JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 102, NO. D1, PAGES 1479-1488, JANUARY 20, 1997
Three-dimensional studies of the 1991/1992 northern
hemisphere winter using domain-filling trajectories
with chemistry
E. R. Lutman, J. A. Pyle, M.P. Chipperfield, D. J. Lary, and I.
Kilbane-Dawe
University of Cambridge, Centre for Atmospheric Science, Cambridge, England
J. W. Waters
Jet Propulsion Laboratory, California Institute of Technology, Pasadena
N. Larsen
Danish Meteorological Institute, Copenhagen, Denmark
Abstract. We describe a new and computationally efficient technique for global
three-dimensional modeling of stratospheric chemistry. This technique involves in-
tegrating a photochemical package along a large number of independent trajectories
to produce a Lagrangian view of the atmosphere. Although Lagrangian chemical
modeling with trajectories is an established procedure, this extension of integrating
chemistry along a large number of domain-filling trajectories is a novel technique.
This technique is complementary to three-dimensional Eulerian chemical transport
modeling and avoids spurious mixing caused by low resolutions or diffusive trans-
port schemes in these models. We illustrate the technique by studying the chlorine
activation in the Arctic winter lower stratosphere. A photochemical model was
integrated along large ensembles of calculated trajectories between 20 and 100 mbar
for the 1991/1992 winter in order to produce a three-dimensional chemical picture.
Large amounts of chlorine was activated at low altitudes ($0 to 100 mbar) as well
as altitudes near 50 mbar. This activated air was well contained at all levels, with
little indication of mixing into lower latitudes. Model results for early January
1992 were compared to daily Microwave Limb Sounder (MLS) C10 observations at
465 K. The structure and evolution of the activated chlorine was well reproduced,
giving faith in the technique, although absolute modeled C10 amounts were smaller
than the MLS data. A larger number of domain-filling isentropic trajectories were
also run at 475 K to produce a higher-resolution picture of vortex evolution in late
January 1992. The model successfully reproduced the wave breaking events which
characterized this period causing transport of activated air to lower latitudes.
Introduction
Severe ozone depletion has been observed in the
Antarctic spring as a result of enhanced chlorine radi-
cals produced when polar stratospheric louds (PSCs)
form in the cold polar vortex. A gradual decrease in to-
tal ozone has also been seen in the northern hemisphere
from 1965 to 1986 [World Meteorological Organization
(WMO), 1990]. In midlatitudes, 4 to 8% ozone de-
creases have been observed during the 1980s [Stolarski
et al., 1991]. The climatologies of the northern and
southern hemispheres are significantly different. Arctic
Copyright 1997 by the American Geophysical Union.
Paper number 96JD00698.
0148-0227/97/96JD-00698 $09.00
dynamics are more variable leading to more distortions
of the vortex, and temperatures are generally warmer
than in the Antarctic, leading to fewer PSC occurrences.
Despite these differences it is clear that the Arctic polar
vortex is generally, at some stage of the winter, primed
for ozone depletion. For example, observations from the
Microwave Limb Sounder (MLS) onboard the Upper At-
mosphere Research Satellite (UARS) show substantial
chlorine activation in the 1991/1992 vortex [Waters e!
al., 1993], as large as in the south.
Three dimensional modeling of stratospheric chem-
istry has made dramatic progress in recent years. Off-
line three-dimensional models, forced by meteorologi-
cal analyses, are powerful tools for interpreting a wide
range of atmospheric observations (for example Lefevre
½t al., 1994). However, the high resolution needed by
these models to faithfully reproduce the chemistry and
1479

1480 LUTMAN ET AL.: DOMAIN-FILLING TRAJECTORIES WITH CHEMISTRY
dynamics of, for example, the polar region makes three-
dimensional models costly to run and produces a large
amount of output.
This study demonstrates a new technique which can
be used to model stratospheric chemistry in three di-
mensions and in particular the polar vortex. Large
ensembles of trajectories with chemistry are used to
calculate the three-dimensional evolution of chemical
fields. We illustrate the use of the technique by consid-
ering the evolution of enhanced active chlorine, C10
(= C1 + C10 + 2C1202), during December 1991 and
January 1992. A day-by-day comparison of model C10
with MLS observations in early January is presented. A
period when the vortex became disturbed in late Jan-
uary is investigated by using domain-filling trajectories
to study possible transport of activated air to lower lat-
itudes.
Method
The first step in the procedure is to calculate the
Lagrangian trajectories of a large number of particles
in the region of interest. Then, g.iven a chemical ini-
tialization, the chemical evolution along a single tra-
jectory can be calculated using a chemical box model.
A three-dimensional (height, latitude, longitude) chem-
ical picture on a certain day can also be produced by
integrating a large number of trajectories on particular
levels. Using the ensemble of endpoints, a complemen-
tary picture to other Eulerian global chemical transport
models is obtained.
This method has certain advantages over traditional
Eulerian global models. There is no mixing between
air parcels and the results are not dependent on prob-
lems with diffusion and artificial mixing associated with
the resolution and advection schemes of some global
models. The integrations are also cheaper than global
model integrations, allowing sensitivity studies to be
performed (for example Lutman ½! al. 1994).
