Investigation on the Absorption Bands Around 3.3 μm in CRISM Data

Recently, the methane seepage detected by Mars Sample Laboratory (MSL) in the Gale crater area during 2013 was conrmed by methane detection by the Planetary Fourier Spectrometer (PFS). While analyzing NIR-IR CRISM data on a site in Oxia Planum area, in the view of a future comparison with data that will be collected by the Rosalind Franklin rover onboard the ExoMars2022 mission, a 3.3 μm absorption was noted in some pixel spectra. Since methane, like other hydrocarbons, shows absorptions in the range 3.1-3.6 μm, we begun to study this band in CRISM data to explore the possibility to look for seepages on Mars surface. The datasets chosen for this study, aside the site in Oxia Planum area, include some sites of observations on Gale Crater and other sites in Nili Fossae area. We used the Planetary Spectrum Generator to simulate CRISM spectra of the different sites, with the diverse concentrations of CH4 spikes. These simulations served to establish the relation between concentration and methane band depths, as seen by CRISM spectrometer. Then, mapping the Modied Gaussian Model t on CRISM data, we extracted the band parameters of the absorptions in the 3.3 μm spectral region. Aside rare, suspected absorptions, an artifact was highlighted. Therefore, we have set a threshold on the depth to consider as depth of potential true absorptions, on the basis of the standard deviation (s) of absorption depth map. Finally, we concluded to favor as potential absorptions: distribution of clusters of pixels in the band mapping not vertically stacked and a threshold value >μ (average)+5σ (standard deviation) of the depth map. These threshold values set the lower limit for each observation on the methane concentration potentially detectable by CRISM. The threshold value varies from one observation to another, in a range between 0.0136-0.0237, that would correspond in a range of lower limit concentrations of 180 and 600 ppbv. We found interesting cluster of pixels which spectra overcome the imposed threshold. We still consider that part of them could still be a kind of unknown artifact. Nevertheless, the aim of this paper is to show that CRISM data can show potential absorptions of methane in such quantities that in some observations are compatible with the order of the methane spikes effectively detected in literature. Even if this work does not conrm nor deny the occurrences of methane seepages in the investigated images it shows a possible method for assessing a condence limit in the detection of this band in CRISM data.


