STRATAQ: A three-dimensional Chemical Transport Model of the stratosphere

,


Introduction
Numerical models are commonly used to help quantify chemical and transport processes in the atmosphere.The formulation of these models varies according the type of problems to be addressed.For example, the so-called "0 dimensional" (0-D) or "box" models are used to simulate a great number of chemical reactions at one point in the atmosphere (e.g.Danilin et al., 1994;Kawa et al., 1992;Ramaroson et al., 1992;Redaelli, 1997).Such an approach that is particularly suited for the use along the lagrangian trajectory of one or multiple traveling air masses allows one to concentrate the computational resources into a highly detailed parameterization of chemical and microphysical reactions.Opposite to those "chemistry-intensive" models, multi-dimensional "transport-intensive" models concentrate on following the evolutions of quasi-conservative tracers with a very simple parameterization of chemistry (e.g.Cunnold et al., 1975;Shindell et al., 1998).Such models, usually three-dimensional (3-D) with a different choice of the vertical coordinate, have been successfully used to study transport processes over various time scales.In between those extremes, there are one-dimensional models that study the evolution of a single air column, but usually with a crude chemical parameterization (e.g.Chameides and Cicerone, 1978), and two-dimensional (2-D) models that study the atmosphere averaging over one of the spatial coordinates, usually the longitude (e.g.Brasseur et al., 1990;Ko et al., 1993;Visconti, 1987).This latest approach has been proved to be a good compromise between the two extremes and has been used to study stratospheric chemistry with good detail in both chemical and microphysical parameterizations.However, those models cannot be used either to analyze specific time chemical pictures or to study single measurements, especially when the high non-zonal properties of the circulation play an important role, as in the presence of a highly disturbed polar vortex.Until recently, the development of global 3-D models with extensive chemistry has been inhibited by a lack of computer resources.This is particularly true for those types of 3-D models where the chemistry is included directly into the dynamical code (the so-called GCMs) (Austin et al., 2000(Austin et al., , 2001;;Steil et al., 1998;Hein et al., 2001).Chemical Transport Models (CTMs) are instead 3-D models where the transport is performed "off-line", meaning that those models use meteorological parameters from an external GCM, or assimilated winds and temperature provided by meteorological agencies for the period of interest, to force the chemical calculations.Such an approach is less computationally expensive, because it avoids the internal dynamical calculations and allows one to have the complete 3-D picture of transport, still concentrating on a highly detailed parameterization of the chemistry.Moreover, this approach allows for the comparisons of model output with observations at a chosen time, because it constraints the results to a "real" specific meteorological condition.This is particularly important when trying to study a great number of sparse single measurements, concentrated in a limited time period, such as during a measurement campaign.In this case, a highly detailed chemical parameterization and a good resolution of non-zonal transport are required.This type of model has performed well in reproducing experimental data and day-to-day chemical and dynamical variability of the stratospheric and tropospheric constituents (e.g.see Rood et al., 1992;Lefevre et al., 1994;Chipperfield et al., 1993).
This paper describes a global off-line three-dimensional chemical and transport model developed at the University of L'Aquila.There are several reasons to develop such a model besides the scientific competition.One reason is the involvement of our group in a few international measurement campaigns that require the use of both CTM and lagrangian trajectory modeling (MacKenzie et al., 2000;Carli et al., 2000).Another reason is the growing opportunity to obtain global or synoptic data on atmospheric components from satellites such as ENVISAT.The more attractive development in this case deals with assimilation techniques from which both models and the completeness of the data will benefit.The general belief is that models need to be tested and this is the purpose of this paper: to show how the model may reproduce some of the available measurements of the atmospheric composition.As discussed clearly in Oreskes et al. (1994), the outcome of such a comparison may not be the basis for rejecting or accepting the models.In fact, in the Earth Sciences, the verification or validation of the models poses a serious methodological problem.On the other hand, if the models fail to reproduce the data it may be a sign that they are faulty and need improvement that may come along with future applications.In addition, a few CTMs are involved in assessment works (i.e.predictions for IPCC, etc.) and still some of them have not been subjected to such an exercise.This work has to be read with this limitation in mind, as a tool that can be improved by being devoted mainly to help analyze experimental data.
The paper is organized as follows: a description of the model structure is given in Sect.2; the formulation used to calculate transport and the chemistry scheme are then presented.In Sect.3, the characteristics of the performed experiments are described and the results are compared with a number of observations to assess the model performance.In the final sections the results are discussed and summarized.