Although the idea of modeling stratospheric chem-
istry along trajectories is well established (for exam-
ple, Austin et al. 1987), running multiple trajectories
with chemistry is a new approach to three-dimensional
modeling. This technique is complementary to other
state of the art approaches, such as contour advection
(for example, Waugh ½! al. 1993), or domain-filling tra-
jectories with no chemistry (for example, Fisher et al.
1993).
Trajectory Calculations
The trajectories used in this study were calculated
using a model described by Chipperfield ctal. [1995].
We used European Centre for Medium-Range Weather
Forecasts (ECMWF) analyzed winds with a spectral
resolution of T42. We calculated both isentropic tra-
jectories and full three-dimensional trajectories where
the vertical motion was obtained from the ECMWF
analyses. The three-dimensional trajectories used in
the first section were integrated from November 26,
1991. The domain-filling 475 K isentropic trajecto-
ries used in the second section were integrated from
December 23, 1991. Phot0chemical trajectory calcula-
tions in the stratosphere have not traditionally been run
for longer than about !0 days, usually because of the
isentropic approximation employed in trajectory calcu-
lations. However, Sutton [1994] has shown that the
features revealed by long-duration, three-dimensional
\"domain-filling\" trajectSries, calculated after data as-
similation of meteorological analyses, exhibit a coherent
structure which agrees well with other, independently
derived quantities, for example, potential vorticity.
Photochemical Models
In this paper we have used two different chemical box
models (model A and model B). Model A has a more
detailed chemical scheme while model B is more compu-
tationally efficient, allowing more trajectories to be in-
tegrated. The two photochemical models are described
below.
Chemistry in Model A
The gas phase chemistry is, with the addition of
bromine species, an extension of that used by Lary and
Pyle [1991] and is described thoroughly by Lary [1991].
To save computational expense, the family approach is
used in integrating the chemical tendencies. The chem-
ical families included in the model are O (0(1D) +
O(aP) + Oa), NO (N + NO + NO2 +NOa), ClO (C1
+ C10 + 2C1202), HO (H + OH + HO2), and Br (Br
+ BrO). The model also integrates separately the reser-
voir species N205, HNO3, HO2NO2, H202, C1ONO2,
HC1, HOC1, BrC1, BrONO2, HBr, and OC10 as well
as some source gases (N20, H20 and CH4) and some
CFCs. The gas phase photochemical and kinetic data
are mostly taken from DeMote et al. [1990] unless oth-
erwise stated. The time integration is performed using
a fourth-order Runge-Kutta method with adaptive time
step, after Press et al. [1992]. The model incorporates
a detailed radiation scheme for the calculation of pho-
tolysis rates. A detailed treatment of diffuse radiation
is included. The photolysis scheme is based on the work
of Meier et al. [1982], Nicolet et al. [1982], and Ander-
son [1983], extended to describe the radiation field to
zenith angles up to 97 ø [Lary and Pyle, 1991]. The
temperature dependence of the HNOa photolysis cross
section [Burkholder et al., 1994] is also included.
A detailed microphysical scheme Larsen [1991] is used
to calculate PSC surface areas available for heteroge-
neous reactions based on Toon et al. [1989]. The follow-
ing reactions are available on polar stratospheric clouds
with specified sticking coefficients, 7, for type 1 PSCs
taken from WMO [1990].
(R1) C1ONO2(g) + HCI(s)
(R2) C1ONO2(g) + H2(s)
(R3) N2Os(g) + HCI()
-- C12(g) + HNO3(s)
- HOCI(g) q- HNO3(s)
- C1ONO(g) q- HNO3(s)

LUTMAN ET AL- DOMAIN-FILLING TRAJECTORIES WITH CHEMISTRY 1481
(R4) N2Os(g) + H20(s) - 2HNO3(s)
(R5) HOCI(g) + HCI(s) - C12(g) q- H20(s)
The reactions are treated as being of first order in the
loss of the gaseous molecule. Reaction (R5) is included
[Hanson and Ravishankara, 1991] and given the stick-
ing coefficient suggested by Abbatt and Molina [1992]
for \"H20-rich nitric acid trihydrate (NAT).\" The reac-
tions (R2) and (R4) also occur on sulphate aerosol in
the model with an enhanced surface area outside the
vortex and background values inside the vortex appro-
priate for 1991/1992. A sticking probability of 0.1 for
reaction (R4) is used and the sticking probability for
reaction (R2) is calculated as a function of temperature
and reaches a maximum value of 0.1 at around 195 K
[Hanson and Ravishankara, 1991].
Chemical Initialization for Model A
Ideally, the individual trajectories should each use
a unique chemical initialization appropriate for the
starting location. Model B is initialized individu-
ally from three-dimensional model chemical fields (see
later). However, with model A a simpler approach was
adopted. Model A was initialized as shown in Table
1, according to an in- or out-of-vortex categorization of
the trajectory starting points (judged by the steepest
gradients in potential vorticity, [Braathen et al., 1992]).
The trajectories were then integrated for a long period
to overcome the effects of the crude initialization on the
short-lived species. This approach is particularly suited
to modeling chlorine activation since in the presence of
PSCs the chlorine reservoirs react rapidly and so chem-
ical initialization is not critical.
