Journal of Geophysical Research: Atmospheres

Processes that account for the ozone maximum at the mesopause

Authors


Abstract

[1] The presence of a maximum in ozone density and mixing ratio in the mesopause region has been known for several decades although the measurement database is still quite limited. This ozone layer is formed in the vicinity of the maximum in atomic oxygen number density. In this paper, we present simulations of the ozone maximum from a three-dimensional dynamical chemical model. Ozone variability is dominated by the diurnal cycle in ozone, which fluctuates between low concentrations in sunlight and high concentrations in darkness. However, the diurnal variability also has strong contributions from atmospheric tides and from slow changes in chemical concentrations following sunrise and sunset. In this study, we find that the magnitude of the ozone secondary maximum is closely tied to the temperature. The very low temperatures at the mesopause accelerate the formation of ozone and inhibit the loss. This factor and the location of the atomic oxygen density maximum both contribute in approximately equal measure to determining the altitude of the ozone layer. The magnitude of the nighttime ozone maximum is sensitive to the eddy and molecular diffusion rates primarily through the influence of these processes on the concentration of hydrogen.

1. Introduction

[2] The maximum in ozone mixing ratio in the upper mesosphere and lower thermosphere (MLT) is known as the secondary maximum. Its existence has been known since the 1970s [Evans and Llewellyn, 1972; Hays and Roble, 1973; Miller and Ryder, 1973]. The nighttime mixing ratio of ozone in the MLT is comparable to that found in the stratospheric maximum (around 10 ppm). Daytime mixing ratios are substantially smaller but are significantly higher than seen in the middle and upper mesosphere.

[3] Satellite observations have given global coverage of ozone in the upper mesosphere, although many of the available observations stop short of the altitude of the mixing ratio maximum. Daytime measurements were made by the Solar Mesospheric Explorer (SME) [Thomas et al., 1983, 1984a, 1984b] at a fixed local time around 3 pm. The SME observations indicated that there was a semi-annual variation in the upper mesospheric ozone mixing ratio: a finding that has not been confirmed by other studies. Solar occultation measurements, which give profiles at sunrise and sunset, were made by HALOE [Brühl et al., 1996], ATMOS [Gunson et al., 1990], and ORA [Fussen et al., 2000]. Full daytime local time coverage was obtained from the High Resolution Doppler Imager (HRDI) [Marsh et al., 2002]. The HRDI observations showed a diurnal variation in the ozone mixing ratio. Analysis by Marsh et al. [2002] indicated that tidal transport and photochemistry both contribute to the variations. Despite the enormous increase in data, the details of the ozone secondary maximum have been difficult to establish because global nighttime ozone measurements in this region are only now becoming available. Two shorter missions were able to obtain ozone in both day and night: ATLAS [Bevilacqua et al., 1996] and CRISTA [Kaufmann et al., 2003] although these data do not always resolve the structure of the ozone mixing ratio peak. Nighttime ozone measurements from GOMOS (E. Kyrölä et al., unpublished manuscript, 2005) averaged over one year indicate that the mixing ratio maximum is between 90 and 100 km, while the density maximum is between 86 and 92 km. Day and night ozone densities are also now being measured by the Sounding the Atmosphere by Broadband Emission of Radiation (SABER) on the Thermosphere Ionosphere Mesosphere Energetics and Dynamics (TIMED) satellite and by the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) on the Envisat satellite but validation of these two data sets is not complete.

[4] The basic reasons for the existence and diurnal variability of the secondary maximum are understood. Photolysis of molecular oxygen by solar radiation in the Schumann-Runge bands and Herzberg continuum produces atomic oxygen; ozone (O3) is formed throughout the middle atmosphere from the 3-body associative reaction of O and O2. Early models of an oxygen-only atmosphere [e.g., Shimazaki and Laird, 1970] indicate a steady decrease of ozone density through the middle atmosphere. Photochemical reactions involving hydrogen and oxygen species give a significant reduction in the odd oxygen (atomic oxygen plus ozone, denoted Ox) concentration through a broad region from the stratopause to the mesopause [Bates and Nicolet, 1950]. The addition of the hydrogen chemistry leads to low ozone densities through the middle mesosphere. Near the mesopause, photolysis of O2 in the Schumann-Runge bands produces large amounts of Ox, giving a maximum in ozone despite the continuing loss due to chemical reactions involving hydrogen.

