Descent and mixing in the northern polar vortex inferred from in situ tracer measurements

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 17, NO. D2, 8285, doi:1.129/21jd961, 22 Descent and mixing in the northern polar vortex inferred from in situ tracer measurements Eric A. Ray, 1,2,3 Fred L. Moore, 1,2 James W. Elkins, 1 Dale F. Hurst, 1,2 Pavel A. Romashkin, 1,2 Geoffrey S. Dutton, 1,2 and David W. Fahey 3 Received 19 June 21; revised 5 November 21; accepted 6 November 21; published 11 October 22. [1] In situ measurements from the Lightweight Airborne Chromatograph Experiment (LACE) and the Airborne Chromatograph for Atmospheric Trace Species (ACATS-IV) taken during the SAGE III Ozone Loss and Validation Experiment (SOLVE) are used to examine the descent and mixing in the northern stratospheric polar vortex. LACE was flown twice on an in situ balloon platform, in November 1999 just after the vortex formed and in March 2 near the end of the vortex lifetime. The aim of this paper is to use simple models of vortex transport to try to explain the changes seen in the long-lived tracers measured by LACE between the two vortex flights. It is the high precision and long-term accuracy of the observations that allow small differences in the data from a number of species obtained on different flights to be used to infer transport processes. Changes in tracer profiles as a function of height or potential temperature can be attributed, to first order, to descent in the vortex. A calculation which used six tracers shows that the total descent was a strong function of potential temperature, from 2 K in the middle stratosphere to 5 K in the lower stratosphere over the nearly 4-month period between flights. Observed changes in long-lived tracer-tracer correlations require a mixing process to have occurred since descent alone cannot change the correlations. Two simple models of mixing are used to examine the tracer correlation changes. One model simulates entrainment of midlatitude air across the vortex edge with relatively efficient mixing within the vortex. A second model simulates the effect of differential descent and subsequent efficient mixing within a vortex that is entirely isolated from the midlatitudes. It is shown that the differential descent and mixing calculation produces results which are much more consistent among all the tracer correlations compared to the results from midlatitude entrainment. The unique sensitivity of SF 6 as a tracer of mesospheric air makes it an outlier in both the descent and mixing calculations. The inclusion of a small fraction of air from the mesosphere is necessary to bring the SF 6 mixing and descent calculations into agreement with the other tracers. The conclusions regarding descent and descent rates in the vortex are consistent with other modeling studies. The results indicate that mixing of midlatitude air into the winter vortex is not a significant contributor to the observed ozone changes in the 1999/2 season. INDEX TERMS: 341 Atmospheric Composition and Structure: Middle atmosphere constituent transport and chemistry (3334); 3334 Meteorology and Atmospheric Dynamics: Middle atmosphere dynamics (341, 342); 3349 Meteorology and Atmospheric Dynamics: Polar meteorology; 34 Atmospheric Composition and Structure: Middle atmosphere composition and chemistry Citation: Ray, E. A., F. L. Moore, J. W. Elkins, D. F. Hurst, P. A. Romashkin, G. S. Dutton, and D. W. Fahey, Descent and mixing in the northern polar vortex inferred from in situ tracer measurements, J. Geophys. Res., 17(D2), 8285, doi:1.129/21jd961, Climate Monitoring and Diagnostics Laboratory, National Oceanic and Atmospheric Administration, Boulder, Colorado, USA. 2 Cooperative Institute for Research in Environmental Sciences, University of Colorado, Boulder, Colorado, USA. 3 Aeronomy Laboratory, National Oceanic and Atmospheric Administration, Boulder, Colorado, USA. Copyright 22 by the American Geophysical Union /2/21JD Introduction [2] The SAGE III Ozone Loss and Validation Experiment (SOLVE) has provided a comprehensive set of measurements within the northern vortex during the winter season in order to study ozone depletion in this region. As part of SOLVE, the Lightweight Airborne Chromatograph Experiment (LACE) was flown twice SOL 28-1

2 SOL 28-2 RAY ET AL.: DESCENT AND MIXING IN THE NORTHERN POLAR VORTEX aboard the Observations of the Middle Stratosphere (OMS) in situ balloon gondola. The first flight took place soon after the vortex formed in November and the second flight occurred near the end of the vortex lifetime in March. The long-lived trace gases measured by LACE and the timing of the balloon flights were chosen to examine the large-scale transport over the course of the vortex lifetime. The main focus of the SOLVE mission was to derive how much chemical loss of ozone took place in the vortex. Without accurate knowledge of transport processes, changes in ozone due to transport could be attributed incorrectly to in situ photochemical loss processes. The LACE measurements from SOLVE have been used in several other studies in which chemical ozone loss was the main focus [e.g., Richard et al., 21; Rex et al., 22; Salawitch et al., 22]. This paper focuses on the transport in the vortex which can be inferred from the LACE measurements. [3] Many previous studies have made use of trace gas measurements to infer transport characteristics of the polar vortex such as descent estimates [e.g., Traub et al., 1995; Abrams et al., 1996]. Other studies of this type have used correlations of long-lived tracers with a shorter lived species such as ozone. This correlation has been used in attempt to remove variations in shorter lived species due to transport in order to derive either chemical ozone loss [e.g., Proffitt et al., 199; Müller et al., 1997] or denitrification [e.g., Sugita et al., 1998; Popp et al., 21]. These studies have primarily made use of one or two long-lived trace gases, such as N 2 O, which are inert in the polar winter stratosphere to account for dynamical changes, such as descent. Among the few studies to examine mixing and descent in the vortex from measurements of several long-lived tracers are those based on ATMOS data [e.g., Michelson et al., 1998, 