The major stratospheric final warming in 2016: dispersal of vortex air and termination of Arctic chemical ozone loss

Abstract. The 2015/16 Northern Hemisphere winter stratosphere appeared to have the greatest potential yet seen for record Arctic ozone loss. Temperatures in the Arctic lower stratosphere were at record lows from December 2015 through early February 2016, with an unprecedented period of temperatures below ice polar stratospheric cloud thresholds. Trace gas measurements from the Aura Microwave Limb Sounder (MLS) show that exceptional denitrification and dehydration, as well as extensive chlorine activation, occurred throughout the polar vortex. Ozone decreases in 2015/16 began earlier and proceeded more rapidly than those in 2010/11, a winter that saw unprecedented Arctic ozone loss. However, on 5–6 March 2016 a major final sudden stratospheric warming ("major final warming", MFW) began. By mid-March, the mid-stratospheric vortex split after being displaced far off the pole. The resulting offspring vortices decayed rapidly preceding the full breakdown of the vortex by early April. In the lower stratosphere, the period of temperatures low enough for chlorine activation ended nearly a month earlier than that in 2011 because of the MFW. Ozone loss rates were thus kept in check because there was less sunlight during the cold period. Although the winter mean volume of air in which chemical ozone loss could occur was as large as that in 2010/11, observed ozone values did not drop to the persistently low values reached in 2011. We use MLS trace gas measurements, as well as mixing and polar vortex diagnostics based on meteorological fields, to show how the timing and intensity of the MFW and its impact on transport and mixing halted chemical ozone loss. Our detailed characterization of the polar vortex breakdown includes investigations of individual offspring vortices and the origins and fate of air within them. Comparisons of mixing diagnostics with lower-stratospheric N2O and middle-stratospheric CO from MLS (long-lived tracers) show rapid vortex erosion and extensive mixing during and immediately after the split in mid-March; however, air in the resulting offspring vortices remained isolated until they disappeared. Although the offspring vortices in the lower stratosphere survived longer than those in the middle stratosphere, the rapid temperature increase and dispersal of chemically processed air caused active chlorine to quickly disappear. Furthermore, ozone-depleted air from the lower-stratospheric vortex core was rapidly mixed with ozone rich air from the vortex edge and midlatitudes during the split. The impact of the 2016 MFW on polar processing was the latest in a series of unexpected events that highlight the diversity of potential consequences of sudden warming events for Arctic ozone loss.

constraints, and parameterizations (Molod et al., 2015;Takacs et al., 2016). Assimilated fields are provided on a 0.625 • × 0.5 • longitude/latitude grid with 72 hybrid σ-pressure levels. Here we primarily use the wind, temperature, and potential vorticity fields provided in the "M2I3NVASM" file collection (Global Modeling and Assimilation Office (GMAO), 2015), the set of dynamically consistent fields obtained after the IAU step; these fields are provided at the full model resolution at 3-hour intervals (8 times per day).

MLS Data
The Earth Observing System (EOS) Aura satellite was launched in July 2004, in a 98 • inclination orbit that provide data coverage from 82 • S to 82 • N latitude on every orbit. Aura MLS measures millimeter-and submillimeter-wavelength thermal emission from the limb of Earth's atmosphere. Detailed information on the measurement technique and the Aura MLS instrument is given by Waters et al. (2006). Vertical profiles are measured every 165 km along the suborbital track and have 10 a horizontal resolution of ∼200-500 km along-track and a footprint of ∼3-9 km across-track. In this study we use version 4 (v4) Aura MLS N 2 O, HNO 3 , H 2 O, HCl, ClO, and O 3 measurements from Arctic winters spanning 2004/05 through 2015/16. The quality of these data is described by Livesey et al. (2015a). Vertical resolution is about 2.5-3 km for O 3 , 3 km for H 2 O, HCl, and ClO, 3-5 km for HNO 3 , and 5-6 km for N 2 O in the lower to middle stratosphere, and about 5 km for CO in the middle stratosphere. Single-profile precisions are approximately 0.03-0.1 ppmv, 0.2-0.3 ppbv, 0.1 ppbv, 0.6 ppbv, 13-20 ppbv, 15 and 16 ppbv for O 3 , HCl, ClO, HNO 3 , N 2 O, and CO, respectively, and 5-15% for H 2 O. The v4 MLS data are quality-screened as recommended by Livesey et al. (2015a). For daily maps, MLS data are gridded at 2 • latitude by 5 • longitude using a weighted average around each gridpoint of 24 hours of data centered at 12:00 UT.
Equivalent latitude (EqL, the latitude that encloses the same area between it and the pole as the corresponding potential vorticity, PV, contour, Butchart and Remsberg, 1986) and scaled PV (sPV, scaled as in Dunkerton and Delisi, 1986;Manney 20 et al., 1994) are used in the analysis described below. These quantities, as well as temperatures from MERRA-2, are obtained at MLS locations from an updated version of the MLS derived meteorological products (DMPs) described by Manney et al. (2007). MLS data are interpolated to isentropic surfaces using temperatures from MERRA-2.

