled in by linear interpolation. The extent of these regions varied greatly from year to year, being anywhere from a span of 10 ø latitude to one of 35 ø latitude. For NH fall 1995 the 0ø-35øN latitude region was filled in by interpolation. Overlain are black dots which represent the final latitude-altitude locations of the back trajectories, indicating where the lower stratospheric spring parcels within the vortex originated. This plate shows that the parcels originated at a wide range of altitudes, for which methane mixing ratios varied from less than 0.2 to over 1.6 ppmv. Air which descended from the upper stratosphere and lower mesosphere was characterized by low methane mixing ratios, while lower-altitude air that was entrained into the vortex from low and middle latitudes was characterized by high methane mixing ratios. The ensemble means of the reconstructed spring methane mixing ratios were then compared with the observed HALOE spring values. The HALOE spring mixing ratios were interpolated to 650 K for the NH and 700 K for the SH and then vortex averaged. The data were time averaged using a 30-day window centered at April 1 for the NH and October 1 for the SH. Because the NH vortex breaks up near April 1, the wide time window is required to produce vortex data for each year Plates 4a and 4b show the normalized pdf's of the computed

8 33,492 ROSENFIELD AND SCHOEBERL: ORIGIN OF POLAR VORTEX AIR ililllllljll lllllljllllll!lljlllllllll L,;' I I... I,, uo!10m:l uo!10..q i o., g o o Igo c o c uo!loej4 o o o uo!loej4! I... I I O. t o oo,-:. o c c; o c o o o o o o o UO!lO;J.-I uo!1oeld

9 _ ROSENFIELD AND SCHOEBERL: ORIGIN OF POLAR VORTEX AIR 33, I... I... I !... I... I... O. o oo O uo!toe.-i uop, oejd uo!ioe d uoli0e.]:l lllllllfllltl llillllllfllllllllllll!111 j It) I I...! ] o. LIo!loeJd! c; c; c; c o uo! oe d

10 33,494 ROSENFIELD AND SCHOEBERL: ORIGIN OF POLAR VORTEX AIR Arctic Spring Methane at 650 K,. i i ß Antarctic Spring Methane at 700 K i i, i,, ß Computed Mean + Observed Mean ß Year Figure 2. Interannual variability of the spring ensemble discrepancies between the computed and observed spring mixmean computed methane for the NH at 650 K and the SH at ing ratios. The total error for HALOE CH 4 in the hpa 700 K. Also shown are the HALOE spring mean values within region is estimated to be 15% [Park et al., 1996]. This correthe vortex at the same theta values. sponds to errors of roughly ppmv in observed spring methane in the K region. Another source of error is the uncertainty due to the poor coverage of spring HALOE spring methane mixing ratios for the NH and SH, respectively. data within the vortex in some years. This is particularly true of The "all-parcels" distributions represent admixtures of air de- the SH winters of and the NH winters of scending from high altitudes and air from lower altitudes which has been entrained into the vortex. Also shown are the pdf's for parcels in population 1, originating in the mesosphere and upper stratosphere, and population 2, originating in the middle through In both 1997 and 1999, there were only 3 days, and in 1998 only 1 day, when HALOE measurements existed within the SH vortex. However, there seems to be no correlation between the number of observational data points and lower stratosphere. Vertical lines indicate the computed and the degree of agreement with the computations. In the and observed means. The computed mean values for the allparcels case correspond to an average of both populations and lie between the individual population means. These pdf's can be compared with the potential temperature pdf's of the back trajectories shown in Plates 2a and 2b. For years in which there 1997 and 1998 SH, years with poor HALOE coverage, the difference between computed and observed means is not exceptional. Likewise, in the NH springs of , agreement is relatively good, while in the NH spring of 1997, a year with good HALOE coverage (22 days), agreement is poorer. are relatively large fractions of low mixing ratios, for example, in the NH winter, Plate 2 shows that population 1 predominates. For years in which the distribution of mixing 4. Descent Over Shorter Time Periods ratios is skewed to larger values, for example, the NH and the 1994 SH, Plate 2 shows that population 2 predom- In the above, we examined the characteristics of polar descent over the 7-month fall-to-spring period. We have carried inates. The means of the springtime observations correspond out some additional computations in order to study descent to admixtures of air which has experienced descent from high altitudes and air from lower altitudes which has been entrained into the polar regions during vortex formation. The degree of agreement between the computed and observed means is an indication of how well the trajectory calculations representhis mixing. over the