1999]. These studies noted the different correlation curves between long-lived tracers in the early and late vortex compared to the midlatitude correlations and qualitatively described mixing and descent in the vortex, which could cause the observed differences. [4] Our knowledge of the large-scale transport in the vortex comes primarily from global general circulation model simulations, analyzed meteorological data and satellite observations [e.g., Fisher et al., 1993; Manney et al., 1994]. The vortex formation takes place in the fall as sunlight recedes from the polar region and large diabatic cooling occurs. This cooling results in rapid diabatic descent in the polar region which draws air from a large range of altitudes, including the entire depth of the mesosphere, down into the polar stratosphere [Fisher et al., 1993; Plumb et al., 22]. The most rapid descent in the high latitudes occurs at the highest altitudes in the stratosphere in the Northern Hemisphere fall and early winter [Rosenfield et al., 1994]. The descent slows in the middle and lower stratosphere and as the vortex forms the high latitudes become more isolated from the surrounding midlatitudes. This rapid descent of air and isolation of the vortex establishes a distinct air mass which evolves over the course of the winter. Year to year variability in the amount of planetary wave activity during the winter can lead to differences in certain vortex characteristics [Dahlberg and Bowman, 1994; Waugh and Randel, 1999] such as how permeable the edge of the vortex is to midlatitude air. The variability and relative permeability of the vortex will have strong effects on the composition of the vortex for a species, such as ozone, that is sensitive to how much midlatitude air is entrained or the temperature of the vortex. Thus, ozone is observed to have significant interannual variability in the northern vortex [Rex et al., 22]. But a number of long-lived trace species are insensitive to temperature and some, such as SF 6, are also only moderately sensitive to entrainment of midlatitude air. The characteristics of these tracers in the vortex will be most sensitive to the source of the air brought into the vortex at the time of its forming and to how much mixing occurs in the vortex. These long-lived trace species should have relatively consistent features in the vortex from year-to-year. An example of this consistency can be seen by comparing SF 6 measurements in the northern vortex from several different years. Measurements of SF 6 from the 1992 polar vortex shown by Harnisch et al. [1999] are very similar to the polar vortex and June 1997 vortex remnant measurements from LACE [Moore et al., 22] when adjusted for SF 6 growth. [5] The suite of in situ long-lived trace gases measured by LACE in the northern vortex allows a quantitative look at large scale transport over the lifetime of the vortex. We will examine descent, mixing of air into the vortex from midlatitudes, and mixing within the vortex, as well as the feedback among these processes. The emphasis of this paper is to describe transport processes in the vortex that are self-consistent with the changes in both tracer vertical profiles and tracer-tracer correlations observed in the LACE measurements over the course of the winter. In doing so we attempt to establish a consistent picture for the long-lived tracers measured by LACE and Airborne Chromatograph for Atmospheric Trace Species (ACATS-IV). These tracers have a variety of atmospheric lifetimes and different distributions at the latitudes and altitudes from which air in the vortex originated. Thus, the measurements place a powerful constraint on describing the properties of large-scale vortex transport. 2. Measurements [6] Most of the measurements used in this study are produced by the LACE instrument. LACE is a lightweight, three-channel gas chromatograph which is designed to operate in ambient pressure on a balloon gondola [Moore et al., 22]. LACE has been part of the Observations of the Middle Stratosphere (OMS) program since its inception in During the SOLVE mission LACE took measurements of CFC-11, CFC-12, halon-1211, SF 6 and N 2 O every 7 s and CH 4, CO and H 2 every 14 s. The characteristics of the trace species used in the calculations in this paper are shown in Table 1. The temporal resolutions of the measurements correspond to vertical resolutions of roughly 3 and 6 meters at the typical balloon ascent and descent rates of 3 5 m/s. The width of the temporal distribution of air sampled at injection is roughly 1 2 s so the measurements represent a spatial average over 3 1 m vertically. LACE took measurements up to altitudes of roughly 32 km with a range of precisions from 1 to 4%. The two LACE flights during SOLVE took place on 19 November 1999 and 5 March 2 from the Swedish Space Corporation s balloon facility at Esrange, Sweden

3 RAY ET AL.: DESCENT AND MIXING IN THE NORTHERN POLAR VORTEX SOL 28-3 Table 1. Characteristics of the Trace Gases Measured by LACE Which Are Used in the Descent and Mixing Calculations a Trace Gas Chemical Formula Stratospheric Precision, Precision Lifetime, years b % c (Mixing Ratio) Nitrous Oxide N 2 O ppb Methane CH ppb CFC-12 CCl 2 F ppt CFC-11 CCl 3 F ppt halon-1211 CBrClF ppt Sulfur Hexafluoride SF d ppt a Precisions of 1 2% are required to resolve the changes between the early and late vortex tracer-tracer correlations used in the calculations. Local lifetimes in the dark vortex are much longer than the stratospheric lifetimes quoted below. b Stratospheric lifetime ranges for all molecules except SF 6 taken from Volk et al. [1997]. c Precision is given as a percent of the CMDL surface average mixing ratio. d SF 6 stratospheric lifetime range taken from Moore et al. [22]. (68 N). Goddard Space Flight Center DAO analyzed temperature and potential vorticity data indicated that both flights were well inside the coldest region as well as poleward of the region of strongest potential vorticity gradient which suggests the flights were in the inner vortex at all levels in the stratosphere. [7] Additional measurements come from the ACATS-IV instrument, which is a four channel gas chromatograph designed to fly in the pressurized Q bay of the ER-2 aircraft [Elkins et al., 1996; Romashkin et al., 21]. ACATS-IV measures the same species as LACE but provides a horizontal sampling of the vortex at ER-2 cruise altitudes of 16 to 2 km in contrast to the vertical sampling by LACE. The two ER-2 aircraft deployments during SOLVE took place in late January-early February and in late-february to mid- March, with each deployment consisting of six flights based out of Kiruna, Sweden (68 N). The two instruments use independent calibration gases that are referenced to common standards produced at CMDL, which allows reliable comparisons to be made between their measurements. 