[5] Ozone in the upper mesosphere is much more abundant at night. The photolytic destruction rate is extremely fast (100s) for ozone, which is optically thin in the upper mesosphere. The nighttime chemical lifetime is about an order of magnitude longer. The relative size of the ozone maximum is therefore larger during night. Photochemical equilibrium models also predict that the OH concentration, which is involved in the chemical cycle that destroys ozone, is higher at night [Allen et al., 1984]. This leads to ozone in the upper mesospheric region below the maximum that is very low at night as well as during day.

[6] Although photochemical considerations alone explain the existence of an ozone maximum, the magnitude and time dependence of ozone in these simple models is not correct. Vertical transport in the form of diffusion can help [Shimazaki and Laird, 1970; Thomas and Bowman, 1972] but is not sufficient for a realistic simulation.

[7] In this paper we look at the processes that control the secondary maximum in a dynamical-chemical numerical model of the middle atmosphere. In particular, we address the following questions. How important are molecular diffusion and eddy diffusion in the overall structure of this ozone layer? What contribution is made by tidal transport? How important is the sensitivity of chemical reaction rates to temperature? What is the ozone sensitivity to atomic oxygen and hydrogen?

[8] The conclusions show that the low temperatures found in the mesopause are very important in the location and magnitude of the ozone maximum. Diffusion of atomic oxygen and hydrogen from the thermosphere also have important impacts. Upward transport of water (the primary source of atomic hydrogen) by advection and diffusion makes a strong contribution to the ozone amount and variability.

2. Model Description

[9] The numerical model used in this study is a new version of the ROSE model. A dynamics-only version of this model has been used before and is described in Smith [2003] and Smith et al. [2004]. The model solves the three-dimensional primitive equations on a sphere as a function of time. The dynamical equations are solved using a leapfrog scheme on a fixed grid; the current resolution includes 36 points in latitude, 32 in longitude, and 64 vertical levels.

[10] In this version of the model, the upper boundary has been extended well into the thermosphere to avoid upper boundary effects for tides and other waves. The model domain extends from the lower stratosphere (90 hPa) to 21 scale heights, which corresponds to a geometric altitude of about 188 km. The vertical resolution is 0.3 scale height (approximately 2.1 km in the middle atmosphere). A sponge layer in which the basic dynamical fields, including temperature, are damped toward MSIS climatology [Hedin, 1991] is introduced gradually above 130 km. This further reduces boundary impacts on waves. In addition, this relaxation is necessary to keep the temperature in the observed range since the model does not include thermospheric processes such as Joule heating and cooling by the 5.3 μm NO emission.

[11] In the model integrations described in this paper, the radiative forcing is partially interactive; it uses the three-dimensional temperature computed by the model but daytime densities of radiatively active gases are prescribed from observations. This ensures that dynamical fields are identical in various model integrations, enabling the isolation of photochemical processes. Several processes are involved in the conversion of absorbed solar energy to heat. Most of the absorption is done by molecular oxygen and ozone; absorption by other molecules is very important for the chemistry of trace species but involves only a small amount of the total energy. A significant part of the absorbed energy is used for photolysis. That part of the energy that goes into photolysis is not available until the products recombine. Even at recombination, some of the energy goes into excitation of vibrational or rotational excited states and can be radiated out of the region without ever affecting the thermal structure. The ultraviolet heating and infrared cooling are solved using the algorithm of Zhu [1994], which accounts for non-local thermodynamic equilibrium. Chemical heating from 7 exothermic reactions is included [Smith et al., 2003] and is an important heat source in the mesopause region. The airglow loss is accounted for by applying the efficiency factors of Mlynczak and Solomon [1993] to the solar and chemical heating rates.

[12] The most important of the sub-grid processes is the propagation, breaking and dissipation of gravity waves. Momentum deposition by a spectrum of waves is represented using the algorithm developed by Hines [1997a, 1997b]. The effect of stationary topographically-forced gravity waves is included using a parameterization based on Lindzen [1981].

