Impacts of Heterogeneous Chemistry on Vertical Profiles of Martian Ozone

Abstract We show a positive vertical correlation between ozone and water ice using a vertical cross‐correlation analysis with observations from the ExoMars Trace Gas Orbiter's Nadir and Occultation for Mars Discovery instrument. This is particularly apparent during L S = 0°–180°, Mars Year 35 at high southern latitudes, when the water vapor abundance is low. Ozone and water vapor are anti‐correlated on Mars; Clancy et al. (2016, https://doi.org/10.1016/j.icarus.2015.11.016) also discuss the anti‐correlation between ozone and water ice. However, our simulations with gas‐phase‐only chemistry using a 1‐D model show that ozone concentration is not influenced by water ice. Heterogeneous chemistry has been proposed as a mechanism to explain the underprediction of ozone in global climate models (GCMs) through the removal of HO x . We find improving the heterogeneous chemical scheme by creating a separate tracer for the HO x adsorbed state, causes ozone abundance to increase when water ice is present (30–50 km), better matching observed trends. When water vapor abundance is high, there is no consistent vertical correlation between observed ozone and water ice and, in simulated scenarios, the heterogeneous chemistry has a minor influence on ozone. HO x , which are by‐products of water vapor, dominate ozone abundance, masking the effects of heterogeneous chemistry on ozone, and making adsorption of HO x have a negligible impact on ozone. This is consistent with gas‐phase‐only modeled ozone, showing good agreement with observations when water vapor is abundant. Overall, the inclusion of heterogeneous chemistry improves the ozone vertical structure in regions of low water vapor abundance, which may partially explain GCM ozone deficits.


X -2 :
This Supporting Information contains further information on the methodology for the cross-correlation analysis and the 1-D model chemical scheme used in the analysis. Text S1 describes the data filtering used for the observed ozone and water ice profiles and covers Tables S1 and S2. Text S2 describes the cross-correlation methodology in more detail.
Text S3 explains the chemical rates used in the heterogeneous scheme and the two model comparisons undertaken to validate the 1-D model. Figures S1 to S5 are covered in Text S2.

Vertical Profiles: Units And Data Filtering
This section applies to the vertical cross-correlation in Section 2.2 of the manuscript.
Water ice profiles are retrieved in ppm using pressures and temperatures from the Global Environmental Multiscale Mars (GEM-Mars) GCM (Liuzzi et al., 2020), while ozone is retrieved as number density (number of molecules / cm 3 ) (Patel et al., 2021). To keep the ozone and water ice data as consistent as possible and to reduce any additional errors arising from using two GCMs, both data are converted to parts per million by volume (ppmv) using the temperatures and pressures from the same GCM.
The temperatures and pressures from the Open University modelling Group Mars GCM (MGCM) dataset are used for the conversion. The MGCM is utilised for the investigation, as it has been run with data assimilation of MCS temperature and dust, and thus is optimised for this analysis. The ozone dataset is converted from the retrieved unit, molecular density, to ppmv via the ideal gas law, using data from the MGCM.
Water ice profiles are retrieved in parts per million (ppm) using temperatures and pressures from the Global Environmental Multiscale Mars (GEM-Mars) GCM (Liuzzi et October 14, 2022, 4:23pm : X -3 al., 2020). Therefore, in order to convert them into ppmv with MGCM temperatures and pressures, the water ice dataset is first converted into number density using the GEM-Mars GCM data, before the data is converted into ppmv with MGCM data.
Once converted into comparable units, the water ice dataset is filtered to remove data with high uncertainty (ozone filtering is described in the main analysis). The mean and median relative errors are used as they give a general summary indication of the distribution of errors. Table S1 shows the different levels of filtering applied to water ice, given the mean and median relative error, while Table S2 shows the results of the sensitivity analysis when implementing different minimum abundance requirements for the ozone and water ice profile pairs. The minimum of 6 datapoints per profile is already included in the profile pair count.

Cross-Correlation: Methodology
This section describes the cross-correlation analysis used in Section 2.2 of the manuscript. The general standardised discrete correlation is given by Chatfield (1983): where X and Y are random variables, σ is the standard deviation of those variables, and the covariance is given by: where µ is the mean and N is the total number of observations. A standard correlation makes the assumption that the dependent variable (e.g. ozone) is correlated at the same point of the independent variable (e.g. water ice). For example, if x i and y i are two single variables from datasets X and Y , then a standard correlation only tests the relationship at point i. Two variables may have a lagged correlation, as is often the case in time series where the effects of X on Y are delayed. A variation on the standard correlation to account for this is cross-correlation and a standardised version of this is given by Chatfield (1983): where z is an array of lags which Y iterates through and the limits for the summations change for each lag. The results of a standardised correlation are bound between −1 ≤ cor X,Y (z) ≤ 1, where 1 is a perfect positive correlation, -1 is a perfect negative correlation, and 0 indicates no relationship. Correlation tests work with the assumption that any potential relationship between the variables is linear. While water ice and ozone may not fit this assumption, the analysis will provide a guidance to the patterns between the variables.
The results are then tested using a Student's T-test with a two-tailed test at a significance level α = 0.05. The significance test assumes that the correlation follows a normal distribution and has a mean of zero and a standard variation of 1, which is a suitable assumption as the correlation values are standardised. Correlation values are converted to critical values using equation 4, which is dependent on the number of datapoints in the correlation, n i .
October 14, 2022, 4:23pm where t i is the critical value and r i is the correlation for the i th occultation.
For a universal comparison, the critical values are then changed into p-values; p-values less than α are deemed statistically significant. The maximum correlation for each occultation is defined by the lowest p-value, min(p i<α ), which is the most significant value.  The largest discrepancy between the two model runs is between 30-50 km. In the MGCM simulation ,ozone abundance is between 0.1-1 ppmv, while it is < 0.1 ppmv in the 1-D MPM simulation throughout the sol. The ozone features visible in the 1-D MPM at these altitudes (30-50 km) do arguably appear in the MGCM, but at a higher altitude (> 50 km). This is likely due to the water ice distribution, as in the 1-D MPM a layer of water ice cloud forms between 25-45 km, and has very little diurnal variation. In contrast, water ice in the MGCM forms in the late evening and persists through the night before subliming into water vapour the following morning. The presence of water ice causes the adsorption of HO x , which decreases the HO x abundance. In the MGCM, the heterogeneous scheme converts the adsorbed HO x species to oxygen and water vapour, which can then be recycled to influence the ozone formation or destruction. At night, the recycled oxygen can combine with molecular oxygen to form ozone, increase the ozone abundance, while during the day, the recycled water vapour can be photolysed to produce  Figure S1. Abundance of the main chemical species affecting ozone taken at 1200 hours, latitude 0 • S, L S = 5 • with the 1-D MPM (left) run with the new heterogeneous chemistry scheme and (right) the old heterogeneous chemistry.