Solar Cycle Modulation of Nighttime Ozone Near the Mesopause as Observed by MLS

Eleven‐year solar cycle variations of nighttime ozone near the secondary ozone maximum layer in the mesosphere are analyzed with Aura Microwave Limb Sounder (MLS) observations since 2004, fully covering solar cycle 24. Produced primarily from the recombination of molecular oxygen (O2) with atomic oxygen (O) transported from the lower thermosphere, the mesospheric nighttime ozone concentration is proportional to atomic oxygen density [O], which itself is modulated by ultraviolet (UV) solar cycle variations. MLS nighttime ozone data and UV data at 240 nm from Solar Radiation and Climate Experiment (SORCE) Solar‐Stellar Irradiance Comparison Experiment (SOLSTICE) show a positive correlation over the solar cycle. Nighttime O3 and nighttime carbon monoxide (CO) distributions are highly correlated with each other with similar seasonal and solar cycle variations, because both [O3] and [CO] depend strongly on UV photolysis and are modulated by Eddy diffusion in this region. Nighttime ozone correlates strongly with temperature, with a generally positive correlation, except at high latitudes during boreal winter.


Introduction
The presence of an abundant ozone layer in the upper mesosphere and lower thermosphere (UMLT) is known as the secondary ozone maximum, next to the maximum ozone layer in the stratosphere (Smith & Marsh, 2005, and references therein). While extensive studies have been carried out to search for evidence of interannual variation of stratospheric O 3 associated with the 11-year solar cycle, both from observations and models (Aquila et al., 2016;Ball et al., 2019;Li et al., 2016;Maycock et al., 2016Maycock et al., , 2018Merkel et al., 2011;Swartz et al., 2012), the secondary ozone maximum layer has been less explored than the stratospheric ozone.
While the radiation, dynamics, and chemistry in the UMLT are largely driven by solar radiation, solar cycle-driven UV irradiance changes are expected to significantly modulate ozone abundances within the secondary ozone maximum layer. Substantial variation in solar UV irradiance, larger than several percent over the 11-year solar cycle (Lee et al., 2016), directly impacts ozone chemistry, because the major single source of secondary maximum ozone production is photochemistry of oxygen species from the Chapman cycle (Chapman, 1930). This ozone production photochemistry involves reaction between atomic oxygen (O) and molecular oxygen (O 2 ), where M indicates the total density of the atmosphere.
The major photochemical reactions for ozone destruction are: Photolysis of ozone [(R2)] from solar irradiance in the Schumann-Runge bands (176-192.6 nm) and Herzberg continuum (200-240 nm) is known as the largest ozone loss mechanism in the secondary ozone maximum region (Smith & Marsh, 2005). Ozone is mostly destroyed via photolysis by solar radiation [(R2)] into atomic oxygen [O] and molecular oxygen [O 2 ], but these oxygen species also produce ozone through the recombination process [(R1)]. During the day, reaction [(R2)] dominates the loss process and the ozone concentration under chemical equilibrium is determined by where the k 1 values represent rate coefficients for reactions [(R1)]. Daytime ozone has a very short lifetime because the J 2 loss term is much larger than the k 3 and k 4 loss terms. During nighttime, when [(R2)] is absent, reactions with atomic oxygen [(R3)] and atomic hydrogen [(R4)] are the only significant sinks for ozone in the mesosphere. As shown in Smith and Marsh (2005), [(R3)] is a slow reaction, and the nighttime ozone loss is dominated by [(R4)]. Recent study has shown that the reaction of O + O 3 also plays an important role in O 3 loss process near the mesopause (Zhu & Kaufmann, 2018). A similar assessment was found by Allen et al. (1984), showing in the upper mesosphere above~80 km the nighttime ozone is largely balanced by reactions [(R1)] and [(R4)]. Hence, in chemical equilibrium, to first order, the UMLT nighttime ozone abundance can be deduced from Thus, the amount of nighttime ozone under chemical equilibrium is not only proportional to [O] and [O 2 ] but also inversely proportional to atomic hydrogen density [H]. Under chemical equilibrium, primarily through the temperature dependence of reaction [(R1)], an anticorrelation between ozone and temperature is expected (Brasseur & Solomon, 2005) It is important to put the odd oxygen chemistry into perspective of the UMLT dynamics as the photochemical lifetimes are competing with the dynamical lifetimes associated with these species . The chemical lifetime of odd oxygen increases from 10 4 s in the upper mesosphere to 10 8 s in the lower thermosphere, whereas the dynamical time scale of the eddy diffusion coefficient (k zz ) is generally~10 6 s. This leads to a transition from chemical to dynamical control in terms of odd oxygen variability with a crossover at~85 km. Under chemical control, as expressed by equation (2) Garcia and Solomon (1985) emphasized the role of eddy diffusion in modulating odd oxygen variability in the UMLT region. With weaker diffusion/mixing (lower k zz ) in the UMLT, odd oxygen resides longer in the lower thermosphere, allowing it more time to accumulate. In the case of stronger diffusion/mixing, more odd oxygen is subject to chemical loss at a lower altitude, leading to a lower concentration in the UMLT. The coupled dynamical and chemical influences are able to explain the SAO variations seen in [O] and [O 3 ], as the UMLT k zz is generally smaller near the equinox. Such dynamics-chemistry coupling has been further explored in recent studies (e.g., Dikty et al., 2010;Garcia et al., 2014;Smith & Marsh, 2005;Tweedy et al., 2013).