STRATAQ chemical transport model
STRATAQ is a three-dimensional off-line stratospheric chemical transport model that extends from 0.6 km to about 52 km in altitude.The model uses a sigma-pressure vertical coordinate and has a vertical resolution varying from about 1-1.5 km until 20 km and up to 2.5 km in the higher stratosphere.The horizontal resolution is 5 • longitude by 2 • latitude.The behaviour of the chemical species is represented by the continuity and transport equation which separately includes the contribution of transport by the large-scale circulation and the tendency due to the chemistry.

Dynamic scheme
In the CTM chemical species are advected with a semilagrangian transport code, using a Van Leer transport scheme (Van Leer, 1974;Allen et al., 1991;Lin et al., 1996) that is a monotonic, upstream-biased differencing scheme.The Lagrangian calculation, characteristic of upstream-biased schemes, makes the method ideal for mass conservation.The major advantage of the scheme is its ability to maintain constituent correlations.In fact, a monotonic grid-point transport scheme applied to 3-D constituent transport does not generate new maxima or minima and, therefore, does not require a filling algorithm.However, the major disadvantage of the scheme is that upstream calculation can be a source of numerical diffusion.The algorithm used makes the diffusion self-limiting and scale-dependent.
The equations are discretized with height on 28 σ levels (see Table 1).The transport scheme uses pure σ vertical coordinates defined as: where p is the pressure and δp = p sf c − p top being p top , the pressure at the top of the model grid box and p sf c the surface pressure.UKMO three-dimensional grid-point analyses are gridded on 18 pressure levels from 1000 hPa to 0.316 hPa on a 2.4 • latitude by 3.75 • longitude global grid.The daily analyses (at 12:00 UTC) are generated using a data assimilation system with data from operational meteorological observations; this system is a development of the scheme used at UKMO for operational weather forecasting, which has been extended to cover the whole stratosphere (Swinbank and O"Neill, 1994).
The vertical wind component is simply obtained in the transport scheme used by integrating the hydrostatic mass continuity equation.