midwinter-to-spring period in the lower stratosphere ( K). This time period and altitude region are of particular interest to those using computed descent in estimations of polar ozone loss over this time period. The calculations were done for the NH January through March of each of the years and the SH July through September of each of The interannual variability of the computed and observed the years Backward trajectories were initialized in mean CH 4 mixing ratios is summarized in Figure 2. In the Arctic, computed CH 4 mixing ratio means are within spring with the parcels at 450 K ( hpa), on the same latitude-longitude grid as above. ppmv of the observations, except for spring 1997, when the computed mean is 0.4 ppmv larger than observed. This may indicate that the amount of mixing occurring during the winter has been overestimated in the computations. Recall that vortex formation was very late for the winter. In the , , and winters, however, there is an apparent underestimate of the amount of mixing. In the SH the agreement between computed and observed means is better than 0.2 ppmv for the years 1992, 1995, and In the years 1993, 1994, and 1999, however, differences are as high as ppmv. The calculations appear to have overestimated the amount of mixing in the years 1993 and Although the trajectory model does a credible job of computing typical mean springtime methane mixing ratios in both hemispheres, the interannual variability is not well reproduced. An exception is the trend in mixing ratios between 1998 and 2000 in the Arctic, which is well reproduced by the model. In the year 1999 the very low computed mean CH 4 mixing ratio in the Antarctic seems suspect. Plate 2b shows that for this year, in addition to the altitude mode above 2000 K, there is a prominent mode at K which does not appear in any of the other years. A problem with the ozone data used in the upper levels of the UKMO analysis starting in January 1998 (R. Swinbank, private communication, 2000) occurred that year, and this apparent anomaly may be related. In 1998 our calculationshow an altitude mode at K (Plate 2b) which may also be related to this problem. However, the mean values shown in Figure 2 for this year do not appear exceptional. Errors in the HALOE methane data will contribute to the The parcel locations at various times of the integration are shown in Plate 5 for the NH in 1996 and the SH in In contrast to the 7-month back trajectory calculations discussed

11 ROSENFIELD AND SCHOEBERL: ORIGIN OF POLAR VORTEX AIR 33, Arctic 3 Month Back Trajectory 'l i i Apr I 500 _ Mar I _ Feb 1 Starting Point _ Jan I I! Latitude Antarctic 3 Month Back Trajectory i i! I "- 550 Oct I 500- ' '." --.:,,.,.,. _., '" ' '... " StaAing Point 400 Jul 1 350,,,, Latitude Sep I Aug 1 Plate 5. Parcel locations at various times for the 1996 NH and 1993 SH 3-month backward trajectories. above, in which middle-latitude air was mixed into the polar winter due to meteorological disturbances in regions during vortex formation, there has been virtually no January and mid-february. These authors also ran calculamixing into the vortex of low- and middle-latitude air in the tions for the Antarctic 1987 winter and found no intrusions 1993 SH during this 3-month time period. This is characteristic into the vortex. In another study, Waugh et al. [1994] of all the SH winters studied and is consistent with the rela- showed, using contour advection techniques, that air was tively undisturbed nature of the Antarctic vortex. For the NH ejected from the lower stratospheric Arctic vortex into midin 1996 the plate shows that there has been only a small latitudes during January and February 1989 and from Deamount of mixing into the vortex of low and middle latitude cember 1991 to March A major warming occurred in during the midwinter-to-spring period. This was also true of February 1989, and a minor warming occurred in late Janthe 1993, 1995, 1998, and 2000 NH winters. During the three uary winters of 1994, 1997, and 1999 a significant amount of We thus conclude that for the SH and for undisturbed years lower-latitude air was mixed into the vortex because of dis- in the NH, descent in the lower stratosphere during the midturbed conditions. A major warming occurred in February winter-to-spring period as determined by tracking constant 1999, and there were disturbances in February 1994 and mixing ratios, as done by Schoeberl et al. [1995], would approx- January imate descent computed from back trajectory calculations. For Of note are the contour advection calculations on the 450-K disturbed years in the NH, however, tracking constant mixing isentropic surface of Plumb et al. [1994]. These showed intru- ratios would overestimate the actual descent as determined by sions into the lower stratospheric Arctic vortex during the back trajectory calculations.