In addition, ionization by energetic particle precipitation, primarily by high-energy electrons with energies greater than 100 keV in the auroral regions, is suggested as another short-term mesospheric ozone depletion mechanism (Andersson et al., 2014;Turunen et al., 2016). By enhancing the HO x and NO x productions, large solar photon events can also induce significant ozone loss in the UMLT (Jackman et al., 2008(Jackman et al., , 2011. Diurnal variations of mesospheric ozone have been explored by Marsh et al. (2002), using day-night differences observed by the high resolution Doppler imager and by Dikty et al. (2010), using daytime measurements from Sounding of the Atmosphere using Broadband Emission Radiometry (SABER). The mesospheric ozone climatology and diurnal differences between nine satellite data sets have been documented in great detail by Smith et al. (2013). It was pointed out that potential sampling biases from different ozone observations and models could result in improper diurnal variation of mesospheric ozone (Studer et al., 2014).
There is a positive ozone response to the 27-day solar rotational variation from a number of studies (Gruzdev et al., 2009;Hood & Zhou, 1998;Ruzmaikin et al., 2007;Thiéblemont et al., 2018). The 11-year solar cycle response of ozone in the middle atmosphere has been also explored by many authors (e.g., Crooks & Gray, 2005;Dhomse et al., 2016;Haigh et al., 2010;Hood et al., 1991;Huang et al., 2008;Soukharev & Hood, 2006;Tang et al., 2018). Despite many efforts to explore different data sets and analysis methods, the amplitude of variations associated with the solar cycle and the response mechanism of mesospheric ozone to UV variation are still not well characterized. In the lower mesosphere, the observed ozone responses to solar forcing from previous works do not agree in magnitude nor in phase with respect to the solar cycle (Beig et al., 2012;Merkel et al., 2011).
Characterizing solar cycle ozone variations and their coupling with atmospheric chemistry and dynamics has been a challenge because ozone measurements inferred from photochemical reactions are often prone to errors in other photochemically active species used in the ozone retrievals. Scarcity of reliable long-term observations and lack of independent validation of ozone data in the upper atmosphere have led to many controversial views on ozone variability.
Long-term observations from two satellite instruments, SOlar Radiation and Climate Experiment (SORCE, 2003-present) and Microwave Limb Sounder (MLS, 2004-present) on Aura, now enable us to examine the direct and indirect responses of ozone in the UMLT region to 11-year solar cycle variability. MLS operates with a regular limb scan up to~90 km, observing directly through the secondary ozone peak in the upper atmosphere. While the profile from individual scans is noisy, MLS daily mean radiances are quite sensitive to small ozone changes in the mesopause region. In this paper, we discuss nighttime variations of MLS ozone and their relationships to 11-year UV variations at the lower end (i.e., 0.002 hPa) of the secondary ozone maximum layer.

Microwave Limb Sounder (MLS)
We use MLS version 4.2× (V4.2×) data (Froidevaux et al., 2008) for daily nighttime ozone, at the 0.002 hPa level. We refer the nighttime ozone as MLS observations for solar zenith angles larger than 90°from either ascending or descending orbits. The ascending nodes of the orbit (i.e., when the spacecraft is moving toward the north) cross the equator at 1:45 ± 15 p.m. local time, and the descending nodes of the orbit (i.e., when the spacecraft is moving toward the south) cross the equator at 1:45 ± 15 a.m. Since the orbital period of the Aura spacecraft is approximately 100 min, the local time of MLS observation changes within ±25 min from the ascending or descending nodes depending on latitudes (https://aura.gsfc.nasa.gov/scinst.html).