Vortex and temperature diagnostics
To investigate the potential for polar chemical processing and ozone loss during the 2015/16 winter, we use a standard set of 25 polar processing diagnostics calculated from MERRA-2 data. We primarily make use of diagnostics described by Lawrence et al. (2015), including minimum temperatures (T min ), the volume of air with temperatures below polar stratospheric cloud (PSC) existence thresholds as a fraction of vortex volume (V P SC /V V ort ), maximum PV gradients, and the area of the polar vortex in sunlight (or sunlit vortex area). All of these diagnostics are calculated from the 12:00 UT temperature and potential vorticity fields provided by MERRA-2 interpolated to isentropic surfaces between 390 and 580 K (approximately 120 to 30 hPa,30 or 14 to 24 km). Further discussion of these diagnostics and their significance to polar chemical processing can be found in Manney et al. (2011) and Lawrence et al. (2015).
Our analysis makes use of a detailed characterization of the 2015/16 stratospheric polar vortex, particularly during the period of time when the vortex split into multiple offspring. We use the CAVE-ART (Characterization and Analysis of Vortex Evolution using Algorithms for Region Tracking) analysis package, which was developed to comprehensively describe the state of the polar vortex throughout the winter season. A paper describing the full details and implementation of CAVE-ART is in preparation (Lawrence and Manney, 2016); in short, CAVE-ART uses image processing and region tracking algorithms to 5 objectively identify any number of vortex regions and track their positions through time. CAVE-ART identifies vortex regions based on altitude-dependent contours of sPV that we specify as being representative of the vortex edge. These sPV values are selected using climatological profiles of sPV spanning 25 isentropic levels between 390 and 1800 K to identify the sPV value at each level that coincides best with maximum sPV gradients from the MERRA reanalysis. Once CAVE-ART identifies individual vortex regions, it filters out those below a specified area threshold; except where otherwise noted, we use herein an 10 equivalent latitude threshold of 84 • , which is an area of roughly 0.5% of a hemisphere. All remaining regions are then tracked through time using the full time resolution of the meteorological data (eight times per day for MERRA-2) until the regions fall below the area threshold, or in some cases merge with another region. CAVE-ART also calculates and saves diagnostics at every timestep that describe the position, size, and strength of each region. These diagnostics include 2-D moment diagnostics such as aspect ratios and centroids (e.g., Matthewman et al., 2009;Mitchell et al., 2011), vortex areas, average altitudes, and 15 vortex-edge windspeeds.
Such detailed characterizations are particularly useful during vortex-split SSW events wherein the resulting offspring vortices can vary in size and strength in ways that ultimately influence polar processing. For example, a preliminary version of CAVE-ART was used by Manney et al. (2015a), to show how the early January 2013 vortex-split SSW was responsible for accelerated ozone loss in January 2013. For the current paper, the CAVE-ART characterization is particularly important be-20 cause, as will be shown, during the 2016 MFW, the vortex rapidly weakened and briefly split into three offspring vortices at some levels. The capability to track more than two offspring vortices is, to our knowledge, currently unique to CAVE-ART, as other methods in the literature rely on moment diagnostics that can, at best, delineate between only two regions (Mitchell et al., 2011). We have included a supplementary animation that shows the evolution of the polar vortex during the March 2016 MFW, which illustrates the CAVE-ART characterization of the vortex split. 25

Transport and Mixing Diagnostics
EqL time series of MLS data are produced using a weighted average of MLS data in EqL and time, with data additionally weighted by measurement precision (e.g., Manney et al., 1999Manney et al., , 2007. All vortex averages of MLS data shown use the altitudedependent sPV values derived for CAVE-ART to identify the edges of the vortex or vortices (Section 2.3). For averages in multiple vortices, the sPV from the MLS DMPs is first used to determine whether the MLS measurement location is within any 30 vortex. Those points that are within a vortex are then marked with the labels for individual regions provided by CAVE-ART to identify which of multiple vortices they are inside. Vortex averages are shown here for "bulk" (all MLS measurements with sPV greater than the altitude-dependent threshold), "sum" (the sum of all the regions with area greater than the 84 • EqL cutoff used in the CAVE-ART runs), and for individual vortices identified by CAVE-ART. Averaging improves MLS precisions to values smaller by a factor of about 10 to more than 100 over the single-profile precisions listed in Section 2.2 for EqL gridded and vortex averaged fields.. sPV gradients and effective diffusivity (K eff ) as a function of EqL calculated from MERRA-2 PV are shown as "global" (that is, characterizing amounts averaged around EqL contours) diagnostics of mixing. Gradients of sPV as a function of EqL provide a measure of the strength of the vortex edge as a transport barrier averaged over each day and all vortices (Manney et al., 2011, 5 2015b, and references therein). K eff is expressed as log-normalized equivalent length, i.e., the length of a tracer contour with respect to the contour of minimum length that would enclose the same area; high (low) values thus reflect complex (simple) structure in tracer (here PV) contours and indicate strong (weak) mixing (e.g., Nakamura, 1996;Haynes and Shuckburgh, 2000;Allen and Nakamura, 2001). The magnitudes of K eff values depend strongly on the resolution of the PV fields used in the calculations, but K eff distributions from MERRA-2 agree morphologically with those calculated from other analyses 10 and reanalyses. Similarly to sPV gradients, the gradients of long-lived trace gases on isentropic surfaces as a function of EqL indicate the strength of the vortex edge transport barrier. We use the EqL/time gridded MLS fields to calculate these gradients.
