Modeling the Frozen-In Anticyclone in the 2005 Arctic 2 Summer Stratosphere

16 Immediately following the breakup of the 2005 Arctic spring stratospheric vortex, a tropical 17 air mass, characterized by low potential vorticity (PV) and high nitrous oxide (N20), was 18 advected poleward and became trapped in the easterly summer polar vortex. This feature, 19 known as a "Frozen-In Anticyclone (FrIAC)", was observed in Earth Observing System 20 (EOS) Aura Microwave Limb Sounder (MLS) data to span the potential temperature range 21 from ~580 to 1100 K (~25 to 40 km altitude) and to persist from late March to late August 22 2005. This study compares MLS N 2 0 observations with simulations from the Global 23 Modeling Initiative (GMI) chemistry and transport model, the GEOS-5/MERRA Replay 24 model, and the VanLeer Icosahedral Triangular Advection isentropic transport model 25 to elucidate the processes involved of is divided into


I Introduction
propagation of these waves (Charney and Drazin, 1961) and the flow becomes nearly zonally symmetric.Hess (1991)  7 The situation cannot be entirely described by kinematic processes, however.Hess and Holton (1985) and Hess (1991) describe the process whereby the tracer anomalies are initially correlated with the winter polar vortex identified by potential vorticity (PV) anomalies, or "vortices".The vortices, which are resistant to weak wind shear, protect the tracer anomalies 11 from stretching into elongated streaks and mixing irreversibly.The vortices eventually decay due to radiative effects, thereby decorrelating PV from the chemical tracer.The chemical 13 tracer anomalies are then advected passively and eventually homogenize with the ambient air due to shear-enhanced mixing.This two-stage process allows tracer structures to survive the 15 vigorous final warming process and the spin-up of the summer vortex.Waugh and Rong (2002) examined the interannual variability of coherent PV structures that remain following the breakup of the Arctic polar vortex, and they found that their longevity depends critically 18 on the timing of the breakup.In early breakup years (February and March), the vortex 19 remnants survive for around two months, while in late breakup years (late April and May), the vortex remnants disappear quickly.In contrast to the protecting influence of vortices, evanescent planetary waves in the lower stratosphere (below -25 km) can enhance tracer structure in the summer (e.g., Hoppel et al., 1999;Wagner and Bowman, 2000).The tracer patterns in this region are not simply "frozen-in", but have a dynamical source.However, in the middle-to-upper stratosphere (above -25 km), where the feature discussed in this paper resides, these waves are not likely to play a large role.hereinafter M06) describe how anomalies of high N20 and low H2O were pulled from the tropics to high latitudes and became embedded in an anticyclone that formed in late March.
5 These chemical tracer anomalies persisted throughout the summer, circling westward around the pole until late August.M06 searched for FrIAC-like signatures in PV fields for other years and found several possibilities in 1982, 1994, 1997, 2002, and 2003.Global maps of long-lived chemical tracer fields are unavailable for verification except in 2003.Indeed, a second FrIAC event was observed in PV and Michelson Interferometer for Passive 10 Atmospheric Sounding (MIPAS) methane data during the summer of 2003 (Lahoz et al., 11 2007).This FrIAC developed during the mid-April 200') final warming and lasted (in the chemical tracer field) until August.Lahoz et al. (2007) examined the FrIAC using polecentered cross-sections along the MIPAS orbit tracks.They identified a W-shaped pattern in 14 the methane field, which was caused by high values at polar latitudes that countered the generally downward and poleward sloping isopleths.Although this pattern was initially established by the poleward advection that resulted in the FrIAC, diabatic processes in the 17 summer may have helped to reinforce the feature.