The values of total inorganic chlorine, Clu, could
be a little high. For example, Schmidt et al. [1994]
calculated, using measurements of the chlorofiuorocar-
bons (CFCs), only a little over 3 ppbv within the lower
stratospheric polar vortex during the European Arctic
Stratospheric Ozone Experiment (EASOE), while Oel-
hal et al. [1994] measured 3.3 ppbv of C1ONO2 deep
inside the vortex in mid-March 1992. This approach,
in which only in- or out-of-vortex cases are considered,
will also oversimplify the structure on any particular
surface. Nonetheless, it should allow us to identify the
impact of processing .by PSCs and aerosol, the main
objective of this study.
Chemistry in Model B
Model B uses the chemistry scheme from the TOM-
CAT three.-dimensional CTM [Chipperfield et al., 1993].
It has a fairly detailed de.scription of the O, NOy,
ClOy, BrOy and HO families as well as longer-lived
tracers (for example, CH4 and N20). PSCs are allowed
to form in the model when the temperature drops be-
low the threshold temperature according to Hanson and
Mauersberger [1988]. Reactions (a2), (a4), and (a5)
occur in the model on both PSCs and a volcanic sul-
phate aerosol distribution. The box model was initial-
ized chemically according to three-dimensional chemical
fields calculated for the starting day by the TOMCAT
model. Model B has the advantage of being more com-
putationally efficient (since it was designed for use in a
three-dimensional model ) and thus allows more trajec-
tories to be integrated in a shorter time than model A.
The main difference between model B and model A is
that model B does not contain a microphysical scheme.
Note that the heterogeneous chemistry schemes in
both models are based around the transformation from
sulphate aerosol to NAT at a critical temperature. Re-
cent ideas, including the possibilities of sulphuric acid
tetrahydrate (SAT) or supercooled ternary solutions
(STS), are not included here.
Chemical Initialization for Model B
Model B was initialized directly from output of the
three-dimensional TOMCAT model for the appropriate
day.
Meteorology of the 1991/1992 Winter
The meteorology of the 1991/1992 (EASOE) winter
has been described in detail by Naujokat et al. [1993]
and Farman et al. [1994]. The vortex formed be-
tween late November and early December at upper lev-
Table 1. Trajectory Initializations at Different Poten-
tial Temperature Levels Relative to the Vortex Edge
(ppbv)
Level 400 K 475 K 550 K
Position In Out In Out In Out
O: 2.4E+3 2.4E+3 2.4E+3 2.4E+3 2.4E+3 2.4E+3
H20 5.0E+3 5.0E+3 5.0E+3 5.0E+3 5.0E+3 5.0E+3
CO 1.8E+l 1.8E+l 1.8E+l 1.8E+l 1.8E+l 1.8E+l
CH 8.0E+l 6.0E+2 5.0E+2 1.0E+3 8.0E+l 8.0E+l
N20 1.5E+l 1.1E+2 9.0E+l 1.9E+2 1.5E+l 1.1E+2
NOy 18.9 13.5 12.14 7.0 18.6 13.5
Clx 0.2 0.2 0.2 0.2 0.2 0.2
C1ONO2 0.4 0.4 0.4 0.4 0.4 0.4
HC1 2.3 2.3 2.7 2.3 2.3 2.3
Cly 3.2 3.1 3.5 3.1 3.1 3.1
Bry 4.0E-3 4.0E-3 8.0E-3 8.0E-3 8.0E-3 8.0E-3

1482 LUTMAN ET AL.: DOMAIN-FILLING TRAJECTORIES WITH CHEMISTRY
els. Temperatures dropped during late December and
early January in the lower stratosphere. Temperatures
were frequently low enough for PSC formation in the
core of the jet stream [Farman et al. 1994], allowing
a large amount of air to be processed heterogeneously.
In this work the extent of the heterogeneous processing
is investigated. A blocking ridge over the Atlantic and
NW Europe affected the structure of the polar vortex
which became distorted in late January with two dis-
tinct lobes of the vortex forming over Greenland and
Europe. PSC processing stopped comparatively early
in the winter due to a strong minor warming in late
January.
Results
Evolution of C10 in Late December and Early
January using Model A
Model A, with its full microphysical scheme, was inte-
grated along 1500 three-dimensional trajectories spread
over the range of 20 to 100 mbar. We investigated the
altitudinal distribution of C10 (C1 + C10 + 2C1202)
from November 26, 1991 to mid-January 1992.
December. No chlorine enhancement was ob-
served in early December. This is in contrast to mea-
surements by MLS [Waters et al., 1993] which observed
moderately enhanced C10 from December 7, 1991, on-
wards with high values notably around December 14,
1991. An increase in chlorine due to PSC processing
was also calculated by Douglass et al. [1993] in a three-
dimensional model simulation of early December. How-
ever, the critical temperature they assumed for PSC
formation was about 2 K warmer than that calculated
by Hanson and Mauersberger [1988]. These tempera-
tures were used to \"compensate for a slight warm bias
in the assimilation temperature fields.\" The Hanson
and Mauersberger [1988] critical temperature may any-
way be too high [Carslaw et al., 1994]. However, in
situ measurements on the ER-2 aircraft [Toohey et al.,
1993] showed only slightly enhanced C10 on December
12, 1991 (with a maximum of 0.4 ppbv at 20 km at
65øN). The absence of enhanced C10 in the trajectory
model integrations suggests either that the heteroge-
neous chemistry scheme is inappropriate; that the tra-
jectories missed the areas of cold temperatures, or that
small variations in temperature causing PSCs to form
were missed by the ECMWF temperature analyses. By
December 14 (not shown) our integrations contained a
few points of enhanced chlorine (0.7 ppbv), although the
majority of trajectories did not show any activation.
