The role of water-vapour photodissociation on the formation of a deep minimum in mesopause ozone

. A one-dimensional atmospheric photochemical model with an altitude grid of about 1.5 km was used to examine the structure of the global mean vertical ozone proﬁle and its night-time-to-daytime variation in the upper atmosphere. Two distinct ozone layers are predicted, separated by a sharp drop in the ozone concentration near the mesopause. This naturally occurring mesopause ozone deep minimum is primarily produced by the rapid increase in the destruction of water vapour, and hence increase in HO x , at altitudes between 80 and 85 km, a region where water-vapour photodissociation by ultraviolet radiation of the solar Lyman-alpha line is signiﬁcant, and where the supply of water vapour is maintained by methane oxidation even for very dry conditions at the tropospheric-stratospheric exchange region. The model indicates that the depth of the mesopause ozone minimum is limited by the e(cid:129)ciency with which inactive molecular hydrogen is produced, either by the conversion of atomic hydrogen to molecular hydrogen via one of the reaction channels of H with HO 2 , or by Lyman-alpha photodissociation of water vapour via the channel that leads to the production of molecular hydrogen. The ozone concentration rapidly recovers above 85 km due to the rapid increase in O produced by the photodissociation of O 2 by absorption of ultraviolet solar radiation in the Schum-ann-Runge bands and continuum. Above 90 km, there is a decrease in ozone due to photolysis as the production of ozone through the three-body recombination of O 2 and O becomes slower with decreasing pressure. The model also predicts two peaks in the night-time/daytime ozone ratio, one near 75 km and the other near 110 km, plus a strong peak in the night-time/ daytime ratio of OH near 110 km. Recent observational evidence supports the predictions of the model.


Introduction
We have examined the role of H 2 O photodissociation on ozone concentration in the upper atmosphere with particular emphasis on the region between 70 and 95 km altitude, where we expect water-vapour photodissociation by the solar Lyman-alpha line to be important.As was shown by Lewis et al. (1983), the photodestruction rate of water vapour by Lyman-alpha peaks within the layer between 70 and 95 km.We have therefore used a 1D photochemical model with a ®ne pressure grid (corresponding to intervals of about 1.5 km at these altitudes) in order to isolate processes that might have small characteristic length scales.Our present study is motivated by the wealth of recent satellite concentration data of atmospheric species important to atmospheric ozone photochemistry that are available from such data bases as UARS/HALOE, ISAMS, MLS, ATMOS and SME with ground-based and in situ data avaliable for validation.In this initial study we compare our model predictions with some typical vertical ozone pro®les obtained by ATMOS (Atmospheric Trace Molecule Spectroscopy experiment on Spacelab 3) and HALOE (Halogen Occultation Experiment launched on the Upper-Atmospheric Research Satellite, UARS, in September 1991 as part of the Mission to Planet Earth, MTPE, Program).Figures 1  and 2 show typical vertical ozone pro®les obtained by HALOE at sunrise and sunset, respectively.Ozone values with altitude are given in the HALOE data base along with a quality factor which represents the statistical error in the measurements.An estimate of the error, taken here to be the quality divided by four (see Overview of HALOE Data Quality Status in the data base), gives an indication of the upper limit of the measurements when added to the given mean value.Above 70 km, the estimated upper limit starts to deviate from the mean value.However, both the mean pro®les and the estimated upper limit pro®les indicate that there is a deep minimum generated near the mesopause.The depth of the minimum is variable with minimum mixing ratios usually below 0.1 ppmv.Our model computations show that the deep mesopause ozone minimum is created by the reaction of ozone with the active hydrogen species HO x (H, OH and HO 2 ) that are produced by the photodissociation of water vapour, and that the variability of the depth of the minimum depends on the eciency of the conversion of H to H 2 through the reaction of H with HO 2 , and on the production of molecular hydrogen during Lyman-alpha photodissociation of water vapour.