3. Descent in the Vortex [8] The trace gases measured by LACE have very long photochemical lifetimes inside the polar vortex, so the changes in trace gas profiles between the two SOLVE OMS balloon flights must be interpreted as due to diabatic descent and horizontal mixing. In this section a first order estimate of the integrated descent in the vortex between the times of the two LACE flights in November and March is obtained assuming mixing is negligible. The method uses vertical profiles separated in time and the simple assumption that tracer mixing ratios within each air parcel are constant. Descent is calculated as the change in altitude or potential temperature of tracer mixing ratios between the two LACE flights. Calculations of descent both in terms of altitude and potential temperature were performed but only the potential temperature results are shown. It is necessary to start with a calculation of the descent since the mixing calculations require an estimation of the descent in the vortex. The mixing calculation will have an effect on the tracer profiles so the descent calculation will be adjusted following the mixing calculation as described in section 5.1. These calculations will be performed on six independent tracers and we look for a common result among all of the tracers. [9] Profiles of N 2 O as a function of potential temperature from the two OMS balloon flights are shown in Figure 1. It is clear that the sampled air has descended by a considerable amount between the 19 November 1999 and 5 March 2 flights. The lines through the data points in each profile are smooth fits which were done for each species used to calculate the total descent between flights. The fits for each species were divided into equal mixing ratio intervals and the difference in potential temperature between the two flights was calculated for each mixing ratio interval. The total descent calculated for six trace gases measured by LACE along with a root mean square weighted mean descent profile are shown in Figure 2. The potential temperature in this plot is from the first flight on 19 November so that the descent starts at this value and the total descent is the amount shown on the x axis. The error bars for each species are calculated based on the variance of the data about the fits to the profiles in quadrature with the uncertainty in the data. The calculation is performed for each species from 4 K up to the altitude at which the mixing ratio is near zero. Thus, at the lower levels all the species can be used to obtain the descent values while above 65 K only SF 6 and CH 4 can be used in the calculation. [1] Several interesting features of the initial calculated descents shown in Figure 2 are worth noting. The first feature is the good agreement of the descent inferred from measurements between the different species excluding SF 6. At nearly all levels the error bars for each species overlap Potential Temperature LACE SOLVE N 2 O Mixed N 2 O (ppb) Figure 1. Profiles of N 2 O as a function of potential temperature from LACE for the two OMS balloon flights during the SOLVE mission. Dates in the figure legend are in the format yyyymmdd. Solid lines through the data are fits used in the calculation of descent in the vortex between the flights. Also included is a representation of a well-mixed early vortex profile (dashed line) from the differential descent mixing calculation. This profile is used to redo the descent calculation Approximate Altitude (km)

4 SOL 28-4 RAY ET AL.: DESCENT AND MIXING IN THE NORTHERN POLAR VORTEX Potential Temperature on 11/19 Potential Temperature on 11/ Total Descent (K) From 11/19-3/5 a) No Mixing 5 1 Descent (K) 15 Descent (K) from 11/19-3/5 2 SF 6 CH 4 N 2 O CFC12 CFC11 h1211 Weighted Mean b) With Mixing 25 Figure 2. Vertical profiles of total descent in potential temperature between 19 November 1999 and 5 March 2 in the polar vortex. Descent estimates are shown for six different species measured by LACE as well as a weighted mean of all the species except SF 6. The potential temperature in this plot is from the first flight on 19 November so that the descent starts at this value and the total descent is the amount shown on the x axis. The descent values in (a) do not include the effects of mixing (section 3) while the descent values in (b) include the effects of mixing discussed in sections 4 and 5. See color version of this figure at back of this issue. and the total spread in descent is less than 2 K. This is a remarkably good agreement and suggests that the assumption of descent as the cause of the changes in tracer profiles in the vortex is correct to first order. A second feature to note is that at most levels the descent calculated from SF 6 is significantly larger than that calculated from the other species. If we assume that the descent calculated from all the other species is correct then some process which has not been accounted for in this calculation is responsible for lowering SF 6 mixing ratios relatively more between flights than for any of the other species measured. The long-lived trace gas SF 6 is an outlier among the tracers used in the calculations in this section. This common theme will be described further in the sections on the mixing calculations. A third feature of note is the structure of the descent as a function of altitude. The descent increases as a function of theta by more than a factor of 4 from 4 K to 75 K. This increase in descent with increasing height is consistent with mass conservation and the decrease of density with height. There is also some structure in the descent which is seen in most of the species. A small dip in