Stratospheric chemistry scheme
Variation in the chemical species and families due to transport and chemistry are treated separately in the model.Each chemical species and each family is first advected and then chemically integrated.
The photochemical package used in the CTM is based on a photochemical box model, developed by Redaelli (1997), to be used for short time integration along lagrangian trajectories (Redaelli et al., 1998).The original 0-D model has been updated with more recent values for chemical constants, introducing a "family" approach and strongly simplifying the heterogeneous calculation to be less computationally expensive.
The chemical package includes 43 chemical species, 89 homogeneous gas-phase chemical reactions, relevant to the stratosphere and troposphere, and 33 photolytic reactions.No additional tropospheric-only related detailed chemistry is included in the scheme.
A family approach has been employed and a number of short-lived species have been combined together into families.In this way it is possible to save computer time and memory.The concentrations of species included in the families are calculated assuming photochemical equilibrium conditions.HO x species (H, OH, HO 2 ) and CH y O x species (CH 3 , CH 3 O 2 , CH 3 O, CH 2 O) are not explicitly integrated but considered in a steady state.In fact, photochemical lifetimes of these species are very short throughout the stratosphere and much shorter than dynamical time scale and so their abundance will not depend directly on transport.N 2 , O 2 , H 2 are assumed to have fixed mixing ratios.H 2 O evolution is roughly parameterized in the tropospheric levels, as a function of thermodynamical parameters.No surface emissions of NO x are considered in the model, while CH 4 , N 2 O and CFCs are constrained at ground and top level according to mixing ratio values used for the initialization.Total inorganic bromine and chlorine are also prescribed in the model using fixed correlation with N 2 O.For the experiment described in Sect.3, a constant total chlorine loading of 3.6 ppbv and a total bromine loading of 20 pptv has been used, following Chipperfield (1999).
To further reduce computational cost, some long-lived species are not chemically integrated but only transported.All the chemical species contained in the model are listed in Table 2, along with a description of the chemical families.
To evaluate photoanalysis rates J we use a "look-up table" calculated by a radiative one-dimensional model based on the δ-Eddington approximation (Joseph et al., 1976).The model includes 26 vertical levels from the ground to about 70 km and takes into account 68 spectral intervals.The radiative scheme takes into account Rayleigh and Mie scattering and includes the Allen-Frederick parameterization of Schumann-Runge bands (Allen and Frederick, 1982).Solar flux values and absorption cross sections for the chemical compounds are taken from DeMore et al. (1997).The resulting threedimensional look-up table for J values (which has coordinates of latitude, pressure altitude, and solar zenith angle with 2 • step) is calculated off-line for clear-sky conditions, as a 10 day average.For the calculations, climatological zonal O 3 data (Pitari et al., 1998) are used.The temperature data are taken from UKMO analyses.Such precomputed J values are then interpolated for each time step at each model grid point.The chemical mechanisms and rate constants are given in Tables 3 and 4. Chemical rates are also calculated using UKMO temperature data.
The model uses the semi-implicit symmetric (SIS) method to integrate the chemical continuity equations (Ramaroson   , 1992).Noted Q i , the photochemical source term for specie i, as: where P i and L i are, respectively, the production and loss terms, and the defined two vectors y and Q whose components are, respectively, the concentrations n i and the photochemical-source terms Q i .In the SIS method, the chemical continuity equation for a species i is developed to obtain: where: y = y(t m+1 )−y(t m ), and J m is the Jacobian matrix: with i = 1, 2, . . .N and j = 1, 2, . . .N being N the number of the species of the photochemical system.The SIS method has the desired properties of an integration scheme: it is accurate, stable and relatively fast compared to other methods for solving stiff systems.The time step used in the CTM for the chemistry is 15 min.Although the chemistry scheme used is stable for longer time steps, such a short time step value is necessary to adequately resolve the diurnal cycle of species.
A treatment of heterogeneous processes, occurring on the surface of stratospheric background aerosols, NAT (Nitric Acid Tetrahydrate) and ICE particles, is included in the model.The scheme does not contain any detailed microphysics, but it does distinguish between those different basic heterogeneous surfaces.In the presence of liquid aerosols and PSCs, the following six heterogeneous reactions are assumed to occur on the particle surfaces: where (s) and (g) stand for solid phase and gaseous phase, respectively.The reaction probabilities γ for these reactions are taken from experimental data as reported in DeMore et al. (1994) and Chipperfield (1999).Temperature and relative humidity-dependent sticking coefficients are calculated from the model H 2 O, T , and HCl according to Hanson and Ravishankara (1994) for reactions (R3) and (R4) and Hanson et al. (1994) for reaction (R5) aver liquid aerosols.The γ used are summarized in Table 5.The rate of chemical reaction is then T is the temperature, [M] is the atmospheric density (cm −3 ) and P is the pressure expressed in hPa.* Three-body reaction rates calculated with:  (1994); * 2 : γ value calculated on-line following Hanson et al. (1994) calculated using the tabulated γ values along with the surface particle area density and the mean kinetic speed of the reacting molecule in a gaseous phase (Rodriguez et al., 1989).Note that Cl 2 and ClNO 2 are assumed to exist only when produced through heterogeneous reactions and then, in the presence of sunlight, photodynamically converted into ClO x and NO x molecules.The particle surface area density is evaluated without introducing any detailed microphysics.The presence of NAT and ICE PSCs is predicted on the basis of simple thermodynamic considerations, as a function of temperature and HNO 3 and H 2 O local partial pressure, respectively.These aerosol particles are typical of polar processes because they may grow at low temperatures (typically below 215 K).NAT are assumed to form when they are thermodynamically possible, by evaluating the HNO 3 saturation pressure calculated using the model temperature and mixing ratios of H 2 O and HNO 3 with the algorithm of Hanson and Mauersberger (1988).Similarly, ICE PSCs are supposed to exist when H 2 O partial pressure is bigger than the saturation pressure of H 2 O over ice, evaluated using the Tetens equation (Murray, 1967).The excess of H 2 O and HNO 3 is removed from the gas phase when saturation occurs and the available surface area density in the PSC region is estimated assuming a radius of 1 and 10 µm for NAT and ICE, respectively, following Danilin et al. (1994).Sulfate aerosols are currently included using a zonal-mean aerosol climatology (WMO, 1991).
The model also includes a simple parameterization for sedimentation of PSC material and for the consequent denitrification and dehydratation effects on the stratosphere.Once they have been formed, NAT and ICE are sedimented with a mean speed of 15 m/day and 1.5 km/day, respectively, following Lefevre et al. (1994).Condensed species are then returned to the gas phase, depending on the ambient temperature, in the lower levels of the model where they have moved, affecting in such a way the vertical distribution of HNO 3 and H 2 O.At this stage HCl removal due to the incorporation into PSC particles has not been taken into account.