The ozone data are aggregated into 43 latitude bins between 82°N and 82°S, because Aura MLS sampling does not cover the regions poleward of 82°N/°S latitude. The vertical resolution of O 3 is from 5 to 7 km over the 0.01 to 0.001 hPa domain. The recommended topmost pressure level for the ozone for scientific use is 0.02 hPa (https://mls.jpl.nasa.gov/data/v4-2_data_quality_document.pdf). The 2σ estimate of systematic uncertainty is~0.2 ppmv or 50% at 0.02 hPa. Although the recommended pressure level for ozone studies is 0.02 hPa, there is still some MLS sensitivity in the uppermost mesosphere (Froidevaux et al., 2008). The levels up to 0.002 hPa contain valuable information on mesospheric ozone (Livesey and Froidevaux, 2019, private communication). The MLS ozone data at this level have been studied previously by Andersson et al. (2014) and Hocke (2017), showing that they are scientifically useful even above the validated single-profile height levels.
The MLS mesospheric ozone is retrieved from an emission feature at 235.71 GHz. As shown in Figure 1, MLS resolves this emission line spectrally with a 25-channel filter bank with variable bandwidths between 6 and 96 MHz. In the mesosphere, the O 3 emission mostly resides in the 6-MHz center channel (Channel 13), while the radiances from Channels 1-3 and 23-25 can be used to estimate the background from calibration residuals or other variations that may change with time and orbit. The difference between Channel 13 and the background radiances constitutes the mesospheric O 3 radiance signal, which contributes the most to MLS mesospheric ozone VMR retrievals. The daytime radiance (not shown) is nearly zero at upper mesospheric heights. Since the nighttime O 3 emission near the mesopause is very weak, a careful removal of the instrument background radiance is critical to observe weak solar cycle variations.
To further evaluate the MLS sensitivity to mesospheric nighttime ozone abundance and its variation with respect to the solar cycle, we average three 96-MHz wing channels (Channels 1-3 and 23-25) at tangent heights above 80 km and subtract this signal from the Channel 13 radiances to obtain the background radiances at about 88 km. Figure 2 shows a time series of the extracted nighttime ozone radiances at 88 km for 2004-2019. The daily ozone radiances show strong annual and semiannual variations at 88 km around the equator. The solar cycle amplitude is obtained from the 11-year period fit after the seasonal cycle is removed. In this case, the regressed 11-year peak-to-peak amplitude is 90 mK with an uncertainty of 5.9 mK. This clear solar cycle signal in radiances agrees very well with variations in MLS retrieved ozone at 0.002 hPa, providing further confidence in the results of mesospheric ozone analyses presented later in this paper.
The residual of the fit is ±0.4 K and is mostly from the measurement noise. There may be some contributions from atmospheric variability that is not seasonal. The MLS Level 2 algorithm employs the subtraction of a varying background from the wing channels (Froidevaux et al., 2008). Therefore, the measurement uncertainty is a combined error from the center channel radiance and the estimated background radiance.
MLS mesospheric ozone data can be valuable especially for long-term ozone variation studies. The MLS standard O 3 product acquired from 235.71 GHz radiances shows no significant temporal drift, in comparison to other ground-based measurements (Hubert et al., 2016).
We have also analyzed MLS mesospheric carbon monoxide (CO) and temperature (T) data (Livesey et al., 2017), since mesospheric CO is a good dynamics tracer due to its long lifetime. The typical single-profile precision of MLS V4.2× CO varies from 0.02 ppmv at 100 hPa to 0.2 ppmv at 1 hPa, and 11 ppmv at 0.002 hPa, with vertical resolution of 4, 3, and 9 km, respectively. MLS temperature in the mesosphere is retrieved primarily from bands near O 2 spectral lines at 118 GHz. The precision of the MLS V4.2× temperature measurement is ±3.6 K, and the bias is ∼3 K between 0.01 and 0.001 hPa. In the mesosphere between 0.01 and 0.001 hPa, the MLS temperature profiles have a vertical resolution of 8-13 km, a precision of 2.2-2.5 K, and a 4-8 K cold bias.