the total descent is seen at roughly 48 K with a gradual increase above this up to 55 K and a slightly larger increase with theta from 55 K to 7 K. Above 7 K the descent is nearly constant which is consistent with the shape of the tracer profiles near 6 K on the March flight. It can be seen in Figure 1 that there is a small increase in N 2 O above 53 K on the March flight. This feature is thought to be due to a small amount of midlatitude or vortex edge air which was mixed into the inner vortex. Tracer-tracer correlations (not shown) reveal that the measurements in this region deviate a small amount from the smooth March vortex correlations which is evidence that a mixing event of some kind caused these features. Since the air which was at K in November descended roughly 2 K by March this corresponds to the region of the March flight influenced by a small mixing event. Thus, the calculation is likely biased toward low descent values above 72 K. [11] In the next section mixing and its effect on the trace gas profiles is discussed followed by a return to the descent calculation in section 5.1. Since the modified descent calculation is thought to be closer to the actual descent than that calculated in this section, comparison to other descent estimates are deferred to section Mixing Inferred From Long-Lived Tracer Tracer Correlations 4.1. Background [12] Tracer-tracer correlations of long-lived species in the stratosphere exhibit compact relationships due to relatively rapid horizontal transport compared to the timescales for vertical transport and photochemistry as described by the global diffuser model [Plumb and Ko, 1992]. The global diffuser model is valid so long as barriers to horizontal transport in the stratosphere do not exist. The model most closely approximates the atmosphere in the midlatitude winter surf zone where air is well mixed along isentropic surfaces. Three regions of the stratosphere, the midlatitude surf zone, the tropical pipe [Plumb, 1996] and the polar vortices [Plumb et al., 2], all establish distinctly different tracer-tracer correlations due to their isolation from each other. Figure 3 shows measurements of CFC-11 versus CFC-12 from several flights of LACE at different latitudes. The tropical data are from 7 S, the midlatitude curve represents an average of several midlatitude flights and the vortex data are from the March 2 LACE SOLVE flight as well as several ACATS flights from February and March 2. The 3 June 1997 flight data from LACE in Fairbanks, Alaska are also plotted because they show the result of mixing between two isolated regions. This flight went through several remnants of the polar vortex and the mixing between the vortex air and midlatitude air is evident from the data which fall on

5 RAY ET AL.: DESCENT AND MIXING IN THE NORTHERN POLAR VORTEX SOL 28-5 CFC-11 (ppt) LACE and ACATS CFC-11 vs CFC-12 1 Tropics (7S) Vortex - LACE Vortex - ACATS Summer High Lat Midlatitude Avg Mixing Lines 2 3 CFC-12 (ppt) 4 5 Figure 3. LACE and ACATS-IV CFC-11 versus CFC-12 from tropical, midlatitude, and vortex flight locations. The 3 June 1997 OMS balloon flight in Fairbanks, Alaska demonstrates how the correlation curves can reveal mixing between vortex and midlatitude air masses. Dashed lines are shown between points in the summer data set to illustrate mixing. See color version of this figure at back of this issue. lines between the midlatitude curve and the vortex curve. Further evidence from tracer correlations for mixing between the different regions of the stratosphere has been seen in ATMOS data from early November 1994 [Manney et al., 2]. Mixing lines between tropical, midlatitude and the protovortex CH 4 N 2 O correlations are seen at this time in the polar stratosphere. [13] MClearly the midlatitudes and vortex have distinctly different tracer-tracer correlations but it is not well known how and when in the course of the vortex lifetime this separation in the correlations evolves. As the vortex forms, air from mid and high latitudes are not separated by any barriers to mixing so the air which enters into and becomes the vortex is assumed to start with midlatitude correlations. We know that the polar vortex is a region of relatively rapid net descent compared to the surrounding midlatitudes due to the isolation of the vortex from regions of slow descent in lower latitudes. The rapid descent itself will not change tracer-tracer correlation curves, but any horizontal mixing of air between regions of the atmosphere which have different vertical profiles of tracers, such as the midlatitudes and polar vortex, can potentially change the shape of correlations. We refer to air masses which are on the same horizontal level but descended from different altitudes as differentially descended air parcels. Horizontal mixing between differentially descended air parcels occurs across a distance in tracer-tracer space which is dependent on the amount of differential descent for tracers with vertical gradients. Plumb et al. [2] refer to this mixing across large distances in tracer-tracer space as anomalous mixing. Anomalous mixing between two points on a linear portion of a correlation curve will result only in a weighted average point on the existing curve. Therefore, even with the occurrence of anomalous mixing, a tracer-tracer correlation curve will not be changed unless the mixing occurs across a region of curvature in the correlation (see mixing lines in Figure 3). This points out the importance of measuring trace gases with a range of photochemical lifetimes because correlations between tracers with different lifetimes are curved at different altitudes and thereby provide a more complete picture of the mixing than the use of a single pair of long-lived tracers. [14] Correlation curves of the long-lived trace gases measured by LACE from the two SOLVE OMS balloon flights, one in the early vortex and one in the late vortex, show small but statistically significant differences between the flights and both vortex curves are substantially shifted from that of a midlatitude profile. In this paper we focus on the changes in the observed correlations between the early and late vortex flights of LACE and do not attempt to explain the changes between the