The role of water-vapour photodissociation
The role of the HO x species (H, OH and HO 2 ) in destroying ozone near the mesopause was investigated in detail by Allen et al. (1984), who stressed the key role played by reactions which convert the active-hydrogen HO x species to the inactive-hydrogen species H 2 and H 2 O.The subsequent reaction rate measurements of Keyser (1986) for the reaction of H with HO 2 and estimates of the branching ratios which lead to either active or inactive hydrogen species highlighted the role of this reaction in controlling ozone concentrations near the mesopause.The reaction of H with HO 2 has three reaction channels (e.g.Keyser, 1986;DeMore et al., 1994) where the ®rst which leads to the production of the active hydrogen species OH is dominant (branching ratio 0.9).It is the strength of the third channel which limits the role of water-vapour photodissociation in ozone reduction through the conversion of atomic hydrogen to inactive molecular hydrogen.According to Keyser's work, the third channel's branching ratio was determined from knowledge of the total reaction rate and the branching ratios of the ®rst two channels.Given the uncertainties in the branching ratios of the ®rst two channels and the fact that these were determined for the temperature range 245±300 K, Keyser concluded that a more exact study of the third channel and its temperature dependence was warranted, having found no signi®cant temperature dependence for the total reaction rate and branching ratios in this temperature range.Clearly, measurements typical of the low-temperature (below 200 K) conditions near the mesopause are needed.
Photolysis of water vapour by Lyman-alpha can also result in the production of active and inactive hydrogen species.The two channels for the photodissociation of water vapour by Lyman-alpha are where the branching ratio a for the ®rst channel is about 0.75 and for the second about 0.25 according to Stief et al. (1972), while according to Banks and Kockarts (1973), the direct production of H 2 from the photodissociation of water vapour is a small fraction of the process of Lyman-alpha photodissociation leading to the production of the HO x species.

The model
The 1D photochemical model used comprises two radiation transfer submodels (Vardavas and Carver, 1984a;Strobel, 1987), a radiative-convective equilibrium submodel (Vardavas and Carver, 1985) and a chemical kinetics-diusion submodel (Vardavas, 1984;Vardavas et al., 1990).One radiation submodel simulates the transfer of solar visible and near-infrared plus terrestrial infrared ¯ux through the atmosphere.These ¯uxes are used to compute net heating-cooling in the radiativeconvective model which computes the vertical atmospheric temperature pro®le needed for the computation of reaction rates.Alternatively, one can specify a vertical temperature structure and use the radiative-convective model to compute the radiation ®elds.The other radiation submodel simulates the atmospheric transfer of solar uv from 0.01 lm to Lyman-beta, to take into account reactions of the ionic species leading to the formation of NO in the lower thermosphere (Siskind and Rusch, 1992), and from Lyman-alpha to 0.35 lm, visible, and near infrared up to 1 lm allowing for pure absorption, Rayleigh and cloud Mie multiple scattering and surface re¯ection (Vardavas and Koutoulaki, 1995).The solar ¯ux is used to compute photodissociation rates using 206 wavelength-grid points.The photodissociation rates depend strongly on O 2 uv absorption in the highly structured Schumann-Runge (S-R) bands (Lewis et al., 1994;Allen and Frederick, 1982).The O 2 S-R bands aect both the uv radiation ®eld, and hence the photodissociation rates of other species, and the photodissociation of O 2 .The incoming solar ¯ux spectrum pro®le at the top of the atmosphere is normalized to a solar constant of 1367 Wm À2 .The solar zenith angle, cloud cover, and ground albedo are speci®ed.Reaction and photodissociation rates (DeMore et al., 1994) are used in a system of 182 reactions, to simulate the vertical transport of 53 species based on a semiempirical expression for vertical eddy diusion.The diusion coecient was taken from Brasseur and Solomon (1984) up to 100 km, while above that the eddy diusion coecient was taken to decrease rapidly (e.g.Strobel, 1989;Roble, 1995) and molecular diusion becomes important.The bi-molecular diusion coecients were taken from Vargaftik (1975).Equilibrium vertical pro®les of species concentrations were computed using diurnal averaging by computing the daytime and night-time species concentrations (Turco and Whitten, 1978).Near the ground, emission and deposition processes are included (Wuebbles, 1981).Water vapour at the surface and in the troposphere is computed from a speci®ed relative humidity pro®le and the tropospheric temperature structure.
The model divides the atmosphere into a ®ne grid of 100 altitude (pressure) levels, to avoid pressure-dependent changes in temperature and composition (Vardavas and Carver, 1984b).Constant-¯ux boundary conditions are enforced at the top of the atmosphere, and also at the ground for most species rather than ®xed concentration boundary conditions, to allow the species concentrations to obtain the values dictated by the photochemical and physical processes.This does not exclude the possibility of specifying a surface ¯ux which can be included through the emission or deposition processes.
The model was used to generate a global mean ozone pro®le (re¯ecting mid-latitude conditions) by setting the solar zenith angle to 60 , cloud-cover fraction to 0.5, surface albedo to 0.1 (mostly ocean), a surface relative humidity of 0.8 and the US Standard Atmosphere temperature structure.The chemistry was set to a 1960 standard without CFCs and total Cl x of 1 ppbv (Johnston et al., 1989).