The diagnostics of mixing and transport barriers described above represent averages around EqL contours, and thus give information on bulk mixing properties; for example, the strength of the transport barrier at the EqL of the vortex edge is an estimate of that barrier averaged over the the entire length of the edges of all vortices present at that time. To examine 15 regional mixing (e.g., variations along the edge of a vortex, or differences between individual vortices), we use the function M (hereinafter referred to as M ) to give a synoptic picture of the strength of the vortex transport barrier prior to, during, and after the 2016 MFW. M is a Lagrangian diagnostic calculated from parcel trajectories that has been used to highlight processes related to transport and mixing in geophysical fluid flows (Mendoza and Mancho, 2010;de la Cámara et al., 2012;de la Cámara et al., 2013;Smith and McDonald, 2014). The formal definition of the function M is as follows: consider a point 20 in an n-dimensional space defined at an initial time t 0 by the general coordinates (x 1,0 , ..., x n,0 ). If a parcel is initialized at this point and advected by the background velocity field ( dxi dt ), then the function M at this point is then defined by the integral equation M (x 1,0 , ..., x n,0 , t 0 ) = t0+τ t0−τ n i=1 dxi dt 2 1/2 dt. This is the Euclidean arc length of the trajectory traced out by the parcel in the time interval [t 0 − τ, t 0 + τ ]. If a grid of such points and parcels is constructed, then a field of M can be defined by calculating the above integral for each point within the grid. For our application, we instead calculate a field of M 25 by using zonal and meridional winds in a trajectory code to advect parcels initialized on a regular longitude/latitude grid. We then calculate M by summing up the distances between parcel locations at successive times assuming that these locations are connected by great circle arcs.
We use the core of the Lagrangian trajectory diagnostic code described by Livesey et al. (2015b) to calculate parcel advection using a fourth-order Runge Kutta scheme. The trajectories we use here are calculated via integrations with a fifteen minute 30 timestep from MERRA-2 winds. We initialize parcels on a 1.25 x 1.00 degree longitude/latitude grid (a grid spacing twice the size of the MERRA-2 native grid) poleward of 20 • N. The calculations of M are based on isentropic trajectories that are carried out for 15 days forward and backward (i.e., τ = 15 days, for 30 days total) from 00:00 UT on the initialization date.
M has been used before to study transport in Earth's polar vortices. de la Cámara et al.  Smith and McDonald (2014); namely, large values of M indicate parcels that were effectively trapped in a transport barrier for most of the trajectory timeline, whereas small values of M indicate the opposite with parcels that were more prone to mixing. We also use an area diagnostic similar to that in Smith and McDonald (2014), obtained by calculating the area enclosed by contours of M for the entire grid, and expressing these as an equivalent latitude, which we denote by "M - 10 EqL". Although EqL is most commonly used to describe the area within PV or tracer contours (e.g., Butchart and Remsberg, 1986;Allen and Nakamura, 2003) (as PV-based EqL is used herein), we have found that examining M and M -EqL together facilitates understanding of how the size, strength, and sharpness of the vortex edge transport barrier evolve with time.
We also use the trajectories described above to explore the origins and fate of air from the polar vortex during its breakup.
We use the CAVE-ART identification of vortex regions to "tag" parcels that were initialized inside each valid and distinct 15 vortex region. This allows us to examine the full history of parcels with respect to their original confinement within materially separated vortex regions. Similar trajectories were calculated using the full 3D trajectory code with diabatic motions, and only extending backwards from 12:00 UT each day, for use in reverse domain filling (RDF) calculations initialized with MLS data; the MLS fields used for these initializations are the gridded map fields described above. in fact, the vortex remained strong and relatively cold after the period shown here, and ozone continued to drop, reaching a minimum of ∼1.5 ppmv in late April (e.g. Manney et al., 2011;WMO, 2014). While the vortex was exceptionally strong 30 and vortex-averaged ozone loss unprecedented in the 2010/11 winter/spring, the size of the vortex during much of the winter (through late February), and the portion of it exposed to sunlight, were both less than average (Figure 1d).