midlatitude and early vortex correlations. For the former, we presume that these are a remnant of the processes which induced the much larger changes from midlatitude to early vortex curves. Figure 4 shows many of the tracer-tracer correlations from the two LACE flights. All of the correlations from the 5 March flight are shifted toward the concave side of the correlations from the 19 November flight, with the exception of the SF 6 correlations. Correlations with SF 6 are mostly shifted to the convex side of the early vortex curve which again makes SF 6 an outlier as seen previously in the descent calculations. Error bars representing the accuracy of each measurement point are not shown because they are not much larger than the size of the symbols in the plots. [15] A change in the correlation curves as observed requires some type of anomalous mixing to have occurred between November and March in the polar vortex. Two types of such mixing will be explored in this paper. The first type of mixing results from the entrainment of midlatitude air across the vortex edge into the inner vortex. This process entails mixing between two distinctly different correlation curves and is constrained to occur on potential temperature surfaces. The second type of mixing takes place entirely within the vortex and assumes the vortex is completely isolated from the midlatitudes. We also assume that differential descent of air in the vortex occurred during the vortex formation and/or occurred in the vortex after the November flight. This second type of mixing takes place only between points on the vortex correlation curve which evolves over the lifetime of the vortex and is also constrained to occur on constant height/potential temperature surfaces. These are two extreme cases and the analysis will try to rule out one or the other mixing mechanism by requiring self-consistency among all the tracer-tracer pairs measured by LACE. The range of tracers and the fact that there are no local loss or production mechanisms for these tracers provide strong constraints on any single mixing mechanism that is to explain all of the observed changes Mixing of Midlatitude Air Into the Polar Vortex [16] The edge of the polar vortex acts as a barrier to most horizontal transport of air from the midlatitudes [McIntyre and Palmer, 1984]. Thus, the vortex is mostly isolated from the surrounding midlatitudes and evolves over the

6 SOL 28-6 RAY ET AL.: DESCENT AND MIXING IN THE NORTHERN POLAR VORTEX LACE SOLVE Tracer-Tracer Correlations CFC-11 (ppt) CFC-11 (ppt) N 2 O (ppb) CFC-12 (ppt) halon-1211 (ppt) halon-1211 (ppt) N 2 O (ppb) CH 4 (ppb) CFC-11 (ppt) CH 4 (ppb) N 2 O (ppb) CH 4 (ppb) 15 N 2 O (ppb) CFC-11 (ppt) SF 6 (ppt) SF 6 (ppt) 4.5 Figure 4. Tracer-tracer correlation curves from the two LACE SOLVE flights. Dates in the figure legend are in the format yyyymmdd. Data from the second flight are significantly shifted from the first flight toward the concave side of the curve suggesting mixing of some type occurred. See color version of this figure at back of this issue. course of the winter in an environment of stronger net descent compared to the midlatitudes. Figure 5a shows profiles of LACE N 2 O as a function of potential temperature from the two SOLVE balloon flights as well as from previous midlatitude flights. Also included are ACATS-IV N 2 O measurements from two transit flights to Kiruna during SOLVE (9 January and 16 March 2) which primarily consist of mid and high latitude data

7 RAY ET AL.: DESCENT AND MIXING IN THE NORTHERN POLAR VORTEX SOL 28-7 Midlatitude and Vortex N 2 O Potential Temperature (K) ACATS Jan 2 ACATS Mar 2 CFC-11 (ppt) Vortex Vortex 235 Midlat ACATS Midlat 2 Midlat Fi t N 2 O (ppb) N 2 O (ppb) Figure 5. (a) Vertical profiles of N 2 O in the vortex from the two SOLVE LACE flights as well as in the midlatitudes from several previous flights. Also included are ACATS-IV data from the 2 and 23 January and 12 March vortex flights and two midlatitude flights during SOLVE. (b) CFC-11-N 2 O correlation curves from the two SOLVE flights as well as a previous midlatitude flight from LACE and the ACATS- IV midlatitude data from SOLVE. See color version of this figure at back of this issue. outside the vortex. Because N 2 O is destroyed globally in the mid to upper stratosphere, these profiles clearly show the large difference in descent between vortex and midlatitude air. [17] Most large-scale mixing in the stratosphere occurs on nearly horizontal surfaces of constant potential temperature (isentropes) due to breaking of planetary waves. These breaking waves effectively mix air along isentropes, as occurs in the winter midlatitude surf zone. Isentropic mixing is known to occur periodically between midlatitude and polar vortex air, due to planetary waves which break near the edge of the vortex for instance [Polvani and Plumb, 1992]. The mixing of midlatitude air with vortex air will occur across a large distance in tracer-tracer space due to the large difference in tracer values on an isentrope that crosses the vortex edge. This anomalous mixing has been shown by Plumb et al. [2] to have a large effect on tracer-tracer correlations. Plumb et al. [2] demonstrated that an entirely new correlation curve is produced in the vortex due to this type of mixing and that without anomalous mixing of some type the midlatitude and vortex correlation curves would be the same. The mixing of midlatitude air into the vortex is illustrated schematically in terms of both an altitude versus latitude cross section (Figure 6a) as well as a tracer-tracer correlation (Figure 6b). The schematic in Figure 6a shows the isopleths (lines of constant mixing ratio) in the vortex as lower relative to those in midlatitudes at the same altitude. Horizontal mixing across the vortex edge creates air parcels which populate the line between points A and B in Figure 6b where A lies on the midlatitude correlation and B on the vortex correlation. The resulting mixed air will lie somewhere along the line between A and B depending on the relative weighting of midlatitude versus vortex air in the mixture. [18] We attempt to simulate this type of mixing by performing a simple time evolution of mixing of midlatitude air into the vortex