Introduction
Understanding the impact of dynamical processes such as sudden
stratospheric warmings (SSWs) e.g., on Arctic ozone is
key to interpreting the observed interannual variability and better
quantifying
polar ozone evolution . The stratospheric
circulation distributes ozone far from its photochemical production region in
the tropics e.g.,. The global
distribution of ozone is largely controlled by a balance between advection by
the stratospheric overturning circulation, rapid isentropic stirring and
mixing that follows planetary Rossby wave breaking, and chemical sources and
sinks e.g.,. A recent example
of this seasonal balance was provided by , showing that
reflective or absorptive wave conditions in the winter stratosphere result in
lower or higher seasonal concentrations of Arctic ozone through adjustments
in transport and chemical reaction rates. The extreme and transient nature of
the dynamical forcing that triggers SSWs alters this balance. Driven by an
abrupt growth of wave activity e.g.,, SSWs produce global changes in the middle
atmospheric circulation that affect polar temperatures – and thus ozone
depletion – and impact tracer transport and mixing
. In a recent study, used reanalysis
data and model output from the Whole Atmosphere Community Climate Model
(WACCM) to provide a composite view of the changes in transport and mixing
properties of the flow during the life cycle of SSWs. They found that after
the onset of SSWs, the residual circulation remains weak as a result of
suppressed wave driving, but enhanced mixing nonetheless persists in the
lower stratosphere for over 2 months. This study also showed a clear
temporal offset between wave forcing and mixing; zonal-mean eddy fluxes of
potential vorticity (PV) decay shortly after the SSW onset, while diffusive
PV fluxes (in equivalent latitudes) remain active several weeks after.
The study found these anomalies in transport and mixing to be stronger
and more persistent for those warming events that occur during a polar-night
jet oscillation (PJO) event . Note that the
notion of PJO events in the present paper is similar to that in
and in the sense that PJO
events are associated explicitly with SSWs (i.e., sufficiently deep SSWs).
This differs from the perspective of , who saw the PJO as a
low-frequency stratospheric mode of variability that sometimes phase locks
with SSWs.
Case studies for several Arctic winters based on the combined use of
observations and Lagrangian transport models highlight the wide range of
inter-event variability and the sensitivity of polar chemical processing to
the different dynamical conditions. used a Lagrangian
transport model and ozone data from the Aura Microwave Limb Sounder (MLS) to
estimate chemical ozone depletion in the polar vortex for the 1991/92 through
1997/98 boreal winters. They found large interannual variability in the
timing and spatial patterns of ozone depletion due to variability in the
position of the vortex and dynamical processes. combined
satellite observations of ozone from POAM (Polar Ozone and Aerosol
Measurement III) and MIPAS (Michelson Interferometer for Passive Atmospheric
Sounding) with simulations of the Chemical Lagrangian Model of the
Stratosphere (CLaMS) to study the 2002/03 Arctic winter. They found that
strong wave events associated with the 2003 SSW may have increased tracer
transport and enhanced chemical ozone destruction in the polar vortex and its
surroundings. The SSW in January 2009 was one of the strongest on record
e.g.,. Both and
highlight the enhanced isentropic mixing of trace gases
across the vortex edge after the onset of the 2009 SSW, in agreement with the
composite analysis of . Another example of the impact
of a sudden warming on Arctic ozone is the 2015/2016 Northern Hemisphere
winter, which was one of the coldest in the polar stratosphere in recent
years. Intense ozone loss developed in February 2016 favored by the low
Arctic temperatures, but was abruptly terminated by a sudden warming in early
March that became one of the earliest final warmings on record
. used reanalysis
meteorological fields to integrate the Global Modeling Initiative (GMI)
chemistry transport model for the 2005–2015 boreal winters and estimated
that winters with SSWs before mid-February have about one-third the depletion of
winters without SSWs. However, the cold, undisturbed vortex conditions of
December 2012, and the subsequent vortex split of early January that produced
unusually long-lasting offspring vortices subject to high sunlight exposure,
led to exceptionally high ozone depletion in January 2013
.
The present study aims to provide a quantitative evaluation of the changes in
Arctic ozone induced during SSW events, using the Whole Atmosphere Community
Climate Model (WACCM). The use of output from a free-running state-of-the-art
chemistry–climate model facilitates the evaluation of the separate
contributions of transport–mixing and chemical processes to the ozone
variations during the dynamically active boreal winter stratosphere, which is
always a challenge for observational studies . The
240 years of WACCM output will also provide the statistical robustness that
the relatively short observational record lacks. We will show that the ozone
field in WACCM shares many features with observations of the Microwave Limb
Sounder (MLS) on the Aura satellite, and with reanalysis. The evaluation of the
different terms of the zonal-mean ozone continuity equation in geometric
latitude on isentropic levels, combined with the analysis of irreversible
mixing diagnostics in equivalent latitude, will show that ozone anomalies
during SSWs are mainly controlled by dynamical processes in the middle to lower
stratosphere. In addition, sudden warmings that occur during a PJO event have
stronger dynamically induced ozone anomalies that persist around 1 month
longer than warmings without a PJO event.