By December 27 the situation had changed consid-
erably. Areas of enhanced active chlorine (C1Ox) are
calculated at all altitudes, with most of the high C10
situated between 40 and 80 mbar but a few points over
0.5 ppbv between 80 and 100 mbar. At 40 to 60 mbar
(Plate 1) a large area of high chlorine was situated over
the Arctic and extending over Russia (see figure caption
for an explanation of how the figure is constructed).
Large areas of chlorine over 0.5 ppbv are found with
some points greater than I ppbv.
January. The C1Ox evolution during January is
presented in Plate 2 between 40 and 100 mbar for Jan-
uary 5, 12, and 22. Note that because of descent of
the three-dimensional trajectories over the course of the
run, the resolution of the picture was degraded by Jan-
uary 22 (see figure caption). The highest values of 2
ppbv are produced at 40 to 60 mbar inside the vortex
on January 5. Note that considerable amounts of ac-
tivation were produced at low altitudes (the 80 to 100
mbar level). The area of high activation is well con-
tained at all levels, with little indication of mixing into
lower latitudes even at low altitudes. No evidence is
seen of the activated chlorine observed by Waters et al.
[1993] outside the vortex (Figure 1).
Although the number of trajectories used to construct
Plate 2 is limited, the apparent containment of the vor-
tex at low altitudes is consistent with the contour ad-
vection studies of Norton and Chipperfield [1996]. They
showed that large interannual variability occurs in the
mass of vortex air transported to lower latitudes. Dur-
ing 1991/1992 the vortex was well contained except for
an event at 475 K in late January (see below).
By January 22, 1992, the vortex at 475 K and 550 K
developed into a \"kidney\" shape [Braathen et al., 1992],
with one arm extending over northern Canada and the
other extending over Russia. The highest C1Ox mixing
ratios between 60 and 80 mbar reflect this behavior and
hence are contained within the vortex. Highly activated
air is seen over the Caspian Sea (Plate 2). The distor-
tion of the vortex on January 22 at all levels means there
is highly activated air at low latitudes inside the vor-
tex, where photolysis rates are faster. However, the net
ozone loss integrated along the trajectories on January
22 (not shown) does not show large loss inside the vor-
tex. There is no net ozone loss on January 22 greater
than 15% inside the vortex despite the high levels of
C10. This is because the simultaneous presence of acti-
vated chlorine and exposure to sunlight only occurs for
a short time; there is rapid instantaneous ozone loss in
late January which is halted by the decay of C10.
There is no evidence in these runs of activated air
breaking off the vortex near 180øE around January 20
as indicated by other trajectory runs and other studies
(see higher-resolution trajectory studies performed for
this period below).
Evolution of C10 From January 5 to 13, 1992
The period January 5 to 13, 1992, is investigated in
further detail in this section. The evolution of C10 in
the lower stratosphere calculated from the trajectory in-
tegrations is compared to measurements by MLS [Wa-
ters et al., 1993]. The C10 measured by MLS is pre-
sented at 465 K at approximately local noon (Figure
1). Since C10 has a large diurnal cycle, comparison
between model C10 at 1200 UT and C10 measured at
local noon is unsatisfactory. For this reason the model
C10 is also presented every day at approximately local
noon (Plate 3).
To compare with the version 3 MLS data on the 465
K potential temperature level (Figure 1), modeled C10
is presented in Plate 3 for the range 425 to 500 K (this

1.50
1.40
1.30
1.20
1.10
1.00
0.90
0.80
0.70
0.60
0.50
0.40
0.30
0.20
0.10
Plate 1. C10 (ppbv) on December 27, 1991, at 40 to 60 mbar calculated using model A. The
C10 values are shown at the location of the individual trajectories.
124.
,/ ß
' L
,- .....
, \"k '-' /t7. ' '- / ',. . / [::
I ø-ø /
I 0.0 /
Plate 2, Time and altitude variation of 010, (ppbv) from model A during January 1992 pro-
jected onto hree pressure intervals. Yirst row; P= 40 to 60 mbar. Second row; P= 60 to 80
mbar. Third row; P= 80 to 100 mbar. Days presented are January 5, 12, and 22, 1992. Because
of descent during the course of the run, there are approximately 100 trajectory endpoints per plot
on January 22, 1992, compared to approximately 200 endpoints per p]o on January 12, 1992.
The C]O field is calculated by projecting the C10 taken from the trajectory end points onto a
single ]eve] and interpolating between the end points.