SORCE SOLSTICE
The SOlar-Stellar Irradiance Comparison Experiment (SOLSTICE) on SORCE measures ultraviolet (UV) solar irradiances from 115 to 320 nm with a resolution of 0.1 nm, an absolute accuracy of better than 5%, and a relative accuracy of 0.5% per year Snow et al., 2005). The SOLSTICE measurements are made of a pair of identical spectrometers, SOLSTICE A and SOLSTICE B. Each instrument independently measures ultraviolet solar irradiance in two intervals: far ultraviolet at 115-180 nm and mid-ultraviolet at 180-320 nm.

Nighttime Ozone Climatology
The vertical profile of ozone VMR in the UMLT is shown for the tropics in Figure 3. The MLS nighttime ozone exhibits a secondary peak near the mesopause (~0.001 hPa). At the tropics, the secondary maximum amount of nighttime ozone is~6 ppmv, and this amount is comparable to~60% of the ozone amount at the stratospheric maximum (~10 ppmv). Daytime ozone VMR is substantially smaller than nighttime ozone in this region but is significantly higher than that in the middle and lower mesosphere.
The 15+ years of Aura MLS ozone and CO observations provide new insights on the interannual variation of ozone associated with chemistry and dynamics in the middle atmosphere. There is a strong correlation  Figure 4. The figure shows MLS nighttime ozone climatology at 0.002 hPa pressure level in different latitude bands. Even though the MLS observations cannot give a full coverage of the secondary maximum ozone layer in the UMLT above 0.001 hPa, MLS ozone clearly shows the lower part of the maximum layer in the upper mesosphere. The ozone VMR at 0.002 hPa reaches up to 6 ppmv in high latitudes and up to 7 ppmv near the tropics at this level. However, the plot also shows low amount of ozone (below 2 ppmv) during the local summer season.
In the high-latitude (60°N/°S-82°N/°S) regions, MLS nighttime ozone VMR shows a large annual cycle superimposed on a semiannual cycle, with an early winter maximum and an early summer minimum. The ozone climatology also displays hemispheric differences between SH (Southern Hemisphere) and NH (Northern Hemisphere), with slightly more ozone in the NH than in the SH during the early winter maxima. In each case, the MLS nighttime ozone climatology shows a semiannual oscillation (SAO) with two seasonal maxima, in early winter (October/November in NH and April/May in SH) and in early spring (February/March in NH and August/September in SH). The peak in early winter is slightly higher than that in early spring.
In the tropics, the seasonal variation of nighttime O 3 shows an obvious SAO with two equinoctial peaks during April and October, with a slightly larger peak in April. Similar to those at high latitudes, the SAO seasonal minima during solstice show~2 ppmv ozone VMR, approximately one third of the equinoctial maximum values. The amplitude of the SAO in the tropics corroborates the monthly O 3 densities estimated from 1D model simulations by Gattinger et al. (2013). Their simulated nighttime O 3 densities and the observed Envisat/GOMOS (Global Ozone Monitoring by Occultation of Stars) O 3 densities are approximately three times larger during the SAO maximum in April, compared to the densities during SAO minimum in August.
Overlaid red curves in Figure 4 denote MLS nighttime CO climatology during the same period. The seasonal variations of nighttime ozone are quite similar to those of CO, indicating similar dynamical and photochemical processes associated with atomic oxygen. Details of the CO climatology have been described by Lee et al. (2018).

Interannual Variation of Nighttime Ozone
MLS nighttime zonal mean ozone VMR time series are shown in Figure 5 for (a) northern high latitudes (60°N-82°N), (b) southern high latitudes (60°S-82°S), and (c) the tropics (12°S-12°N) at 0.002 hPa. The ozone in SH high latitudes (Figure 5b) shows a similar annual cycle as in the NH, but with approximately a 6-month phase shift versus that from the NH, as already shown in Figure 4. The ozone VMR time series shows the SAO cycle in the tropics, with two almost equal maxima.