The remainder of the paper is organized as follows.
Section describes the model runs, the observational
data, and reanalysis used, and the diagnostics employed.
Section presents and discusses the results, and
Sect. gives the main conclusions.
Data and methods
Model output and data
WACCM is a global chemistry–climate model developed at the National Center
for Atmospheric Research (NCAR) and can be used as the atmospheric module of
the Community Earth System Model (CESM). The version used in this study,
version 4 , has a horizontal resolution of
2.5∘× 1.9∘ longitude–latitude and 66 levels in
the vertical with the top at about 140 km in altitude. A few updates from
include a new chemistry module with revised heterogeneous
chemistry and changes to the orographic
gravity wave parameterizations that significantly reduce the Antarctic cold
pole bias . We use daily averaged fields from four members of
an ensemble of 60-year climate simulations (a total of 240 years; each
ensemble member only differs in slightly different initial conditions of the
atmospheric state) originally designed for the Chemistry–Climate Model
Initiative (CCMI) . These runs are forced
with observed sea surface temperatures and external forcings for the period
1955–2014 (i.e., the CCMI REF-C1 configuration), and the quasi-biennial
oscillation is nudged by relaxing the stratospheric tropical zonal winds
towards observations .
Two observational datasets and one reanalysis product are used for validation
purposes. The ozone mixing ratio from the Stratospheric Water and Ozone Satellite
Homogenized (SWOOSH) dataset for the period 1984–2017 is
used for a seasonal cycle comparison with the model. This monthly-mean merged
ozone product combines observations from a set of satellite instruments
(SAGE-II/III, UARS HALOE, and UARS and Aura MLS) after a homogenization
process to account for inter-satellite biases and to minimize artificial
jumps in the record seefor more information. For
comparisons of the evolution of Arctic ozone during SSWs, we use the daily
averaged zonal-mean ozone mixing ratio from Aura MLSv3, which covers the
period September 2004–July 2012 on a 7.5∘ latitude grid, and daily
output of ozone mixing ratio from the European Interim Reanalysis (ERAI),
produced by the European Center for Medium-Range Weather Forecasts
, for the period 1979–2012 on a
1∘× 1∘ longitude–latitude grid.
determined that the observation minus analysis residuals
of ozone in ERAI are typically within ±5 % in the region of the ozone
mixing ratio maximum at 10 hPa and above, but larger up to around 20 % in
the lower stratosphere.
Diagnostics of ozone transport
We will use daily-mean WACCM output to evaluate the continuity equation of
zonal-mean ozone concentration on isentropic levels
e.g.,:
∂tO‾3=S‾*-a-1v‾*∂ϕO‾3-Q‾*∂θO‾3︸mean advection+σθ‾-1(acosϕ)-1∂ϕ(Mϕcosϕ)+∂θMθ︸eddy transport-σθ‾-1∂tσθ′O3′‾,
where O3 denotes the ozone mixing ratio, S is the net ozone
tendency due to chemistry (chemical production minus loss), (v,Q) are the
meridional and cross-isentropic velocities (Q is the diabatic heating
rate), σθ≡-g-1∂θp is the isentropic
density, a is the Earth radius, ϕ is latitude, θ is potential
temperature, and t is time. The vector M=(0,Mϕ,Mθ)=(0,-(σθv)′O3′‾,-(σθQ)′O3′‾) is the eddy
flux term, whose divergence can be interpreted as an eddy transport term.
Overbars denote the zonal mean, primes denote departures from it, and stars denote
mass-weighted, zonally averaged variables (X‾*=σθ‾-1σθX‾). Note that the
second and third terms in the right-hand side of Eq. ()
represent advection (isentropic and cross-isentropic, respectively) by the
zonal-mean overturning circulation. Also, the last term is usually quite
small and will not be shown, but it has been taken into account to compute
the balances.
The eddy transport term is frequently used as an estimate of the two-way
mixing effect of Rossby wave breaking on tracer concentrations to
distinguish it from the mean advective transport by the residual circulation
e.g.,. However, this eddy transport term, computed as
the divergence of the eddy tracer flux, does not completely separate the
irreversible two-way mixing and it can include a component of reversible
transport see. Furthermore, there can be eddy
transport of chemical species that is irreversible in the absence of wave
breaking. This can occur when the waves are dissipated thermally, or when the
chemical lifetime of a species changes along the wave trajectory i.e.,
chemical eddy transport; see. To explore in more
detail the role of irreversible mixing in ozone tendencies during
SSWs, we evaluate the normalized equivalent length squared (hereafter simply
equivalent length) for ozone :
ΛeqO3(ϕe,t)≡a2|∇θconstO3|2(∂ϕeO3,e)-2,
where × represents the area average between
consecutive tracer contours, and O3,e is the ozone mixing
ratio in equivalent latitude ϕe (EqL) coordinates
. This coordinate system assigns the area A
enclosed by a given tracer contour to a circle of latitude (i.e., the
equivalent latitude) that is the boundary of the polar cap with the same
area, A=2πa2(1-sinϕe). Λeq is
proportional to the effective diffusivity κeff
(Λeq=κeff/κ), with κ being a constant diffusion parameter that depends on
the model's spatial resolution and the hyperdiffusivity scheme employed. The
equivalent length (or the effective diffusivity) quantifies the changes in
microscale diffusion due to the irreversible elongation of tracer contours
mainly caused by large-scale Rossby wave breaking and subsequent stirring. In
EqL coordinates, the continuity equation for ozone will therefore be given
by
∂tO3,e=-(acosϕe)-1∂ϕeFdO3cosϕe+(diabatic and chemical terms),
where FdO3=-a-1κeff∂ϕeO3,e is the horizontal diffusive flux of
ozone in EqL. Estimating the explicit value of FdO3
is challenging since the constant diffusion parameter κ of the model,
which enters in the definition of κeff, is unknown.