[13] Vertical diffusion includes contributions from three processes. (1) A Richardson-number dependent vertical diffusion coefficient adapted from that described by Williamson et al. [1987] is applied to horizontal momentum and potential temperature. This diffusivity is beneficial for model stability but has little impact on the atmospheric structure in the MLT. (2) Eddy diffusion due to breaking gravity waves is a product of the gravity wave parameterizations. The magnitude of diffusion in the Hines [1997a, 1997b] gravity wave parameterization depends on the adjustable parameter Φ6, which in the model is set to 0.25. This parameter controls the magnitude of the diffusion coefficient relative to the gravity wave drag. For both the Hines [1997a, 1997b] and Lindzen [1981] parameterizations, an effective inverse Prandtl number of 0.3 is used. (3) Molecular diffusion of heat and trace gases is included with the parameterization of Banks and Kockarts [1973]. According to this parameterization, the diffusion rate DM for species i is given by

equation image

where Mi and Mg are the masses of species i and the background air in amu; ρ is the background density; T is temperature, and DM is in SI units. Molecular diffusion also includes a displacement of gases relative to one another, according to their masses. This is represented as a vertical diffusion velocity wM.

equation image

where Hi and Hg are the vertical scale heights of species i and the background air and αt is −0.38 for atomic and molecular hydrogen and zero for other species. This parameterization is valid for minor species diffusing through a background; it does not hold up where major species are in diffusive equilibrium. Tests indicate that application of this approximate form in the ROSE model is justified because molecular separation of the main atmospheric gases does not occur to any significant extent below 110 km.

[14] Horizontal dynamical diffusion is provided by the filter described by Shapiro [1971]. There is, in addition, a truncation of zonal wavenumbers greater than two at the latitudes nearest the poles (±87.5°) in the dynamical prognostic variables. There is no horizontal diffusion or spectral truncation of chemical species.

[15] The lower boundary conditions on horizontal winds, temperature and geopotential change with each timestep. These values are interpolated from once-daily NCEP analysis data beginning in 2002 and continuing through the end of 2003. These have been filtered to include only zonal mean and wavenumbers 1–4; shorter scales do not propagate into the middle atmosphere. Amplitude and phases of the diurnal and semidiurnal tides at the lower boundary are taken from the Global Scale Wave Model [Hagan et al., 1999] for the appropriate time of year. Upper boundary conditions are diurnal averages derived from the MSIS model [Hedin, 1991].

[16] The model solves for the concentrations of 29 chemical species containing reactive oxygen (O, O(1D), O2, O3), hydrogen (H, H2, OH, HO2, H2O, H2O2), nitrogen (N, NO, NO2, NO3, N2O, N2O5, HNO3, HNO4), chlorine (Cl, ClO, HOCl, HCl, ClONO2, CF2Cl2, CFCl3) and carbon (CH4, CH2O, CO2, CO). N2 is held constant below 100 km. In the thermosphere, its concentration is determined as a residual from the net density of the atmosphere and the densities of other constituents. The seasonally varying lower boundary conditions on species is taken from the model of Brasseur et al. [2000]. The upper boundary condition for species with significant thermospheric concentrations is taken from the MSIS model. Both upper and lower boundary conditions are imposed by specifying the volume mixing ratio (vmr) at a grid point beyond the model domain.

[17] The model chemistry solver is described in Marsh et al. [2003] and all of the reactions are listed in that paper. The chemical species are split into two categories. The first includes species that react very rapidly in the middle atmosphere: O, O3, H, OH, HO2, N, NO, NO2, Cl, and ClO. The concentrations of these are solved simultaneously using a Newton Raphson solver which is iterated until it converges. For the remaining species, the time-dependent chemical behavior is determined using a Gauss-Seidel iteration. Species from each of the two solvers are used in the other, and the pair is iterated several times. O(1D) is determined from the assumption of photochemical equilibrium. Most of the reaction rate coefficients are taken from the compilation by Sander et al. [2003]. The exception are several rates not available in that report, which are obtained from Campbell and Gray [1973] and Roble [1995].

[18] Variation of photolysis rates with solar zenith angle and overhead ozone column amount are calculated from a look-up table. The overhead molecular oxygen is based on the mean model value. The table is computed from the TUV (tropospheric ultra violet) model [Madronich and Flocke, 1998] over the wavenumber range 120–735 nm. Note that this includes all radiation that penetrates to the MLT region and below. However, the photolysis of O2 by radiation shorter than 120 nm is not included. As a result, the model would calculate concentrations of O that are too small in the thermosphere. To account for this effect, the ratio between O and O2 above 115 km is constrained to follow the MSIS empirical model. Note that this adjustment will ensure that the impact of molecular diffusion on the balance between O and O2 is accounted for as well. Below 115 km, the O to O2 ratio is determined self-consistently by the photochemistry, transport, and diffusion included in the model.

[19] Since ion chemistry is not included, the model underpredicts the production of NO in the lower thermosphere. The additional NO is specified in the model by using the empirical model based on SNOE observations [Marsh et al., 2004].