In early January 2013, temperatures abruptly rose far above those at which chlorine can be activated (and hence chemical ozone loss occur) during one of the strongest "vortex-split" SSWs on record (Manney et al., 2015a); however, the exceptional cold prior to that event (Figure 1a,b), and exceptional exposure of the vortex (offspring vortices) to sunlight in December (January) (Figure 1d) led to denitrification comparable to that in 2011 ( Figure 2b) and the largest early winter chlorine activation and ozone loss on record (Figure 2d,e,f) (Manney et al., 2015a). and near or at record lows from late December though January. The period of over a month, from late December through early February, with temperatures below the ice PSC threshold was unprecedented in the Arctic, where the previous record rarely shows more than a few contiguous days below this threshold and never more than about three weeks (Figure 1a and 15 Figure 3c,d). The long period of temperatures below the ice PSC threshold led to a much greater degree of dehydration than has been seen before in the Arctic: Compare the evolution of H 2 O in 2016 in Figure 2c with the small decrease in H 2 O seen in 2011 when there were separated periods in late January and February of about one and three weeks' duration, respectively, with temperatures below the ice PSC threshold. The presence of large ice PSCs also can lead to greater denitrification than the presence of (typically smaller) NAT or liquid PSC particles alone (e.g., Wofsy et al., 1990;Hintsa et al., 1998;Santee 20 et al., 1998;Lowe and MacKenzie, 2008;Dörnbrack et al., 2012;Wohltmann et al., 2013), and is consistent with the extreme denitrification evident in Figure 2b. Figure 1d shows that the 2015/16 vortex was not only larger than usual, but also had a larger area than usual receiving sunlight during January through mid-March. The steep decrease in ClO began nearly a month earlier than that in 2011. 10 V N AT /V V ort (V P SC /V V ort calculated for nitric acid trihydrate PSCs and integrated over a winter) is a diagnostic commonly used to indicate the overall potential for polar processing and ozone loss (e.g., Rex et al., 2004;Tilmes et al., 2006). The polar processing potential in 2015/16 estimated using this diagnostic ( Figure 3a) was nearly identical to that in 2010/11. Another measure, the number of days integrated over the lower stratosphere on which activation can occur (Figure 3b; e.g., Manney et al., 2011, Supplementary Information) shows values similar to the largest previously observed excepting 2011. Similar di-15 agnostics for the volume and duration of air below the ice PSC threshold (Figure 3c,d) indicate much greater potential for dehydration (and denitrification) than in any previously observed Arctic winter. It may be argued that more extensive denitrification in 2015/16 enhanced the ozone loss potential because of its effect of slowing chlorine deactivation (e.g., Douglass et al., 1995;Santee et al., 1996Santee et al., , 2008Waibel et al., 1999;Davies et al., 2002). Thus, the critical factor resulting in less ozone loss than in 2011 was the much earlier increase in temperatures and vortex breakup in 2016. In contrast to 2011 and 2013, during which evidence of renitrification was seen above 400 K (e.g., Sinnhuber et al., 2011;Arnone et al., 2012;Manney et al., 2015a, b), sequestration in PSCs and denitrification led to depleted gas phase HNO 3 ( Figure 4b) extending below 400 K in 2015/16. Sequestration in ice PSCs, and evidence of dehydration (in that low vortex H 2 O lingered well beyond the period with temperatures below the ice PSC threshold) is apparent in Figure 4c from about 420 K 30 to above 550 K. Extensive chlorine activation is apparent from about 400 K up to above 600 K (Figure 4d,e), an upper extent comparable to that in the Antarctic. The upward tilt of ozone contours (Figure 4f) at levels from below 400 K to above 600 K beginning in early January indicates sufficient chemical ozone loss to exceed the replenishment by descent. This signature extends until mid-February at the higher levels, early March near 500 K, and continues into April at the lowest levels shown.
In the following, we focus on the evolution of the vortex and trace gas transport during the MFW on the individual isentropic surfaces marked by horizontal lines in Figure 4. 850 K (∼31 km, estimated from CAVE-ART vortex averaged altitude over the winter) is shown to represent the middle-stratospheric regime where the vortex decay is very rapid. 490 and 550 K (∼20 km and 22 km, respectively) represent the two regimes in the lower stratosphere with significantly differing vortex evolution during the MFW. These lower stratospheric levels are near the maximum (490 K) and top (550 K) of the region of chemical processing. March. The vortex area shrinks steadily through the winter, even as the vortex edge transport barrier strengthens and mixing outside the vortex increases (weaker sPV gradients, higher K eff ). This is consistent with the climatological development of the Aleutian anticyclone, intensified mixing in the surf zone, decreasing vortex area, and accompanying strengthening of PV and 15 tracer gradients along the vortex edge (e.g., McIntyre and Palmer, 1984;Leovy et al., 1985;Butchart and Remsberg, 1986;Harvey et al., 2002). Lower sPV gradients and higher K eff at midlatitudes in February indicate increasing activity in the surf zone (as has been previously reported, e.g., Haynes and Shuckburgh, 2000;Allen and Nakamura, 2002) consistent with the spreading of higher H 2 O values out from the vortex edge region. CO, because of its extremely strong gradients across the vortex edge, provides a sensitive indicator of export of vortex air, and indicates periods of such enhanced transport in mid-February 20 and early to mid-March. After the MFW began, the rate of vortex shrinkage accelerated rapidly, with the area enclosed within a transport barrier (sPV gradient maximum, K eff minimum) approaching zero by the end of March. The H 2 O and CO values show only slightly weakened gradients across the vortex edge in its final days, suggesting that most of the air in the remnants of the vortex was well confined within them until they disappeared.
