Ozone in the Upper Stratosphere and Mesosphere

A detailed photochemical model of the upper stratosphere and mesosphere is compared with three extensive sets of ozone observations: Atmospheric Explorer-E backscattered ultraviolet experiment (BUV), Nimbus-4 BUV, and rocket flights from Wallops Flight Center (ROCOZ). The Nimbus-4 and rocket observations are most sensitive to ozone between 30 and 50 km, whereas observations from AE-E measure the abundance of ozone up to 70 km. The photochemical model accurately reproduces the observed relationship between BUV intensity and local solar zenith angle, although the absolute calibration on AE-E appears to be in error. The AE-E observations and the model both exhibit a morning-afternoon asymmetry, with more ozone in the morning owing to the build up of HOx species in the afternoon. Seasonal changes in atmospheric temperature produce an annual maximum in tropical mesospheric ozone during June-July-August. The amplitude of the observed effect is somewhat larger than calculated by the model. Some problems appear to remain with the presently accepted kinetic rates for HOx species in the atmosphere.

A primary question which must be answered is whether these three data sets provide a consistent picture of ozone in the upper stratosphere and mesosphere.
The analysis of the complete set of data from Atmospheric Explorer-E (BUV) and a limited set of Nimbus-4 data [Heath et al., 1979] is presented in section 3, where differences between the two data sets are discussed with attention to seasonal variations. Results from the theoretical model are compared in turn with the ozone concentrations of the midlatitude ROCOZ data and with the satellite albedos from the tropics.
A major goal of this research is the development of an objective photochemical model which, as accurately as possible, reproduces all aspects of the observed distribution of ozone. In section 4 the overall validity of this photochemical model is critically examined. The key uncertainties in the model are pointed out, and weaknesses or inconsistencies among the observational data for ozone are discussed.

PHOTOCHEMICAL MODEL
The photochemistry which controls the abundance of ozone in the upper stratosphere and mesosphere can be described by about 50 kinetic and photolytic reactions as given in Table 1 The rates of chemical reactions which produce or destroy odd-oxygen are given in Figure 3. The only significant source of odd-oxygen above 30 km is the photolysis of molecular oxygen [Chapman, 1930].   [Burrows et aL, 1977;Chang and Kaufman, 1978] and is much smaller than the values at attoospheric pressure reported by DeMote [1979], k17 10 -'ø cm a s-'. The measurements made at a pressure of 2-4 tort employed inert carrier gases with at most traces of molecular oxygen. If we accept DeMore's hypothesis of a pressuredependent reaction rate, then the present kinetic data do not determine rates applicable in the atmosphere.
Because (R17) is the dominant path for odd-hydrogen recombination below 75 km, a change in k17 has the same effect on model calculations as a change of opposite sign in H:O, the primary source for odd-hydrogen. In particular, the standard kinetic model with k17 = 4 x 10 -• cm a s -• and 2 ppm H:O would produce the same HO•, distributions below 75 km as a model with DeMore's rate k17 •'• 1 x 10 -•ø cm a s-' and 5 ppm H:O.
The rate constant for (R15) has been measured only at room temperature [Hack et al., 1978[Hack et al., , 1979, and thus the temperature dependence adopted here [Baulch et al., 1972] is not well determined. This reaction (R15) is most important to the HO• budget near the extremely cold mesopause (~ 190øK). If  I 1111 I  I  I I I I III I  I  I I\l IIII  I  I  I I I I I   0+0   ß  H '•,  \  ODD- where P•, L• and V. q•, are the local production, local loss, and divergence of the flux for species i, respectively (cm -3 s-l). For short-lived species P and L will be much larger than Consequently, the calculations presented here assume that transport of radicals is negligible, and the flux divergence term in (1) is neglected for all species. The local production and loss terms in (1) involve products of concentrations of other chemical species. The coupled set of continuity equations is accordingly nonlinear. The inverse Euler technique is used to solve (implicitly) for species concentrations at each new time step. The mass of any closed system is conserved to 1 part in 10 lø. Time steps of the order of 1 hour are used to model the 24-hour day with finer temporal resolution near sunrise and sunset. Steady state conditions are achieved by an acceleration procedure which ensures that the net change in concentration of all species is zero after a 24hour model day [Wofsy, 1978].

c. Solar Radiation and BUVAlbedos
The radiation field is calculated at each time step during the model day for the true zenith angle of the sun. The model includes absorption by molecular oxygen and ozone as well as molecular scattering. The attenuation of the direct solar flux is calculated by integrating through the atmosphere in the direction of the sun. The calculation takes into account both sphericity and changes in local solar time along the path. Refraction is not included. This direct solar flux is the source function used to calculate the diffuse radiation in an inhomogeneous scattering atmosphere which is assumed to be plane parallel [Prather, 1974]. The method allows the calcu-  The backscattered radiance from the atmosphere may be calculated from the time-dependent distribution of ozone.

