Global Impact of the Antarctic Ozone Hole' Dynamical Dilution With a Three-Dimensional Chemical Transport Model

A study of the Antarctic ozone hole has been made with a three-dimensional chemical transport model (CTM) using a linearized photochemistry for ozone. The tracer model uses the winds and convection from the Goddard Institute for Space Studies general circulation model (8 ø x 10 ø x 23 layers). The general circulation model (GCM) develops an Antarctic circumpolar vortex in early winter with strong westerlies that reverse in austral spring; the circulation compares favorably with the observed climatologies. A 4-year control run of the CTM with annually repeating winds produces ozone distributions that compare reasonably with the observed climatology. We examine different linearizations of the ozone chemistry and show that the calculated column ozone is sensitive to the chemical time constants in the lower stratosphere. ha separate numerical experiments a hypothetical Antarctic ozone hole is induced on September 1 and on October 1; the CTM is integrated for 1 year with a linearized model that assumes standard photochemistry, not including the heterogeneous reactions and unusual chemistry associated with formation of the ozone hole. The initial depletion, assumed to be 90% of the 03 poleward of 70øS between 22 and 200 mbar, amounts to about 5% of the total 03 in the southern hemisphere. As the vortex breaks down and the ozone hole is dispersed, significant depletions to column ozone, of order 10 Dobson units (3%) occur as far north as 40øS during austral summer. One year later, only 30% of the original depletion remains, mostly below 100 mbar and poleward of 30øS. The October 1 initialization is continued for a second year, the ozone hole being reinduced 1 year later with the same parameterization. The cumulative effects from the year before are noticeable but add only 20% to the depletion. A budget analysis for the southern high-latitude stratosphere (10-350 mbar x 31ø-90øS) indicates the ozone hole is replenished equally by photochemical regeneration and by reduced transport of ozone into the troposphere, with a lesser fraction being filled in by an increased flux from the tropical stratosphere. altitudes and to mid-latitudes Ozone stratospheric chemistry and transport into the troposphere deficits of order 6% are found in the stratosphere between 50 where atmospheric and surface reactions destroy ozone. The and 200 mbar as far north as 40øS; deficits of order 4% reach time constant to replace ozone in this stratospheric box is 1.2 30øS; and part of the original depletion is seen in the years.

1 initialization is continued for a second year, the ozone hole being reinduced 1 year later with the same parameterization. The cumulative effects from the year before are noticeable but add only 20% to the depletion. A budget analysis for the southern high-latitude stratosphere (10-350 mbar x 31ø-90øS) indicates the ozone hole is replenished equally by photochemical regeneration and by reduced transport of ozone into the troposphere, with a lesser fraction being filled in by an increased flux from the tropical stratosphere.  Stolarski et al., 1986]. Concern over this depletion has led to several major expeditions in order to understand the chemistry of the Antarctic stratosphere: National Ozone Expedition [see Solomon et al., 1987] and Airborne Antarctic Ozone Experiment [see Tuck et al., 1988;Anderson et al., 1988;Tuck et al., 1989]. These field studies have implicated high levels of chlorine as being associated with, and possibly as being an immediate cause of the springtime ozone decline. Global satellite observations of the column abundance of ozone since circulation model has been demonstrated by Cariolle and Ddqud [ 1986]. On the basis of observations we impose an Antarctic ozone hole and then follow the dynamical dilution of the ozonedepleted air for the next 12 months. We assume that the unusual chemistry responsible for formation of the ozone hole has disappeared and no longer influences the subsequent evolution. Similarly, we assume that the general circulation is unchanged, and thus we do not include the predicted feedback mechanism by which the presence of the ozone hole helps to isolate further the Antarctic vortex [Kiehl et al., 1988]. The initial ozone deficit is conserved by dynamical transports and can be eliminated only through photochemical regeneration in the stratosphere or mixing into the lower troposphere where ozone is effectively removed near the surface.
The 3D CTM calculations presented here show that transport of ozone-poor air from the Antarctic vortex results in measurable decreases to column ozone (>1%) extending to 30øS during austral summer. Approximately 70% of the initially prescribed ozone deficit is replenished through stratospheric chemistry by the end of the year. Year-to-year accumulation of residual deficits is shown to play a minor role at best in the rapid decline of Antarctic ozone each October, as observed over the past decade. These results are similar to 2D model simulations also presented at the Polar Ozone Workshop .
The dynamical and chemical tracer models are described in section 2. The linearized chemistry is based on observed ozone, temperature and trace gas climatologies for the present atmosphere. Alternate treatments of the linearization and the resulting ozone distributions for standard climatologies are presented in section 3. The CTM is derived from the general circulation model (GCM) for the middle atmosphere developed at the Goddard Institute for Space Studies (GISS) [Rind et al., 1988a, b]. The Antarctic meteorology (winds, temperatures) from the GCM is described in section 4 with an emphasis on the breakup of the winter vortex. Results from the simulated evolution of the ozone hole are given in section 5, and conclusions in section 6.