1 2 Observational Data 2 2.1 Meteorological Data 3 The meteorological dataset used for the dynamical fields in this study is the Goddard Earth 4 Observing System Version 5. 1 0 (GEOS-5) analysis from NASA'-s Global Modeling and 5 Assimilation Office (GMAO), described by Reinecker et al. (2008).GEOS-5 uses the 6 Gridpoint Statistical Analysis method of Wu et al. (2002), a 313-Variational system, and a six-7 hour analysis window.The interface between the observations and the Global Circulation 8 Model (GCM) is performed using the incremental analysis update (IAU) approach (Bloom et 9 al., 1996), which avoids shocking the model, thus producing smoother analyses.GEOS-5 analyses are provided on 72 model levels from the surface to 0.01 hPa (75 km), on a 1/2' latitude by 2/3' longitude grid.The GEOS-5 PV and geopotential height are interpolated vertically to six isentropic surfaces for use in this study (580, 650, 740, 850, 960, and 1100 K).The Modern Era Retrospective-Analysis for Research and Applications (MERRA) reanalysis is also used for the Replay simulation described in Section 2.3.MERRA is a reanalysis from 1979 to the present, which uses the GEOS-5 data assimilation system (Version 5.20) throughout, providing a consistent analysis that includes online bias correction for satellite radiance observations (see http://gmao.gsfc.nasa.gov/merrafor details).8 version of the flux form semi-Lagrangian numerical transport scheme (Lin and Rood, 1996).
9 The meteorological fields in GMI are updated every 3 hours.The "G5Aura" simulation that is presented in this study uses the 'Combo' chemical mechanism.'Combo' combines the stratospheric chemistry described in Douglass et al. (2004) with tropospheric chemistry from the Harvard GEOS-CHEM model (Bey et al., 2001), as discussed in Strahan et al. (2007).
The horizontal resolution is 2.0' latitude by 2.5' longitude, while the vertical resolution in the region of interest for this study (approximately 30-3 hPa) is -1.1-1.5 kin.Replay also uses the Combo chemistry mechanism, but transport is computed using the GEOS-5 advection core, a newer version of the flux form semi-Lagrangian numerical transport than used in GMI.
One noticeable difference that impacts this study is that the GEOS-5 advection core has an improved implementation of polar transport compared with the current GMI transport scheme via use of a smaller polar mixing cap (details are provided in Section 4.3).Replay uses 20 GEOS-51MERRA analyzed meteorology, and also differs from the GMI in that it reads in 21 analyzed fields every 6 hours and then recomputes the anal ysis increments and physical 22 parameterizations every 30 minutes to update the meteorological fields. ?9 I grid uses non-equilateral triangles.Defining the triangle center for non-equilateral triangles is ambiguous.We decided to use the circumcenter, defined as the intersection of the perpendicular bisectors of the sides.This definition allows the center-to-side differences to be easily calculated in determining the gradients.Attempts were made to limit the variation in triangular size by shifting the locations of the nodes upon iteration.The current configuration using 983040 triangles has a mean center-to-side distance of 9996 km, with a minimum distance of 6,147 km (38% smaller) and maximum distance of 11,814 km (18% larger).The 8 nominal resolution is defined as twice the mean center-to-side distance or -20 km.The second difference is that the limiting function used in Putti et al. (1990), the so-called 10 "minmod" function, was replaced with the "superbee" function (Roe, 1985).The superbee 11 limiter provides less numerical diffusion, allowing sharper tracer gradients to be maintained 12 (Sweby, 1984).The VITA code is driven by offline winds (GEOS Version 5. 1 0) at 24-hour increments interpolated linearly in time and space to isentropic surfaces and to the circumcenter of each triangle.We first examine the spin-up of the 2005 FrIAC using PV and geopotential height in order to highlight dynamical processes.Figure 2 shows Northern Hemisphere PV contours at 850 K, overlaid with 10 hPa geopotential height (black lines), for select days during March 2005.On 1 March, the winter polar vortex (identified by high PV) is somewhat elongated and displaced from the pole due to a strong Aleutian high (identified by black "H").There are two additional anticyclonic centers (identified by white "H"s) visible on this day with closed height contours and low PV; one is off the coast of western Africa and the other is over Asia.