Initialization
Results of short time CTM integrations can depend critically on the initialization of the chemical fields.Our model can be initialized using zonally symmetrical tracer fields taken from two-dimensional models.In this case, following Douglass et al. (1990), long-lived tracer fields in the winter lower polar stratosphere are initially constrained to follow the morphology of the polar vortex using potential vorticity and potential temperature coordinate (conservative for adiabatic and frictionless motion) (e.g.see Redaelli et al., 1994, and references therein), while short-lived chemical species are brought into equilibrium with the 3-D initial meteorological fields performing a 2 day CTM integration with a constant temperature field and without transport (Chipperfield et al., 1993).Alternatively, full chemical output from another CTM long-term simulation, performed using analyzed winds and referred to the same period, can be used.Note that such a long-term simulation is not so strictly dependent on initialization conditions.When such fields are available, this can be the best choice, since both the 3-D long-lived and short-lived tracer fields are already constrained to the "real" meteorological situation from which we start our integration.

Model experiment
The STRATAQ CTM was integrated from 2 January until 20 March 1997.The UKMO data set was used for transport calculation, as described in Sect.2.1.The model was initialized using output fields from long-term integration performed with the low resolution version of the 3-D SLIM-CAT CTM (Chipperfield, 1999), relative to 2 January 1997, at 12:00 UTC.Winter 1996-97 has been chosen for its remarkable dynamical situation, along with the availability of sparse chlorine, nitrogen and O 3 data, also showing very low O 3 values in late March/April.The integration started in the period when minimum temperatures in the vortex started to show values below the PSC formation threshold, and continued to cover the ozone hole formation period, up to the dates when record minimum ozone column values where reported by the satellites.
The 1996-97 Northern Hemisphere polar vortex was very strong, as well as cold and symmetric, which is somewhat similar to what is usually found in the Antarctic spring vortex (Coy et al., 1997).Record low temperatures were set for most of March, and the minimum temperature remained below the type I PSC formation temperature from the middle of January through most of March.The later persistence of lower temperature possibly prolonged heterogeneous processing into a period of greater exposure to sunlight than in other years.In addition, the vortex remained substantially intact in the lower stratosphere into early May, exhibiting unprecedent longevity.Due to those unusual features, the winter 1996-97 is an excellent example of the impact of both dynamical and heterogeneous reactions on the chemistry of the stratosphere.
Model outputs are shown here and compared with available global observations by MLS and TOMS instruments for selected days.