Diurnally averaged pro®les
In Fig. 3 our model results (curve a b c) for the diurnally averaged ozone pro®le computed using the H HO 2 branching ratios of Keyser and a photodissociation branching ratio a 1 are compared with the pro®le of Roble (1995), which used solar minimum radiation ¯ux conditions and branching ratios taken from Brasseur and Solomon (1986).Computations with the branching ratios given in Roble showed no signi®cant dierence from the ozone pro®le of curve a b c.On removing the third channel of the reaction of H and HO 2 , the mesopause ozone minimum deepens considerably (curve a b, while removal also of the second channel leads to a further deepening of the minimum (curve a).Our model's deep minimum, obtained by removing the channels that convert the active H species to inactive molecular hydrogen and water vapour, is in better agreement with the observations.In Fig. 4 are shown the ozone pro®les obtained using all three channels for the reaction of H with HO 2 for dierent values of the Lyman-alpha photodissociation branching ratio a.The pro®le with a 0, which corresponds to photodissociation that leads to the production of inactive molecular hydrogen, exhibits a very weak ozone minimum at the mesopause.The pro®le with a 1, which corresponds to photodissociation that leads to the production of active atomic hydrogen, exhibits a deep minimum in ozone concentration at the mesopause.Clearly the reactions that produce molecular hydrogen instead of active atomic hydrogen limit the depth of the ozone minimum at the mesopause.
Our model shows a strong rise in the H concentration near the mesopause, rising to a value of 1X3 Â 10 À5 ppmv, in keeping with other models (Strobel, 1972;Brasseur and Solomon, 1984) and gives an H escape rate of 2X8 Â 10 8 molecules/cm 2 s, close to the estimates of others (Carver, 1981;Kasting et al., 1979).Our model H 2 O pro®le, obtained using a global mean surface Figure 6 shows the ozone pro®les obtained for dierent water-vapour pro®les for progressively drier tropopause conditions obtained by setting the surface relative humidity equal to 0.8, 0.4 and 0.1, using channel a only for the reaction of H HO 2 and a 1.The water vapour mixing ratio in the stratosphere and mesosphere is maintained by methane oxidation even for very dry conditions in the tropospheric-stratospheric exchange region.The corresponding mesopause ozone deep minimum is thus made slightly shallower by the drop in the water-vapour mixing ratio whose peak value in the upper atmosphere goes from 5.7 to 2.9 ppmv, when the surface relative humidity is reduced from 0.8 to 0.1.

Night-time/daytime ratios
In Fig. 7 are given the daytime and night-time model pro®les for the channel-a-only case with a 1, and these are compared with: the ATMOS sunrise and sunset observations (Gunson et al., 1990)   The night-time-to-daytime variation of the ozone concentration generated by the model is in agreement with the mid-latitude variation (with a weak seasonal dependence) observed by Connor et al. (1994).The ozone night-time/daytime concentration ratio predicted by the model is given in Fig. 8, along with the corresponding ratios for the HO x species and O.The ratio for ozone is in very good agreement with the observations up to 70 km, with the night-time mixing ratio remaining nearly constant at about 1.5 ppmv between 60 and 70 km, also in agreement with the observations.The model predicts two distinct peaks in the ozone night-time/daytime ratio, one near 75 km and another near 110 km, plus a sharp narrow dip at about 82 km, which corresponds to a region where the nighttime/daytime ratios of OH and HO 2 exhibit very sharp peaks, resulting in more eective night-time destruction of ozone by these two species.