12 33,496 ROSENFIELD AND SCHOEBERL: ORIGIN OF POLAR VORTEX AIR 5. Summary and Conclusions The importance of accurately estimating polar ozone loss has motivated us to reexamine the question of polar vortex descent and the degree of isolation of polar from midlatitude air. We have computed 7-month forward and backward 3-D trajectories between fall and spring for the through NH and the 1992 through 1999 SH stratosphere. We have also carried out fixed descent computations in which the parcels were fixed at their latitude-longitude locations. The calculationshed new light on aspects of transport in the polar regions and indicate that polar vortex descent is most accurately estimated by back trajectory calculations. The forward trajectory model computed descent is always less than the unmixed descent because the trajectory parcels are moved to lower latitudes and experience lower cooling rates. The forward calculations show more descent in the NH than in the SH because of the larger cooling rates. Unlike the forward calculations, however, the backward calculations show a complex distribution. One population of parcels originates in the upper stratospheric and mesospheric low and middle latitudes in the fall and moves poleward before descending, while the remaining parcels originate lower in the stratosphere at low and middle latitudes in the fall and are mixed into the higher latitudes during vortex formation. Because of this complex parcel distribution the net amount of descent experienced over the entire winter period, as determined by the back trajectories, will generally be less than that computed by forward trajectories. Since the back trajectories indicate that much of the air can come from lower altitudes than would be implied by the forward calculations, using a comparison between prewinter and postwinter tracer profiles to estimate the amount of descent over this period would result in larger values of descent than computed by back trajectories. Of note is the fact that model results shown by Fisher et al. [1993] do not show a population of parcels which have been mixed into the vortex from middle latitudes in the middle stratosphere, although they show some mixing in below hpa ( K). Eluszkiewicz et al. [1995] also found little mixing into the vortex of midlatitude air. This may be a result of weak wave activity in these models. As a sensitivity test, we have repeated the forward and backward trajectory calculations for the NH winter using winds and temperatures from the NCEP analysis. These calculations lead to the same general conclusions we have discussed above. We find more mixing into the vortex of midlatitude air with the NCEP data. This is expected, since the NCEP data are at a lower resolution than the UKMO data. spring values from HALOE. Our results indicate that methane observed in the springtime lower stratosphere is a mixture of low mixing ratio methane coming from the upper stratosphere and mesosphere and higher mixing ratio methane coming from lower altitudes. There is fairly good agreement between calculated and observed typical mean mixing ratios. Differences are generally less than 0.2 ppmv in the NH and 0.4 ppmv in the SH. However, the interannual variability is not well represented by the model. We have also studied polar descent in the lower stratosphere over a shorter time period, from January through March for each of the years in the NH and July through September for each of the years in the SH. Trajectory calculations beginning at 450 K within the spring vortex and running back to midwinter show that little or no mixing of parcels from outside of the vortex occurs in the Antarctic, and in the Arctic in undisturbed years. Thus, except for disturbed years in the Arctic, tracking constant mixing ratios of tracers as was done by Schoeberl et al. [1995] should allow for a reasonable approximation of the amount of descent computed by back trajectories in the lower stratosphere during the midwinter-to-spring period. Acknowledgments. We thank R. Alan Plumb and an anonymous reviewer for helpful comments. Chou, M.