The volcanic geology of Morella Crater, Ganges Cavus and Elaver Vallis, Mars


 Mars contains a large number of yet unexplained collapse features, sometimes spatially linked to large outflow channels. These pits and cavi are often taken as evidence for collapse due to the release of large volumes of pressurized groundwater. One such feature, Ganges Cavus, is an extremely deep (~ 6 km) collapse structure nested on the southern rim of Morella Crater, a 78-km-diameter impact structure breached on its east side by the Elaver Vallis outflow channel. Previous workers have concluded that Ganges Cavus, and other similar collapse features in the Valles Mariners area formed due to catastrophic release of pressurized groundwater that ponded and ultimately flowed over the surface. However, in the case of Ganges Cavus and Morella Crater, I show that the groundwater hypothesis cannot adequately explain the geology. The geology of Morella Crater, Ganges Cavus and the surrounding plains including Elaver Vallis is dominantly volcanic. Morella Crater contained a large picritic to komatiitic lava lake (> 3400 km3), which may have spilled through the eastern wall of the basin. Ganges Cavus is a voluminous (> 2100 km3) collapsed caldera. Morella Crater, Ganges Cavus and Elaver Vallis illustrate a volcanic link between structural collapse, formation and potential spillover of a large lake, and erosion and transport, but in this case, the geology is volcanic from source to sink. The geologic puzzle of Morella Crater and Ganges Cavus has important implications for the origins of other collapse structures on Mars and challenges the idea of pressurized groundwater release on Mars.


Introduction
The surface of Mars exhibits many large, complex collapse features. Since they were rst discovered nearly 50 years ago, multiple hypotheses have been put forward to explain their formation. Many or most of these ideas have a common thread, which is that collapse of the Martian surface occurred due to removal of ground ice or groundwater .
A connection between surface collapse and ground ice or groundwater is reasonable. Mars is a cold planet with a porous crust where ground ice is predicted to be stable at various depths 2 , and in some cases, is directly observed even today 3 . Jumbled blocks of disrupted terrain termed "chaos" are spatially linked with out ow channels. These channels are seen by some as analogous to glacial outburst oods on Earth that form due to rapid and catastrophic melting of glacial ice, sometimes associated with warming climate and other times associated with subglacial volcanism [4][5][6][7][8] . On Mars, potentially contentious elements of the aqueous hypotheses include the existence, rapid release and recharge of vast quantities of groundwater and ground ice 2 .
The link between ground ice, groundwater and surface collapse on Mars remains enigmatic. On Earth, catastrophic oods are never associated with signi cant surface collapse, but are sometimes associated with volcanism 9,10 . On Earth, many steep-sided collapse features result from removal of subsurface material physically (e.g mining) or more often chemically (e.g. dissolution, cave formation, karsting, etc.) 11 . Though lava tubes exist on Mars 12 , they are characterized by pit chains and are much smaller volume than the collapse features associated with chaos terrain. The formation of caves by dissolution has been previously proposed 6,13 , but a more recent picture of the subsurface geology and mineralogy shows rocks susceptible to large-scale dissolution (e.g. carbonates and salts) are not present in large enough thickness or volume to facilitate surface collapse by subsurface dissolution 14 .
On Mars, many examples of major surface collapse have no association with channels or other evidence for aqueous processes, and likely have volcanic/magmatic origins 15 . Yet in some areas of Mars, a temporal and spatial link between jumbled, chaotic terrains and extremely large channel forms is undeniable 16 . There may be a causal link between collapse and formation of out ow channels, be it volcanism 17 or groundwater release 18 . The idea of catastrophic collapse by groundwater release remains widely cited and accepted for Mars, but there are no known examples of an analogous type of large-scale surface collapse caused by groundwater on Earth or other planets. Caldera formation can however form a wide range of complex collapse features of comparable structure, size, depth and surface expression 19 . Channels of the scale of Martian out ow channels attributable to formation by owing water have not been observed on Earth or other planets. However, some large channels of similar dimensions are observed on the Moon and Venus, where erosion by water is impossible 17 .
This paper investigates the geology of the Xanthe Terra region, and focuses primarily on the geology of Morella Crater, Ganges Cavus, and the Elaver Vallis out ow channel (Fig. 1). Morella Crater is a 70 kmdiameter structure that occurs at 308.6E, 9.5S, directly south of Ganges Chasma (Fig. 1). The crater is breached on its east side by Elaver Vallis, a ~ 180 km long suite of out ow channels. Previous authors have concluded that Elaver Vallis formed during a mega ood sourced from release of pressurized groundwater in Morella Crater 20 . This work challenges the idea that Ganges Cavus and other collapse features in the Xanthe Terra region formed by catastrophic release of groundwater and further explores the hypothesis that surface collapse was instead driven by magmatism.