However, we will employ the equivalent length
ΛeqO3 instead of the effective diffusivity
κeff to compute FdO3 as they
basically contain the same information. This will not affect our results
since we are interested in the anomalies of this diagnostic during SSWs (and
not in its absolute value). Also, note that there is no horizontal
(isentropic) advection term (either by the mean flow or by the eddies) in
Eq. () since it is embedded in the tracer-based coordinate
system. The only horizontal process involved in the evolution of ozone in EqL
is the first term in the right-hand side, which represents the mixing-induced
ozone tendency.
Methodology
Sudden stratospheric warmings are identified in ERAI and WACCM applying the
widely used criterion of . The day when the zonal-mean
zonal wind at 60∘ N and 10 hPa turns negative is set as the central
warming date, provided that it occurs between November and March (i.e.,
midwinter warmings), the separation from the previous central date is longer
than 20 days, and the wind returns to positive values for at least 10
consecutive days before 30 April. In the 34-year period of ERAI we
identify 23 SSWs (0.68 yr-1), while in the 240 years of WACCM
simulations we have 152 SSWs (0.63 yr-1). In the MLS period
(September 2004–July 2012), we have six events as identified with ERAI.
We classify SSWs depending on whether or not they occur during PJO events.
These events are identified following the procedure of
. Briefly, the PJO classification is carried out in terms of
the first two empirical orthogonal functions (EOFs) of daily-mean
polar-cap-averaged (70∘–90∘ N) temperatures over the
middle-atmospheric column, which both present a vertical dipole structure
. A PJO event is identified when the temperature anomaly
(as projected onto these two EOFs) maximizes at a height of approximately
60 hPa, so long as it is sufficiently strong seefor further
details. Consequently, SSWs that occur during PJO events
will have a strong signal in the lower stratosphere, but note that the
identification criterion does not explicitly consider the persistence of the
anomalies. We find that 70 SSWs occur during PJO events (hereafter PJO SSW)
in WACCM, while 82 are not linked to PJO events (hereafter nPJO SSW).
The methodology followed consists of constructing composites of the fields as
a function of latitude or altitude, centered on the SSW central date. The
daily anomalies are calculated as the difference between the daily value and
the daily climatological average (smoothed with a 10-day running mean). The
statistical significance is assessed applying a two-tailed Student's t test
to compare the composite mean of SSWs and the climatology. We use N-1 degrees
of freedom, N being the number of SSWs included in the composite, and a
confidence level of 99 % (i.e., α=0.01). Each SSW event has been
assumed to be independent to estimate the degrees of freedom.
Climatological seasonal cycle of zonal-mean ozone (ppmv) north of
30∘ N at different pressure surfaces in the lower and
mid-stratosphere for (a, d, g) SWOOSH data,
(b, e, h) ERA-Interim, and (c, f, i) WACCM output.
Results
Annual cycle of ozone in observations, reanalysis, and WACCM
We first compare the Northern Hemisphere seasonal cycle of ozone in SWOOSH
(1980–2017), ERAI (1979–2012), and WACCM (240 years).
Figure shows the seasonal evolution of zonal-mean ozone
mixing ratio at 10, 70, and 100 hPa (or the nearest levels available) for the
three data sources. As expected, in all datasets the latitudinal gradients
have opposite signs in the lower and middle stratosphere
e.g.,: ozone mixing ratio over the Arctic is smaller
than in midlatitudes at 10 hPa, and the opposite is true at 70 and
100 hPa. Also, at 70 and 100 hPa the seasonal cycle is characterized by
maximum values in winter and minimum values in summer (consistent with the
overturning circulation seasonality), while at 10 hPa the minimum values
occur in autumn over the polar cap. WACCM and ERAI present very similar
values at the three levels shown, and there is good agreement with SWOOSH at
10 hPa. In the lower stratosphere (70 and 100 hPa) WACCM and ERAI agree
well with observations, although both the model and reanalysis present mixing
ratios around 10 % larger than SWOOSH in winter over the Arctic, which
matches the findings of .