1484 LUTMAN ET AL.: DOMAIN-FILLING TRAJECTORIES WITH CHEMISTRY
0.0 0.5 1.0 1.5 2.0 CIO / ppbv
Figure 1. Microwave Limb Sounder version 3 C10 (ppbv) between January 2 and 13, 1992, at
465 K. (Figure taken from Waters et al. [1993]).

LUTMAN ET AL.- DOMAIN-FILLING TRAJECTORIES WITH CHEMISTRY 1485
i 2.50 2.30 2.10
I .o
I .7o
I .sø
1.10
0.70
0.30
Plate 3. C10 (ppbv) calculated using model A between January 5 and 13, 1992, for 425 to 500
K at local noon. There are approximately 290 trajectory endpoints per plot.
Plate 4. C10 (ppbv) at 475 K for January 16 to 24, 1992, using chemical model B. There are
1280 endpoints per plot.

1486 LUTMAN ET AL.: DOMAIN-FILLING TRAJECTORIES WITH CHEMISTRY
range is used to obtain an adequate number of data
points, approximately 290 points, per figure). On most
of the days shown, the shape of the area of the modeled
maximum C10 is remarkably similar to that of MLS
although the absolute magnitudes are lower. It should
be noted that the magnitude of the version 3 C10 data
shown in Figure 1 needs to be reduced by 8% (J.W. Wa-
ters, personal communication, 1995); see errors discus-
sion below. The structure inside the region of enhanced
C10 is also well reproduced. From January 5 the ar-
eas of maximum model C10 and maximum MLS C10
increase in size. On some days the MLS C10 extends
farther west than the maximum model C10 (although
some enhanced C10 is still present). The reason for this
is uncertain. It may be due to the overall levels of en-
hanced chlorine being smaller in the model than in the
measurements. It may also be due to the error in output
times as described in the errors discussion below. Al-
ternatively, a warm bias in the ECMWF temperatures
could mean the model misses some PSC regions. There
are cold areas west of the activated chlorine, which may
explain why MLS C10 extends farther west than the
calculated C10 (see errors discussion below).
From January 9, 1992, the area of enhanced C10 ob-
served by MLS moved to the east; this is also seen in
the calculated C10. The area of high C10 remains dis-
placed to the east for the remainder of the period of
investigation in both the model and the MLS C10 maps
and the calculated C10 reproduces the structure of the
MLS C10 extremely well during the remainder of the
integration. For example, on January 11 in both cases
the highest values of C10 are found around 20øE. There
is a secondary maximum in both cases from 90øE, ex-
tending across the pole, due to thermal decomposition
of the dimer, and also a small patch of C10 around
270øE in both cases.
The qualitatively good agreement in terms of loca-
tion and evolution of the features in the modeled C10
fields and the MLS data gives confidence that the tech-
nique of domain-filling trajectories with chemistry can
be used to generate three-dimensional chemical fields.
The model also agrees well with the high-resolution (20
x 2 ø) simulations of œefevre t al. [1994] during the same
period. This has been achieved by integrating chem-
istry in effectively only 290 boxes compared with 180x90
needed for a single-level Eulerian 2øx2 ø model. As the
trajectories evolve independently, reducing the number
does not degrade the chemical evolution of those that
remain, unlike reducing the number of grid boxes in an
Eulerian model. Chipperfield et al. [this issue] com-
pare results from domain-filling trajectories with high-
resolution Eulerian model and also find that the trajec-
tory calculations produce consistent features.
Sources of Error Present in MLS and in the
Model
In this section we mention some sources of error
present in the version 3 MLS C10, and sources of er-
ror in the model which will affect this comparison.
There are some problems with the absolute values of
the version 3 MLS C10 data set. As previously men-
rioned, the magnitude of the version 3 MLS C10 shown
in Figure 1 is overestimated and needs to be reduced
by 8%. (J.W. Waters, personal communication, 1995).
The MLS C10 data values may also be a little too high
in areas below the PSC temperature threshold, because
of an incorrect treatment of the HNO3 in these areas.
There is some degree of error present in both the
MLS and the model output times. The local solar times
of the MLS measurements vary with latitude due to
the precession of the satellite's orbit. Some small error
may be present in both the absolute value of the model
C10 and its precise location since the model does not
output C10 at exactly local noon. This is because, for
the large number of trajectories being processed, data
are output by the model at a constant time step (to
limit model output). Thus the C10 concentration is
selected from each trajectory when the end of the time
step falls at some local time between 1200 and 1259
LT. The maximum possible error in the output time is
1/24 day = 0.0417 day. The possible rror in C10 near
local noon due to this maximum variation of 0.0417
day is very small. The variation in longitude may be
somewhat larger, a possible maximum error in longitude
at these high latitudes (around 70øN) is estimated as
between 1 ø and 3 ø . The error in latitude is expected to
be considerably smaller.
So far, our comparison between model and MLS data
has been only qualitative. A quantitative comparison
between model and MLS C10 would be hindered by the
errors mentioned above, and since in this work we are
describing a novel technique, and illustrating it with
MLS data, a quantitative comparison is not included.