UARS MLS data
The microwave Limb Sounder (MLS) on the UARS satellite made a measurement of O 3 , ClO and gas-phase HNO 3 in the Arctic on selected days in December 1996, late January and February 1997 and April 1997 (Santee et al., 1997;Manney et al., 1997).Version 4 of the MLS gridded data obtained from the NASA/GSFC DAAC archive are used here, and stereographic maps relative to the 45 hPa vertical level are shown, compared to the CTM data relative to the 11th model level, corresponding to 44.4 hPa.To examine data in light of the relative dynamical situation, UKMO temperatures at 50 hPa are also shown for the chosen days in Fig. 1.We have overlaid on the temperature maps the 195 and 192 K temperature contours to approximately represent various PSC existence thresholds and also contours of UKMO isentropic PV at 475 K, to indicate the approximate extent of the polar vortex.
Figure 2 shows model outputs for 28 January 1997, 26 days after the model initialization, corresponding to the resuming of MLS observation of northern high-latitudes.By that date, the stratospheric temperature at 475 K had been below 195 K for 25 days and below 192 K for 12 days (Coy et al., 1997), allowing for the formation of PSC particles and heterogeneous activation of chlorine.This is confirmed by the enhanced ClO present in the sunlit portion of the vortex in MLS.CTM local noon data satisfactorily reproduce the satellite ClO day time picture.However, features such as the three peak structure around 90 W, present in ClO, such as in HNO 3 MLS data, are not reproduced in the CTM fields.In addition, CTM reports higher ClO data values if compared to MLS, but this should be primarily related to the use of local noon CTM fields in comparison to MLS data relative to the day side of the orbit.For this reason, CTM ClO local data should be considered as the maximum value expected for the selected day.Uniformly high HNO 3 mixing ratios were observed inside the vortex by MLS, and reproduced by CTM.HNO 3 capture by PSC is not active due to local temperatures that exceeded 195 K, causing any PSC to disappear.
A different situation is shown in Fig. 3 for 24 February, (53 days after the CTMz initialization), where a large circumpolar zone of depleted gas-phase HNO 3 is reported by MLS and well reproduced by CTM.This zone is strongly correlated with the inner part of the vortex with temperature values below 195 K, where the model forms PSCs that capture HNO 3 in solid phases.Significant MLS ClO enhancement extended over a large horizontal area, filling a larger portion of the vortex, as reproduced by the CTM field.
The CTM qualitatively reproduces the deactivation of active chlorine ClO x (Cl+ClO+2Cl 2 O 2 ) and the formation dur- ing the spring period of ClONO 2 removed during the winter period into the vortex region by heterogeneous reactions.
Figure 4 shows CTM modeled fields for 24 February and 20 March 1997.In particluar, the CTM simulates the socalled ClONO 2 "collar" (left column), which is repeatedly found in experimental data (e.g.Blom et al., 1995) and modeled by CTMs (e.g.Chipperfield et al., 1997), principally due to an in situ release of NO 2 (e.g. from HNO 3 ) that causes the gradual photochemical recovery of PSC processed air, leading the active chlorine in the air parcels to decay into ClONO 2 .While no satellite data are available for the chosen period for comparison, the values and structure of the modeled field seem to be consistent with literature data.In particular, the collar evolution is the one typical of the Northern Hemisphere, where the region of high ClONO 2 at the vortex edge tends to extend and fill the whole vortex, with ClONO 2 becoming the major chlorine species in late spring.
A different recovery regime would be expected in the Southern Hemisphere (Douglass et al., 1995), where ClO in the vortex decays directly into HCl, due to stronger denitrification.Figure 4 shows a ClONO 2 collar, with mixing ratios in the range of about 0.5-2.0parts per billion in volume (ppbv), that seems to be bounded on the poleward side by regions not temperatures, but also by of a large-scale dynamicallyinduced reduction linked to an anticyclonic disturbance that developed near the pole in the upper troposphere during March (Lefevre et al., 1998).