Geomorphology
Ganges Cavus is a deep, asymmetric, steep walled pit located on the southern edge of Morella crater ( Fig. 1). The southern rim of the cavus has an elevation of 2000 m, which is 1000 m higher than the northern rim of the cavus. The hummocky, irregular oor of the cavus is tilted toward the north where it reaches its greatest depth at -4000 m elevation. A conservative estimate of the current volume of the depression is > 2100 km 3 . The oor units are composed largely of light-toned, smooth, hummocky deposits, but other parts of the oor contain irregular boulders that span an array of colours in HiRISE false colour data. Thermal inertia values of materials in the lower slopes of the cavus are ~ 500 and values for the hummocky oor deposits are ~ 800, suggesting the presence of rocky material . The elevation of the oor of the Ganges Cavus (-4000 m) is approximately the same as the elevation of the oor of Ganges Chasma, the southern margin of which is located < 10 km from the northern rim of Morella Crater (Fig. 1).
The ~ 180 km-long Elaver Vallis out ow channel seemingly originates at a breach in the eastern wall of Morella Crater (Figs. 1 and 2a). It consists of two compound channels, a main northern channel and shorter southern channel. In longitudinal cross section, note that the maximum depth of the northern channel is ~ 400 m and the maximum depth of the southern channel is ~ 250 m. However, one critical point that has never been addressed is the fact that a cross section along the channel centre, in both cases, shows that the topographic high point of the channel occurs near the midpoint of the channel along its length. In other words, the elevation of the channel oor of the midpoint of Elaver Vallis is > 250 meters higher than both the origin and terminus of the channel. Elaver Vallis is disrupted by chaos terrain, which composes the topographically higher ground near the midpoint of the channel (Fig. 2). The chaos occurs in multiple distinct patches (Fig. 2), with a total area of approximately 2000 km 2 and depths below the surrounding plains of ~ 500 m. The chaos blocks rise to ~ 100-500 m above the oor of the chaos unit.
The plains surrounding Elaver Vallis are mapped as Middle to Late Noachian undifferentiated units by Tanaka et al. (2014). Higher resolution views show that they are volcanic plains containing NE-SW trending fractures (Fig. 2). A ~ 11 km-diameter, 700 m-deep irregularly shaped, at-oored depression might be a volcanic vent (See Supplementary Materials). The depth/diameter ratio for this feature is high for any crater, and extremely high for anything but a youthful crater unmodi ed by erosion (Michalski and Bleacher, 2013). But the depression is likely Hesperian, and therefore it is unlikely to be an impact crater based on morphometrics.
Smaller channels are also observed within Morella crater (Fig. 3a). These consist of a few nearly straight (low sinuosity) channels up to ~ 35 m length that ow into Ganges Cavus. The channels occur within gently sloping valleys, but the channels themselves occur in positive relief approximately 5-15 m above the adjacent terrain.