between 19 November and 5 March as air parcels descend in the vortex. The mixing is constrained to occur isentropically and we assume a constant rate of entrainment at a given potential temperature through the entire period. The calculation is also constrained to mix parcels which are either on the midlatitude or evolving vortex correlation curves. This temporal evolution of mixing is performed for each tracer-tracer pair of long-lived species measured by LACE. Smooth fits through the vortex (V) and midlatitude (M) profiles as a function of theta (q) for tracer 1 of the tracer-tracer pair are performed to give mixing ratios c V1 (q) and c M1 (q), respectively. The midlatitude fit, c M1 (q), is assumed to be constant throughout the period of the calculation. The tracer-theta fits are shown as the lines through the data in Figure 5a for the example of N 2 O (tracer 1 ). Fits are also made through the vortex and midlatitude tracer-tracer correlation curves to give c V2 (c V1 ) and c M2 (c M1 ) as shown for N 2 O versus CFC-11 (tracer 2 ) in Figure 5b. [19] By comparing the fits in Figures 5a and 5b it is clear that the midlatitude tracer-tracer correlation is much more compact than the midlatitude tracer-theta correlation. It is for this reason that we cannot perform the calculation for each tracer individually in tracer-theta space. This is an important

8 SOL 28-8 RAY ET AL.: DESCENT AND MIXING IN THE NORTHERN POLAR VORTEX Mixing of Midlatitude Air Into the Vortex (a) Vortex Altitude χ 1 χ 1 (b) A B χ 1 A B Midlat Pole Latitude χ 2 Mixing Within the Vortex (c) Vortex (d) χ 1 A Altitude A B χ 1 B Midlat Pole Midlat Latitude χ 2 Figure 6. Schematic representations of the mixing of midlatitude air into the vortex (a,b) and the mixing of differentially descended air within the vortex (c,d). For the midlatitude mixing case, tracer isopleths in the vortex (horizontal lines in (a)) are lower compared to those in midlatitudes due to the larger mean descent in the vortex. The near vertical solid lines in (a) and (c) define the boundaries of the vortex. Horizontal mixing across the vortex edge depicted by the line between air parcels A and B, will mix air between the distinct midlatitude and vortex tracer-tracer correlation curves as shown in (b). The resulting mixed air will lie somewhere on the line between A and B in tracer-tracer space depending on the relative amount of midlatitude and vortex air in the mixture. For the case of mixing within the vortex, tracer isopleths are assumed to be horizontally inhomogeneous (curving solid lines in (c)) due to differential descent in the vortex. The line between air parcels A and B represents horizontal mixing which will act to smooth the gradients. In tracer-tracer space (d) the mixing will occur between points A and B on the same evolving vortex correlation curve. point because we want to insure that the calculation simulates mixing between the observed compact midlatitude and vortex correlation curves. The large spread in midlatitude tracer-theta relationships makes it highly unlikely that any chosen tracer-theta relationships for two tracers would fall on the observed midlatitude tracer-tracer correlation. The result of a mixing calculation simply using tracer-theta relations for each tracer would be unrealistic since it would essentially be the equivalent of each tracer being mixed into the vortex from a different physical location in the midlatitudes. The spread in tracer-theta relationships in midlatitudes is due to the meridional gradient in tracers on theta surfaces. By relating the midlatitude tracer-theta fits of each tracer through the midlatitude tracer-tracer correlation we constrain the tracers to be mixed from the same physical location in midlatitudes into the vortex. The spread in midlatitude tracer-theta relations are taken into account in the error estimates described below. [2] Vortex profiles of both tracers, c Mix1 and c Mix2,are evolved in time by mixing a fraction of midlatitude air, determined by f(q), into the vortex each time step. The mixing fraction f(q) is assumed to be constant in time but variable in theta. The general time dependence of the tracer mixing ratios inside the vortex is given by, c Mix1 ðq; t þ t Þ ¼ c Mix1 ðq; tþþf ðþ q ðc M1 ðq; tþ c Mix1 ðq; tþþ ð1þ The specific method we use is to start with the initial measured vortex tracer profiles, c V1 (q(t i )) and c V2 (q(t i )). These profiles are descended based on the calculated descent rates in section 3 and mixed with midlatitude air

9 RAY ET AL.: DESCENT AND MIXING IN THE NORTHERN POLAR VORTEX SOL 28-9 Accumulated Fraction of Midlatitude Air in the Vortex 7 Potential Temperature (K) 6 5 N 2 O-CFC11 N 2 O-h1211 CH 4 -CFC12 CH 4 -CFC11 CH 4 -h1211 CFC12-CFC11 SF 6 -N 2 O Midlatitude Fraction of Air in Vorte x Figure 7. Vertical profiles of the total accumulated fraction of midlatitude air in the vortex calculated as described in the text from several tracer-tracer pairs. See color version of this figure at back of this issue. as they descend. The equation for the final vortex profile of tracer 1 is thus given by, c Mix1 q t f ¼ X tf t¼t iþt f ðqðþ t Þðc M1 ðqðþ t Þ c Mix1 ðqðþ t ÞÞþc V1 ðqðt i ÞÞ ð2þ where t i and t f are the initial and final times, 19 November and 5 March in this calculation and t is the time step which is 5 days for the results shown here. The initial vortex mixed profile is equal to the fit of the 19 November profile, c MIx1 (t i )=c V1. The potential temperature as a function of time is given by, q t f ¼ X tf t¼t iþt dq ðqðt tþþ t dt þ qðt i Þ where dq/dt(q) is a descent rate which is estimated from the total descent calculated in section 3 and is assumed to be constant in time. A description of how the descent rate is calculated is given in section 5.1. The equation for the second tracer s mixed vortex profile is obtained through the correlation between the two tracers. Thus, c Mix2 q t f ¼ X tf t¼t iþt f ðqðþ t Þ ðc M2 ðc M1 ðqðþ t ÞÞ c Mix2 ðqðþ t ÞÞþc V 2 ðc V1 ðqðt i ÞÞÞ ð4þ where c Mix2 (t i )=c V2 (c V1 ). The calculation for the second tracer is also performed on constant potential temperature ð3þ surfaces and is linked to the first tracer by the fit to the midlatitude correlation curve. This guarantees that the