Introduction
The great part of the missions on Mars is focused on the research of past/present life on the Mars surface and subsurface. In these last ten years a number of steps forward have been made in the knowledge of Mars environments that could constitute proxies for the development of primitive lifeforms.
In this sense, the ndings related to methane continue to be intriguing because on the Earth, part of it, formed by microbes from the domain Archaea in anoxic conditions during the Proterozoic age (Pavlov et al., 2003). Nevertheless, methane, like other hydrocarbon compounds, can result from abiogenic geological processes (Etiope et al., 2011(Etiope et al., , 2013. For the methane gas these can be serpentinization of basalts, (Oze and Sharma, 2005), gas absorbed in the regolith ( To date, several instruments both orbiting and onboard rovers have detected methane on Mars surface. Since 2003, several puzzling detections of methane in the Mars atmosphere were done (Atreya et al., 2007). For example, from Earth: using the Cryogenic Near-IR Facility Spectrograph (CSHELL) at the Infrared Telescope Facility (IRTF) and the Gemini ground telescopes, Mumma et al., (2004), detected localized points with > 250 ppbv of methane, a concentration that was then corrected at 45 ppbv, comparing data with PFS measurements (Mumma et al., 2009) For what concerns spikes: the detection of Mumma et al., (2009), that observed a strong release of methane up to 50 ppbv during the Northern Hemisphere summer of 2003. Their observations ranged up to 3 years in which they observed a progressive decreasing in the methane mixing ratio over the three years and a substantial variation according to latitude. They concluded that the occurring of this strong release of methane was limited spatially and temporarily.
Later, other spatial-temporal investigations with other instruments con rmed this conclusion. All these ndings provide interesting constraints on the occurrence of methane on Mars.
In fact, summarizing the results of the previous observations, the source of Mars methane should be spatially restricted but also temporarily restricted with potential sources in form of seepages (Yung et al., 2018): from micro-seepages, (Etiope et al, 2015;Moores et al., 2019), mini-seepages (Etiope et al., 2015) to macro-seepages, (Oehler et al., 2017). Furthermore, the different amounts detected in spikes Table 1, seasonal oscillations and non-detections are compatible with geological seepage dynamics that involve changes in gradient of pressure and in the permeability of rocks .
The variability of the amount of methane observed in these measurements and the estimated time for methane sequestration by photochemistry and oxidation that spans from months to years (Mumma et al., 2009), allow to hypothesize also that the plume release observed was recent at the time of the measurements.
Since the seasonal and local variation of the amount of methane suggests that the seepages could have an extension of meter to km scale, in this work, we explored the chance to nd clues of C-H compounds in data of CRISM. In fact, hydrocarbons show absorptions in the IR range between 3.1 and 3.6 µm and CRISM has an IR range up to 3.92 µm with a spatial resolution capable of investigating the surface at a  (carbon dioxide) using data from space and ground-based telescopes. Furthermore, also CO2 ice shows a strong absorption at 3.3 µm (Hansen et al., 1997). However, the phase diagram of CO2 for Mars shows that at average temperatures around − 50 °C, at average latitudes, and pressure around 6 millibar, CO2 should be present in gas form (Longhi 2006).This research focused on signatures of hydrocarbons in the IR range on the surface of Mars. Oheler and Etiope, 2017, point out that due to the transient nature of the methane detected on Mars and the uncertainty on its lifetime in the Martian atmosphere, the only method for studying methane seepages on Mars surface is placing probes on the ground at xed positions. This would allow constant sampling of the gas methane to determine uxes and for minimizing the effects of isotopic fractionation of CH4 which is important to ascend to its origin. However, if macro seepages could be, at rst, localized from remote sensing data, it would be possible to plan a landing mission for taking ground instruments and measuring methane uxes directly on these macro seep sources. Therefore, we started this work to search potential methane spikes with these premises: 1) In June 2019, estimated CH4 abundance on Mars surface is about 20 ppbv by SAM-TLS (AbSciCon 24-28 June 2019), and it is unknown the distance of Curiosity from the source of the detected spike.
2) Detection of plumes from ground telescopes during 2003 of about 40 ppbv (Mumma et al., 2004(Mumma et al., , 2009) were integrated on a wide area. We hypothesize that this abundance on a broad area would potentially mean a greater concentration in the source sites.
3) The time of survival of CH4 in atmosphere spans from hours to 3 hundred years (Dartnell et al., 2012). 4) Since August 2012, Curiosity detected only two methane spikes during 2013 and 2019. Assuming that these sudden increases of methane concentration are sporadic, in our work, we were looking for CH4 absorption that eventually would correspond to spikes of CH4 in the scene, i.e. a concentration of methane greater than the values found for the background of some tens/hundreds of ppb. Consequently, we expected to nd eventually few featured pixels/none in the greater part of the investigated images.
Literature data related to methane showed that a background does exist due to probable seepages that could be activated by different processes ( Beside methane, the feature at 3.3 µm is also typical of other more complex C-H compounds, the Polyciclic Aromatic Hydrocarbons (PAH), (Tokunga et al., 1991). Like methane, also PAH's can be originated by degradation of organisms (Mckay and Gibson, 1996). In this case, the eventual detection would be related to the time of the single observation with lower chances to nd them again in eventual missions. This, for two reasons: nature of PAH's which origins and eventual relation with the environment could be only studied by in situ chemical facilities, and due to the short time, about 3 days, (Dartnell et al., 2012 ) of surviving of PAH's exposed at UV rays on surface.
Moreover, also ethane shows absorptions at 3.3 µm, but the only data are from Krasnopolsky, (2011), with IRTF-CSHELL that placed an upper limit of ethane in Mars atmosphere of 0.3 ppb. It is similar to background values of methane concentration on Mars surface. We did not search, nor expect to detect spectral features in 3.3 µm region at such concentrations.