Vortex Distortion in Late January 1992 Using
Model B
There is currently much debate concerning the mech-
anisms responsible for midlatitude ozone loss. Possible
causes may include in situ loss due to chlorine activation
on sulphuric acid aerosol, or loss due to a temporary
displacement of the vortex bringing air with enhanced
chlorine over mid latitudes. Alternatively air with en-
hanced chlorine may be expelled from the vortex (the
\"flowing-processor\" theory), particularly during times
of distortion, or from low altitudes where the vortex is
less well contained.
In mid-January the lower stratospheric vortex as
shown by ECMWF analyses was a single, coherent en-
tity, slightly disturbed by planetary waves. By January
20 a ridge over the Northeast Atlantic Ocean, associated
with persistent tropospheric blocking, had distorted the
vortex. The ridge continued to develop and the analy-
ses show the development of two distinct vortex centers.
The ECMWF analyses indicate that the vortex regained
its initial coherence after January 24.
Vortex distortion makes exchange of vortex air with
midlatitudes highly probable. Since exchange of air
between vortex and midlatitudes is one possible cause
of midlatitude ozone loss; this period was examined in
more detail using model B. This model was used since
it is more computationally efficient and allows a higher
number of trajectories to be run, increasing the spatial

LUTMAN ET AL.' DOMAIN-FILLING TRAJECTORIES WITH CHEMISTRY 1487
resolution of the results. The possible expulsion of ac-
tivated air and intrusion of midlatitude air during this
highly unstable period was investigated.
The 1280 isentropic trajectories were run at 475 K in
this study, concentrated in middle and high latitudes.
The forward trajectories were started on a grid from
80øN to 30øN (excluding 70øN) at a resolution of 5 ø
and at a longitude interval of every 2.81 ø. The results
may be compared to Plate 2. Because of the much
larger number of trajectories used in this integration,
the results are at a higher spatial resolution than shown
in Plate 2.
C10 taken from the trajectories is plotted for Jan-
uary 1992 at 475 K (Plate 4). The vortex is full of
activated air, with values of up to 1.4 ppbv. From Jan-
uary 17 at about 150øE, activated C10 is seen outside
the vortex. During the whole period of the integration
the \"blob\" of highly activated air remained near the
edge of the vortex around 180øE. This streamer was
also observed by Waugh et al. [1994] who used high-
resolution contour advection with surgery (CAS). The
blob remained in high latitudes, and the activated chlo-
rine relaxed back to background levels. This blob of ac-
tivated air was not seen in Plate 2. The area of the acti-
vated air inside the vortex began to distort on January
19, 1992, and this distortion increased over the next
few days with one limb extending over Greenland and
Canada and the other limb extending over the European
sector. By January 20 the activated air had distorted
into a kidney shape. The limb over the European sector
extended further on 21 January and on January 22 to
23, low activated air appeared to be entrained into the
vortex between the two limbs. The entrainment feature
was also reported by Pyl½ ½t al. [1994]. This feature
was not seen in Plate 2, our low-resolution run, again
being missed by the fewer (three-dimensional) trajec-
tories. Plumb ½t al. [1994] used CAS to estimate that
the vortex in late January included a few percent of
intruded air following this event.
Conclusions
We have described a new approach to three-
dimensional global chemical modeling consisting of inte-
grating a photochemical box model along a large num-
ber of domain-filling trajectories. This procedure pro-
vides a three-dimensional Lagrangian picture of the at-
mosphere which complements that obtained from a tra-
ditional three-dimensional Eulerian model. Moreover,
this technique avoids some problems associated with
low-resolution Eulerian models concerned with spuri-
ous numerical diffusion and excess mixing between grid
boxes.
We have illustrated the technique of domain-filling
trajectories with chemistry by studying chlorine activa-
tion in the Arctic lower stratosphere. A photochemical
model was integrated along multiple / domain-filling
trajectories during the 1991/1992 northern hemisphere
winter. Results from the model were compared to MLS
C10 observations during early January. The model re-
produced the location and evolution of C10 very well
providing support for the fidelity of the technique.
The results (presented in Plates 2 and 4) show little
indication of the vortex acting as a processor. Consid-
erable active chlorine is produced at low altitudes; how-
ever, it does not appear to be exported to midlatitudes.
When the vortex undergoes a major distortion (Plate
4), modest amounts of air appear to be exported. This
is in agreement with high-resolution dynamical studies
contour advection studies [Waugh et al., 1994].
Despite the length of the trajectories being longer
than that generally considered to be ideal (approxi-
mately 10 days), their behavior was consistent with
meteorological analyses. The density of trajectories re-
quired to reconstruct an adequate picture of the atmo-
sphere depends on the meteorological situation. While
a medium resolution run reproduced the observed MLS
C10 well when the vortex was stable in early January,
during an unstable period, such as late January 1992,
the number of trajectories must be increased to achieve
an adequate picture. In our experiments we increased
the number of trajectories by a factor of 10.
The domain-filling trajectories with the chemistry
method is a novel approach complementary to running
global transport models. In a Lagrangian run the reso-
lution of the chemical picture is dependent only on the
number of trajectories which are run and the interpola-
tion between the endpoints for graphical display. The
high resolution of the chemical picture which may be
obtained by running many trajectories is ideal for ex-
amining highly detailed structures such as vortex dis-
tortions. They are a useful and inexpensive way of mod-
eling the chemical evolution of the winter, allowing sen-
sitivity studies to be performed much more easily than
in three-dimensional models (for example, Lutman ½t
al., 1994).