CHEMICAL TRANSPORT MODEL
The chemical transport model solves the continuity equations for a set of chemically reactive tracers over a 3D grid. This task involves computing the changes in tracer distribution due to advection by winds, convective and diffusive mixing (troposphere only), surface emissions and uptake, and in situ chemical production and loss. Details of the CTM are described by Prather et al. [1987], who documents the nine layer tropospheric version of the CTM. The major difference between the tropospheric version and the 21-layer stratospheric model used here is in the vertical structure described below. The stratospheric CTM has been used before in a study of meteoric sources [Prather and Rodriguez, 1988] and in the NASA 2D intercomparison [Jackman et al., 1989].
The CTM uses the wind fields from the 23-layer stratospheric version of the GISS general circulation model [Rind et al., 1988a, b]. A simulated 1-year climatology of the stratosphere and troposphere from the GCM are stored on history tapes, containing 8-hour averages of the wind and pressure fields plus monthly averages of temperatures and convective patterns. The CTM shares the grid resolution of the stratospheric GCM shown in Table 1; it stores and recalculates the total mass of air and tracer within a volume or "grid box," designated in (longitude, latitude, pressure) coordinates by (I, J, L) and defined by the boundaries given in Table 1. The horizontal resolution of the CTM is 7.8 ø in latitude and 10 ø in longitude; vertical resolution within the stratosphere is approximately 5 km.
The CTM computes advective transport of a tracer using an upstream method that conserves the second-order moments of the tracer distribution subject to the constraint of positive definite tracer concentrations [second-order moments (SOM) with limits: Prather, 1986]. The SOM method conserves tracer to arbitrary (computer limited) accuracy, is highly accurate and nondiffusive but requires nine moments of the tracer distribution to be stored within each grid box. The additional expenditure in storage is more than compensated for by the increase in effective spatial resolution. The SOM method partially compensates for the coarse vertical resolution of the GCM, is able to maintain steep vertical gradients in tracer mixing ratio of more than a factor of 10 between adjacent layers but cannot make up for the poor resolution of the tropopause and lower stratosphere in this version of the GCM.

Odd Oxygen and Chemical Steady
where [P-L] denotes net production (vol/vol mixing ratio per second), including both production and loss terms. Overall there are more than 30 reactions involving odd oxygen species that contribute directly to the net production [DeMore et al., 1987]. The value of [P-L] can be calculated rigorously under given stratospheric conditions, but the assignment of individual reactions to production (P) or loss (L) is often ambiguous. For example, the reaction NO + 0 3 ---) NO 2 + 0 2 where O represents either form of atomic oxygen. The primary reaction responsible for generating the ozone layer, photolysis of molecular oxygen, is clearly assignable to production of odd oxygen. where X may be either NO or C1. For these pathways, loss of odd oxygen will not respond in a simple linear manner to changes in ozone. We use observed climatologies to define profiles for the trace gases and families: 03, 0 2, CH 4, H20, H 2, CO, NOy (= HNO 3 + NO + NO 2 + NO 3 + N205 + HNO 2 + HO2NO 2 + C1ONO2), C1 x (= HC1 + C10 + C1 + C12 + HOC1 + C1ONO 2 + (C10)2), Brx (= Br + BrO + HBr + HOBr + BrONO 2 + BrC1). The values for 03 and temperature (and hence, O 2 and N 2) are taken from standard climatologies for the monthly, zonal means: 12 months by 10 ø in latitude by 2 km in altitude [see McPeters et al., 1984]. These quantifies define the ultraviolet radiation field and hence the key photolysis rates. The profiles of the trace gases are based on a single average of mid-latitude measurements plus model results and are shifted in altitude according to the tropopause height appropriate for a given latitude and month (see Table 2, also Logan et al. [1978]).