mixing occurs between tight tracer-tracer correlations in the evolving mixed profiles. [21] The final mixed vortex correlation given by c Mix1 and c Mix2 is compared to the measured 5 March correlation curve. An iteration is performed to obtain the best fit to the 5 March correlation curve by adjusting f(q) to minimize the distance in tracer-tracer space between the mixed and measured correlation curves. The iteration is performed a maximum of 8 times but typically convergence to the final solution occurs in 2 3 iterations. The accumulated fraction of midlatitude air mixed into the vortex, F, is calculated from the fraction mixed in at each time step, f, by F q t f ¼ X tf f ðþ q ð1 Fðqðt tþþþ: ð5þ t¼t iþt Profiles of F as calculated from several tracer-tracer pairs are shown in Figure 7. Error bars on the calculated F take into account the precision of the measurements as well as the uncertainty in the midlatitude profiles due to atmospheric variability. For the error due to measurement precision, the precision of each tracer was projected onto the mixing line between midlatitude and vortex correlation curves. The projected precisions for the endpoints of the mixing lines are added in quadrature to obtain an error estimate on the ability to fit the final measured correlation curve. Uncertainty in the midlatitude profile due to atmospheric variability is esti-

10 SOL 28-1 RAY ET AL.: DESCENT AND MIXING IN THE NORTHERN POLAR VORTEX mated by using two fits to the midlatitude measurements, as shown in Figure 5a, which roughly span the measurements. The calculation is performed for both of these fits with the average value used for the final values of F. The spread in the calculated fractions using the two midlatitude profiles is added in quadrature with the errors estimated from the measurement uncertainties to obtain the total error estimates shown in Figure 7. [22] The calculated midlatitude fractions F have several features of note. First, is that the values of F including error bars do not overlap for each tracer-tracer pair at each theta level. Thus, above 55 K and below 45 K the mixing of midlatitude air into the vortex cannot simultaneously account for all the changes in the measured tracer-tracer correlations. A second feature is that the calculated fractions decrease with increasing height up to roughly 5 K where most of the calculated fractions are zero. At 4 K, which is roughly the bottom of the vortex, 15 25% of the air in the vortex could have been entrained from midlatitudes over the course of the winter based on the spread in the calculated fractions. Above 55 K the fractions calculated from SF 6 N 2 O are much larger than those calculated from the CH 4 CFC 12 correlation. Below 55 K the SF 6 N 2 O fractions are zero and in fact at these levels it is not possible to move the SF 6 N 2 O correlation curve toward the late vortex curve with any amount of midlatitude mixing into the vortex. [23] Another indication of how well the calculation is able to account for the tracer-tracer correlation changes is the goodness of fit of the mixed correlation with the measured late vortex correlation. The goodness of fit is estimated by the difference between the calculated and measured late vortex correlation normalized by the measurement error (Figure 8). All non-sf 6 tracer pairs produce good fits (values less than 1) to the final vortex correlation curve but all SF 6 tracer pairs have a significant potential temperature range over which a fit cannot be made. So SF 6 is an outlier in this calculation in that even when a fit to the late vortex correlation curves can be made the calculated fractions are much different than for other tracer-tracer pairs at similar theta levels. [24] These results are for one extreme case where it was assumed that all of the mixing which caused the changes in the tracer-tracer correlation curves from the early to late vortex was due to mixing with midlatitude air. Thus, the calculated fractions are meant to represent an upper bound on midlatitude air in the vortex. The calculation shows that this type of mixing can simultaneously only account for the changes among all the tracers over a limited range of potential temperatures near 5 K where the calculation suggests near zero mixing. This implies that there must have been another mixing mechanism which occurred at most theta levels, and perhaps at all levels, to account for all of the tracer-tracer correlation changes. The next section will examine another mixing mechanism as an alternative to mixing of midlatitude air into the vortex Mixing of Differentially Descended Air Within the Vortex [25] The formation of the polar vortex is known to be a time and place of rapid diabatic descent [e.g., Rosenfield et al., 1994]. Thus, during vortex formation air descends from a large range of altitudes in a fairly short time. In fact, it has been shown in modeling experiments that the entire mass of Potential Temperature (K) Goodness of Fit to Late Vortex Correlation Curves 1 2 the mesosphere descends into the polar vortex each winter [Fischer et al., 1993; Plumb et al., 22]. It is likely that this rapid descent of air inside the vortex is not homogeneous with respect to vortex equivalent-latitude coordinates. That is, air in different regions of the forming vortex will have descended somewhat more or less than air in surrounding regions at the same altitude. Several modeling and observational studies have found spatial variability in the descent of air in the vortex [e.g., Manney et al., 1999; Pierce et al., 22]. For the computations performed here we do not assume any particular distribution of differential descent within the vortex, only that air on similar horizontal levels within the vortex has descended different amounts. [26] The effect of differential descent on tracer isopleths in the vortex is illustrated schematically in Figure 6c. We assume that some amount of local horizontal mixing occurs as air differentially descends during vortex formation. This mixing will act to reduce horizontal gradients in tracers and thereby evolve tracer-tracer correlations from their original midlatitude shape toward that of the late inner vortex. However, the timescale for mixing is likely not fast enough, due to the lack of any substantial horizontal mixing mechanism such as wave breaking, to mix out all of the differential descent in the forming of the vortex. As the differential descent slows in the early vortex, the horizontal mixing becomes more effective at smoothing out the tracer gradients caused by differential descent. Evidence for the relative smoothness of tracer profiles in the late vortex compared to the early vortex can be seen in Figure 5 for both LACE and ACATS-IV N 2 O measurements. As shown in Figure 6d, the horizontal mixing of differentially descended air entirely within the vortex can occur across a large range in tracer-tracer space. Therefore, this is a form of anomalous mixing which can result in the change of tracer-tracer correlation curves. 