Climatological seasonal cycle of the different terms in the ozone
continuity equation (Eq. ) averaged over the Arctic
(70–90∘ N) at (a) 850 K, (b) 500 K, and (c) 400 K. WACCM
output.
Figure shows the contribution of each term in
Eq. () to the simulated seasonal cycle of the ozone budget
in WACCM, averaged over 70–90∘ N on the isentropic levels
of 850 K (∼ 10 hPa), 500 K (∼ 60 hPa), and 400 K (∼ 100 hPa). Note the transient eddy term (last term in Eq. ) is
usually very small and not shown here. At 850 K isentropic eddy transport
and net chemical loss nearly balance each other, particularly from February
to May, and vertical advection makes a small contribution in autumn and
winter. As a result, the tendency is a small residual relative to these two
competing effects. The polar middle stratosphere constitutes the transition
layer above which ozone is chemically controlled and below which it is
dynamically controlled e.g.,. This is evident in
Fig. b, c, where the ozone budget terms are
displayed at 500 K and 400 K. At 500 K, chemical destruction is still
relevant in spring and summer, but the shape of the ozone seasonal cycle is
mainly determined by the seasonally varying cross-isentropic advection and
isentropic eddy transport (although the chemical sink in late spring and
early summer delays the ozone minimum to midsummer). Downward motion in
winter increases ozone over the pole, while isentropic eddy transport works
against it, smoothing out the ozone meridional gradients. At 400 K, the
chemical term is practically irrelevant, and the seasonal budget of ozone is
completely controlled by the competing effects of cross-isentropic advection
and isentropic eddy transport. The good agreement in the ozone seasonal cycle
between WACCM and observations, as well as the reproduction of well-known
features in the ozone budget, allows us to explore
the driving mechanisms of ozone changes during the lifetime of SSWs using
WACCM in the next subsections.
Composite evolution around the SSW central day as a function of potential
temperature of ozone concentration anomalies (ppmv) averaged over 70–90∘ N
for (a) MLS, (b) ERAI, and (c) WACCM; (d) similar
composite but averaging over EqL 70–90∘ N in WACCM.
Black dots denote statistically significant anomalies (two-tailed Student's t test,
α=0.01).
Note that the statistical test has not been performed for MLS due to the small sample of
SSWs. The approximate pressure level is indicated on the right axis.
Changes in polar ozone during SSWs
Figure a, b, c show the composite anomalies of
ozone in MLS, ERAI, and WACCM, respectively, averaged over the Arctic
(70–90∘ N) as a function of time lag with respect to the SSW
central date (Fig. d will be discussed later).
The three panels show very similar behavior despite the variety of datasets
and years covered (note that the MLS composite is based on only six events).
The Arctic ozone mixing ratio is enhanced (0.5–0.6 ppmv) at levels at which ozone
decreases poleward (>550 K) and reduced (0.2–0.3 ppmv) at levels at
which
it increases poleward (<550 K). MLS and ERAI show a sharper growth of the
anomalies from lags -5 to 0 days
(Fig. a, b), while in WACCM the growth is more
gradual over the course of the 10 days preceding the central date
(Fig. c). At positive lags there is a slow
“descent” of positive ozone anomalies in the mid-stratosphere, and ozone
returns to pre-warming values faster in the upper than in the lower
stratosphere.
Composite evolution, centered on the SSW central date, of the
anomalies of the different terms in the zonal-mean ozone continuity equation
(Eq. ) (ppbv d-1) as a function of potential temperature,
averaged over 70–90∘ N, for WACCM. Black dots indicate
statistically significant values (two-tailed Student's t test,
α=0.01).
We can now take full advantage of WACCM meteorological fields and investigate
the driving mechanisms of these anomalies during SSWs by evaluating the
different terms in the zonal-mean ozone budget equation
(Eq. ). Figure
shows the anomalies of the most relevant terms of Eq. (),
including the ozone tendency
(Fig. a), the isentropic and
cross-isentropic mean advection
(Fig. c, d, respectively), the
isentropic eddy transport
(Fig. e), and the chemical
production minus loss (Fig. f).
The cross-isentropic eddy transport and eddy transient terms (the last two
terms in Eq. ) are very small and will not be shown. Note
that instead of showing the residual,
Fig. b displays the ozone
tendency (∂tO‾3|i) that results from the
sum of all the terms in the right-hand side of Eq. (), and
it can be compared with the direct calculation of the ozone tendency in
Fig. a. There is a relatively good
agreement between the direct and “indirect” calculations of the ozone
tendency below ∼ 700–800 K
(Fig. a, b). However, some
discrepancies appear at θ>800 K, which are likely due to
uncertainties in the calculation of the eddy transport term (note that
periods of large discrepancy between ∂tO‾3
and ∂tO‾3|i at θ>800 K, such
as at lags 45–60 and 75 days, coincide with
periods of very large anomalies of isentropic eddy transport at those levels
in Fig. e). This in turn can be
due to differences in the numerical formulations between the model transport
scheme and our offline diagnostics. Another source of discrepancy is that in Eq. () we
do not include the effects of numerical
diffusion, and vertical diffusion due to the gravity wave parameterization in
WACCM, which are presumably non-negligible in the middle to upper stratosphere.
