5 Discussion
5.1 Column abundance
In this work, the MCD was used to provide a priori values for the column abundance retrievals with the uncertainty set equal to the abundance. With a posterior-to-prior error ratio analysis, we demonstrated that the synergy injects a significant amount of information to the retrieval, and obtains highly robust column abundances. The climatology presented here displays a water vapor cycle consistent with established literature, both in terms of magnitude and seasonal and meridional variations. Abundances are observed to peak near the edge of the seasonal frost cap in spring, forming an annulus of vapor encircling and following the retreating seasonal cap into the early summer. The appearance of the water annulus is consistent with the proposed mechanism for the seasonal cycling of the water in the Northern hemisphere, in which vapor subliming from the seasonal water frost annulus re-condenses on the surface of the retreating CO2 cap. The water decrease poleward of the annulus is observed consistently for all seasonal intervals in our composite average, yet annual variations have been previously reported (Pankine et al., 2010).
Although the overall behavior is well known and also agrees well with the MCD model, significant differences do exist. The synergy column abundances deviates most prominently from the MCD in terms of absolute value with significantly lower abundances, particularly in the summer NH. The observed northern sublimation maximum is 30% lower than MCD estimates, and the sublimation season onset itself is observed to occur later in time. In the SH, the model and observations are in better agreement, and similar to what was reported by Clancy et al., (Clancy et al., 2017) using CRISM occultation data, who also found that retrieved water vapor abundances matched MCD model estimates better in the SH than in the NH. The synergy yields slightly higher values in the southern early summer, resulting in a somewhat asymmetrical relationship between the synergy and MCD, where the synergy finds a lower summer peak in the NH, but a larger peak in the SH.
When compared to previous works, the synergy northern maximum abundance was quite consistent with PFS, SPICAM and the revised TES abundances of 60-70 pr-μm (Fouchet et al., 2007; Pankine et al., 2010; Trokhimovskiy et al., 2015), while CRISM obtained a slightly lower sublimation peak in MY 28 and 29 of around 50 pr-μm (Smith et al., 2009). Although the synergy finds a smoothed average of around 50 pr-μm at 75°N and Ls=105°-120°, some local and transient instances of abundances up to 100 pr-μm occur. Observations from the Limb and Nadir Observation channel of the NOMAD instrument on the ExoMars TGO satellite agree well with the synergy in terms of seasonal variations, however, the northern maximum obtained by the synergy is significantly higher than those found by NOMAD for the corresponding time and place (just above 30 pr-μm) (Crismani et al., 2020).
The southern maximum coincides in time with previous results, but the large asymmetry between the NH and SH maxima observed by SPICAM and CRISM is not as prominent in the synergy dataset (see Figure 6, where a few very high column abundances are observed), as the northern maximum is normally a factor of 2 higher than the southern peak for the corresponding season (Figure 10). On average, the synergy finds a southern maximum of ∼33 pr-μm, significantly higher than SPICAM. It should be noted that the location where the largest SH abundances were observed were at latitudes not captured by previous TES and PFS studies. It should also be pointed out that observations in the south polar region are much sparser than elsewhere, and measurements from several years are binned together, whereas the observations of the north polar region are abundant and mostly from MY 27. Smith (2004) found that the year-to-year variations can be as high as 10 pr-μm, and might thus explain why we observe instances of high vapor abundances in the south.
Outside the summer maximums, the synergy again is most similar to SPICAM and PFS, and agrees very well also with NOMAD. During Ls=0°-50°, the mean low latitude (0°-30°N) CIA was 7-8 pr-μm for the synergy, SPICAM, PFS and NOMAD, and ~5 pr-μm for CRISM. Later, during Ls=150°-180° for the same latitudes, the mean abundances were 13-15 pr-μm for the synergy, SPICAM, NOMAD and CRISM, ~12 pr-μm for PFS.
The difference between the synergy and other datasets is most likely due to differences in calibration and data processing techniques, even though diurnal variations cannot be excluded. For example, NOMAD samples local times from 08:00 to 16:00, anf PFS covers local times into the late afternoon. TES sampled the equatorial region and mid latitudes around 14:00 and 02:00, with only data captured during the 10:00-14:00 range being used to assemble the revised dataset presented by Pankine et al., (2010). No evidence supporting diurnal variations have yet been uncovered using OMAGE or SPICAM (Maltagliati, Montmessin, et al., 2011; Trokhimovskiy et al., 2015), and in the synergy, any diurnal variations are lost in the averaging process as PFS and SPICAM cover a broader time interval. Crismani et al., (2020) found no evidence for substantial diurnal variation in the total dayside water vapor column, thus the plausibility of diurnal variations causing such a large spread in column abundances is still considered unlikely.