The Asian high remains stationary over the next several days, while the African high moves eastward (counter clockwise) and merges with the Asian high, resulting in a strong Asian high on 7 March.During this time, the Aleutian high weakens and the vortex moves back towards the pole.The Asian high travels slowly eastward over the next three days and is approaching the Aleutian high on 10 March.Together with the winter polar vortex and anticyclones, the stratosphere develops a strong wave 2 pattern in the geopotential height on 1 3) March, characteristic of the preconditioning phase of many major sudden warmings (Andrews et al., 1987).
similar to the events of 13-18 March.A large tongue of low PV air stretches across northern Asia on 25 March, indicating significant poleward transport.A portion of this tongue is entrained into the Aleutian high so that by 28 March, the newly merged anticyclone is a coherent entity at high latitudes that is vying for the polar position with the weakening main polar vortex remnants.During this time, the extratropical wave motion is subsiding and the FrIAC becomes a well-established feature in the polar circulation (identified by white "H" on 28 March).April from 580-1100 K (Figure 7), which can be compared with the PV plots in Figure 4.
N20 anomalies exist in the MLS data at all levels, although at 1100 K the data become rather noisy.Neither GMI nor Replay has a coherent positive anomaly at 580 K and 1100 K, but 5 both show coherent anomalies from 650-960 K.The position and extent of the anomalies 6 agree with the MLS data, but in each case the maximum N20 values in GMI and Replay are 7 too low.This low bias is due to differences in background N20 gradients.Figure 8 plots 8 zonal mean N20 from 650 to 960 K, averaged over 1-10 March, the period preceding the 9 poleward flow events that set up the FrIAC.The MLS data show higher N20 near the equator 10 than the simulations, with Replay showing the largest low bias.Since the FrIAC develops 11 from relatively unmixed tropical air, we would expect from these plots that the peak GMI and 12 Replay values will be low.The VITA simulations in Figure 7 that are initialized with MLS observations nicely capture the location and magnitude of the N20 anomalies at all levels, suggesting the spin-up process mainly involves isentropic flow.Complicated swirling structure is observed at 850 and 960 K, highlighting the anticyclonic circulation of the FrIAC.16 The success of the simulations in reproducing many of the fine-scale features in the polar stratosphere attests to the high quality of both the implemented transport scheme as well as 18 the assimilated wind fields from GEOS-5 and MERRA. 1 around the pole.On 15 May, the PV anomaly is very weak, and by 30 May has completely 2 disappeared.The PV anomaly therefore lasted approximately two months after the winter 3 polar vortex breakup in late March, consistent with expectations for a late-March breakup of 4 the northern winter vortex (Waugh and Rong, 2002).
5 Figure 10 presents the MLS, Replay, and VITA N20 evolution at 850 K for the same days as 6 shown in Figure 9.For this phase, VITA was reinitialized on I April 2005 with MLS data.
7 This re-initialization allows a better direct comparison for this period by removing much of 8 the fine-scale structure that was generated in VITA during the polar vortex breakup (see 9 Figure 5).Also, this reinitialization removes any biases that develop due to lack of vertical motion in VITA.The maps for I April show high N20 air in the anticyclonic vortex.From this point on we will continue to refer to the region of high N ' O as the Frozen-In Anticyclone (FrIAC), although as seen in Figure 9, the feature isn't always identified by anticyclonic circulation.From 1-7 April, the FrIAC moves eastward and poleward, with anticyclonic flow that spins off several narrow streamers of high N20 air as produced in the VITA simulation.
From 7-14 April, the FrIAC moves directly across the pole and becomes centered over northern Greenland.This episode provides a useful test of the numerical representation of cross-polar flow, which has posed problems for global models based on regular latitude/longitude grids (Williamson, 2007).Early versions of GMI imposed a well-mixed polar cap in order to dampen effects of noisy assimilated wind fields at the pole (Allen et al.,   1991).Although present meteorological fields are far less noisy than those available to Allen et al. (1991), the GMI the polar cap still extends from 87' to the pole and is 6' 0660 km) across.The Replay transport scheme applies the polar cap only over one grid cell, extending from 89' to the pole, thereby decreasing the size of the cap by a factor of 9, substantially reducing numerical diffusion associated with the larger polar cap.
the pole.Only a very slight drop (-I%) is observed in this run, consistent with the nearly constant peak in the MLS data.In hindsight, these results are to be expected, since implementation with a polar cap of 660 kin is unable to resolve a feature -2000 km across.