The chemical model used in these calculations
The The projected steady state mixing ratio is based on a linear extrapolation about the observed state and is therefore not likely to be the true steady state value. For any mixing ratio f, we calculate the net production of ozone from

Our derivation of f* and T depends on how we choose to interpret the derivative d[P-L]/df.
We have examined three different forms for the linearization of the ozone chemistry, designated X, Y and Z. Chemistry X assumes that stratospheric ozone is everywhere in chemical steady state (fo = f,) and that photolysis of molecular oxygen (reaction (8)) is the only source of odd oxygen (T = fo/p o). These assumptions will clearly introduce errors in the lower stratosphere. Nevertheless, chemistry X has a certain simplicity in that the model depends only on the values of ozone and temperature (i.e., photolysis of 02) and not on the assumed distribution of trace gases or other details of the chemistry.
Chemistry Y uses the full set of photochemical reactions to calculate the value of [P-L] ø and assumes that primary production of odd oxygen, po, can be defined by reactions (8)-(10) above. We assume that this production term is invariant with respect to changes in the local concentration of 03 , and that the loss term, varying linearly for small perturbations about fo, can be def'med from the residual  Figure ! shows the time scale for ozone (T) during January using both chemistries X and Z. The values of T derived from chemistry X correspond to the ozone replenishment times used by Solomon et al. [1980] which are much slower than the actual chemical time scales for ozone.  We use the 1D photochemical model, as described above, in conjunction with the assumed climatology for ozone, temperature and trace gases to calculate tables off* and T that are averaged vertically over each stratospheric layer in the CTM. The steady state mixing ratios for ozone (either projected or actual using chemistry Y and Z) have a problem that is shared with most current photochemical models: predicted values are 10-30% below those observed over the altitude range 1-10 mbar. This discrepancy represents a fundamental, systematic difference between models and measurements that cannot be resolved here. These differences occur in the upper stratosphere where the time constant to achieve a steady state is rapid, and thus the observed values should represent a true photochemical steady state which is not perturbed by transport. We have chosen therefore to use the observed mixing ratios as the projected steady state values where the time constant, T, is less than 7 days and to interpolate smoothly between fo and f* for T greater than 7 days.
The net chemical change for ozone in each grid box of the CTM is based on the appropriate entry from these tables, fo3(t) -fo3(0) -(Of* -fo3(0)) (1 -e -t/r) (18) and is approximated for each 4-hour time step (T in seconds) as fo3(t) = fo3(0) + -fo3 (0) These control runs are in approximate steady state, in so far as the mean annual cycle of ozone repeats to within about 1%. In this section we examine the ozone climatology produced by the different chemical models (X, Y and Z). We compare with the Solar Backscatter Ultraviolet (SBUV) climatology used to developed the linearized chemistry for ozone. In the upper stratosphere our predicted 03 should match the observed values, since chemical time constants are rapid and transport of ozone is unimportant. In the lower stratosphere, however, the ozone distribution is determined by a balance between chemical tendencies and transport, and the comparison provides a good test of our model. The Dobson map of observed column ozone is shown in Figure 2, and ozone mixing ratios from the SBUV climatology for January and April are given in Figure 3. The control run with chemistry-X produced concentrations of ozone in the lower stratosphere that were much smaller than observed in the tropics and yet much larger than observed at high latitudes. Calculated column ozone and profiles are given in Figures 4 and 5  We wish to emphasize here that our ability to reproduce the observed distribution of ozone, given that distribution as input to the chemical model, is not trivial. Ozone in the middle and lower stratosphere is not in photochemical steady state. What we are testing here is a combination of CTM transport and photochemical tendencies for ozone in the lower and middle stratosphere. We show here that the photochemical tendency of ozone below 10 mbar is a critical element in determining It should be noted that in these calculations the chemistry used in the CTM is not an ab initio calculation of ozone concentrations from first principles. As noted above, such models still have problems with the upper stratosphere. In summary, our errors in the simulation of ozone with chemistry Z could be due either to incorrect transport in the CTM, such as too rapid upwelling in the tropical lower stratosphere, or to errors in the chemical model. This question cannot be answered with the current experiments but will be examined in subsequent CTM simulations using other chemical tracers such as CH 4, CFC13, CF2C12, N20 and NOy.