Discussion
From the examination of the O 3 data shown, it seems that during the progress of the winter, as the O 3 losses increases, so does the amplitude of the O 3 column difference, with the model underestimating the O 3 column value in the polar vortex in late March.We can suggest no certain cause for this behaviour, but some speculation can be made.Since the low O 3 column values recorded in March are closely linked both to chemical and dynamical effects (Lefevre et al., 1998), the disagreement between the model and the observations during late winter can be due to an inadequacy of the model in correctly reproducing any of these effects.
To quantify chemical ozone loss we can consider the relationship between O 3 and a practically inert tracer, CH 4 (Proffitt et al., 1993;Muller et al., 1997a).Inside the vortex, this relation should not be altered by transport processes due to the limited exchange of air between the vortex and midlatitudes, and should be preserved throughout the lifetime of the vortex in absence of chemical processes.Thus, any deviation from the initial (i.e.corresponding to late November) CH 4 /O 3 relationship over the winter must be due to chemical changes.Figure 6 shows this relationship between CH 4 and O 3 volume mixing ratios in the Northern Hemisphere stratosphere for the winter 1996-1997, as modelled by STRATAQ, by considering grid points inside the vortex and in the potential temperature range of 400-650 K.In comparing CTM data for November 1996 (corresponding to "early" vortex unperturbed conditions and obtained with a 10 day CTM run) with CTM data from March 1997 (corresponding to a "late" vortex period), ozone reductions are evident through all the height ranges considered.In particular, the model simulates chemical destruction resulting in local ozone losses of up to 85% over the potential temperature range of 450-500 K, corresponding approximately to a CH 4 concentration in the range of 0.6-1.0ppmv.To estimate the ozone mixing ratio, which is expected in the absence of the chemical removal of ozone, we use the CH 4 /O 3 relationship in the "early" vortex to obtain a proxy for the ozone mixing ratio O 3 = f (CH 4 ) that would be expected in the vortex if chemical processes were absent (Muller et al., 1997b).Such an O 3 proxy, derived from a polynomial fit to November data, can be compared to CTM O 3 fields in late spring, to identify chemically removed O 3 reproduced by CTM despite variations caused by dynamical processes.The accumulated O 3 reduction between 30 November 1996 and 20 March 1997 is then integrated over the potential temperature range of 400 to 650 K to estimate the depletion in column ozone.The resulting ozone column loss reflects variability inside the vortex, showing chemical ozone losses up to 80-90 dobson units.Such values are about 10% higher than those derived for the same period This result, together with the rather good performance of the model in reproducing, the timing and the extent of chlorine activation due to PSC air processing, as deducted from model to data comparison, even without a very detailed microphysical parameterization, suggests that we should concentrate instead on the possible inadequacy of the model dynamic.For example, the model description of horizontal and vertical transport could be in error.However, the model, forced by meteorological analyses and including an accurate and mass-conserving dynamical scheme, seems to reproduce fairly well the observed horizontal structure of ClO and HNO 3 for the studied cases in strong correlation with the corresponding large-scale dynamical and meteorological situation (i.e. the position and the extent of the temperature minima and of the polar vortex).A significant systematic error in horizontal transport seems to be at least unlikely.An error in the vertical transport could instead explain part of the discrepancy.In particular, diabatic downward transport in the model that is to weak would cause a lower mixing ratio O 3 to be carried from higher altitudes, leading to an underestimation of column ozone values and ozone in the lower stratosphere.Evidence for this could be added by validating the behaviour of the model CH 4 , i.e. treated almost as an inert tracer in the CTM. Figure shows profile comparisons between the CTM and the UARS Halogen Occultation Experiment (HALOE) CH 4 data, for 31 January, 24 February and 20 March 1997, at middle latitudes.Unfortunately, no vertical profiles of CH 4 are available inside the vortex for the studied winter period, due to the UARS orbit (Muller et al., 1997b).There is a clear difference between the various comparisons, with the model first underestimating, at almost every given altitudes, the observed CH 4 profiles at the end of January, while showing reduced differences in late February, and an overestimation in late March, between 5 and 50 hPa.This may indicate a clear systematic tendency of the model to underestimate the vertical downward transport, although other causes cannot be excluded.In order to check this hypothesis better, the model output would need to be systematically compared with a great number of observations of longlived tracers, measured at adequate times and positions, and with good vertical resolution and coverage.No such data are available, at least for the studied winter.However, a general estimate of the vertical transport simulated by CTM can be obtained by examining directly the vertical velocity values as calculated by the model's internal transport code.Average vertical velocity values in the CTM, inside the vortex area on the 50 hPa level, are of about −0.1 cm/s over the January-February period, consistent with data from literature (e.g.see Proffitt et al., 1993).But values for March show a small average upward motion component inside the vortex, which is quite unlikely during such a winter period.This outcome clearly shows that vertical velocity is a crucial factor in our model that deserves further studies.
In addition to dynamical errors, there could be of course other possible causes for the identified data/observation discrepancies, each with different and cumulative contributions to the resulting fields.For example, we may suspect an incorrect estimation of the tropospheric contribution to the O 3 total column, likely to be caused by the poor details in the tropospheric chemistry that can be of particular importance in the peculiar meteorological situation of the late 1997 winter, showing low temperature in the stratosphere, which is frequently linked to high tropopause.Clearly, further work is needed to resolve this issue.