5.1 Partitioning index
The strongest motivation for the use of a spectral synergy retrieval approach is to access information on the vertical distribution of water vapor. We have shown that during the polar cap sublimation periods, the magnitude of the near-surface vertical confinement matches model predictions quite well, though discrepancies in the meridional partitioning gradient are significant. For both hemispheres the vertical partitioning remains high and fairly constant (±0.2) for all seasons and latitudes, while displaying a wave-like behavior. Poleward of the polar cap edge however, the hemispheres differ. In the south the partitioning index is observed to drop for all seasonal intervals except for during mid spring. In the north the PI seems to be decreasing at first between 70° and 80°N, and then rapidly increases beyond the polar cap edge, especially so for mid and early spring. This polar cap behavior is well reproduced by the global climate model used to construct the MCD, except during spring for both hemispheres.
The largest differences in MCD and synergy vertical confinement in the northern hemisphere are found at mid-latitudes after Ls=150° (see Figure 9). The column abundance, which never exceeds 20 pr-μm, agrees best with the MCD in this region (though still the synergy finds a lower value), while the obtained synergy partitioning was more than 50% higher than model estimates. This might be indicative of less water escaping through the hygropause than what is estimated in the MCD. For Ls=135°-150°, Figure 11 shows that the MCD and synergy are quite consistent for high latitudes, both finding a PI of 0.7 at 70°N. Further equatorward in the drier low latitudes, where model and synergy agree quite well with regard to column abundances, the partitioning differs significantly. The model suggests the confinement decreases monotonically, reaching a PI=0.4 at 20°N, while the synergy maintains a strong confinement, obtaining a PI of ~0.7 at 20°N, having barely changed despite a drastic reduction in the total water column. This could suggest that the circulation incorporated in the current model at low latitudes is too strong, causing the MCD partitioning to decrease more quickly towards the equator. The difference could also possibly be due to diurnal “breathing” of the regolith, actively exchanging water with the atmosphere and thus maintaining a near-surface layer.
Overall, the synergy finds a more variable vertical partitioning than what the model suggests, which corresponds well with results from solar occultations observations with SPICAM (Maltagliati et al., 2013). This demonstrates that the synergy is particularly useful at mid to low latitudes where atmospheric dynamics influence the vertical partitioning, and over the polar regions where seasonal variations in the vertical partitioning are large and not reproduced by the model. It would be of great interest to compare the synergistic partitioning with high resolution vertical profiles from for example the solar occultation instruments NOMAD and ACS on TGO. This will be included in future work, although as mentioned earlier, the ability of these instruments to probe the water vapor content in the very low atmosphere is not always present. As the southern hemisphere normally has a higher dust loading than the north, conditions are most favorable in the north high latitudes. At low latitudes where we observe large differences between synergy and model, continuously high dust loading will also make direct comparisons between synergy and TGO difficult.
6 Conclusions
Presented here are the results from a spectral synergistic retrieval method applied to water vapor nadir measurements from PFS and SPICAM sampled over seven Martian years. The synergy produces a highly reliable water vapor climatology with geographical and temporal patterns consistent with established literature. When compared to the LMD MCD, the synergy tends to retrieve lower total column abundances, in absolute differences the deviance is biggest for the northern summer sublimation peak, while in relative terms the most significant discrepancies are found at mid latitudes. In the southern hemisphere the synergy and MCD correspond very well. Other differences of note include timing and latitudinal extent of the sublimation onset, which occurs earlier in the MCD, and extends much further equatorward. The synergy finds very comparable column abundances to previous works using single spectral domain approaches with SPICAM and PFS (Fouchet et al., 2007; Trokhimovskiy et al., 2015), somewhat higher values than CRISM (Smith et al., 2009), and slightly lower than TES (Pankine et al., 2010; Smith, 2002).
The ability to extract information on the vertical distribution of water vapor from nadir observations is a unique capability of the spectral synergy approach. The synergy is unable to produce a vertical profile of fine resolution, but it can set reliable constraints on the partitioning of the water column, differentiating between the near-surface content below 5 km and the rest of the column. Significant differences between the vertical partitioning over the north and south hemispheres are revealed, where the southern hemisphere exhibits a generally weaker confinement coupled with a stronger seasonal dependence and latitudinal variations than in the north. The near-surface confinement from the synergy overall differs from the MCD especially at low and middle latitudes where the synergy finds a stronger near-surface confinement than MCD estimates. The synergy also finds that the meridional spread of this strong confinement is larger than what the model suggests, with a strong confinement far south of the polar region. There appears to be no clear connection between a peak in total column abundance and the amount of vertical partitioning. In general, the synergy finds that the vertical confinement is subject to rapid and local variations, and can change significantly even while the total column abundance remains stable, or remain stable while the column abundance varies.
We have shown that by combining two separate spectral intervals, within which water vapor possesses diagnostic features, increased robustness is brought to the retrieval of column abundances as well as additional information about the vertical content, as compared to the commonly used single-interval retrieval approach. The combination of more accurate column abundances and constraints on the vertical distribution is essential for our understanding of the processes that control the distribution and transport of volatiles in the lower atmosphere.
Considering that current knowledge of the water distribution in the lowermost layer of the atmosphere is mainly based on GCMs, the comparison between the synergy partitioning results and the predictions of the MCD is of particular interest. The significant discrepancies between the two indicate that our understanding of the physics that shape the vertical distribution of atmospheric water on Mars is incomplete.