Data And Methods
The scope of this work is to search in hyperspectral data acquired by CRISM over selected areas spectral absorptions that can be linked to the presence of methane. To this end, the considered CRISM cubes are processed to obtain the map of 3.3 µm absorption depth, center, and width. Afterwards, the spectra with potentially true absorptions are selected, based on the statistics of the depth map at 3.3 µm and excluding those pixels clearly related to artifacts. Finally, we did a simulation with Planetary Spectrum Generator (PSG) tool to obtain a relation that links spectral properties (i.e. absorption depth) to CH4 concentrations.

Areas Selection
We considered three areas: area of Oxia Planum, in the view of the upcoming Exomars 2022 mission (Vago et al., 2017), to compare the results of this work with data collected on Martian surface by the Rosalind Franklin rover (Voosen, 2017); area of Gale crater in which the increase of methane was proven from orbiter and on ground; area of Nili Fossae which mineralogy is compatible with methane formation (Wray and Elhmann, 2011) and in which the abundance 40 ppbv of methane was estimated from ground telescopes (Mumma et al., 2009). The three areas chosen for the research are indicated with stars in  The elaboration of CRISM data followed two processing chains: the rst follows the steps implemented in the CRISM Analytical Toolkit (CAT) as a package in Envi software. This rst kind of processing is necessary to remove atmospheric and photometric effects and to remove instrument artifacts. Speci cally, the cubes are corrected in re ectance (I/F), (Murchie et al. 2007a(Murchie et al. , 2009b, then, I/F data are divided for the cosine of the solar incidence angle and nally the atmospheric contribution is removed using the so called Volcano Scan method. In the volcano scan method, an atmospheric transmission spectrum is derived from observations at the base and top of Olympus Mons (Mustard et al., 2005).

Processing of 3.3µ m absorption
Then, for what concerns the peak around 3.3 µm, another chain of processing was created. Each CRISM I/F cube was processed through a procedure that, as rst, removes spectral spikes and results in a hyperspectral cube in which the range of I/F is between 0.0 and 0.3. After this step, a destriping process is applied according to the procedure described in De Angelis et al., 2015. This last processing undergoes another processing for searching the 3.3 µm absorption.
For each pixel in the scene, only the portion of spectrum in the range between 3.2 and 3.4 is considered.
Since we are considering a narrow range the continuum was removed by subtracting a linear function passing through the re ectance value of the rst and last point of the range.
After continuum removal, a Modi ed Gaussian Model (Sunshine, 1990) function was tted to obtain the map of the absorption parameters: band center, depth, width, bias. A speci c routine allows to set a threshold for each of the band parameters computed and to print those spectra within such threshold. In order to nd the detection limit of methane concentration in CRISM data we calculated the statistics of depth values on the absorption map in the range between 3.2 and 3.4 µm. To take into account all the noise sources and the variability of the different CRISM scenes considered for this work, the standard deviation must be calculated for every observation.
Since in the images there are unknown artifacts, for each image, a threshold was set at µ + n*σ, where µ and σ are respectively, the average value and the standard deviation of depth map. From the analyses of data, we found that value of n = 5 for σ is a good compromise for avoiding false positives. This threshold on the 3.3 µm depth set the lower concentration limit for detection of methane through CRISM in real data.

Mars surface modeling
In order to see if the estimated quantities of methane by literature data would be observable by CRISM we used the Planetary Spectrum Simulator (PSG). The PSG for surface modeling using CRISM data combines a realistic Hapke scattering model and the capability to ingest a broad range of optical constants allowing to accurately compute surface re ectances and emissitivities (Villanueva et al., 2018).
To simulate spectra of Mars surface for each CRISM observation, the PSG tool requests some parameters: date and season of the observation, the position of target and the geometry of the view, the atmosphere and surface properties, the characteristics of the observing instrument, such as altitude of observation, spectral range and resolution. The noise was not simulated but directly computed on each CRISM observation.
Through the PSG tool, simulating each CRISM observation, we found the relation between increasing methane concentration and the band depth at 3.3 mm. Then, using CRISM data, we have converted the calculated threshold on depth into lower limits of methane detection, for each observation. Through the investigated data we found that the position of artefact absorption is variable spatially and spectrally between 3.34 and 3.4 µm. To avoid false positives, we considered only those pixels in clusters that did not show a distribution along the columns of the images. Despite this precaution, there might still be false positives in the selected pixels.