Summary
STRATAQ is a global three-dimensional chemical and transport model of the stratosphere suitable for short-term integration to study transport and chemical evolution related to particular meteorological conditions in a given region of the stratosphere.The model includes detailed stratospheric chemistry of O x , NO y , HO x , ClO y and BrO y , described with 89 homogeneous and 6 heterogeneous chemical reactions, 33 photochemical reactions, and provides the distribution of 43 chemical species.The behaviour of species is represented by the continuity and transport equation which includes the contribution of transport by the large-scale circulation and the tendency due to the chemistry.The continuity equations of the chemical species are solved numerically on a 5 • longitude × 2 • latitude horizontal grid and over 28 σ levels.The transport is calculated using a Van Leer-type scheme in a semi-lagrangian flux form.The chemical scheme is written for chemical families and a semi-implicit symmetric method is used to integrate the chemical equation.A state-of-theart model has been described and the results from the arctic winter 1996/97 simulation have been shown and compared to satellite data.

Fig. 1 .
Fig. 1.False color maps of UKMO temperature at 50 hPa at 12:00 UTC on (from top to bottom): 31 January, 24 February and 20 March 1997.Superimposed in black on the maps are the 195 and 192 K temperature contours to approximately represent various PSC existence thresholds.Superimposed in white are the 2.8 • 10 −4 Km 2 Kg −1 s −1 UKMO isentropic PV contours at 475 K, as representative of the steepest PV gradient, to indicate the approximate edge of the vortex.

Fig. 2 .
Fig. 2. False color maps of HNO 3 (left column) and ClO (right column) mixing ratio at about 45 hPa, on 31 January 1997, from UARS/MLS data (bottom row) and as modelled with STRATAQ CTM (top row).Values are in ppbv.Only data for the dayside of the orbit are shown for MLS, while CTM data are plotted at local noon.

Fig. 7 .
Fig. 7. Relationship between CH 4 and O 3 volume mixing ratios in the Northern Hemisphere stratosphere for the winter 1996-1997 as modelled by STRATAQ CTM.The data encompass potential temperature ranges from 400 K to 650 K, but exclude points outside the vortex.Plus signs indicate CTM data for 30 November 1996, while black dots indicate CTM data for 20 March 1997.The black solid line shows the relation derived from the November data.

Table 1 .
Sigma levels of the model

Table 2 .
Chemical species contained in the model

Table 3 .
Photolytic reactions in the model

Table 4 .
Chemical reaction mechanism and kinetic rates