I--• dr ß &(•) ß P(% •r -•) exp [-• -%1 (2)
In this case, I is the monochromatic intensity of light scattered in the direction of the zenith, Fo is the incident solar flux, P is the scattering phase function for natural light, 0 is the solar zenith angle, & is the single scattering albedo, ß is the vertical extinction, and •, is the extinction path in the direction of the sun [see Chandrasekhar, 1960]. One may define the dimensionless B UV albedo px (•) = 4•rlx (•)/fox The incident light is assumed to be unpolarized, and the scattered intensities of different polarization states have been combined. Only first-order scattering is included in (2). Higher-order scattering may be included rigorously [Prather, 1974] but was found ti• be of little importance for wavelengths less than or equal to 2922 A. In these cases the solar flux does not penetrate significantly below 30 kin, and the vertical scattering depth is at most 0.01. The calculations assume that the only source of scattered light is due to Rayleigh and Raman scattering by the molecules of the background atmosphere (N2, 02, CO2). The mean Rayleigh scattering cross section per molecule is given by [Penndorf, 1957] where A is the wavelength (pm), Ns is the molecular density at 15øC (cm-3), and ns is the index of refraction for air at 15øC [Edlen, 1953]. and are given in Table 3 The detailed diurnal behavior of ozone with altitude is displayed in Figure 6. Significant day-to-night variations in the concentration of ozone occur above 50 km because the daytime atomic oxygen density is comparable to or greater than that of ozone. The daily cycle of ozone above 70 km exhibits a complex structure with variation of the ozone concentrations in excess of a factor of 2 over the sunlit portion of the day. The rise in ozone and total odd oxygen begins approximately 20 minutes after sunrise, driven by photolysis of molecular oxygen ((R38)) in the Schumann-Runge bands. Photolysis of H20 at Lyman-alpha, the source of odd hydrogen above 65 km, is strongly peaked about local noon. The HOx concentrations maximize in the afternoon, and the ozone density is thus asymmetrical about noon. An increase in the concentration of mesospheric water depresses the mean concentration of ozone and causes the peak daytime concentration to occur earlier in the day. In this photochemical regime, odd oxygen will also be sensitive to changes in solar Lymanalpha. Such variations are known to occur [Hinteregger, 1979] over the solar rotation (-30 days) and over the solar cycle (-11 years; see section 3g). Noontime densities of OH and H for the standard model are shown in Figure 7. Also shown is the decrease in OH for 2 ppm H:O and the increase in OH corresponding to solar maximum.

OZONE OBSERVATIONS a. BUY Satellites
Remote sensing of the earth's atmosphere by satellite provides the only practical means for global monitoring of the     (10øN-20øN,  10øS-10øN, 20øS-10øS) and also into morning and afternoon. The period of observations is from December 1975 through March 1977. The average profile for ozone from 32 to 58 km as reported by Wright [Krueger and Wright, 1979] is shown in Figure 15 along with two theoretical profiles of noontime ozone. The modeled distributions represent the mean of profiles computed for the January, April, July, and October 30øN atmospheres [Cole and Kantor, 1978]. The standard photochemical model diverges from the observed profile above 40 kin. How-  Table 1 for 30øN atmospheres with both 2 and 5 ppm volume mixing ratio of H20.

Fig. 9. Monthly statistics for AE-E observations at 2555 •,. The number of acceptable 2555 A observations in each month of AE-E BUV operations is broken into three latitude bands
(30 days) and to the solar cycle (11 years). Flux variations during the latter two phenomena depend strongly on wavelength as shown in Figure 18. The present analysis considers only the direct effect of flux variations on the photochemistry and assumes an unchanging atmospheric structure.
The 11-year solar cycle may contribute to the systematically different BUV albedos observed by AE-E (solar minimum) and Nimbus-4 (solar maximum). Data on the variation of solar flux previous to cycle 21 are sparse and are sometimes contradictory [Simon, 1978]. Such data from solar cycle 20 have been used in previous studies [Penner and Chang, 1978 (Table 1) for a range of H20 mixing ratios (2-5 ppm) are given for all wavelengths at all angles (vertical bars).