Surface mineralogy
Both thermal infrared emission (THEMIS and TES) and near infrared/short-wave infrared re ectance (CRISM and OMEGA) data detect olivine within crater oor deposits (Fig. 3b-c). THEMIS daytime DCS images colour stretched with bands 8, 7, and 5 as R, G, and B, respectively show olivine occurrences as purple 28 , and provide a reliable way to map olivine occurrence on Mars in the thermal infrared (Fig. 3b).
OMEGA global olivine index maps 29 and CRISM multispectral index maps 26 indicate the presence of olivine in the same locations (Fig. 3c). The OMEGA spectral indices suggest that the olivine is Mg-rich or ne-grained, based on comparison to laboratory spectra. High thermal inertia values (> 600) are not consistent with the ne-grained scenario, indicating the olivine is indeed Mg-rich. The precise Fo# of the Mg-rich olivine mapped with OMEGA and CRISM is di cult to constrain without detailed gaussian modelling 30 , but the olivine standard used to produce the maps of Martian olivine by Ody et al. (2013) is nearly pure forsterite.
The spectral detections of olivine-bearing materials are further constrained with TES data 31 . Using data of a single orbit (called "ick" in this case), which minimizes difference potentially attributable to dynamic atmospheric conditions, the detection of olivine in Morella crater ll deposits is clear (See supplementary data). A simple approach to constraining the olivine mineralogy is to plot the position of (Mg,Fe)-O-Si absorptions observed in the mid-infrared as a function of Mg# (Mg-content), as measured in the lab; the results here suggest that the Mg# of the olivine is ≥Fo68 (See supplementary data). In addition, the wavelength position of the major Si-O surface can be used to estimate the SiO 2 content of the rocks. In this case, the wavenumber of the Si-O stretching in the olivine-bearing rocks is ~ 930 cm −1 , which corresponds to an approximate composition of the volcanic rocks ~ 44% SiO 2 (See supplementary data).
In other words, the rocks are olivine-rich and relatively silica-poor, near the boundary of ma c-ultrama c composition.
The collapse of Ganges Cavus has resulted in exceptional exposures of the Morella Crater oor units. The south-facing, northern wall of Ganges Cavus reveals a > 1 km-thick succession of erosion-resistant rocks capped by a relatively thin (~ 100m) covering of re-worked materials (Fig. 4a). CRISM infrared data draped onto HiRISE image data show that the olivine-bearing deposits are continuous and > 1 km-thick (Fig. 4b); they are not just a surface veneer. These olivine-rich deposits represent a widespread crater-ll unit that is exposed throughout Morella crater. It is potentially thinnest in this location, which is located near what was once the southern wall of Morella crater and might be substantially thicker elsewhere in the basin. Even if a conservative 1-km-thick average is assumed, it still suggests that a lower estimate of 3200 km 3 of olivine-rich volcanic material lls the basin.
THEMIS thermal inertia data of the same south-facing, northern wall of Ganges Cavus show variation in TI within the olivine-bearing unit. Layers are observed, with TI values that vary by 60-100 TI units from layer to layer (Fig. 5). This observation is important for two reasons: 1) because it shows that multiple olivine-rich units are present and 2) because it shows that the volcanic processes that produced the olivine-rich unit were cyclical and potentially therefore sustained for some period of time. In this scenario, the lower TI values might correspond to fractured, eroded ow tops or lava lake surfaces, or potentially interbedded volcaniclastics.
Olivine-bearing materials within the oor of Morella crater (Fig. 6a) are composed of olivine-rich (Fig. 6b), fractured blocks of bedrock (Fig. 6c). As noted by ), this material is morphologically similar to rock exposures within the oor of Syrtis Major caldera (Wray et al., 2013). THEMIS thermal inertia data indicate that these materials have values of ~ 325-475, consistent with a mixture of bedrock and sandy particulates.
The interior of Morella Crater contains several arcuate scarps, which form terraces with at surfaces several km-wide (Fig. 7). The terraces contain olivine-rich deposits of similar morphology and composition to those on the oor of the basin, but the terraces are perched approximately 200 m above the oor. The olivine-rich deposits have high thermal inertia (500-700) indicative of rocky materials. It appears that this unit must have either been deposited effusively, and the adjacent oor was later structurally down-dropped by hundreds of meters, or that the material was emplaced through air-fall.
The oor of Ganges Chasma, shows enrichment in olivine, including within a sedimentary apron at the mouth of Elaver Vallis (See supplementary materials). Of course, volcanic materials would be mobilized in the event of a catastrophic ood by water, but it is notable that the olivine in relatively concentrated in those deposits.

Synthesis
The idea of mega ooding on Earth was originally considered outrageous, and acceptance of this concept was achieved only through perseverance and accumulation of undeniable evidence 9 . In the case of Mars, the idea of mega ooding has been more readily accepted . Indeed, it is sensible that melting of permafrost could result in release of large volumes of water capable of erosion. But several facts remain: 1) there is no known case of catastrophic eruption of huge volumes of groundwater capable of eroding such large channels from the subsurface on this planet or any other and the modelled dynamics of such rapid groundwater release require some potentially unrealistic physical scenarios 36 ; 2) all known examples of catastrophic ooding on Earth can be linked to melting of ice dams in surface environments; 3) there are indeed cases of large channel erosion, of comparable scale those observed on Mars, found on the Moon and Venus where erosion by liquid water is recognized to be actually impossible 17,37 . Given these facts, a volcanic origin for out ow channels deserves serious consideration, as has been proposed in multiple works by Leverington and colleagues. Here I focus largely on one region (Xanthe Terra) and speci cally one channel system (Elaver Vallis), but this setting has implications for other locations on Mars.