4 The Replay polar cap is much smaller relative to the size of the FrIAC and therefore better resolves the feature.In the remainder of the paper we focus on the Replay simulation, since 6 the FrIAC is significantly "washed out" in the GMI simulation after this event.Note that the 7 VITA simulation advects the FrIAC over the pole undiminished as there is no "pole-problem" with the triangular grid.
During this time the easterly summer jet is accelerating (see Figure 1), so that around 14 April 10 the FrIAC starts its westward march around the pole, reaching 180'E longitude by 28 April and back near the Greenwich Meridian on 5 May (Figure 10).Even though the feature is advected westward around the pole, it still exhibits local anticyclonic rotation, with streamers 13 of high N20 air drawn off equatorward (see Replay and VITA results for 28 April).After the second polar crossing on 5-9 May, the FrIAC continues to circle the pole with a rotation period of approximately 10-15 days, maintaining a central position at latitude around 70- 80°N. 1 The anticyclonic phase of the FrIAC is marked by both dynamical (local coherent anticyclone and low PV) and chemical (high N20) signatures.The dynamical properties act to protect the 3 chemical tracer anomaly from the shearing effects of the wind.This is similar to the protective nature of the winter polar vortex remnants that were discussed in Hess and Holton 5 (1985) and Hess (1991).In the case of the 2005 FrIAC the dynamical signature lasted approximately two months, consistent with expected timescales of radiative damping.In the next section we examine the fate of the chemical signature of the FrIAC as it becomes exposed to the shearing effects of the summer vortex.

Modeling the Shearing phase: June-August
The coherent structure of the main body of the FrIAC observed during the anticyclonic phase 11 suggests that the horizontal flow is nearly in solid body rotation (SBR), at least at the high latitudes where the FrIAC develops, in agreement with Piani and Norton (2002).SBR occurs 13 when there is no meridional wind and the zonal wind is proportional to the cosine of latitude 14 (U(9) = Uq cos(g), where U,q is the equatorial wind speed and (P is latitude).In Figure 13, the 15 GEOS-5 zonal mean wind at 850 K is plotted for select days from April through August, 16 along with zonal wind for SBR with different periods of rotation (10, 20, 30, 40, and 50 17 days).In late April the flow is approaching SBR at high latitudes, with period of around 10 18 days (called SBR IO for short), while on May 10, the wind is very close to SBR IO from about 19 75-85'N, similar to the latitudinal extent of the FrIAC.The flow does not follow SBR equatorward of 70'N in late May, consistent with the high-N 2 0 streamer that develops (see 21 Figure 10).From 30 June-20 July, the rotation more closely follows SBR 20 from about 65-90°N.By 10 August, the flow deviates from SBR with slower winds than required at polar 23 latitudes, suggesting that significant horizontal shearing of the FrIAC will occur in August.
As the winds reverse to westerly the SBR rotation completely breaks down, as seen on 25 August.These wind analyses show that quasi-SBR occurs at high latitudes thoughout most of May, June., and July at 850 K, but breaks down completely in mid-to late-August.A high-27 latitude feature in the tracer field would thus be expected to survive the flow over several 28 months, barring effects of vertical shear (examined in more detail below). 19The vertical structure of the FrIAC for MLS and Replay is shown in Figure 14  Further work is necessary to elucidate whether these results are consistent with this anticyclonic vortex in the polar stratosphere.As the PV anomaly decays in late May, however, the FrIAC starts tilting and weakening, as observed on 28 May, with even further tilting occurring in June, described in more detail below.From 15 July to 30 August, MLS shows the FrIAC to be slowly dissipating (Figure 15), 2 mixing with the background N20 levels.M06 examined SLIMCAT and Reverse-Trajectory (RT) simulations of the FrIAC and found that unrealistic shredding of the feature occurred in 4 the simulation, suggesting that the analyzed horizontal winds (U.K. Met Office in their case) are unrealistically dispersive at high latitudes in summer.However, the Replay results show 6 remarkably good agreement with MLS during this third phase, matching the morphology of 7 the N20 contours well through at least 15 August.This suggests that the MERRA winds and 8 Replay transport scheme are able to reliably capture the transport of the summer middle 9 stratosphere.Complete mixing of the FrIAC in Replay does not occur until late August, when 10 the feature has all but disappeared in the MLS data as well.Note that the elevated N20 observed in MLS data on 30 August near 90'E and just off the pole is a new feature that was 12 not formed by the FrIAC, but the feature at 180' longitude is the final observable remnant of the FrIAC, indicating that complete mixing of the FrIAC may not occur before the winter 14 vortex becomes established in September.That the VITA simulation shows considerable structure in late August suggests that it is largely vertical shear (neglected in VITA) rather than horizontal shear that acts to dissipate the FrIAC.We will attempt to quantify the relative importance of horizontal and vertical shearing effects below.