Spectral investigation on CRISM I/F observations
In table 3, for each selected cluster, we listed the x, y coordinate of pixels with highest value of the depth, band centerthe number of the pixelsthe average value (µ), and standard deviation (σ) of depth of the cluster.
In gures 3,4,5 there are the results of the band minima in the range of 3.2-3.4 for three sites in three investigated areas. In the observation frs0003a896, the depth of the deeper pixel in the cluster (red g. 3) is -0.0123 with a band center at 3.34 µm.
In the observation frs00028346, the depth of the deepest pixel in the cluster (red g. 4) is -0.0200 with a band center at 3.35 µm.
In g. 5 the image frs0002a9b2 is shown. Red pixels indicate the 3.297 µm absorption. The maximum value of absorption depth in the red cluster is 0.045.

Statistics of clusters
Next, to establish a threshold for the depth, we computed the average µ and the standard deviation σ of the depth map of absorptions in the 3.3 µm region, for each dataset.
For each image, the average depth value ranged from − 0.01 to -0.003 and the standard deviation of depth maps resulted from 0.002 to 0.004, Table 4.
Therefore, we have considered as threshold for potential absorptions only depth values greater than µ + 5σ, for each image. Within the considered dataset, the resulting threshold values range from 0.0136 to 0.0237. Simulated spectrum of methane gas on Mars surface The simulation of surface spectra plus increasing content of methane (in ppbv) was computed by the Planetary Spectrum Generator (PSG) tool using different parameters depending on the position respect to the Sun of the investigated CRISM observation site. The use of PSG simulator was intended to estimate the empirical function that link absorption depths, as they would be detected by CRISM, to methane abundances.
As an example, the simulated I/F spectrum, in the 3.2-3.8 µm range of the observation frs0003a896, in Oxia Planum area, shows a weakly visible band absorption at 3.3 µm (Fig. 6) corresponding to the CH4 input value of 100 ppbv. To study how the band can vary in depth according to different concentrations we also simulated: 40, 100, 300, 500 ppbv (Fig. 7) and plot the corresponding depth for each area investigated (Fig. 8, 9, 10). The depth values are calculated as the depth of the minima in the spectrum absorption in the range 3.2-3.4 µm.
As seen, each CRISM observation of this work is collected during a different year, season and time ( Table   2, Fig.11). However, the three plots of the increasing simulated CH4 band absorption vs depth in the simulated spectra, show a general agreement among the depths, indipendently from the season and year of observation. Hence, in general, we can say that the depths of the absorption band at 3.3 µm of methane, would correspond to about 0.008 for 100 ppbv for all the sites. Only in the case of the observation frs00041a28, the concentration of 100 ppbv corresponds to a deeper value of absorption (depth=0.012). In summary, we can say that looking at the plot of simulated observations (Fig. 8, 9,10), the values of depths correspond to a range of methane concentration between 180 and 600 ppbv, depending on the considered site (Table 5).

Discussion
The simulations for each site (Fig. 8, 9, 10) show that the depths corresponding to same concentration of methane do not change much from site to site. For example, 100 ppbv corresponds to a range of depths from 0.008 to 0.012.
In these simulations, one exception is the observation frs00041a28 in Nili Fossae area which shows greater depths with respect to the other sites. This could be related to the viewing geometry or season/hour or a combination of these variables, see Fig. 12.
The depths of the deepest pixels in these featured clusters in CRISM observations show different values that goes from − 0.013 to -0.057 (Table 3).
To avoid, as more as possible, misinterpretation of false absorptions due to unknown artifacts the threshold for depth to consider was set at µ + 5σ of depth maps. The thresholds range from 0.0136 to 0.0237 (Table 4).
Comparing the plots of the simulated depths with the threshold derived from the depth map statistics, we could say that, for concentrations lower than 180-600 ppbv, depending on the considered site, would increase the chance to nd a false absorption.