Discussion of volcanic processes
The geology of Morella Crater, Ganges Cavus and the surrounding plains is dominantly volcanic 38 (Fig. 8). The crater is lled with layered, olivine-rich units that may include both effusive and pyroclastic deposits. The crater oor contains fractured, light-toned, olivine-rich units that are either fractured lavas or pyroclastic units. The walls of Morella Crater basin contain terraces with olivine-rich unts that are of similar composition to those on the oor of the basin. The terminus of Elaver Vallis contains olivine-rich fan deposits, and the oor of Ganges Chasma is olivine rich in general. The plains around Morella crater do not have strong mineralogical signatures of olivine, but they do contain smooth, fractured plains that are interpreted as volcanic throughout the region 38 (Fig. 2b and Fig. 8).
Questions about hydraulic head and groundwater recharge are non-issues for the volcanic hypothesis. Magma can build pressures to greater elevations than groundwater, and in any case there is abundant evidence that volcanism did in fact occur as "break out" ood type deposits in the oor of Ganges Chasma immediately to the north and Eos Chasm to the south. In fact, the chasma all contain abundant evidence for olivine-rich materials 39 (Fig. 8).
Collapse of Ganges Cavus can be explained in the volcanic model simply by removal of magma (Fig. 9).
Edwards (2008) estimates that ~ 10 5 km 3 of olivine-rich lava was erupted on the oor of eastern Ganges and Eos Chasmas. In western Ganges Chasma (closer to the study area), olivine-rich materials cover > 10 4 km 2 of the oor and in western Eos, approximately 10 5 km 2 . Assuming a thickness of 100 m to 1 km, this equates to ~ 10 4 to 10 5 km 3 of extruded lava in the vicinity of Ganges Cavus. This volume, in addition to the ~ 10 3 km 3 of volcanic material within Morella Crater greatly exceeds the volume of the collapse feature in Ganges Cavus.
The only major challenge to the volcanic hypothesis is with regard to the formation of Elaver Vallis itself. Is it possible to form such a channel through physical or thermal erosion by lava and/or pyroclastics? As pointed out by Leverington, channels of comparable size and morphology are observed on Venus and the Moon, where they can only be interpreted as volcanic. Observations from the Archean Earth, described below, illustrate that thermal and mechanical erosion can both result in the formation of large-scale channels. In the case of Mars, unknowns in lava composition and temperature (i.e. viscosity) and uncertainties with regard to the composition, porosity and volatile content mean that models of lava channel erosion on Mars are not well constrained.
The structure, composition, and morphology of the collapse features and surface units are similar to well understood caldera structures on Earth. For example, the shape and depth/diameter ratio of Ganges Cavus is similar to that which occurs in the summit caldera of Kilauea and smaller pit crater Halema uma u in Hawaii (~ 0.15) (Fig. 10). Further, the hummocky and fractured textures observed within the olivine-rich oor material in Morella Crater (Fig. 6c) are strikingly similar to the lava units within Kilauea Iki in Hawaii (Fig. 10). The features in Kilauea Iki formed through cooling of a lava lake, resulting in columnar jointing at the scale of ~ 4-5 m 40 . The fractures and resulting polygonal pattern in Morella Crater occur at the same spatial scale, which likely re ects similarities in cooling rate 41 .
Morella Crater likely contained a lava lake in the late Noachian or Early Hesperian. High stands of olivinerich material on the walls of Morella Crater might have been deposited when the lake was in ated to higher elevations, or they could represent airfall material from cyclic explosive activity. The lava lake might have spilled over the eastern rim of the crater, resulting in the formation of Elaver Vallis by thermal and/or mechanical erosion of the substrate by picritic or komatiitic lava.