Discussion
Our interpretation of the strong ozone reduction in the mesopause region and of the night-time/daytime ratios is as follows.
In the stratosphere below about 50 km, the destruction of ozone is balanced by three-body recombination of O and O 2 , so that the equilibrium that is rapidly maintained during daytime is unaltered at night-time.Note that the night-time/daytime ratio for O is negligible up to about 70 km.
Between 50 and 70 km, direct photolysis of ozone in the Hartley-Huggins bands reduces the ozone daytime concentration, while the night-time mixing ratio is maintained fairly constant with altitude (Fig. 7) through (eddy) diusive equilibrium with the atmospheric layers from below.The result is a rise in the night-time/daytime concentration ratio of ozone as shown in Fig. 8.There is also some rise in the night-time concentration of HO 2 due to photodissociation shielding at night-time (Fig. 8).Above 60 km, there is a rapid rise in daytime H with altitude through the photodissociation of water vapour by Lyman-alpha and ultraviolet radiation in the region of the Schumann-Runge bands of oxygen.This is accompanied by a rise in the daytime concentration of the other HO x species, OH and HO 2 .
Between 70 and 80 km, there is a rise in night-time H accompanied by an increase in the other HO x species, as diusion from above rises the H night-time/daytime ratio towards unity, which occurs at about 85 km.This results in a decrease in the night-time/daytime ratio of  Connor et al. (1994) ozone between 70 and 80 km.The very rapid rise in night-time H between 80 and 85 km results in a rapid rise in night-time OH and HO 2 , as the latter is shielded at night from photodissociation, with an associated sharp drop in the night-time/daytime ozone ratio below unity.In this region, atomic oxygen rises rapidly due to the photodissociation of molecular oxygen by absorption of solar ultraviolet radiation by the Schumann-Runge bands and continuum.As in the case of H, through diusion from above, the night-time/daytime ratio of O rapidly rises to unity at about 85 km.This results in two processes.
First, the rapid rise in O between 80 and 90 km produces a rapid rise in ozone concentration both during daytime and night-time (Fig. 7), sucient to counteract the eects of the decreased production of ozone with altitude via the pressure-dependent threebody recombination of O and O 2 .Above 90 km, there is a rapid decrease in daytime ozone with altitude due to photolysis, accompanied by a drop in night-time ozone, as the fall in pressure results in a decrease in the rate of ozone production via the three-body recombination of O and O 2 .
Photolysis produces a rise in the night-time/daytime ozone ratio, with a peak at about 110 km, as shown in Fig. 8. Secondly, since HO 2 in this region is created by three-body recombination of H and O 2 and is destroyed by reaction with O, and both H and O have night-time/ daytime ratios of unity above about 85 km, the nighttime/daytime ratio of HO 2 also falls to unity.The drop in the night-time/daytime ratio of HO 2 initially produces a drop in the ratio of OH, but this is followed by a rise in the night-time/daytime ratio of OH as ozone reacts with H to produce OH in this region.The OH night-time/ daytime ratio therefore follows the variation in the ozone night-time/daytime ratio.
We have performed some preliminary sensitivity tests to evaluate the response of the ozone deep minimum near the mesopause to changes in some of the atmospheric properties.Reduction of the reaction rate constant for the reaction of HO 2 with O to produce OH and O 2 by 70% of its JPL94 value (DeMore et al., 1994), as proposed by Clancy et al. (1994), did not change the structure of the minimum and second ozone layer, but did increase the ozone concentration in the layers from 40 to 70 km, in keeping with their results.We have also made a preliminary examination of the eect of the thermospheric rise in temperature on the ozone concentration and night-time/daytime ratio pro-®les by setting the temperature equal to the minimum mesopause value of 186 K.The eect was to remove the second peak in the ozone night-time/daytime ratio near 110 km and to replace it by a continuous rise from about 100 km.The thermospheric rise in NO was also found not to aect signi®cantly the structure of the second ozone layer.Variation of the ozone mesopause minimum between solar cycle minimum and maximum [using a uv ¯ux variation from Vardavas (1987)] was found to be weak, although both the stratospheric and lower thermospheric peaks in ozone concentration were found to be enhanced at solar maximum as expected.

Summary
Our model computations indicate that there are two distinct ozone layers in the atmosphere, separated by a sharp drop in the ozone concentration near the mesopause.The depth of the mesopause ozone minimum is limited by the eciency with which inactive molecular hydrogen is produced, either by the conversion of atomic hydrogen to molecular hydrogen via one of the reaction channels of H with HO 2 or by Lyman-alpha photodissociation of water vapour via the channel that leads to the production of molecular hydrogen.Recent satellite observations exhibit this deep minimum in mesopause ozone.Both laboratory reaction rates between H and HO 2 at low temperatures and more accurate ozone measurements above 80 km are required better to quantify the depth of the minimum and its variation.

Fig. 1 .
Fig. 1.A sample of HALOE ozone pro®les showing mean (®ne line) and estimated upper limit (thick line) pro®les at sunrise in the northern hemisphere

Fig. 2 .
Fig. 2. A sample of HALOE ozone pro®les showing mean (®ne line) and estimated upper limit (thick line) pro®les at sunset in the northern hemisphere

Fig. 3 .
Fig. 3. Model global mean diurnally averaged ozone pro®les with and without channels a, b and c of the reaction H HO 2 , using a Lymanalpha water-vapour photodissociation branching ratio a 1 at two locations ± 48 S (sunrise) and 28 N (sunset) ± from 30 April through 6 May 1985; and with the HALOE observations in May 1992 at 49 S (sunrise) and 28 N (sunset).Both the sunset and sunrise observations below 80 km follow more closely the daytime values of the model, while above 80 km both observation sets exhibit the rapid rise in ozone concentration predicted by the model.

Fig. 4 .
Fig. 4. Model global mean diurnally averaged ozone pro®les for dierent values of the branching ratio a for the photodissociation of water vapour by Lyman-alpha, using the branching ratios of Keyser (1986) for the reaction H HO 2 Fig. 5. Model diurnally averaged volume mixing ratio for HO x and H 2 O together with sunrise and sunset H 2 O UARS/ISAMS measurements given in Goss-Custard et al.(1996)  and the ATMOS measurements given inGunson et al. (1990)

Fig. 7 .
Fig. 7. Global mean daytime and night-time averaged ozone pro®les computed by the model are compared with the sunrise and sunset measurements obtained by the ATMOS experiment on Spacelab 3 as given in Gunson et al. (1990) and by UARS/HALOE