Over the 2 weeks prior to the central date (lags
-15–0 days), the isentropic eddy transport leads
off the ozone changes (Fig. e).
This indicates that the initial increase in ozone mixing ratio at negative
lags above and decrease below ∼550 K
(Figs. c
and a) is primarily a consequence
of the growth of planetary waves in the stratosphere that ultimately triggers
the SSW. Other terms of Eq. () make a relatively smaller
albeit significant contribution at this early stage, such as a growing
cross-isentropic advection that increases ozone below 900 K and decreases
ozone above 900 K (Fig. d), negative
anomalies of isentropic mean advection at levels higher than 500 K, and
large negative chemical tendencies above 700 K
(Fig. f) that tend to restore
photochemical equilibrium in response to the dynamically induced ozone
anomalies. In the aftermath of the warming (positive lags), the anomalies of
cross-isentropic advection present a downward-progressing structure
(Fig. d) that leads to a gradual
return to climatological values of ozone below 500 K. Above 500 K in the
middle and upper stratosphere, where wave activity is suppressed in the
aftermath of the warming
e.g.,, reduced
isentropic eddy transport and cross-isentropic advection allow the ozone mixing
ratio (Fig. c) to recover at a much faster rate
(Fig. e), while chemical
tendencies partially counteract these effects.
Composite evolution, centered on the SSW central date and averaged
over equivalent latitudes ϕe=70–90∘ N, of
(a, b, c) anomalies of ozone tendencies (ppbv d-1), and
(a, b, c) the standardized anomalies of mixing-induced ozone
tendencies (see Eq. ). (a, d) All SSW events in
WACCM, (b, e) PJO SSW events, and (c, f) nPJO SSW events.
Black dots indicate statistically significant values (two-tailed Student's t
test, α=0.01).
Extracting the effects of irreversible mixing from the Eulerian-mean eddy
transport term is a challenging task, and no effort will be made to do so here.
Instead, we use the equation for the evolution of ozone in the equivalent
latitude EqL framework (Eq. ), in which the only isentropic
process that modifies ozone is irreversible mixing (see
Sect. ). Figure d shows the
evolution of ozone anomalies averaged over EqL
ϕe=70–90∘ N. The structure of ozone anomalies in EqL
is overall similar to that in geographical coordinates
(Fig. c), but there are details that provide a
somewhat different perspective. The level of maximum positive anomalies in
EqL appears to be located at altitudes higher than 1100 K (while in
geographical coordinates it is located at 700–800 K). But the most
significant difference is that the initial changes in ozone appear around 1
week later in EqL than those over the geographical polar cap at all levels
(compare Fig. c, d). It should be noted that
the average over ϕe=70–90∘ N in EqL encompasses the
interior of the polar vortex, especially at negative lags. Therefore, the
initial ozone increase over the Arctic (in geographical coordinates) above
600 K that starts at a lag of -15 days (Fig. c) does
not happen inside the vortex; otherwise it would have been captured in EqL
coordinates. Additionally, we conclude now that those changes should be a
consequence of reversible isentropic eddy transport in
Fig. e (as opposed to
irreversible mixing) since ozone in EqL (which can only be changed by
nonconservative processes; see Eq. ) does not present
those changes.
To explore this feature in more detail, the left column of
Fig. shows the evolution of the
anomalies of ozone tendency in EqL
(Fig. a) and mixing-induced ozone
tendency (first term in the right-hand side of Eq. )
(Fig. d) averaged over the EqL
ϕe=70–90∘ N during SSWs. Note that the central and
right panels in Fig. will be discussed
in Sect. . The anomalies of the mixing-induced tendency term
have been normalized by the standard deviation at each isentrope since we
cannot compute the absolute value of the diffusive flux of ozone
FdO3 (see Sect. ). A comparison
between Fig. a and d reveals that the
initial changes in ozone in EqL share timing with enhanced irreversible
mixing, which tends to reduce ozone below ∼ 600 K and increase ozone
above ∼ 600 K (Fig. d). The ozone-induced mixing
anomalies persist well after the onset of SSWs, up to a lag of
30 days in the middle and upper stratosphere and up to a
lag of
45 days in the lower stratosphere.
Composite evolution centered on the SSW central date as a function
of EqL of equivalent length ΛeqO3 anomalies
(nondimensional units), for (a, d, g) all SSW events,
(b, e, h) PJO SSW events, and (c, f, i) nPJO SSW events in
WACCM, at 850 K, 500 K, and 400 K as indicated. Black dots indicate
statistically significant values (two-tailed Student's t test,
α=0.01).