[20] Several photolysis cross-sections include a temperature dependence. In the model, a global average temperature profile is used to determine the photolysis cross-sections. We are not aware of any measurements of the temperature dependence of O3 cross-section at MLT temperatures. There is an indication from laboratory studies by Molina and Molina [1986] and others that the cross-section decreases with decreasing temperature; however, the known temperature dependence occurs in a part of the spectrum (277–347 nm) that contributes a very small part of the total ozone absorption in the upper mesosphere.

[21] Chemical transport uses the semi-Lagrangian advection algorithm of Smolarkiewitz and Rasch [1991]. In addition, two diffusive processes are applied to the chemical mixing ratios. Eddy diffusion calculated from the gravity wave parameterization is applied uniformly to all chemical species. A species-dependent molecular diffusion is applied to all species that are important at the mesopause and above. The algorithm is used down as far as the middle mesosphere because of evidence [e.g., Lopez Puertas et al., 2000] that diffusive separation can play a role to altitudes around 75 or 80 km.

3. Structure and Variability of the Ozone Secondary Maximum

3.1. Simulation of the Ozone Secondary Maximum by the ROSE Model

[22] The relative magnitude of the mesopause ozone maximum depends on the manner of defining the abundance. This, and much of the analysis below, is presented for the month of March. Figure 1 shows the simulated day and night ozone concentrations for March in terms of volume mixing ratio (vmr) and log of number density (log[O3]). The averages include all longitudes and all days, but a single local time. The altitude of the maximum is higher for vmr than for log([O3]); by both measures it is higher during night than during day. In this paper we will focus primarily on mixing ratio, which is conserved in displacement. In other words, the mixing ratio of an air parcel will not change due solely to movement of that parcel to higher or lower pressure. Mixing ratio is also the relevant quantity for diffusion; a gradient in mixing ratio is necessary for transport by advection or diffusion.

Figure 1.

Midnight (left) and midday (right) ozone volume mixing ratio (top; units are ppmv) and log10 of number density (bottom; units are cm−3) simulated by the ROSE model for March.

[23] The altitude of a given pressure level varies with latitude, longitude and time. The global average heights of the ozone maximum in the model are 93 km for vmr at midnight; 89 km for vmr at midday; 89 km for number density at midnight and 87 km for number density at midday.

[24] Figures 2 and 3show the annual evolution of ozone vmr at the equator and at 55°N for midnight and midday. An ozone maximum is evident at all times. There is a seasonal variation in both the pressure level of the layer and the maximum mixing ratio. During equinoxes, the altitude of the nighttime maximum at the equator is higher while that of the daytime maximum is lower. This difference reflects the importance of the diurnal tide. The equatorial mixing ratio around equinoxes has more high frequency variability, another result of tidal forces. There are differences in the pressure of the maximum and the seasonal variations between the equator and 55°N. Note that the temporal evolution at different times of the night or day will be different from that shown for midday and midnight due to differences in the phase of the diurnal tide and to evolution of chemical species with photochemical lifetimes of hours.

Figure 2.

Annual evolution of ozone vmr (ppm) at the equator as a function of pressure during night (top) and day (bottom).

Figure 3.

Annual evolution of ozone vmr (ppm) at 55°N as a function of pressure during night (top) and day (bottom).

[25] To illustrate the latitudinal structure of ozone, Figure 4 shows the mixing ratio of ozone at .001 hPa as a function of latitude and local time. In addition to the large day/night differences in ozone vmr, the equatorial ozone also varies with local time in response to the diurnal tide. Note that temperature and atomic oxygen mixing ratio also vary at the equator in a tidal pattern. Nighttime ozone in low latitudes is out of phase with temperature. Atomic hydrogen has little diurnal variability but has a strong latitudinal gradient, increasing from the spring to the fall hemisphere.

Figure 4.

Latitude and local time variation ozone vmr, temperature, atomic oxygen vmr (units are 10−3), and atomic hydrogen vmr at .001 hPa during March.

[26] The very large difference in the ozone between day and night that is evident in Figures 14 indicates that there is a large amplitude diurnal cycle. Figure 5 shows the ozone mixing ratio as a function of local time at the equator and at 55N. Both diurnal photochemistry and tidal transport of ozone and other constituents influence the ozone variations. The diurnal behavior in the upper mesosphere is strongly dominated by photolysis but, as shown in Figure 4, tidal variations make a contribution.