Solar Cycle Variation of Nighttime Ozone
MLS O 3 zonal mean anomaly data show a strong interannual variation during the analysis period (2004-2019) that correlates with the 11-year solar cycle. Shown in Figure 6 are interannual variations of high-latitude ozone anomalies during 2 months near the seasonal maximum (November/December for NH and March/April for SH) at 0.002 hPa, overlaid with SORCE/SOLSTICE UV at 240 nm. In the NH high latitudes (Figure 6a), the nighttime ozone anomaly varies by up to 2 ppmv, showing 1 ppmv less than average ozone amounts during the solar minimum period. Similarly, the ozone anomaly variation in the SH high latitudes (Figure 6b) shows an in-phase variation with the 11-year solar cycle within 1.5 ppmv.
Our result, more than 25% of nighttime ozone variation with a solar cycle correlation at this level, corroborates the Hamburg Model of Neutral and Ionized Components (HAMMONIA) simulations by Beig et al. (2012). Their 2D simulation exhibits a strong positive (∼40-45%/100 sfu) ozone response to solar forcing near 0.0023 hPa (∼90 km) in 40°N/°S-60°N/°S latitude bands. Since the F10.7 radio flux changes~100 sfu over one solar cycle, their solar cycle variation amplitudes (in %/100 sfu) are comparable to our estimate of solar cycle response of ozone. However, our MLS data analysis level, 0.002 hPa, is close to the top of the retrievals and its averaging kernel shows a substantial (~40%) contribution from the modeled a priori at the top level (Froidevaux et al., 2008) which does not vary with solar cycle. This may cause an underestimation of the actual amplitude of the solar cycle response of the current analysis, although it is difficult to estimate how much uncertainty this effect would add.
In the upper mesosphere (0.001 hPa), there exists an in-phase ozone variation within 3 ppmv (50%) versus the solar cycle. In the lower mesosphere below 0.002 hPa, the zonal mean ozone is less than 2 ppm and decreases by 0.5 ppm at 0.005 hPa when UV is increased (not shown). Even though this small amount of solar cycle variation is difficult to assess from the MLS observations, ozone in the lower mesosphere is expected to be decreased with solar cycle when nighttime ozone increases in the upper mesosphere. Since more UV is absorbed by increased ozone above, less UV is transmitted to the lower mesosphere. Reduced UV leads to less O 2 photolysis, and therefore less O 3 is expected in the lower mesosphere.

Correlation Between Nighttime Ozone (O 3 ) and Carbon Monoxide (CO)
CO has a major source from CO 2 photolysis in the UMLT region, that is, In the lower thermosphere, the main loss of CO is through three-body recombination with atomic oxygen, which is a slow reaction: From the mesosphere to stratosphere, a rapid loss process becomes increasingly important through the reaction , Thus, the CO concentration under chemical equilibrium can be expressed by where J is the photo-dissociation rate constant for CO 2 [(R5)] and k is the reaction rate constant. In the UMLT the CO concentration is largely a result of photochemical balance between reactions [(R5)] and [(R6)], meaning that the second term in the dominator above becomes negligible. Like nighttime ozone, CO variability has both dynamical and chemical influences in the upper mesosphere with the crossover height at~75 km , slightly lower than the height of odd oxygen. Therefore, the UMLT CO is largely under dynamical control, which makes it a good dynamical tracer.
Supply of UV provides favorable conditions for CO production with CO building up as the solar radiation increases during solar maximum. CO shows an in-phase relation with the 11-year solar cycle throughout the mesosphere, because the long lifetime of CO in the upper mesosphere allows CO to be transported downward toward the lower mesosphere Lee & Wu, 2013).
As shown in Figure 7, the strong correlation between MLS nighttime ozone and CO VMRs for all latitude bins confirms the dynamical modulation of nighttime ozone in the UMLT region. CO and O 3 are highly correlated in the tropics, in particular. The correlation is relatively low during NH late winter, January and February, when an anomalous amount of CO transport from the thermosphere happens during stratospheric sudden warming (SSW) events. CO has a photochemical lifetime longer than 30 days, due to low amounts of OH during the polar night (Minschwaner et al., 2010), which allows it to be transported to the mesosphere through the strong mesospheric polar vortex during the SSW (Lee et al., 2011). Due to very low concentrations of ozone during JJA (June, July, and August) in the NH, the correlation is nearly zero during that time. The regression coefficients and linear correlation coefficients between ozone and CO, for each month, are summarized in Table 1.