In the lower stratosphere, at 490 K ( Figure 6) the maximum PV gradients align closely with the minimum in K eff and indicate 25 a strong barrier to mixing. The large and strong vortex persists until nearly mid-March, past the start date of the MFW. In early February, maximum K eff in mid-latitudes increases, suggesting that more vigorous mixing in the surf zone extends down into the lower stratosphere (consistent with the results of, e.g., Waugh and Randel, 1999;Harvey et al., 2002). Vortex area suddenly decreased and maximum sPV gradients/minimum K eff decreased/increased immediately after two small offspring split off the vortex (around 13 March, second vertical red line), leaving the larger parent vortex even more distorted (see supplementary spreading of values previously characteristic of the vortex edge throughout the midlatitude surf zone. This is consistent with the common pattern of filaments of air being drawn off the vortex, around and into the anticyclone during this period. HCl shows consistent evidence of some mixing of very low values out of the vortex, with a concurrent extrusion of high ClO, in early February; these signatures are short-lived, since active chlorine transported out of the vortex in filaments rapidly decays via both deactivation and mixing (e.g., Konopka et al., 2003;Tan et al., 1998;Marchand et al., 2004). Additional brief events of   15 Manney et al., 2005), and substantial mixing into midlatitudes is apparent, consistent with the signature in the other species.
While the transport barriers seen in sPV gradients and K eff are weaker after mid-March, a significant maximum and minimum, respectively, remain along the edge of the rapidly shrinking vortex through early April. It is only at this time (apparent around 7 April in Figure 6) that very low N 2 O and O 3 previously confined to the vortex core are seen equatorward (in EqL) of the strong PV gradients, indicating the final decay of the vortex. 20 A somewhat similar evolution is seen at 550 K (Figure 7), with a large vortex bounded by a strong transport barrier into early March, accompanied by increased mixing in midlatitudes in February consistent with filamentation and a more vigorous surf zone. In contrast to 490 K, however, while the vortex area shrinks after the onset of the MFW and vortex split, the maximum sPV gradients remained about as strong as before the split, and K eff continued to show a more pronounced minimum than at 490 K. Consistently, N 2 O and O 3 (and other trace gases, not shown) do not suggest substantial mixing out of the vortex 25 core until mid-April. Note that, similar to 490 K, vortex ozone was also strongly depleted at this level, resulting in very strong gradients along the inner edge of the vortex. At this level, however, the unperturbed morphology of ozone is such that vortex values are generally much lower than those outside the vortex prior to the onset of chemical loss.
The evolution of transport barriers and trace gases in the lower stratosphere is examined in more detail in Figure 8.  The EqL-based view presented above gives a global perspective on the evolution in vortex area and strength during the MFW period. This averaged view of mixing diagnostics shows that a small area of well-confined vortex air lingered through March, but by early April the transport barrier presented by the vortex edge was greatly weakened, and the potential for mixing was high. In the following, we focus on the synoptic evolution of the vortices and regional aspects of transport and mixing during the MFW period.  Figure 12 shows trajectory-based air parcel history maps at 850 K initialized on 12 March, just before the vortex split described above. On this date (column 2), a long filament of significant area had been drawn off the vortex and broken up into two pieces, each large enough to be identified in CAVE-ART as a vortex region. Though these were drawn off the vortex edge on 12 March, that air had been deep within the vortex 14 days before (column 1); this indicates that there was substantial mixing within the vortex itself. After being drawn off, these narrow filaments quickly dissolved, with the air from them being widely  Figure 13 shows similar air parcel history maps at 490 K during and after the vortex split. The air parcels in the vortices on 16 March (about two days after the split, row A) originated within the vortex 12 days earlier, with the parcels in the two small offspring vortices (green and blue) coming primarily from the narrower portion extending south near 30 • E longitude. After the split, most of the air in the blue offspring vortex, which originated near 0 • longitude, remained within a tight confined region for over two weeks, even after a vortex was no longer identified in that region. The 20 March initialization (row B, column 2), 5 in which the blue vortex is the same one as the more persistent vortex region from the 16 March initialization, shows that this offspring vortex retained its identity into early April. The parent vortex (black) began to experience substantial filamentation in late March (row B, column 3). This main vortex was very small and had weakened by early April (row B, column 4), but was still identified as a vortex region on 10 April (row C, column 2) and maintained some coherence into late April (row C, columns 3 and 4). 10 At 550 K (Figure 14), the air parcels in the smaller (green) offspring vortex just after the split (row A) originated 12 days before primarily from a ring of air just inside the vortex edge. Most of the air in this offspring vortex remained coherent through March (row A, columns 3 and 4; row B, columns 2 and 3). Another small and very short-lived (about a day) offspring that broke off on 24 March (row B, column 2, blue vortex) was rapidly sheared out and the air originating in it wrapped around the outside of the parent (black) vortex on 8 April (row B, column 4). By this time the air in the green offspring vortex was starting 15 to lose its coherence, though it can be seen (row C) that the air remained within a relatively confined region through 15 April (row C, column 3) and still showed some coherence on 20 April (row C, column 4).