Figure 5.

Ozone mixing ratio as a function of local time at the equator (top) and at 55°N (bottom) at four pressure levels.

3.2. Processes That Affect the Ozone Concentration

[27] The continuity equation for volume mixing ratio μ of a trace species (in this case ozone) is

equation image

where Kzz is the eddy diffusion rate, P is the production rate, L is the loss coefficient, z is scaled log pressure z = −Hln(p/ps), p is pressure and ps is surface pressure. The molecular diffusion terms DM and wM are defined in equations (1) and (2), respectively. The material derivative of μ, expressed in flux form, is

equation image

This includes the three-dimensional transport in longitude (λ), latitude (ψ), and log pressure; a is the radius of the Earth.

[28] First, we consider photochemical equilibrium, which is a good predictor of the ozone distribution. Under equilibrium conditions, P = Lμ, or μ = P/L. From the photochemistry, there is a single production term for ozone, reaction (R1), which is the associative reaction between O and O2 (see Table 1). There are eight kinetic reactions and one photolytic reaction that destroy ozone. Three of the loss rates are listed in Table 1; the additional reactions of ozone with O(1D), OH, HO2, NO, NO2, and Cl are included in the model but do not play a significant role in the vicinity of the ozone secondary maximum. Figure 6 gives the ozone loss reactions for midnight and midday conditions. From this figure, it is evident that photolysis (reaction (R4)) is larger than the next most important reaction by about an order of magnitude. The photolysis coefficient does not change perceptibly with altitude during all daylit times because ozone is optically thin in this altitude range. To a good approximation, ozone loss by kinetic reactions can be neglected during the day. Even at night, only reactions (R2) (reaction with atomic oxygen) and (R3) (reaction with atomic hydrogen) are significant destroyers of ozone at MLT altitudes.

Figure 6.

Vertical profile of the loss coefficients associated with the main reactions that destroy ozone at midday and midnight, averaged over longitude and latitude for March. The three leading terms (reactions (R2), (R3), and (R4)) are labeled; the remaining curve (long dashed line) is the sum of all other reactions. Units are s−1.

Table 1. Chemical Reactions That Control MLT Ozone
Reaction
(R1)
equation image
(R2)
equation image
(R3)
equation image
(R4)
equation image

[29] From the above, we can write an approximate formula for ozone mixing ratio during day O3(d) and night O3(n), based on equilibrium assumptions. First consider night. The production and loss rates depend on density. With a background number density of ρ, ozone density is O3ρ, oxygen density is Oρ, etc.

equation image
equation image

From the ideal gas law, ρ = p/kT, where p is pressure, k is Boltzmann's constant and T is temperature. On a constant pressure surface, the density is inversely proportional to temperature. Other temperature dependence in equation (6) comes from the reaction rates.

equation image
equation image
equation image

When the temperature is lower, ozone is created more readily from reaction (R1) while at the same time it is destroyed more slowly by reactions (R2) and (R3). All three rates favor higher equilibrium ozone concentrations at lower temperatures. The density on a pressure surface is larger for lower temperature; this also contributes to higher ozone where temperatures are lower.

[30] Figure 7 shows the model ozone and the equilibrium approximation for day and night during March. Two different ratios are shown for night, one based on equation (6) and the other similar except that reaction (R2) is omitted. For both day and night, the equilibrium assumption tends to slightly overestimate the ozone mixing ratio at the maximum.

Figure 7.

A comparison of nighttime (left) and daytime (right) equilibrium ozone and simulated ozone (ppm). The solid lines are the actual ozone in the model; the long dashed line is from equation (5) (night; includes (R2) and (R3)); the short dashed line (night) includes (R3) but not (R2); and the dash-dot line (day) is from equation (10).

[31] The curves in Figure 7 were made at the specific model timesteps of midday and midnight. At these times, photochemical variability is at its smallest; times close to the day/night terminator will have significantly larger departures from equilibrium. (In addition, the chemical balance of ozone in the upper mesosphere shifts somewhat in the perpetual low sun angle at the polar night terminator; this has been described by Marsh et al. [2001] and Hartogh et al. [2004] and will not be discussed further here.) Due to dynamical activity from tides, diffusion and other motion, the ozone will never be in equilibrium even if all other chemical reactions can be safely neglected. An equilibrium ozone that is different from the model ozone could be because dynamical motions are changing the ozone itself or the background temperature or chemical concentrations on a time scale comparable to the photochemical timescale. As evident from Figure 4, the background T and O have large diurnal (tidal) cycles in low latitudes; these changes in the background keep the ozone perpetually away from equilibrium.