Since it is able to resolve the feature during the polar crossings, the N20 anomaly is higher in 2 mid-May than in GMI.The anomaly remains undiminished throughout May, June, and July, 3 and even shows a similar splitting into two anomalies in late July, as seen in the MLS data.In 4 August, the Replay continues to capture the anomaly until it diffuses into the background by 5 the end of the month.Replay contours of 75 and 100 ppbv are overlaid on the MLS contours 6 in Figure 17 for comparison.As seen by the close correspondence with MLS, Replay 7 performs remarkably well at simulating the evolution and decay of the FrIAC.
8 The VITA simulation is also shown for comparison.As discussed previously, VITA tends to 9 generate complex structures in the tracer field that last longer than in the MLS observations, due to better horizontal resolution.The Hovmbller plot shown here uses a composite of VITA runs that are initialized on I March, I April, and I June, respectively, in order to reduce the build-up of these features.Even so, more detailed structure occurs in the VITA simulation than seen in the MLS data or the other simulations.Particularly during August, the VITA simulation shows considerable structure.This highlights the fact that vertical processes (neglected in VITA) are necessary for complete modelling of the FrIAC.
That Replay is able to simulate the remnants of the FrIAC well into August implies that the implemented vertical and horizontal resolution is adequate to capture the details of the feature during the shearing phase, at least to the resolution of MLS.Vertical shear during the June-August period causes significant tilting of the tracer structures that decreases the vertical scale and enhances vertical mixing as explained by Haynes and Anglade (1997).To quantify this scale reduction, we employ a simple model to estimate the time for the vertical scale of the FrIAC to reduce to the implemented grid scale (-I km for Replay).The N20 cross-sections in Figure 14 suggest that we can approximate the FrIAC as a rectangular tracer anomaly with horizontal length (in the zonal direction) H and vertical depth D embedded in a zonal flow with vertical shear (see schematic in Figure 18).We can calculate the vertical depth D'of the anomaly after the top has completely cleared the base by where T is the time elapsed.Using the approximate vertical shear in zonal wind at 75'N and 10 hPa (-0.4 ms-1 km_' or 0.0004 s-1 in late June/early July) and horizontal scale -2000 km we obtain a time-scale of -60 days to reduce the FrIAC to -I km.This suggests that GMI should resolve the vertical features of the FrIAC for around two months during the shearing phase, consistent with results presented here.Similar arguments can be used to estimate the timescale for horizontal shear to reduce the lateral scale of the FriAC to that of the GMI resolution (2').Using the lateral shear in the zonal wind (-0.5' longitude/day/ Olatitude in June and July) 2 along with a zonal width of 100' we estimate that the FrIAC will be resolvable by GMI for -100 days.That Replay simulates the horizontal structure of the FrIAC throughout the June-4 August shearing phase is consistent with these rou gh estimates.t^' 5 The FrIAC provides an excellent case study for examining chemical tracer evolution in weak 6 shear flows.Given its coherent nature, it is easy to discriminate air within the FrIAC over the 7 course of five months.The Replay simulates the structure of the FrIAC in N20 over its entire lifecycle, at least to the resolution of MLS, attesting both to the assimilated wind fields, the 9 numerical transport, and the simulated N20 destruction.

Summary and Conclusions
The general process whereby complicated tracer structures are frozen into the relatively quiescent summer easterly stratospheric flow has been understood for some time.
used general circulation model (GCM) simulations of the final warming of 1979 and Nimbus 7 Limb infrared Monitor of the Stratosphere (LIMB) ozone observations to show gradual homogenization of long-lived chemical tracers with the 5 background over the course of several months.Similar results were obtained with SLIMCAT simulations Of N20 in the 1998 Arctic spring/summer by Orsolini (2001).