Consistent with what was mentioned at the end of the previous paragraph, there
is practically no sign of enhanced mixing at negative lags, confirming the
reversible nature of the Eulerian-mean isentropic eddy transport increase at
negative lags in Fig. e. The
timing and duration of the mixing-induced ozone tendencies are dominated by
the behavior of the anomalies of the equivalent length
ΛeqO3, which are shown in the left column
of Fig. at the 850 K, 450 K, and 400 K isentropes
(the central and right panels in Fig. will be
discussed in Sect. ). The anomalies of
ΛeqO3 during SSWs have a similar latitudinal
structure and evolution to Λeq computed from the PV
field at these levels , which emphasizes that the
evolution of ozone is dominated by the dynamics. Positive anomalies of
ΛeqO3 (enhanced mixing properties) start
at 850 K and a lag of -10 days in the midlatitudes, migrating poleward at positive
lags lasting until a lag of 30 days, and being replaced by a period of weak mixing. As
we move down to lower altitudes the positive anomalies of
ΛeqO3 appear increasingly delayed and
persist for longer than 2 months at 450–400 K.
The study of showed that the response of irreversible mixing to
wave breaking during SSWs is not instantaneous, but extends over several
weeks (as long as 2 months in the lower stratosphere) after the large-scale
wave forcing has decayed. This behavior is reproduced in the comparison of
the zonal-mean isentropic eddy transport of ozone
((acosϕ)-1∂ϕ(Mϕcosϕ)) and the mixing-induced
ozone tendency in EqL
(-(acosϕe)-1∂ϕe(FdO3cosϕe))
in Figs. e
and d, respectively. On the one hand, the
Eulerian-mean eddy transport term increases ozone above and decreases ozone
below ∼ 600 K (i.e., smooths out the horizontal gradients) at negative
lags, and then the anomalies reverse sign tightly following the behavior of
the wave forcing during SSWs e.g.,. On the other
hand, the ozone gradient-smoothing effect of enhanced irreversible mixing
persists long after the wave forcing (and the Eulerian-mean eddy transport)
have declined. Part of this temporal offset may be explained as follows. The
wave forcing (e.g., Eliassen–Palm flux divergence from a transformed Eulerian mean perspective or meridional
eddy PV transport from an Eulerian perspective) distorts PV contours in
geometric coordinates and fluxes PV (and ozone) across latitude circles
(zonal-mean isentropic eddy transport). This is fast, occurs in the week or
two prior to the central dates of the composite, and is reversible (e.g., if
the wave packet propagates through and the contours return to zonal). At this
point the air within the vortex in EqL has not changed so the lack of ozone
anomalies in EqL prior to the central date is consistent
(Fig. d). After the planetary waves break, the
ozone (and PV) contours remain perturbed with smaller-scale motions, giving
rise to slow irreversible mixing. Indeed, the role of nonconservative
processes such as mixing in the aftermath of SSWs is to damp wave activity
and delay the recovery of the vortex, particularly in the lower stratosphere
. Note that this discussion on
reversible versus irreversible transport is based on WACCM results. However,
the fact that ozone anomalies in ERAI and WACCM evolve similarly during SSWs
(Fig. b, c), as well as the resemblance of
ERAI and WACCM dynamics during SSWs , makes us consider
this discussion to be also relevant for ERAI.
As in Fig. , but for PJO SSW and nPJO SSW events in WACCM.
Modulation by PJO events
Recent studies have shown that SSWs that occur during a PJO event (PJO SSW)
experience larger alterations in circulation and temperature than those
warmings that occur without a PJO event (nPJO SSW), particularly in the
recovery phase e.g.,.
Figure shows that PJO and nPJO warmings also
have different signatures in polar ozone. The vertical structure and
evolution of the composite anomalies in PJO and nPJO events, both in polar
cap and equivalent latitude averages (top and bottom panels, respectively, in
Fig. ), are similar to those for all the events
(Fig. c, d). However, the magnitude of the
anomalies is larger (below 500 K ozone anomalies are twice as large), and
their persistence in the aftermath is much longer in PJO than in nPJO SSWs.