3 4 Normalized Goodness of Fit N 2 O-CFC11 N 2 O-h1211 CH 4 -CFC12 CH 4 -CFC11 CH 4 -h1211 CFC12-CFC11 SF 6 -N 2 O SF 6 -CFC11 SF 6 -h1211 Figure 8. Vertical profiles of the normalized goodness of fit to the late vortex correlation curves. The deviations of the calculated correlations from the measured correlations were normalized by the measurement uncertainty. Values much greater than unity indicate a poor fit to the late vortex curve. 5 6

11 RAY ET AL.: DESCENT AND MIXING IN THE NORTHERN POLAR VORTEX SOL Distribution of Differential Descent Used in Mixing Calculation Vortex Area Mesospheric Air Z dz Figure 9. Schematic of the distribution of differential descent used for one altitude in the within vortex mixing calculation. The flight profile is assumed to have sampled air of average descent in the first vortex flight which is represented as dz = in the schematic. A linear difference in descent is then assumed between the areas of the vortex with the most and least descent of air. In the mixing calculation a solution is found for the quantity Z to obtain the best fit to the late vortex correlation curves. The dashed line distributions are examples of how the distribution can be adjusted to obtain the best fits. The calculation in section 4.4 includes a fraction of mesospheric air which is shown in the schematic as a small area of the vortex which has a large value of dz. [27] In this section we perform a simple calculation to determine if the type of mixing described above can account for the changes in the observed tracer-tracer correlations. As in section 4.2, we make the assumption that this type of mixing is the only mechanism which can change the correlation curves. We also assume a basic linear variation in the differential descent as a function of vortex area. A schematic of the possible shapes of the distribution of differential descent used in this mixing calculation are shown in Figure 9 in which the y axis represents vortex area and the x axis represents descent relative to the air sampled by the balloon. Positive (negative) values of dz represent areas of the vortex which descended more (less) than that sampled by the balloon. The total area under the distribution is zero since for simplicity we make the assumption that the first balloon flight sampled vortex air of average descent. We therefore assume equal areas of the vortex have descended both more and less than the sampled vortex air. The quantity Z is the difference in descent between the sampled air and the most or least descended areas of the vortex. Since we assume the variation in dz is linear between more or less descended areas of the vortex, the value of Z determines the shape of the distribution at each altitude. We allow Z to vary as a function of altitude and make an initial guess of Z increasing linearly with altitude, similar to the shape of the vertical profiles of total descent shown in section 3. [28] As in the calculation for mixing of midlatitude air into the vortex, the mixing within an isolated vortex calculation is performed on each tracer-tracer pair separately. One difference between this and the midlatitude mixing calculation is that there is no time dependence in this calculation. This is due to the mixing occurring entirely within the vortex and the fact that we have no information on when the differential descent took place. Consistency of the results to within error bars across all tracer-tracer pairs would signify that the mixing mechanism is supported by the LACE data. The assumption that the balloon sampled air of average descent has some effect on the derived profiles of Z. But since all of the tracers are affected in a similar manner by making a different assumption about the sampled air, and consistency of the results between all the tracertracer pairs is what we are looking for, this assumption should not impact the conclusions we derive. [29] For each tracer-tracer pair a fit is performed for one of the tracers as a function of pressure altitude. We use pressure altitude instead of theta as the vertical coordinate since theta surfaces are nearly horizontal within the vortex. Thus, horizontal mixing within the vortex should mostly occur along constant pressure altitude surfaces. Also, temperature is not measured as accurately, or with the same resolution, as pressure on the OMS platform so theta is less accurate than pressure altitude. The second tracer of the pair is then fit to the first tracer using the correlation curve from the first flight. Profiles of each tracer are then lowered or raised in altitude relative to the measured first flight profiles to represent differential descent areas within the vortex. The differential descent profiles are then mixed horizontally through the relation:, c Mix ðþ z XNA cðz þ dzði; zþþ N A ð6þ i¼1 where c(z + dz(i, z)) represents the differential descent tracer profiles, i is a vortex area index representing equal areas of the vortex and N A is the number of areas within the vortex used to perform the calculation. The linear relationship between dz and the area index i is given by: dzði; z Þ ¼ 2ZðÞ z N A 1 i N A For the results shown here the vortex is divided into five equal area sections so that N A = 5. No assumption is made about the shape of the areas for instance as a function of latitude. Calculations using larger values of N A gave similar results. An iteration is performed to get the best fit to the correlation curve of the second flight. This is done by ð7þ