Thus, compared to the rapid and wide dispersal of vortex air in the middle stratosphere, air from even small offspring vortices in the lower stratosphere maintained some coherence much longer after the vortex split in the lower stratosphere. At all levels, examination of the grey parcels -that is, all the parcels that were outside any vortex on the initialization day -without the 20 overlaid vortex parcels indicates that few of them move into coherent vortex regions. That is, as long as the regions were large enough to be identified in CAVE-ART, they remained mostly devoid of air with extravortex origins. This indicates that the mixing during the vortex break up was largely one-way, with air mixing out of the vortices through filamentation as they eroded and lost their identity. This result is consistent with previous studies of dispersal of air from the lower stratospheric vortex (e.g., Chen et al., 1994;Manney et al., 1994), and with the picture of a shrinking and weakening vortex decaying 25 primarily by erosion into midlatitudes. Figure 15 summarizes how the transport and mixing processes described above affected trace gases in the lower stratospheric vortex. The top panels show the evolution of the vortex areas, and the MLS sampling of those vortices. An abrupt decrease in vortex area immediately followed the vortex split, with the total (sum of all vortices) area decreasing by about 40% and 30% at 490 and 550 K, respectively. This is consistent with the maps shown above and the time evolution shown in the 30 supplementary vortex regions animation. At 490 K, the vortex size decreased more slowly, but steadily, thereafter, to about 3% of the hemisphere by the end of March, and less than 1% of the hemisphere by mid-April. At 550 K, the decay was more step-like, with another fairly rapid decrease in the area to a total of about 3% of the hemisphere in late March (near the time the second (blue) small, but very short-lived, offspring was pulled off and dispersed), followed by a sudden disappearance (that is, no vortex had area greater than about 0.5% of the hemisphere) by 12 April. 35 15 Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-633, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 29 July 2016 c Author(s) 2016. CC-BY 3.0 License.
The MLS sampling of the large, strong vortex in January through mid-February included 500-700 measurements per day, but both its area and the number of measurements in it had dropped somewhat at all levels by 24 February (the start date of the panels in Figure 15). In general, the number of MLS measurements in the vortices closely tracks their area, and there are several MLS measurements in each: For every vortex region identified by CAVE-ART that lasted more than one day, the minimum number of MLS measurements on a day was at least six. This suggests that MLS usually provided relatively 5 unbiased sampling of even small offspring vortices that were just larger than the 84 • EqL cutoff used by CAVE-ART. The number of MLS measurements begins dropping earlier, in the period between the beginning of the MFW and the split, because the vortex shifted farther off the pole to where MLS sampling is less dense. The rate of decrease in MLS measurements in the vortex at 550 K was steeper before the split than at 490 K, consistent with the vortex at that level being shifted farther out into midlatitudes. At 490 K, the steepest decrease in vortex MLS measurements was right around the split date. The minimum in 10 number of MLS measurements shortly after the split (especially apparent at 490 K) is likely related to the fact that the vortex was shifted very far off the pole into midlatitudes and moved closer to the pole, into areas more densely sampled by MLS, in the following several days (see, e.g., Figures 10 and 11).
Vortex edge windspeeds show a deep minimum in the period between the start of the MFW and the split. Windspeeds showed some day-to-day variability after the split, but over all decreased steadily. The minimum just prior to the split arises 15 largely because the vortex had already developed into two separate circulations that were only joined immediately prior to the vortex split by a narrow "bridge" with high PV but low windspeeds. As seen above, the offspring at 490 K were short-lived (about 5 and 7 days for the green and blue vortices, respectively), during which the windspeeds decreased rapidly along their edges. In fact, as seen in Figure 13, a coherent mass of air from the blue vortex persisted into April -represented in Figure 15 by the individual purple points labeled "transient", which mark the days on which the area of this region was larger than the 20 84 • EqL cutoff. The windspeeds around the edge of the parent vortex (black) remained stronger, though generally decreasing, into late April. A somewhat similar picture is seen at 550 K, with the windspeeds around the single offspring vortex (green) being weaker than those bounding the parent vortex, and the parent outliving the offspring; however, the offspring vortex at this level was much longer lived than those at 490 K. The evolution of vortex edge windspeeds is thus consistent with that of the transport barriers seen in the mixing diagnostics (sPV gradients, K eff , and M ) shown above. 25 The evolution of trace gases in the individual vortices is also consistent with the picture of mixing and vortex breakup seen above. At 490 K, N 2 O values were substantially higher in the blue offspring vortex, which persisted slightly longer than the green one, but was still rapidly sheared out into an elongated shape and weakened (as indicated by decreasing windspeeds).
Examination of reverse domain filling (RDF, Sutton et al., 1994) maps initialized with MLS data (not shown) suggests that the rapid N 2 O increase in the blue vortex in the last two days may be an artifact of MLS not sampling the low values in the narrowest 30 part of the vortex as it was sheared out. Figure 13 (e.g., row A, columns 3 and 4) shows rapid and widespread dispersal of the air from both blue and green offspring vortices, but with some of the air from the blue vortex remaining relatively coherent in a small region even after that vortex was no longer defined. H 2 O values were higher in the green offspring vortex because the air in that vortex came from nearer the edge of the parent vortex, rather than from the core where H 2 O was strongly depleted ( Figure 13, row A, column 1). Average H 2 O in the blue offspring vortex was close to that in the parent (black) vortex, consistent 35 16 Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-633, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 29 July 2016 c Author(s) 2016. CC-BY 3.0 License.
with that air coming from somewhat deeper in the parent vortex; this is also consistent with the appearance in Figure 6 of a "path" of low water crossing the vortex edge at the time of the split. Ozone was higher in both the green and blue offspring than in the parent because the air originated in the high O 3 collar near the vortex edge. It was highest in the green vortex because that air came from farther out towards the region of the O 3 maximum (see Figure 6). As was the case for N 2 O, the increase in the blue vortex in the last few days may be exaggerated by MLS sampling "missing" a narrow filament of vortex air.