ANTARCTIC ]VIETEOROLOGY
How well does the model simulate the dynamic climatology of the Antarctic stratosphere, especially in the critical springtime months of September and October? A description of Antarctic meteorology for this time period is available from various studies using satellite observations [e.g., Harwood, 1975;Hartmann, 1976 I  I  I  I  I  I  !  I  I  •  I  I  I  I  I  I  I   - This comparison of zonal fields of stratospheric winds and Eliassen-Palm flux indicates that the transport of chemical tracers averaged over time and longitude (i.e., the residual circulation, see Andrews et al. [1987] in the model may be similar to the transport occurring in the southern hemisphere during the breakup of the Antarctic ozone hole. The agreement of these zonally averaged meteorological quantities does not necessarily imply that the transport of chemical tracers will be accurately simulated in three dimensions during this transient phenomenon when the winds reverse. Indeed the predictions of the dispersal of the ozone hole in this paper provide a test of the transport of trace species in the CTM. In conclusion, our average climatological model for the southern hemisphere may be representative of Antarctic spring, but further, derailed studies and/or simulations of the observations from individual years will be necessary to test the model. February 1 the deficit at 60øS declines to 6% while that at 30øS reaches 1%. Recovery of the ozone deficit continues almost uniformly through austral summer, slowing as winter approaches. By the following spring, column reductions at 60øS are only 2%, and more than two-thirds of the initial deficit has been regenerated by photochemical production of ozone. The calculation initialized on September 1 is almost identical to the October run, demonstrating the lack of sensitivity to these choices in initializing the ozone hole. Parallel model calculations (not shown) using chemistries X and Y produce similar results.

The focus of this paper is a 3D chemical tracer experiment
The cumulative buildup of an ozone hole through successive years of ozone depletion is found to be a second-order effect. Note that 1 month after reinducing the hole for the second year, deficits at 60øS are of order 9.7% compared with 8% the  previous year. By January 1, deficits of order 1% could be It is instructive to examine the processes controlling column found as far north as 25øS, as compared to 30øS in the first ozone in the southern hemisphere, both with and without the year. The average ozone depletion at the end of the second Antarctic ozone hole. We choose to perform a budget analysis year is only 20% larger than that at the end of the first year, for the total amount of ozone in a "box" comprising the lower and for either year, amounts to approximately 0.7% of the stratosphere ( I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I   to replenishment of all but 20% of the initial deficit. An additional 10% of the original deficit (0.7x10 zø kg) remains in the atmosphere but has been transported outside of the southern mid-latitude stratosphere. In particular, we predict significant decreases to upper tropospheric ozone in the southern hemisphere, of order 10% over most seasons, that are attributable to the Antarctic ozone hole.

CONCLUSIONS
In this study we have shown that the 3D chemical transport model produces an adequate, although not ideal, simulation of the observed ozone climatology using a simple, linearized model for chemistry. Inducing an Antarctic ozone hole and then returning the chemistry to its unperturbed state on October 1 gives a subsequent evolution of the ozone deficit that displays some of the observed features of ozone depletion in the southern mid-latitudes (TOMS observations; see also Atkinson et al. [1989]). This simulation predicts chemical regeneration of most of the ozone hole during the subsequent year with only modest residual accumulation after several years of recurring ozone holes.
In order to compare with observations during August and September, we must improve the initialization procedure, allowing for a more realistic generation of the ozone hole. The choice of initializing the ozone hole on October 1 ignores the photochemical destruction of ozone that is observed to occur continuously since the return of sunlight in August. The initialization on September 1 is similarly unrealistic and only gives us some clues regarding the impact of ozone loss earlier in the season. Therefore the simulations shown here fail to account for ozone depletion within the vortex, or its dilution to mid-latitudes, that occurs in late winter and early austral spring.
Parallel work on the chemical propagation of the Antarctic ozone hole [Prather and Jaffe, 1990] has shown that additional ozone depletion is unlikely to occur during the transport of ozone-poor, chemically perturbed air parcels from the ozone hole to the mid-latitudes. Moreover, the transport of this denitrified air will reduce NOy at mid-latitudes by 5-10%, which in mm leads to reduced chemical loss of ozone at midlatitudes and a faster recovery from the effects of dilution of the ozone hole. On the other hand, the dilution 'and mixing of polar-processed air without ozone depletion may lead to additional loss at mid-latitudes. In general, the transport and mixing of chemically perturbed air parcels occurs on spatial scales much smaller than those resolved by the CTM but may be studied with the tracer model by transporting chemically perturbed air as a tracer with potential for ozone depletion.
Several Development of the CTM will move toward a more realistic chemistry for ozone that, of necessity, includes simultaneous transport of ozone, methane, water vapor and the families of odd nitrogen, chlorine and bromine. One immediate application of the linearized ozone chemistry developed here will be an interactive GCM/CTM simulation of the Antarctic ozone hole. In this calculation the transport and chemical replenishment of the ozone hole would be coupled with the radiative forcing and the dynamical response of the GCM.