Good candidates but artefacts
Among the known artifacts of CRISM, there is an optical effect due to out of band leakage in zone 3 of the IR order sorting lter. This leakage peaks appears at 3.4 µm (Murchie et al., 2007). However, this kind of artifacts generates positive signal peaks.
However, we consider that absorptions or a great part of them could represent an unknown artifact. In fact, although we carefully analyzed the noise typical of each image, we have considered that these absorptions could be a new artifact similarly to a probable artifact found in CRISM data at 3.18 µm However, the clusters that satis ed the two criteria we set for potential methane detection in CRISM data, consist of few pixels, 4-15. For each cluster, the value of the pixel with a deepest absorption is considered and listed in table 3. In general, the remaining pixels in the cluster show shallower absorptions. Which means that if these absorptions were methane, the clusters could represent diffusion of gas in the atmosphere from a source point, or a diffusion by a more spread source on the surface.
The concentrations found during this investigation are high respect to previous spikes and plumes detections, but remains in the order of hundreds of ppbv, also we do not have precise information The spectral features we observed at 3.3 mm in CRISM data could be also assigned to magnetic dipole CO2 absorption bands. Nevertheless, some of the differences between our investigation on CRISM data and ACS results stand in the geometry of the scene and location of investigations. The variation of latitude and geometry of the scene correspond also to variation in temperatures and pressures. In this work, we analyzed CRISM data acquired at Nadir whereas data from ACS where collected at solar occultation conditions. Moreover, we focused the investigation on CRISM data at mid latitudes; the ACS spectrometer focused to northern latitudes (> 65°N). However, currently, the new bands of ozone and CO2 magnetic dipole are not integrated in the HITRAN database. Consequently, it is not possible to model the absorption of CO2 and O3 at 3.3 mm with the PSG tool.
Organic matter and PAH's Some clusters that show absorptions at longer wavelengths, could be related to aliphatic hydrocarbons such as methane, as well as other aliphatic compounds that show absorptions at about 3.3-3.6 µm.

Implications for ExoMars2022 and other rover missions
In 2022 the Exomars mission (Vago et al., 2017) will deliver the Kazachock surface platform and the Rosalind Franklin rover on Mars surface that will host several instruments onboard.
On the rover, almost all these instruments will provide data on eventual C-H compounds and organic molecules. Therefore, for what concerns Oxia Planum the results of this investigation would potentially be compared.
Once landed the rover, the Infrared Spectrometer for ExoMars (ISEM) will work coupled with the PanCam camera to select interesting sites for biosignatures. It has a spectral range of 1.15 to 3.3 µm with a spectral resolution 3.3 µm at 1.15 µm and 28 nm at 3.30 µm. As already seen, the range just end at the value of C-H absorption at 3.3 µm, therefore a potential comparison could be done, also looking for absorptions in the 2.2-2.4 µm range of C-H compounds. At micrometric scale, MicrOmega (Micro observatoire pour la mineralogie, l'eau, les glaces et l'activité) -IR will analyse in situ the powder material derived from crushed samples collected by the rover's core drill MaMISS. MicrOmega-IR has an IR range from 0.95 to 3.65 µm in 320 channels of about 8 nm of spectral resolution.
The analyses of samples collected in a depth up to 2 m by MaMISS will be very useful. In fact, either sample extracted will be not so much irradiated and damaged as the surface materials, and this increases the probability to nd organic compounds. Moreover, in case of methane, this will be rapidly detected in the original abundance with respect to methane detected on the surface, which is mixed and/or removed from the Mars near surface atmosphere. This would be potentially possible, searching for other C-H absorptions at also in the region from 2.2 to 2.6 um, range that is characterized by combination and overtone bands (Cloutis, 1989).
Finally, visible and NIR data on crushed samples will be compared with data from Raman Laser Spectrometer (RLS) that will permit the identi cation of minerals and the detection of different organic functional groups to be successively analyzed by the Mars Organic Molecules Analyzer (MOMA), (Rull et al., 2017). One of the major goals of the MOMA analyzer will be to assess whether the potential organic compounds detected are biogenic or abiogenic (Goetz et al., 2018).