5
At 550 K, N 2 O values were consistently higher in the single (green) offspring vortex than in the parent, indicating more extravortex or vortex edge air than in the parent, as shown in Figure 14 (row A, column 1). That air, however, remained largely confined within that vortex after the split (Figure 14, row A, columns 3 and 4), consistent with relatively constant N 2 O mixing ratios, and suggesting little additional mixing. RDF maps (not shown) at this level do not show obvious evidence of MLS measurements missing filaments of vortex air. There was much less dehydration than at 490 K (see, e.g., Figure 4), so vortex 10 values carried into the green offspring vortex were substantially higher than extravortex values, and the anticorrelation seen between N 2 O and H 2 O in that offspring vortex is consistent with this morphology. Low ozone values extended out to the vortex edge at 550 K (e.g., Figure 7), and thus the offspring carried very low ozone values with it. This offspring vortex was long-lived, and, though it shrank to an area too small to be cataloged a few days sooner than the parent, the air within both it and the parent remained coherent into late April ( Figure 14, row C, columns 3 and 4). Higher ozone air was drawn up around 15 the parent vortex later on (e.g., Figures 11, and 14,  further evidence that the air in the offspring vortices was confined by an effective transport barrier as long at those vortices remained coherent. Thus, except in the period immediately surrounding the split, rapidly decreasing ClO and increasing HCl in all offspring resulted primarily from photochemical deactivation. While non-zero, albeit small, values of ClO (e.g., Figure 2) are apparent in the vortex averages through March, the area in which additional chemical loss could occur was small, less than 4% and 2% of the hemisphere at 490 K and 550 K, respectively. 25

Summary and Conclusions
We have analyzed meteorological fields from the MERRA-2 reanalysis and trace gas data from the Aura Microwave Limb Sounder (MLS) to provide an overview of the exceptionally cold 2015/16 winter and a detailed description of the the vortex breakup in a major final SSW ("major final warming" or MFW) that prevented chemical ozone loss from reaching record high values. Our analyses utilized several mixing diagnostics, as well as a new package (CAVE-ART) for characterizing multiple 30 vortex regions.
The 2015/16 Arctic winter was the coldest on record in December through early February. Lower stratospheric temperatures were at or near a record lows from late December into early February, and far below average from December through mid-March. A substantial region of temperatures below the ice PSC threshold was present continuously from late December through early February, far longer than during any previously observed Arctic winter: The winter mean volume of air below the ice PSC threshold was over twice that previously seen. The chemical ozone loss potential, measured by the commonly used metric of volume of air below the chlorine activation threshold, was nearly identical to that in 2010/11 (when unprecedented Arctic ozone loss occurred). The evolution of trace gases from MLS is consistent with the exceptional meteorological conditions: 5 Vortex-wide dehydration was present between about 410 K and 520 K potential temperature, something never before observed in the Arctic. Denitrification was also exceptional, and extensive chlorine activation and chemical ozone loss began earlier than in all but one previous winter.
That lower stratospheric ozone loss did not reach values comparable to those in spring 2011 was primarily due to the occurrence of an MFW beginning in early March 2016. This event had two critical consequences: First, while the the total 10 volume of cold air during the winter was similar to that in 2010/11, that cold period ended significantly earlier in the winter in 2016, when ozone loss was slower due to less sunlight exposure. Second, the sudden vortex breakup in the MFW resulted in rapid dispersal of chemically processed air from the vortex and consequently curtailed chemical processing, which might have lingered for some time if chlorine had remained confined in relatively large intact vortex and thus deactivated more gradually.
The Arctic winter meteorology in 2015/16 was so remarkable that extensive study of numerous processes will be needed to 15 fully characterize its consequences. In this paper we focus on one aspect of this exceptional winter: a detailed description of the event that limited ozone loss to an amount that, while larger than typical in the Arctic, was not unprecedented -the MFW and vortex breakup in early March. The MFW itself was an unusual SSW: The major SSW criteria were fulfilled when the vortex was a single elongated entity displaced far off the pole (a typical "displacement" SSW as defined by Charlton and Polvani, 2007), but a few days later the vortex split over a wide range of altitudes covering most of the stratosphere (behavior typical of 20 a "split" SSW, e.g., Matthewman et al., 2009). Moreover, in a narrow range of levels in the lower stratosphere near 450 K to 550 K, that split was into three pieces.