Covariance between ozone and CO VMR time series is further shown for the tropics (12°S-12°N) in Figure 8. They show similar strong SAO oscillations and interannual variabilities. It was suggested in the early study by Dunkerton (1982) that the SAO is generated by gravity wave momentum, as waves in the easterly and westerly phases are alternatively propagated through the semiannual variation of wind phases near the stratopause. Many authors proposed detailed mechanisms associated with the SAO for equinox and solstice seasons, but vertical advection and eddy diffusion are commonly accepted as major contributing factors for the SAO generation (Gattinger et al., 2013;Qian et al., 2009, and references therein).
The mechanism of eddy diffusion variations, as suggested by Garcia and Solomon (1985), can explain the SAO in CO. They found that eddy diffusion in the lower thermosphere was maximum in the summer hemisphere and had a significant effect on the seasonal variation of atomic oxygen. The major CO loss in the mesosphere is through reaction with OH. Thus, a stronger k zz in the upper mesosphere will lead to lower [CO] by mixing it down to the chemically lossy mesosphere. The eddy diffusion modulation mechanism can also explain the nighttime ozone variations in this region through its relationship with [O] (see the next section).

Temperature Dependence of Nighttime Ozone
If the ozone in the secondary maximum layer were under chemical control, its variation would be sensitive to temperature variations and anticorrelation between ozone and temperature is expected (Hood, 1986; Note. The regression coefficients represent y = ax + b, when x = CO VMR (ppmv) and y = O 3 VMR (ppmv).  Hood et al., 1991;Smith & Marsh, 2005;Tang et al., 2018). The very low temperatures at the mesopause can accelerate the formation of ozone and suppress the ozone loss. It is also suggested that the nighttime secondary ozone layer is enhanced during SSWs due to adiabatic cooling in the polar mesosphere caused by polar descent (Tweedy et al., 2013).
To examine the temperature dependence of nighttime mesospheric ozone, we correlate the MLS daily nighttime means of ozone mixing ratio and temperature at 0.002 hPa. As shown in Figure 10, the temperature dependence of nighttime ozone is not monotonously positive nor negative but changes with location and season, indicating complex roles of temperature in ozone chemistry and middle atmosphere dynamics. Depending on location and time of the analysis, there is a mixture of positive and negative correlations. Strong positive temperature dependence during fall (SON) is found in all three latitude bins. For other seasons, the overall temperature-ozone correlations are generally positive, except northern high-latitude winter.
The observed positive ozone-temperature correlation is consistent with the expectation that the nighttime near-mesopause ozone is largely under dynamical control . The residual meridional circulation near the mesopause, due to gravity wave breaking (e.g., Holton, 1983;Leovy, 1964), leads to descending motions over the winter pole with associated adiabatic heating. This dynamical heating produces higher temperatures in the winter upper mesosphere. Because of the large amount of ozone near the mesopause (i.e., the lower part of the secondary ozone layer), the residual meridional circulation and vertical motions at high latitudes produce higher ozone concentration in the winter hemisphere. On the contrary, ascending motions over the summer pole lead to adiabatic cooling and low temperature anomalies. The ascending motions transport low amounts of ozone in the lower mesosphere to the upper mesosphere in the summer pole. Therefore, the positive ozone-temperature correlation near the mesopause is indicative of dynamically induced variations.
The negative correlation in the northern high-latitude winter, which mostly come from SSWs associated with polar mesospheric cooling, may indicate a period of time when the UMLT ozone is under chemical control, as suggested by Tweedy et al. (2013). Their model simulation shows that during SSWs the mesospheric upwelling induces mesospheric cooling, which is consistent with the idea of elevating the chemically controlled region to a higher altitude. The lack of negative correlation in the southern high-latitude winter implies a much weaker mesospheric upwelling without SSW.

Roles of Atomic Oxygen [O] and Eddy Diffusion
Since [(R5)] is the major source for atomic oxygen and CO in the UMLT, the eddy diffusion variation is expected to modulate both of their concentrations, as suggested in Garcia and Solomon (1985). As seen in the SME [O] observations (Thomas, 1990) and in the SCHIAMACHY observations (Kaufmann et al., 2014), the SABER daily mean [O] values also exhibit a SAO in the tropical upper mesosphere (Smith et al., 2010). The equinoctial maxima at this height are related to the SAO variation in k zz . Because nighttime [O 3 ] in the UMLT is proportional to [O], its SAO is a manifestation of the SAO from [O], and the root cause is from the k zz modulation.