[32] Even though ozone is not precisely in equilibrium, the formula giving the equilibrium structure (equation (6)) can be used to show the impact of various parameters on the ozone vertical structure. Figure 8 compares how the nighttime equilibrium ozone mixing ratio changes with altitude due to each of the contributions in equation (6): density, mixing ratio of O, mixing ratio of O2, mixing ratio of H, and the dependence of reaction rates on temperature. The units are arbitrary: only the vertical variation is illustrated on the figure. The ozone vertical structure is approximately given by the product of the 6 curves. The increase of ozone with altitude in the mesosphere is a result of increase in atomic oxygen mixing ratio and the temperature effects of the two reaction rates; the background density and the mixing ratios of O2 and H act to decrease ozone across the MLT region. It is evident from the righthand panel that the impact of the reaction rates on the vertical structure is comparable to the impact of the atomic oxygen density. For this period (monthly average for March at midnight), the location in the ozone maximum at the equator predicted by reaction rate temperature dependence is identical to that predicted by the maximum in O density. However, this is not always the case; the relative altitudes predicted from the two processes vary with the phase of the tide and can be offset.

Figure 8.

Left panel shows vertical profiles (arbitrary units) of the terms that make up the nighttime equilibrium ozone concentration: k1, O vmr, O2 vmr, atmospheric number density ρ, 1/k3, and 1/H. Right panel shows these grouped into three terms: the combined impact of temperature dependent reaction rates k1/k3, the atomic oxygen number density [O] (=O · ρ) and the other chemical contributions O2/H. The curves show the global mean at midnight except where sunlit.

[33] During the day

equation image
equation image

where image is the photolysis rate (reaction (R4) in Table 1). Figure 9, comparable to Figure 8, shows the relative roles of various environmental factors in the vertical structure of daytime ozone. Although the mixing ratios of atomic and molecular oxygen in the upper mesosphere are similar to those at night, the combined density and mixing ratio effect is significantly different from the nighttime values due to the higher order dependence on density. The impact of the stronger density dependence is to lower the altitude of the ozone maximum relative to that at night. In the middle mesosphere, the structure is affected by the strong diurnal cycle in O. Note also that due to a different temperature structure (opposite phase of the diurnal tide), the reaction rate contribution has a quite different vertical structure. The tidal dependence is strongest near the equator but is present with reduced impact at other latitudes.

Figure 9.

Left panel shows vertical profiles (arbitrary units) of the terms that make up the daytime equilibrium ozone concentration: k1, O vmr, O2 vmr, atmospheric number density ρ, and image Right panel shows these grouped into three terms: k1, [O][O2] (=O · O2 · ρ2), and image The curves show the global mean at midday except during polar night.

4. Processes That Determine Structure of the Secondary Maximum

[34] It is evident from Figures 8 and 9 that the vertical level of the secondary ozone maximum is a result of two factors: the relatively low temperature and the local maximum of atomic oxygen number density. This section determines a quantitative assessment of the importance of these factors.

4.1. Importance of Low Mesopause Temperature

[35] The low mesopause temperature accelerates the ozone production by (R1) and slows the loss by (R3) and (R2). Figure 10 shows the results of a model integration in which the reaction rates were calculated from temperature except where the temperature dropped below 225 K; otherwise, they were calculated using T = 225 K. The magnitude of the ozone maximum is reduced substantially compared to the normal case (Figure 1). The difference is larger during night, when both production and destruction are affected; during day, temperature affects the production but not the photolytic loss rate. This temperature dependence is the reason why early numerical simulations of the ozone density gave lower values than observed; there were not sufficient temperature data to establish the very cold mesopause temperatures and high mesopause altitude. For example, Shimazaki and Laird [1970] used a fixed temperature of about 210 K between the middle mesosphere and the thermosphere, whereas measurements indicate that the minimum temperature ranges between 130 and 170 K [e.g., Mertens et al., 2004]. Several other early studies used temperature profiles from the 1966 U.S. Standard Atmosphere, which gave realistically cold upper mesospheric temperatures (minima ranging from 165 to 202 K, depending on latitude and season), but increasing temperature above 90 km.

Figure 10.

Simulated ozone values when the temperature used in the reaction rates is constrained to be above 225 K, for comparison with Figure 1. Panels show midnight (left) and midday (right) volume mixing ratio (top; units are ppmv) and log10 of number density (bottom; units are cm−3) for March.