Conclusions
The premises of this research were the following: 1) To June 2019, estimated CH4 abundance on Mars surface is about 20 ppbv by SAM-TLS and is unknown the distance of Curiosity from the source of the detected spike.
2) Detection from ground telescopes in 2003 was around 40 ppbv, integrated on a great area.
3) According to literature, the time of survival of CH4 in atmosphere spans from hours to 3 hundred years.

4)
We were looking for CH4 absorption that eventually would correspond to spikes of CH4 in the scene, i.e. greater than the values found for the background of some tens/hundreds of ppb.

5)
Consequently, in our research, we expected to nd eventually few featured pixels/none in the greater part of the investigated images.
As results: Since CRISM is devoted to discovering the mineralogy of Mars surface and not for hydrocarbons detection, it does not collect data time to time, to monitor spectral changes in the same zone periodically.
Furthermore, the comparison with other instruments conceived to detect gas in atmosphere is di cult for different reasons: time of persistence of methane in Mars atmosphere (hours, months, years?), different spatial resolution and/or the non-correspondence of time of observations. Due to artifacts and noise, in this work we cannot con rm the presence of methane seepages in the analyzed datasets. But neither we exclude that absorptions in some pixels can effectively be related to C-H compounds. Overall, we wanted to show a method to exploit CRISM data to search for C-H signatures in areas of Mars surface. The method illustrated in this work could be applied to hundreds of images to explore the chance to nd potential methane macro-seepages.

Declarations
List of abbreviations: Not applicable Availability of data and materials: The dataset analyzed in this work can be downloaded from https://ode.rsl.wustl.edu/mars/, the planetary simulator tool is available at this address https://psg.gsfc.nasa.gov/.     Cluster in frs0003a896. a) frs0003a896 image with a red cluster featured by spectral absorptions at 3.35; b) zoom of the cluster. c) Corresponding absorption and tting with MGM curve to extract spectral parameters.

Figure 3
Cluster in frs0003a896. a) frs0003a896 image with a red cluster featured by spectral absorptions at 3.35; b) zoom of the cluster. c) Corresponding absorption and tting with MGM curve to extract spectral parameters. Figure 4 Cluster in frs00028346. a) frs00028346 image with a red cluster featured by spectral absorptions at 3.35.
b) zoom of the cluster. c) Corresponding absorption and tting with MGM curve to extract spectral parameters.

Figure 4
Cluster in frs00028346. a) frs00028346 image with a red cluster featured by spectral absorptions at 3.35.
b) zoom of the cluster. c) Corresponding absorption and tting with MGM curve to extract spectral parameters. Figure 5 Cluster in frs0002a9b2. a) frs0002a9b2 image with a red cluster featured by spectral absorptions at 3.3. b) zoom of the cluster. c) Corresponding absorption and tting with MGM curve to extract spectral parameters.
Page 29/40 Figure 5 Cluster in frs0002a9b2. a) frs0002a9b2 image with a red cluster featured by spectral absorptions at 3.3.
b) zoom of the cluster. c) Corresponding absorption and tting with MGM curve to extract spectral parameters. Figure 6 PSG tool simulation. Simulated spectrum of frs0003a896 image, input parameters for the surface: 100% abundance Mars spectrum (PSG library) plus 100 ppbv of CH4.  Simulated transmittance spectra of frs0003a896 image, with increasing concentration of CH4 from 40 ppbv to 500 ppbv.

Figure 11
Schematic representation of the position of Mars respect to Sun (Season) during each considered CRISM observation.

Figure 11
Schematic representation of the position of Mars respect to Sun (Season) during each considered CRISM observation.

Figure 12
Green: Radiance spectrum of featured pixels in the cluster of frs0002a9b2 image; Black: Corresponding I/F spectrum. The two spectra show that the absorption feature at 3.3 μm was present in radiance data and was not caused by the I/F calibration pipeline.