In the UMLT, k zz can be strongly modulated by the diurnal tide in the tropics. The diurnal tidal amplitude and k zz are anticorrelated (Geller et al., 1997), and small-scale gravity waves are the major contributor to k zz variations (e.g., Weinstock, 1984). Turbulence and mixing induced by breaking gravity waves are converted to k zz in model parameterization schemes. Using model simulation experiments, Meyer (1999) and Akmaev (2001) found that larger k zz values lead to weaker diurnal tide amplitudes in the UMLT region. ] were dominated by chemical processes, one would expect its 11-year amplitude to be proportional to its background mean. To help for this evaluation, Figure 11 compares the 11-year amplitude of nighttime ozone with its background at 0.002 hPa. Except for the months of April, October, and November, the ratio of solar cycle amplitude over the background nighttime ozone is not a constant across latitude. The dynamical contributions need to be considered to interpret the distribution of nighttime ozone solar cycle variations, as discussed in the previous section.

Conclusions
Our analyses of MLS nighttime ozone in the upper mesosphere at 0.002 hPa, coupled with past work on this topic, indicate that forcing from the 11-year solar cycle clearly affects interannual variations near the secondary ozone maximum, likely via direct photochemical processes as well as by indirect modulation of atomic oxygen transport. The O 3 distribution in the middle atmosphere is driven by vertical transport and eddy diffusion, as summarized by a simple diagram in Figure 12. Vertical transport inside the polar vortex brings anomalously large nighttime O 3 values from the lower thermosphere down to the upper mesosphere in high latitudes. Near the tropics, smaller k zz and the increased lifetime of O 3 lead to small local increases in O 3 during equinox, as shown by the mean O 3 mixing ratios in Figure 11.
1. The MLS nighttime ozone seasonal climatology shows that MLS usefully samples the lower part of the secondary ozone maximum near 0.002 hPa, even though it does not cover the whole secondary maximum ozone layer in the UMLT. In general, there exist strong coherences in seasonal variations between MLS nighttime O 3 and CO in the UMLT. Both nighttime ozone and CO show a large annual cycle, with early winter (October/November in NH and April/May in SH) maxima and minima during local summer (DJF in NH and JJA in SH). They also show the SAO, as a secondary peak during local spring (February/ March in NH and August/September in SH). In the tropics, the nighttime ozone seasonal variation shows the obvious SAO with two peaks during April and October, the April maximum being slightly larger than the later peak. 2. The positive correlation between nighttime ozone and solar UV irradiance at 240 nm indicates an apparent in-phase response of nighttime ozone to solar UV irradiance variation. The change of nighttime ozone over the solar cycle can be as high as 1.5 ppmv during winter. This amount corresponds to a 20-25% change in nighttime ozone change over the solar cycle. 3. MLS nighttime ozone and CO VMRs are highly correlated. Positive correlations between them are shown for each month. Similarities in solar cycle variations between nighttime ozone and carbon monoxide in the upper mesosphere suggest that there also exists a solar cycle in atomic oxygen [O] (which also has a long lifetime). While the distribution of [O] is challenging to measure, nighttime [O 3 ] could be viewed as a tracer for [O] at most latitudes. 4. The temperature dependence of nighttime ozone is not monotonously positive nor negative. The temperature-ozone correlations are generally positive, except at northern high latitudes in winter.
Radiative heating due to ozone may result in regional warming and a positive correlation with temperature.
In this study, we analyzed 15 years of UV data at wavelengths near 240 nm. The solar cycle variation of solar spectral irradiance, including UV, is essential to elucidate wavelength-dependent terrestrial interactions with the Sun. In particular, reliable long-term UV data are a key element in characterizing the long-term changes of atmospheric constituents, that is, oxygen, ozone, and carbon monoxide, in both observational and modeling studies.
The Total and Spectral Solar Irradiance Sensor 1 was deployed with two instruments, Total Irradiance Monitor and Spectral Irradiance Monitor on the International Space Station in late 2017, to continue the total and spectral solar irradiance observations (200-2,400 nm). Absolutely calibrated satellite measurements of spectral solar irradiance are vital to be used as model inputs, since most of the climate models largely depend so far on the modeled solar spectrum. The value of such long-term solar irradiance observations, along with the fundamental middle atmosphere measurements cannot be overemphasized for the successful assessment of Earth's climate change in relation to solar forcing.