-D., and L. Kouvaris, Calculations of transmission functions in the infrared CO2 and 03 bands, J. Geophys. Res., 96, , Coy, L., E. R. Nash, and P. A. Newman, Meteorology of the polar vortex: Spring 1997, Geophys. Res. Lett., 24, , Eluszkiewicz, J., R. A. Plumb, and N. Nakamura, Dynamics of wintertime stratospheric transport in the Geophysical Fluid Dynamics Laboratory SKYHI general circulation model, J. Geophys. Res., 100, 20,883-20,900, Fisher, M., A. O'Neill, and R. Sutton, Rapid descent of mesospheric air into the stratospheric polar vortex, Geophys. Res. Lett., 20, , Fleming, E. L., C. H. Jackman, R. S. Stolarski, and D. B. Considine, Simulation of stratospheric tracers using an improved empirically based two-dimensional model transport formulation, J. Geophys. Res., 104, 23,911-23,934, Kawamoto, N., and M. Shiotani, Interannual variability of the vertical descent rate in the Antarctic polar vortex, J. Geophys. Res., 105, 11,935-11,946, Lahoz, W. A., et al., Northern Hemisphere midstratosphere vortex processes diagnosed from H20, N20, and potential vorticity, Geophys. Res. Lett., 20, , Manney, G. L., R. W. Zurek, A. O'Neill, and R. Swinbank, On the motion of air through the stratospheric polar vortex, J. Atmos. Sci., 51, , Manney, G. L., W. A. Lahoz, R. Swinbank, A. O'Neill, P.M. Connew, and R. W. Zurek, Simulation of the December 1998 stratospheric The comparison of forward and backward trajectorie shows that the two types of computations can give a very different picture of the composition of air within the vortex. The differmajor warming, Geophys. Res. Lett., 26, , ences seen above imply that net winter descent determined by Nash, E. R., P. A. Newman, J. E. Rosenfield, and M. R. Schoeberl, An comparing fall and spring tracer profiles will give erroneous objective determination of the polar vortex using Ertel's potential descent amounts, as defined by back trajectories. The spring- vorticity, J. Geophys. Res., 101, , Nielsen, J. E., R. B. Rood, A. R. Douglass, M. C. Cerniglia, D. J. Allen, time polar lower stratospheric air is not necessarily represenand J. E. Rosenfield, Tracer evolution in winds generated by a global tative of air that was in the fall polar upper stratosphere and spectral mechanistic model, J. Geophys. Res., 99, , lower mesosphere. The present calculations indicate that the Park, J. H., et al., Validation of Halogen Occultation Experiment CH 4 percentage of springtime lower stratospheric air which has measurements from the UARS, J. Geophys. Res., 101, 10,183-10,203, originated in the fall upper stratosphere and mesosphere varies Plumb, R. A., et al., Intrusions into the lower stratospheric Arctic greatly from year to year. vortex during the winter of , J. Geophys. Res., 99, Using the back trajectories, we have reconstructed spring 1105, methane mixing ratios and compared them with observed Rosenfield, J. E., M. R. Schoeberl, and M. A. Geller, A computation

13 ROSENFIELD AND SCHOEBERL: ORIGIN OF POLAR VORTEX AIR 33,497 of the stratospheric diabatic circulation using an accurate radiative transfer model, J. Atmos. Sci., 44, , Rosenfield, J. E., P. A. Newman, and M. R. Schoeberl, Computations of diabatic descent in the stratospheric polar vortex, J. Geophys. Res., 99, 16,677-16,689, Rosenfield, J. E., D. B. Considine, P. E. Meade, J. T. Bacmeister, C. H. Jackman, and M. R. Schoeberl, Stratospheric effects of Mount Pinatubo aerosol studied with a coupled two-dimensional model, J. Geophys. Res., 102, , Russell, J. M., A. F. Tuck, L. L. Gordley, J. H. Park, S. R. Drayson, J. E. Harries, R. J. Cicerone, and P. J. Crutzen, HALOE Antarctic observations in the spring of 1991, Geophys. Res. Lett., 20, , Schoeberl, M. R., and L. C. Sparling, Trajectory modelling, in Diagnostic Tools in Atmospheric Physics: Proceedings of the International School of Physics, Enrico Fermi, Course CXVI, edited by G. Fiocco and G. Visconti, pp , North-Holland, New York, Schoeberl, M. R., L. Mingzhao, and J. E. Rosenfield, An analysis of the Antarctic Halogen Occultation Experiment trace gas observations, J. Geophys. Res., 100, , Sparling, L. C., Statistical perspectives on stratospheric transport, Rev. Geophys., 38, , Swinbank, R., and A. O'Neill, A stratosphere-troposphere data assimilation system, Mon. Weather Rev., 122, , Waugh, D. W., et al., Transport out of the lower stratospheric Arctic vortex by Rossby wave breaking, J. Geophys. Res., 99, , J. E. Rosenfield and M. R. Schoeberl, NASA Goddard Space Flight Center, Greenbelt, MD 20771, USA. (rose@euterpe.gsfc.nasa.gov) (Received January 11, 2001; revised May 23, 2001; accepted June 23, 2001.)