1970-1971 Nimbus-4 BUV and the 1975-1977 AE-E BUV data.
The predominant effects at solar maximum are due to the increased solar radiation at Lyman-alpha and have little effect on the ozone concentrations below 60 km. At solar maximum the odd hydrogen concentrations are greatly enhanced (Figure 7), and ozone is depleted by 25 and 50% at altitudes of 70 and 80 km, respectively. Some caution is necessary since the greater rate of H20 photolysis will enhance the conversion of H20 to H2. Depending on the rate of vertical transport [Keneshea et al., 1979; Allen et al., 1980], the H20 concentration at 80 km may fall below 3 ppm, thus canceling the predicted ozone depletions. Assuming fixed H:O profiles, the model predicts greater albedos at solar maximum, at most 1.5% for the shorter wavelengths near sunset. The effect for solar zenith angles less than 40 ø would be below 0.5%. Thus changes in the solar flux associated with the 11-year solar cycle do not appear by themselves to produce a detectable variation in ozone  The standard photochemical model (5 ppm H20) is described by Tables 1, 2, and 3. The dry model with 2 ppm H20 is also equivalent to the standard model with k•7 = I x 10 -

h. Detailed Comparison with Theory
A detailed comparison of the Nimbus-4 albedos with those from two theoretical models is shown in Table 6. The standard photochemical model appears to underestimate the quantity of ozone and thus overestimates the B UV irradiances by 9-16%. The dry theoretical model provides much better agreement but is still 5% greater than the observed albedos.  (Table   4), but the cross section at this wavelength has been measured several times [Griggs, 1968].
Resonance fluorescence by nitric oxide has been proposed as a source of contamination in the 2555 /• B UV albedos [Guenther et al., 1979]. A substantial fraction of the sunlight at 2150/• is reradiated in the (v', v" = (1, 4) gamma band near 2555/•. Guenther et al. [1979] note that the major contribution is from thermospheric NO and that .it may be as large as 5% of the BUV signal at high latitudes. In the tropics, however, thertoospheric NO densities are typically 5 times less than those observed at high latitudes . Thus the nitric oxide contribution to the tropical 2555/• radiances should be less than 1% for daytime observations. At this point we can compare the ROCOZ data from 38øN with the Nimbus-4 tropical data. Both sets of observations present a consistent pattern which is well described by a standard photochemical model with 2 ppm H20 or, equivalently, with kl7 --1 x 10 -•ø cm 3 s -l. This model reproduces the observed ozone scale height from 35 to 55 km and predicts con- In Table 5  The decrease in B UV albedo with solar zenith angle is reproduced to within 6% by the theoretical models. The standard photochemical model matches the diurnal behavior of the shorter wavelengths, whereas the dry model is more representative of the longer wavelengths. One possibl.e conclusion is that from 50 to 60 km either the H20 mixing ratio increases from 2 to 5 ppm or else the rate constant for OH + HO2 (R17) decreases by the same factor. The anomalous period January 1976 is consistent with a much wetter stratosphere (•8 ppm H20).
The asymmetry in the daytime ozone concentrations shown in Figure 6 leads to a morning-afternoon difference in the backscattered ultraviolet which should be observed by satellite. In Figure 19 the modeled ratio of afternoon-to-morning backscattered light is compared with values from the three periods of AE-E data. Potentially large, systematic shifts in these ratios can be caused by an ephemeris error as small as 3 s. The model predicts less than 1% difference between the corre-   (38øN) show everywhere a summer maximum in ozone density versus altitude. The model which fits the ROCOZ data implies, however, a summer minimum in ozone partial pressure versus total pressure between 40 and 50 km. Thus the July minimum in shortwavelength albedos in the tropics will become a July maximum by northern mid-latitudes. Above 50 km, by any manner of comparison, the Nimbus-4 radiances analyzed here (25øS to 25øN) and the Wallops Flight Center (38øN) ROCOZ data are in harmony with the AE-E data and imply a mesopheric maximum in July from 25øS to 38øN.
While the three sets of ozone data presented in this paper are reasonably consistent and show overall agreement with the model, not all observations of ozone support this view. For example, the several midnight ozone profiles deduced from stellar occultations by the Copernicus satellite [Riegler et at., 1977] do not agree at all with the model results. These nighttime observations cannot be compared directly with the daytime observations except through the photochemical model. In this case, the Copernicus ozone densities are a factor of 2 larger than the densities from any of the models discussed here and remain controversial [Gille et al., 1980]. Another case of disagreement between data sets can be seen by comparison of Watanabe and Tohmatsu's [1976] seasonal ozone dependence with that of Krueger and Wright [1979]. Between 40 and 55 kin, Watanabe and Tohmatsu (31 øN, Japan) report a winter to summer ozone ratio of approximately 2; whereas Krueger and Wright (38øN, Wallops Island) observe a modest excess (10-30%) of summer ozone over winter ozone (density versus altitude). A separate test of the photochemical model is provided by in situ observations of atomic oxygen in the region 60-80 km [Dickinson et al., 1980]. Daytime observations are limited and few (e.g., three profiles at 57øN winter) but are adequately described by the theoretical model with 2 ppm H20 (see Figure 5).