I i.e., long-lived cyclones originating from high latitudes.The FrIAC was first identified in Aura Microwave Limb Sounder (MLS) data from March-August 2005.Manney et al. (2006,

I 5
observed by MLS (polewards of 82'N) are calculated by interpolating over the polar region 2 from neighboring grids.This is necessary for initializing VITA with complete global maps.Results are presented from two different three-dimensional simulations based on the Global 6 Modeling Initiative (GMI) Chemistry and Transport Model (hereinafter, GMI) and the GEOS-7 5/MERRA Replay model (hereinafter, Replay).The GMI advection core uses a modified zonal wind evolution at all levels throughout this height range is very similar to that shown 2 here in terms of the timing of the wind reversals.3Thevertical lines drawn on Figure I relate to the three phases of the FrIAC described in the 4 subsequent sections.The first phase involves the spin-up of the FrIAC during March and early 5 April.The second phase refers to the period where the FrIAC can be identified by a coherent 6 anticyclone (early April-late May).The third phase (June-August) marks the time when the 7 chemical tracer signature of the FrIAC is gradually sheared by the background wind.

Figure 3
Figure 3 illustrates the vertical extent of the FrIAC a week later, on 4 April, using PV maps on multiple isentropic surfaces.Coherent PV anomalies are visible from 580 K to 1100 K (25-40 km), with correlated closed height contours indicating anticyclonic circulation.At 1300 K there appears to be more noise in the PV data and only a weak indication of a low PV anomaly associated with one closed height contour.The FrIAC is observed to be vertically upright at this stage with a horizontal width of approximately 20 degrees latitude (2000 km) across.So at this point, the FrIAC can be described as a coherent vortex -15 km high and -2000 km wide embedded in the stratospheric flow.Later we will show that the FrIAC remains vertically upright throughout April and most of May.
, 'which plots the zonal N20 anomaly at 74'N as a function of longitude and pressure for select days in May and June 2005.On I and 22 May the FrIAC (marked by high N20) is upright, whereas the 1 background N20 contours show significant vertical tilting (westward with height), 2 particularly in Replay.Dynamical studies show that coherent vortices can exhibit resistance to 3 the tilting effects of weak vertical shear.Vandermeirsh and Morel (2002), using a 2 V2 layer 4 quasigeostrophic model, with separate PV anomalies in each layer, show that the PV 5 anomalies have a self-sustaining advective effect on the other layer that keeps the vortex from 6 splitting in weak shear zones.A more detailed study (Jones, 1995), using primitive-equation 7 numerical modeling of a tropical tropospheric cyclone in vertical shear, shows that upper and 8 lower PV anomalies of an initially barotropic vortex rotate about a common center.The 9 effects of this rotation act to oppose the destructive action of the vertical shear on the vortex.

Figure 15
Figure15shows the evolution of MLS, Replay, and VITA N20 at 850 K for select days during June-August 2005.VITA was again reinitialized for this period using MLS data on I June.In early June, the FrIAC exhibits evidence of horizontal shearing.Whereas on I June the main cell of the FrIAC is nearly circular in shape and centered on the Greenwich Meridian near 80°N, by 10 June the feature has elongated in the zonal direction due to weak meridional 19 shear of the zonal wind.Note the long tail that lags behind the main cell of the FrIAC at lower latitudes on 10 June, evident in MLS data and the simulations.By 20 June, the main

Figure 1 .Figure 3 .
Figure 1.(a) Zonal mean zonal wind at 850 K potential temperature for March-August 2005.(b) Zonal mean zonal wind at 850 K potential temperature and 60'N latitude for March-August 2005.Vertical dotted lines indicate transitions between the three phases of the FrIAC (see text for details).

Prow 2 Figure 5 .Figure 6 . 7 .
Figure 6.MLS and Replay curtain plots (along-track cross-section as function of pressure and profile number) for the two orbits shown on Figure 5.
Zonal mean N 2 0 mixing ratios averaged from I -10 March 2005 calculated from MLS observations and GMI and Replay simulations.