[36] All temperatures in the model at .001 hPa are lower than 225 K. Therefore, in the model case just described, there will be no temporal variations in reaction rates at .001 hPa, and in fact through the entire MLT region. Figure 11 shows the ozone from this case. Comparison with Figure 2 indicates that not only is the amount of ozone substantially reduced over the normal case, but also the short-term variability is less. Part of the equatorial ozone temporal variability is a direct result of temperature fluctuations.

Figure 11.

Annual evolution of ozone vmr at the equator as a function of pressure during night (top) and day (bottom) for the model case in which the temperature used in the reaction rates is constrained to be above 225 K.

4.2. Local Production Versus Transport of Atomic Oxygen and Hydrogen

[37] Referring back to Figures 7 and 8, we see that the two factors that determine the altitude of the ozone maximum are response of temperature dependent reaction rates to cold temperatures, as discussed in the previous subsection, and the peak in the number density of atomic oxygen. The photochemical lifetime of atomic oxygen is long at the altitude of the ozone maximum [e.g., Brasseur and Solomon, 1986] so transport is important for its distribution. The global mean vmr at a given pressure {μ} ≡ equation imageequation image μdλdψ can be determined from the general chemical continuity equation.

equation image

Figure 12 shows a comparison of the percentage changes in atomic oxygen vmr in 24 hours due to various processes. Photochemical processes give a net production of atomic oxygen above .0006 hPa (approximately 96 km) and a net loss below that level. This means that, at the altitude of the ozone vmr maximum, production and loss of atomic oxygen balance to zero. In the levels below the ozone maximum, eddy diffusion and resolved advection both act to increase atomic oxygen.

Figure 12.

March average of the percentage change of global mean atomic oxygen over 24 hours due to various processes. The curves give the net photochemical production and loss (P-L), resolved advection, eddy diffusion, and molecular diffusion.

[38] Molecular and eddy diffusion are expressed in a similar form (see equation (12)) but have different coefficients (DM and Kzz) and have opposite effects on the oxygen budget above .001 hPa. The coefficient for molecular diffusion grows rapidly with decreasing pressure and increasing temperature; its net effect in the lower thermosphere is always to increase the mixing ratio of O. In physical terms, the diffusion downward is always larger than the diffusion upward because of the sharp increase in DM with altitude. Since atomic oxygen mixing ratio increases with height, there is a net global increase due to this process. The eddy diffusion coefficient increases more slowly; the impact of this diffusion then depends on the vertical structure of atomic oxygen and of the diffusion coefficient. Where the oxygen gradient is steepest below the mesopause, eddy diffusion, like molecular diffusion, acts to diffuse O downward and to increase the local concentration. However, in the lower thermosphere, eddy diffusion tends to move atomic oxygen from higher to lower density, decreasing its concentration.

[39] Since both the eddy and molecular diffusion are parameterized, they are subject to significant uncertainty. The eddy diffusion in this model is less than 100 m2 s−1 everywhere below 100 km, which is smaller than assumed in some models, so the relative roles of eddy and molecular diffusion cannot be determined from this study.

[40] The conclusion from Figure 12 is that the transport effects of diffusion and advection are comparable or larger in magnitude to the photochemical production and loss at all levels in the MLT region. The change in atomic oxygen is 3–5% per day due to these processes. Pure photochemical estimates give a lifetime of O of about a month at 95 km [e.g., Brasseur and Solomon, 1986], within the same range.

4.3. Impact of Eddy and Molecular Diffusion

[41] The model integrations were altered to determine the impact of diffusive transport on the ozone maximum. Two additional model integrations were performed. In one, eddy diffusion was suppressed for all chemical constituents. In the other, molecular diffusion was suppressed below 125 km and included normally above that altitude; in the levels with no molecular diffusion, both DM and wM were set to zero. Figure 13 compares the global mean O, H and midnight and midday O3 for cases where either eddy or molecular diffusion was suppressed. When there is no eddy diffusion, the atomic hydrogen is significantly less at all levels while the atomic oxygen density profile shifts upward. Suppression of molecular diffusion leads to a much larger concentration of H and a significant upward displacement of the [O] profile. Separate model experiments (not shown) indicate that it is the relative displacement effect of molecular diffusion (wM) that is primarily responsible for the large differences in H, whereas this effect is negligible for oxygen at the altitude of the ozone layer.