DISCUSSION
The three sets of ozone data which are analyzed in this paper present a reasonably coherent view of ozone in the upper atmosphere. The absolute calibration of the AE-E albedos is anomalous, but their diurnal structure and symmetry are consistent with the ROCOZ profiles and Nimbus-4 albedos. The AE-E BUV data are the only set of continuous observations which are sensitive to ozone above 55 km in the tropics. With the renewed availability of the AE-E instrument, the issue of absolute calibration may be resolved by simultaneous overpasses of AE-E BUV and Nimbus-7 SBUV/TOMS.
The short wavelength AE-E data suggest the existence of an annual cycle in tropical ozone above 50 km. There is an unexplained 6% excess of January albedos over July albedos at 2555 A and a possible excess at 2735 A. Several explanations for excess ozone in July may be hypothesized. One approach is to invoke variable water vapor concentrations above 55 km with a minimum in July. Another is to hypothesize large-scale variations in thermospheric ozone between 85 and 105 km (M. B. McElroy, private communication, 1980); these variations would have to be of the order of 1015 ozone molecules cm -2, which is as large as the entire nighttime thermospheric ozone column [Riegter et al., 1977]. Also, ff the reactions O + OH (R4) and O + HO: (R5) depend on temperature, then they might proceed more slowly in the colder July atmosphere above 60 km. For modest activation energies [cf. Logan et at., 1978], the resulting increase in ozone could explain half of the observed excess in the annual cycle. As was noted before [Guenther et al., 1979], resonance fluorescence of nitric oxide contributes to the observed intensities at 2555 A; the 5% annual variation would correspond to a yearly cycle of order 4 x 1015 NO cm -: above 50 km. Since thermospheric NO in the tropics is observed to vary by only 1 x 1014 cm -: , such a cycle would require unusually large variations (•100 ppb) in mesospheric NO [cf. Hudson and Reed, 1979, p. 173].
A major uncertainty in the photochemical model for ozone centers on the kinetic data for the reaction OH + HO: (R17). Measurements at low pressures [Burrows et al., 1977;Chang and Kaufman, 1978] give kl7 •' 4 X 10 -ll cm 3 s -l, while the reaction appears to be much faster at 1 atm [DeMore, 1979], kl7 --• 1 X 10 -lø cm 3 s -l. Excellent modeling of the ozone observations is achieved within this range of kinetic measurements. The ozone data indicate a possible transition in kl7 between 50 and 60 km (0.7-0.3 torr). Similarly, an acceptable model of mesospheric CO observations is possible with the lower value of kl7 near 80 km [Allen et al., 1980]. The sensitivity of the photochemistry to the abundance of H20 makes mesospheric water measurements an essential part of any refinement in the theoretical model. Such observations must be devised not only to determine the mean profile of water vapor but also to detect annual short term variability, for example from meteoritic sources.
The photochemical lifetime of (O q-03) is less than 1 day between 40 and 80 km. Thus the response of ozone to changes in the composition and structure of the atmosphere will be rapid. Global monitoring of ozone in the upper stratosphere and mesosphere provides information not only on long-term trends in the concentration of ozone but also on rapid changes in the composition and structure of the upper atmosphere. To interpret the latter, we must increase our confidence in the photochemical modelling of ozone through continued in situ measurements of ozone by rocket and balloon.
Observations of ozone in the upper stratosphere and mesosphere on a global scale are best accomplished by satellite BUV observations at selected wavelengths between 2200 and 2950 •. The instrument should have a narrow field of view (1'ø-3 ø) and be placed in a circular orbit below 800 kin. The orbit must include timely coverage of the full range of local solar zenith angles at a given latitude. The geometric simplicity of nadir observations does not place stringent requirements on the pointing accuracy, makes the analysis of large data sets straightforward and also allows for rapid observations (1 s) of small-scale horizontal structures (8 km) in the atmosphere. Such an instrument must also be able to take advantage of solar and stellar occultations by the atmosphere in order to measure twilight and nighttime ozone profiles.