The evolution of the different terms in the zonal-mean balance equation
(Eq. ) averaged over the polar cap (70–90∘ N) is
shown in Fig. at the 850 K and 450 K
isentropes. We focus first on the days prior to the onset of the SSW, i.e., at
negative lags. The tendencies of ozone in this period (black line), positive
at 850 K and negative at 450 K, are dominated by the isentropic eddy
transport term (red line). At 850 K the anomalies of isentropic eddy
transport are shorter lived but with higher peak values in nPJO than PJO
events. At 450 K, the eddy transport anomalies are quite similar in PJO and
nPJO events, but the vertical advection (dark blue line) starts to build up
more strongly around a lag of -10 days for PJO than
for nPJO warmings. We focus next on the aftermath of the SSWs, in which the
disparate evolution of ozone between PJO and nPJO events has several
contributors depending on the vertical level. In the mid-stratosphere at
850 K (Fig. a, b), the isentropic eddy
transport (red line) term becomes negative more abruptly in the first days after
the central date in PJO than in nPJO events, transporting more ozone out of
the polar cap. Also, the strong suppression of wave driving in the aftermath
of PJO SSWs in the mid-stratosphere leads to a super-recovery of very cold
polar temperatures e.g.,. The subsequent stronger
positive anomalies of diabatic heating in PJO than in nPJO events
e.g., produce larger negative
vertical advection of ozone (dark blue line) in the former than in the latter
that lasts until a lag of 60 days (Fig. a, b). These stronger dynamical
tendencies result in larger negative ozone anomalies in the 700–900 K layer
starting at a lag of 40 days in PJO than in nPJO
warmings (compare Fig. a, b), which in turn
induces stronger positive chemical tendencies (cyan line in Fig. a, b) trying to restore
ozone to its photochemical equilibrium
(Fig. a, b). In the lower stratosphere at
450 K (Fig. c, d), there are no
significant differences between PJO and nPJO events in terms of intensity and
timescale of the anomalies prior to a lag of 0 days. At
positive lags, the eddy transport anomalies (red line) fluctuate around zero,
and the main difference between PJO and nPJO events is the larger
contribution (around 3 times as large) of vertical advection (dark blue line in Fig. c, d)
to the recovery of ozone values in PJO than in nPJO events.
Composite evolution of the anomalies of the different terms in
Eq. () at 850 K and 450 K, averaged over 70–90∘ N.
(a, c) PJO SSW and (b, d) nPJO SSW. Thick lines indicate
statistically significant values (two-tailed Student's t test,
α=0.01). WACCM output.
Figure e, f show the mixing-induced
tendencies of ozone in EqL coordinates for PJO and nPJO SSWs. Consistent
with the stronger ozone anomalies during the former than during the latter,
the gradient-smoothing effect of mixing is stronger in PJO than in nPJO
events. The central and right panels of Fig.
show that PJO SSWs produce larger and longer-lasting changes in equivalent
length of ozone than nPJO events. Again, the differences between PJO and nPJO
are more pronounced in the lower stratosphere: positive
ΛeqO3 anomalies (i.e., enhanced ozone mixing
properties) are 3 times as large and last over 30 days longer after
PJO SSWs than after nPJO SSWs.
Composite evolution, centered on the SSW central date, of total ozone column
(TOC) anomalies (DU) as a function of latitude. Composite for (a) SSWs in MLS,
(b) SSWs in ERAI, (c) SSWs in WACCM, (d) PJO SSWs in WACCM, and (e) nPJO SSWs in WACCM. Black
dots indicate statistically significant values (two-tailed Student's t test, α=0.01)
(the statistical test has not been performed for MLS data due to the small sample of SSWs).
The impact of PJO and nPJO sudden warmings on ozone concentrations agrees
well with what is expected from the differentiated responses in the advective
overturning circulation and irreversible mixing identified in ERAI and WACCM
by for these two types of warmings. Particularly in
the lower stratosphere, they found that the enhanced mixing and the anomalies
of the vertical component of the overturning circulation were twice as
strong, and lasted 1 month longer in PJO than in nPJO warmings.
presented evidence indicating that this longer
duration of the effects of PJO over nPJO warmings is due to the long
radiative timescales in the lower stratosphere
e.g.,. PJO SSW events are characterized by a deeper
and stronger penetration of the warming into the lower stratosphere, where
radiative relaxation timescales are very slow . Along
with enhanced, long-lasting diffusive flux of PV , they
both work to ensure anomalous circulation and temperature conditions for
longer times after the onset of the event, delaying the recovery of the
vortex.
Response in total ozone column
Several studies have found that interannual variations in winter
extratropical total ozone column (TOC) are well correlated with variations in
planetary wave activity in the lower stratosphere
e.g.,. Planetary wave activity
affects both the mean advection and mixing of ozone, so those correlations
are a simple and useful way of isolating the contribution of dynamics to
interannual variations in TOC. There were also early indications that SSWs
are followed by large increases in polar TOC after SSWs
e.g.,.
To complement the analysis of composite changes of ozone based on ozone
mixing ratios, we have calculated the resulting zonal-mean anomalies of TOC during SSWs using MLS, ERAI, and WACCM
(Fig. , top row).
The three composites in the top row of Fig.
show a significant increase in TOC north of 45∘ N with a maximum
larger than 25 DU (Dobson units) in MLS and 47 DU in ERAI and WACCM starting a few days
before the SSW central date. A TOC reduction larger than 2 DU in MLS and 3.6
DU in ERAI and WACCM appears south of 45∘ N; in MLS the reduction of
TOC is confined to subtropical latitudes. The change of sign of TOC
anomalies at ∼ 45∘ N is approximately coincident with the
climatological position of the maximum latitudinal gradient in WACCM (not
shown) and in observations . As happens with mixing ratio
anomalies, the changes in TOC are present over 40–50 days
(Fig. ) after the SSW onset. In addition, the
bottom row of Fig. shows the corresponding TOC
anomalies for PJO and nPJO SSWs in WACCM. TOC anomalies are stronger and last
longer in the former than in the latter (peak values over the pole around
twice as large, 47 versus 25 DU; note the logarithmic scale), indicating that
PJO SSW events have more profound impacts in total column values through
deeper alterations of the stratospheric circulation and associated transport
and mixing seeand Figs.
and .
noted the fluctuating nature of TOC, locally sensitive to
reversible transport. For instance, the number density increases as air
descending along isentropic surfaces compresses, resulting in higher TOC, and
this descent must be compensated for elsewhere by expansion of air along rising
isentropic surfaces. However, reversible transport is unlikely behind the
long-lasting, north–south dipole pattern in TOC during the life cycle of
SSWs.