Figure 13.

Global mean profiles of midday ozone vmr, midnight ozone vmr, zonal mean atomic oxygen number density, and zonal mean hydrogen vmr averaged over March for three model cases: the normal case; an integration in which molecular diffusion has been suppressed at altitudes below 125 km (no DM); and an integration in which eddy diffusion has been suppressed at all levels (no Kzz).

[42] The nighttime ozone responds more strongly to the differences in hydrogen. Eddy diffusion brings up water, the source gas of hydrogen, from below, which destroys ozone during nighttime. The eddy diffusion rate is a major uncertainty in global middle atmosphere models because of a lack of observational constraints. These model results indicate that the nighttime ozone maximum could serve as a good diagnostic to determine the ranges of this parameter. In contrast, molecular diffusion, which adds oxygen to and removes hydrogen from the MLT region, enhances the ozone maximum. During day, diffusion has no effect on the photolytic loss rate and the loss by reaction with H is minor, so the ozone maximum response to diffusion follows the response of atomic oxygen density. As was the case for oxygen, daytime ozone in the MLT is much more sensitive to the molecular than to the eddy diffusion.

[43] Advective transport has a direct impact on Ox (Figure 12) but also an indirect impact on ozone. The southern hemisphere to northern hemisphere gradient in nighttime ozone seen in March (Figure 4) is a response to a pole-to-pole gradient in hydrogen of the opposite sign. The summer to winter circulation during solstice seasons in the MLT has been bringing water from below into the southern latitudes while transporting hydrogen-poor air downward into the northern latitudes. Such a global asymmetry maximizes at the end of the winter (for example, March). Lower ozone in the fall hemisphere is a response to this transport circulation. The seasonal cycle is also evident in the top panel of Figure 3.

[44] The advective transport (Figure 12) also includes tides. Another model case was performed to determine how much of this advective transport is due to tides. In this integration, tides were suppressed in the model; see Smith et al. [2003] for a description of how tides are removed. The results (not shown) indicate that in the global mean during March, the impact of tides is significantly smaller than the impact of diffusion. However, the tidal impact is locally large at low latitudes where the vertical velocity due to the diurnal tide is largest. It also depends on time of year. Model calculations for April [see Smith et al., 2003] indicate a greater role for tides.

5. Discussion and Conclusions

[45] This paper investigates the processes that are responsible for the maximum in ozone mixing ratio near the mesopause. This feature is known as the secondary maximum and is certainly secondary in number density when compared to the original ozone maximum in the stratosphere. However, it is now evident from observations that the nighttime ozone mixing ratios are about the same as, or even larger than, those at the stratospheric maximum.

[46] The three-dimensional ROSE dynamical-chemical model has been used to simulate the ozone secondary maximum for a period of two years. Additional model cases determine the importance of various processes. The principal conclusions are as follows:

[47] 1. Without the very low temperature at the mesopause, the ozone concentration would be significantly smaller. The increasing temperature with altitude in the lower thermosphere is an important contributor to the sharp decrease in ozone above the mesopause.

[48] 2. Eddy diffusion acts to decrease the nighttime ozone concentration by bringing water up from below. Water is the source for atomic hydrogen, which destroys ozone. Eddy diffusion has much less impact on the daytime ozone.

[49] 3. Molecular diffusion acts to increase the nighttime ozone concentration by moving atomic hydrogen upward, out of the MLT region. Molecular diffusion also increases the concentration of atomic oxygen below about 105 km. During both day and night, the ozone concentration is increased by molecular diffusion through this additional supply of atomic oxygen.

[50] 4. Advective transport of hydrogen-containing species by the mean circulation leads to a seasonal cycle in ozone in the middle and high latitudes.

[51] 5. The role of tides in generating the MLT ozone maximum is small.

[52] Ozone is one of the principal absorbers of ultra-violet solar radiation in the atmosphere. In addition, the exothermic reaction between ozone and atomic hydrogen is one of the dominant heating processes in the upper mesosphere. The variable solar flux due the 11-year solar cycle and other variations will affect ozone and will alter the heating rate and temperature in the MLT. All of these factors point to a need to understand the ozone and its variability throughout the entire atmospheric system. The present modeling paper provides a framework for investigating new global ozone observations.

Acknowledgments

[53] Support for this work was provided by the NASA Office of Space Science. The National Center for Atmospheric Research is operated by the University Corporation for Atmospheric Research under the sponsorship of the National Science Foundation.

Ancillary