Acknowledgments. This work was funded by DG XII
of the CEC under contract STEP-CT91-0139 for the sup-
port of EASOE. ERL was generously supported by NERC
and a Gassiot Award from the Meteorological Office. IK-D
was funded by the Human Capital and Mobility Programme (HCMP) proposal ERB4001GT921327. Themodeling work
described here is part of our UK NERC supported UGAMP
effort. ERL would like to thank A.R. MacKenzie for all his
help in the preparation of this paper.
References
Abbatt, J.P. D. and M. J. Molina, The heterogeneous re-
action of HOC1 + HC1 C12 q- H20 on ice and nitric acid
trihydrate: Reaction probabilities and stratospheric m-
plications, Geophys Res. Left., 19, 461-464 1992.
Anderson, D. E., The troposphere to stratosphere adiation
field at twilight: A spherical model, Planet. Space Sci.,
$1, 1517-1523, 1983.
Austin, J, R. C. Pallister, J. A. Pyle, A. F. Tuck, and A.
M. Zavody, Photochemical model comparisons with LIMS
observations in a stratospheric trajectory coordinate sys-
tem, Q. J. R. Meteorol. Soc., 1I$, 361-392, 1987.

1488 LUTMAN ET AL.' DOMAIN-FILLING TRAJECTORIES WITH CHEMISTRY
Braathen , G. O., F. Stordal, T. Gunstrom, B. Knudsen, and
K. Kløster, EASOE meteorology report, Norw. Inst. for
Air Res., Kjeller, Norway, 1992.
Burkhold%, J. B., R. K. Talukdar, A. R. Ravishankara, and
S. Solomon, Temperature dependence of te HNOa UV
aborptipn cross sections, J. Geophys Res., 98, 22,937-
22,948 1994.
Carslaw, K. S., B. P. Luo , S. L. C!egõ , T. Peter, P. Brim-
hiecombe, and P. J. C.rutzen, Stratospheric aerosol growth
and HNOa gas-phase depletion from coupled HNOs and
ware.r-uptake by liquid particles, Geophys. Res. Lett., 21,
2479-'2482, 1994.
Chipperfield, M.P., D. Carlolie, P. Simon, R. Ramaroson,
and D. J. Lary, A three-dimensional modeling study of
tracespecies in the Arctic lower stratosphere during win-
ter 1989-1990, J. Geophys. Res., 98, 7!99-7218, 1993.
Chipperfield, M.P., J. Kettleborough, and A. Pardaens, The
TOPCAT offline trajectory model, internal report of the
UK universities Global Atmos. Model. Proj., Cambridge,
UK, 1995.
Chipperfield, M. P, E. R. Lutman, J. Kettleborough, J. A.
Pyle and A. E. Roche, Model studies of chlorine deactiva.-
tipn and formation of C1ONO2 colla.r in the Arctic polar
vortex, J. Geophys Res., this iss.ue.
DeMote, W. B., M. J. Molina, S. P. Sander, D. M. Golden,
\" R. F. Hamson, M. J. Kurylo, C. J. Howard, and A. R.
Ravishankara, Chemical kinetics and photochemical data
for use in stratospheric modehug, in Evaluation 9, NASA
JPL Publ. 90-1., 1990.
Douglass, A., R. Rood, J. Waters, L. Froidevaux, W. Read,
L. Elson, M. Geller, Y. Chi, M. Cerniglia, and S. Steen-
rod, A three-dimensional simulation of the early winter
distribution of reactive chlorine in the north polar vortex.
Geophys. Res. Lett., 20, 1271-1274, 1993.
Farman, J.C., A. O'Neill, and R. Swinbank, The dynamics
of the Arctic polar vortex during the EASOE campaign,
Geophys Res. Lett., 13, 1195-I19'8, 1994.
Fisher, M., A. O'Neill, and R. Sutton, Rapid descent of
mesospheric air into the stratospheric polar vortex, Geo-
phys. Res. Lett., 20, 1267-1270, 1.993.
Hanson, D. R., and K. Mauersberger, Vapour-pressures
Of HNOs/H20 solutions at low temperatures, J. Phys.
Chem., 92, 6167-6170, 1988.
Hanson, D. R., and A. R. Ravishankara, The reaction prob-
abihties of C1ONO2 and NaOs in 40 to 75% sulphuric acid
solutions, J._Geophys. Res., 96, 1.7,307-17,314, 1991.
Larsen, N., Polar stratospheric clouds: A microphysical sim-
ulation model, Dan. Meteorol. Inst. Sci. report, 1991-92,
Copenhagen, 1991.
Lary, D. J., Photochemical studies with a three-dimensional
model of the atmosphere, Ph.D. thesis, University of
Cambridge, Cambridge, 1991.
Lary, D. J., and J. A. Pyle, Diffuse radiation, twilight and
photochemistry I, J. Atmos. Chem., 13, 373-392, 1991.
Lefevre, F., G. P. Brasseur, !. Folkins, A. K. Smith, and P.