In the middle stratosphere (exemplified herein by 850 K), mixing diagnostics and MLS trace gases show that by the time of the MFW the vortex had already shrunk and a strong Aleutian anticyclone and vigorous surf zone formed, consistent with climatology. In mid-March, about a week after the MFW began, the vortex split into two very unequal pieces; the larger one 25 rapidly sheared out and dispersed, while a very small coherent remnant of the other remained intact with relatively little mixing into early April. The evolution of MLS CO and H 2 O in the decaying vortices indicates that air within them remained well confined as long as they were intact. Snapshots of the function M show a picture consistent with the the trace gas evolution, in that the vortex transport barrier decayed rapidly after the MFW onset.
The breakup of and dispersal of air from the vortex in the lower stratosphere was slower and more episodic, with largest 30 changes in the short period surrounding the vortex split. Some of the specific consequences of the lower stratospheric vortex evolution (shown here at 490 and 550 K) during the MFW for transport, mixing, and dispersal of chemically processed air are as follows: -At 490 K, two small offspring split off the main vortex in mid-March, but persisted for only about a week.
-At 550 K, the vortex split into two pieces, both of which remained well defined for over a month after the split.
-Mixing increased only slightly after the onset of the MFW around 7 March, but extensive mixing occurred in the few days during and after the vortex split in mid-March.
-Immediately following the split the total vortex area decreased by 30% to 40%, with the largest offspring covering about 4% of the hemisphere, and smaller offspring an additional 1 to 2% of the hemisphere.

5
-Following this period of intensive vortex erosion and mixing, air remained well-confined within the remaining offspring vortices.
-Abundances of MLS N 2 O and O 3 in the offspring vortices at 550 K remained closer to those in the parent vortex than at 490 K, indicating less mixing; this is consistent with the stronger transport barrier after the vortex split seen in the mixing diagnostics at that level, and with the greater persistence of the offspring vortices.

10
-ClO rapidly decayed in the offspring vortices as a result of a combination of rapid deactivation and dispersal of vortex air during the split.
-The evolution of ozone in the offspring vortices was dependent on the region within the parent vortex where the air originated, such that the offspring at 490 K contained higher values characteristic of the collar of undepleted ozone along the vortex edge, whereas at 550 K, low ozone values extended farther out into the vortex edge region and the smaller, but 15 stronger, offspring vortex carried lower ozone than the parent.
-The "function M ," when binned as a function of EqL, evolved consistently with the bulk mixing diagnostics (sPV gradients and effective diffusivity), but also revealed local variations (including relative strength of the offspring vortices, variations in the transport barrier around the vortex edge, and the dissolution of the individual vortices) that are consistent with the synoptic evolution of MLS trace gases. 20 In both the lower and middle stratosphere the mixing following the MFW was primarily via erosion and filamentation of the vortices as long as they remained intact. This resulted in wide dispersal and rapid mixing of air formerly in the vortex, but little extra-vortex air intruding into the vortex regions while they remained well-defined.
The major final SSW in early March 2016 was a remarkable finale to an already exceptional Arctic winter. The results presented here suggest the need for many further studies to assess not only how well the evolution of the vortex and trace 25 gases throughout the 2015/16 winter fits with our current understanding of and ability to model lower stratospheric polar chemical processes, but also provides a unique addition to the already wide variety of natural experiments conducted via the immense variability in Arctic polar vortex evolution, longevity, and breakup. This new information is important for improving our detailed understanding of variations in dispersal of ozone depleted and/or chemically activated air from the vortex and its implications for present and future global ozone distributions. Further studies will include detailed analyses using similar 30 methods to this work comparing the vortex breakup in 2016 with that in other winters, both Arctic and Antarctic. This is particularly interesting given reported differences between years with early and late Arctic final warmings, which have not, in general, accounted for the suddenness of those final warmings (e.g. Waugh and Rong, 2002;Akiyoshi and Zhou, 2007); the 2011 vortex breakup, for example, was very late, but also quite sudden, whereas late final warmings in 2007 and 2008 were more gradual. In contrast to the Arctic, chlorine is typically deactivated well before the Antarctic vortex breakup (e.g., Manney et al., 2005;Santee et al., 2008), but the details and timing of that breakup still have important consequences -not 5 only for local ozone minima over populated areas, but also for dilution of midlatitude ozone (e.g., Ajtić et al., 2004) and for radiative impacts of the Antarctic ozone hole (e.g., Polvani et al., 2011;WMO, 2014). Additional Lagrangian transport and air mass history studies, combined with analyses of Aura data over it's (so far) dozen year mission, will help quantify the fate of activated and ozone depleted air as the polar vortices decay.
In light of the 2012/13 winter, when an exceptionally strong vortex-split SSW resulted in record early winter ozone loss, 10 and the 2014/15 winter, when a very brief, minor SSW resulted in record high vortex ozone values, the importance of the early and abrupt major final SSW in limiting ozone loss in spring 2016 once again emphasizes the complexity of the interactions between these extreme dynamical events and chemical processes in the stratospheric polar vortex. In each of these winters, the SSW events had dramatic consequences that were largely unanticipated. SSW characteristics are also expected to evolve with the changing climate (e.g., Charlton-Perez et al., 2008;McLandress and Shepherd, 2009). We should thus expect the Arctic 15 wintertime meteorology, and its impact on chemical processing, to continue to surprise us in the future, making continued comprehensive monitoring of stratospheric composition a critical priority.