Figures , , ,
and all show that cross-isentropic
advection and isentropic irreversible mixing are the main dynamical processes
that change ozone mixing ratios during SSWs, in varying proportions at
different heights and time lags, and which operate at longer timescales than
the driving wave force.
Summary and conclusions
We have used 240 years of CESM/WACCM climate simulations, run with observed
external forcings and boundary conditions for the period 1955–2014, to
quantify variations in Arctic ozone during SSWs and their driving mechanisms.
Composites of vertical profiles of polar cap (70–90∘ N) anomalies
of ozone concentrations on isentropic surfaces during the life cycle of SSWs
show common features in MLS data, ERA-Interim, and WACCM
(Fig. ). Starting a few days before the SSW
onset, there is a higher ozone mixing ratio at levels at which ozone decreases
towards the pole (roughly above 550 K) and a lower ozone mixing ratio where
ozone increases towards the pole (below ∼ 550 K). Enhanced isentropic
eddy transport is the dominant driver of these anomalies during the onset
period. From a zonally averaged perspective in geographical coordinates, the
imbalance between suppressed eddy transport and reinforced cross-isentropic
advection is responsible for the slow recovery of the ozone field in the
aftermath of the warming that lasts over 1.5 months below ∼ 600 K
(Fig. ).
Based on WACCM diagnostics, we have found substantial differences in the
timing when the ozone anomalies appear in geographical and equivalent
latitude (EqL) averages, which highlight the different dynamical processes
involved. In geographical coordinates the initial polar ozone anomalies grow
around 1 week earlier than in EqL, which indicates the reversible
(conservative) nature in the initial changes of the zonal-mean isentropic
eddy transport. Conversely, ozone anomalies in EqL averages
(ϕe=70–90∘ N) appear and disappear at the same time
as anomalies in irreversible isentropic mixing of ozone, as diagnosed with
the equivalent length ΛeqO3
(Eq. ) and derived diffusive fluxes (Eq. ).
Particularly in the lower stratosphere, where radiative timescales are much
longer than in the upper stratosphere, the gradient-smoothing effect of
enhanced isentropic mixing of ozone persists over 2 months after the SSW
onset (Figs.
and ), contributing to the delay of the Arctic
ozone recovery in the aftermath of SSWs. The clear temporal offset between
enhanced eddy transport of ozone, which operates in the SSW onset, and
enhanced irreversible mixing, which operates in the aftermath of the events,
is in good agreement with recent estimates of eddy transport and mixing of
PV during SSWs .
The large sample of SSWs in the WACCM runs (152 in 240 years) allows a
statistically robust evaluation of different types of SSW, namely those that
are classified as PJO events and those that are not
(PJO SSW and nPJO SSW, respectively). These two types of SSWs are
characterized by different evolutions of polar temperature and zonal winds in
the aftermath of the SSW and different intensity and
duration of the anomalous stratospheric transport and mixing properties
. We have found that polar ozone undergoes larger
variations (anomalies up to 50 % as large) that last longer in PJO than in
nPJO events Fig. . While the evolution of
isentropic eddy transport anomalies does not particularly differ between PJO
and nPJO SSWs, irreversible isentropic mixing of ozone and mean
cross-isentropic advection of ozone (nonconservative effects) are stronger
and persist longer in the aftermath of PJO than in nPJO warmings. These are
manifestations of larger and more persistent circulation anomalies in the
former than in the latter .
The reported changes in ozone mixing ratios also affect total column values
(Fig. ). TOC estimates from MLS, ERAI, and WACCM
present reasonable agreement, with high-latitude increases of ∼ 47 DU
peaking a few days after the SSW onset and subtropical decreases of around
3.6 DU (MLS column ozone has slightly weaker anomalies). The dipole structure
of TOC anomalies lasts around 40–50 days after the SSW onset but is more
persistent in PJO SSWs (around 2 months) than during nPJO SSWs (1 month).
The results of the present study contribute to a better interpretation of the
observed interannual variability in Arctic ozone and a better quantification
of its evolution, with particular emphasis on the effects of irreversible
mixing. However, the impacts of SSWs on the ozone field reach tropical
latitudes as suggested in Fig. . The exploration
of tropical ozone variability during SSWs and its interaction with the quasi-biennial oscillation, will be explored in a future study.