Simon, Chemistry of the 1991-1992 stratospher.ic winter:
Three-dimensional model simulations, J. Geophys. Res.,
99, 8183-8195, 1994.
Lutman, E. R., J. A. pyle, D. J. Lary, A. R. MacKenzie, I.
Kilbane-Dawe, R. L. Jones, N. Larsen, and B. Knudsen,
Trajectory model studies of C10 activation and ozone
loss during the !991/92 northern hemispheric winter, Geo-
phys Res. Lett., !3, 1419-1422, 1994.
Meier, R.R., D.E. Anderson, and M. Nicolet, Radiation field
in the troposphere and stratosphere from 240 to 1000nm,
Gen. Anal. Planet. Space. Sci., 30, 923, 1982.
Naujokat,'B., et al., The stratospheric winter 1991/92, Beil.
Berl. Wetterkarte, SO 18/92, 1993.
Nicolet, M., R. R. Meir, and D. E. Anderson, The radia-
tion field in the troposphere and stratosphere from 240-
100rim' Numerical analysis, Planet. Space Sci., 30, 935-
983, 1982.
Norton, W. A., and M.P. Chipperfield, Quantification f the
transport of chemically activated air from the northern
hemisphere polar vortex, J. Geophys. Res., 100, 25,817,
1996.
Oelhaf, H., T. v. Clarmann, H. Fischer, F. Friedl-Vallon,
C.. Fritzsche, A. Linden, C. Piesche, M. Seefeldner, and
W. Volker, Stratospheric C1ONO and HNO3 profiles ob-
tained inside the Arctic vortex from MIPAS-B limb emis-
sion spectra obtained during EASOE, Geophys. Res. Lett.,
œ1, 1263-1266, 1994.
Plumb, R. A., D. W. Waugh, R. J. Atkinson, P. A. Newman,
L. R. Lait, M. R. Schoeber.1, E. V. Browell, A. J. Simmons,
and M. Lowenstein, Intrusions into the lower stratospheric
Arctic vortex du. ring the winter of 1991-1992, J. Geophys
Res., 99, 10.89-1105, 1994.
Press, W. H., S. A. Teukolsky, W. T. Vetterling, and B.
P. Flann.ery, Numerical recipes in FORTRAN- The art
of scientific computing, 2nd ed., Cambridge Univ. Press,
New York, 1992.
Pyle, J. A., G. D. Carver, and U. Schmidt, Some case stud-
ies of chlorine activation during the EASOE campaign,
Geophys Res. Lett:, 13, 1431-1434, 1994.
Schmidt, U., R. Bauer, A. Engel, R. Botchers, and J. Lee,
The variation of available chlorine, ClOy, in the Arctic
polar vortex derived from regular profile observations of
CClaFa (CFC 12) during EASOE, Geophys Res. Lett., 13,
1215-1218, 1994.
Stolarski, R. S., D. Bloomfield., R. D. McPeters, and J.
R. Herman, Total ozone trends deduced from Nimbus 7
TOMS data, Geophys Res. Lett., 18, 1015-1018, 1991.
Sutton, R., La. grangian flow in the middle atmosphere, Q.
J. R. Meteoro!. Soc., 1œ0, 1299-1321, 1994.
Toohey, D. W., L. M. Availone, L. R. Lait, P. A. New-
man, M. R. Schoeberl, D. W. Fahey, E. L. Woodbridge,
and J. G. Anderson, The seasonal evolution of reactive
chlorine in the northern hemisphere stratosphere, Science,
œ61, 1134-1135, 1993.
Toon, O. B., R. P. Turco, J. Jordan, J. Goodma. n, and G.
Ferry, Physical processes in polar stratospheric e clouds,
J. Geophys. Res., 94, 11,359-11,380, 1989.
Waters, J. W., L. Froidevaux, W. G. Read, G. L. Manney,
L. S. Elson, D. A. Flower, R. F. Jarnot, and R..S. Har-
wood, Stratospheric C10 and ozone from the Microwave
Limb Sounder on the Upper Atmosphere Research Satel-
lite, Nature, 36œ, 597-602, 1993.
Waugh, D. W., Contour surgery simulations of a forced polar
vortex, J. Atmos. Sci., 50, 714-730, 1993.
Waugh, D. W., et al., Transport out of the lower strato-
spheric arctic vortex by Rossby-wave breaking, Geophys.
Res,, 99, 1071-1088, 1994.
World Meteorological Organization (WMO), Scientific as-
sessment of stratospheric ozone: 1990, Global Ozone Res.
and Monit. Project I/VMO Rep. œ1, Geneva, 1990.
M.P. Chipperfield, I. Kilbane-Dawe, D. J. Lary, E. R.
Lutman, and J. A. Pyle, University of Cambridge, Cbntre for
Atmospheric. Science, Department of Chemistry, Lensfield
Road, Cambridge, CB2 1EW, UK
N. Larsen, Danish Meteorological Institute, Lyngbyvej
100, DK-2100 Copenhagen O, Denmark. ß
J. W. W.aters, Jet Propulsion Laboratory, California In-
stitute of Technology, Pasadena, CA 91109
(Received September 2, 1995; revised February 2!, 1996;
accepted' September 21, 1996.)