Comparison to the Archean Earth
Komatiitic lavas occur in Archean and early Proterozoic terranes, but are vanishingly rare on Earth after the early Proterozoic 42 . Though the rocks representing these high-temperature, low-viscosity lavas have been generally metamorphosed to greenschist facies or higher grade, their geochemical records and physical geometries are discernible. From source areas to down-ow regions, Archean komatiite lavas in western Australia show increasing chemical contamination caused by thermal erosion of substrate bedrock, which was melted and chemically incorporated into the ow 43 . Where geometries are mappable, it appears that extremely high ux, giant komatiite ows in the Perseverance area have physically and thermally eroded hundreds of meters into the substrate and may have owed for 10s to 100s of km down-channel 44 . Ten-km scale lava channels, forming erosional troughs tens of metres deep and up to 200m wide are developed associated with the nickel sul de ores at Kambalda 45 . Lava temperature and effusion rate clearly affect the turbidity of ows and their erosive potential, but a signi cant factor controlling erosion by komatiite lavas is the geology of the substrate. As noted by Williams and others, the composition, porosity and water content of the bedrock signi cantly impacts the erosion depth of lava channels. The Kambalda channels, developed on basalt, represent a lower limit on the expected extent of erosion by ultrama c lavas 45 . Erosion of long lava channels on Mars is possible if the substrate is unconsolidated material and/or appreciable volatiles 46 . It is likely that the surface layer of Mars would have contained unconsolidated, volatile-bearing and likely sulfate-bearing megaregolith in the Early Hesperian. This material would have been susceptible to thermo-mechanical erosion by lava.

Convoluted evidence for volcanic and aqueous processes
Numerous authors have proposed that out ow channels in general and Elaver Vallis in particular formed as the result of catastrophic release of groundwater 6,20,47−49 . In general, all of the models depend on the past existence of vast quantities of groundwater pressurized beneath a cryosphere that was catastrophically breached. Though there are no known examples of this type of process on Earth or elsewhere in the Solar System, it is potentially conceivable that such a process could occur at low elevations on a water-rich planet with signi cant pore space, high permeability and an adequate water recharge mechanism. But, Xanthe Terra is among the topographically highest regions of the planet and a recharge mechanism seems untenable even if the highlands were icy with some glacial meltwater to recharge the subsurface 50,51 .
Coleman (2013) proposed that Elaver Vallis was eroded in a week by > 2000 km 3 of water released from an ice-dammed lake in Morella Crater, and ultimately from the subsurface in the Ganges Cavus area. Not only does this imply what are potentially unrealistic pore volumes in the subsurface below the release zone, it also requires the need for signi cant hydraulic head 36 . Coleman points out that a challenge to the subsurface water hypothesis is the need for signi cant hydraulic head in the close proximity to Ganges Chasma, which is as deep as Ganges Cavus and would have likely experienced breakouts along the oor and wall long before head could be built to ood high on the adjacent plains.
In fact, the evidence for an aqueous origin of Elaver Vallis and Ganges Cavus rests entirely in the observation of a channel with a grooved oor and streamlined eroded features. Komatsu et al. point out that Morella Crater contains no direct evidence for a lake having existed (e.g. no deltas, no shorelines, and no lacustrine deposits), though they nonetheless interpret the geology of the area as having formed by catastrophic ooding from release of pressurized groundwater. Komatsu et al also point out that the terminus of Elaver Vallis contains a large fan complex on the oor of Ganges Chasma, but note that they detect no mineralogical evidence for aqueous deposits.
In the aqueous out ow model for Elaver Vallis, the release of 1000s of km 3 of groundwater resulted in collapse of the surface, resulting in the formation of Ganges Cavus. It has even been suggested that vast subsurface caverns might be responsible for the migration of huge volumes of water in the subsurface 6 . Though the lower gravity environment of Mars compared to Earth would result in more pore space to greater depth on Mars compared to Earth 2 , there is no evidence for subsurface caverns or even large amounts of soluble rocks in the subsurface that could plausibly result produce sink hole-type geology on Mars. Spectral imaging of the walls of Ganges Chasma provide a window into the subsurface in the area, and these data show the presence of pyroxene-rich materials indicative of a relatively dense igneous basement. There are no soluble units in the subsurface in this area. Furthermore, even the largest sinkholes on Earth are ~ 4-5 orders of magnitude smaller volume than Ganges Cavus.

Conclusions
Morella Crater is a ~ 75 km diameter basin of likely impact origin, though it could possibly be a caldera. Regardless of its formation mechanism, the basin has been lled with olivine-rich lava in the Late Noachian or Early Hesperian forming a lava lake. Polygonal-fractured textures within the volcanic oor deposits are analogous to cooling joints seen in the lava lake at Kilauea Iki and other similar contexts. Olivine-rich shelves preserved along the basin walls might represent airfall volcaniclastic deposits but are most easily explained in the context of a high-stand of the lava lake.
Ganges Cavus is a deep, steep, voluminous (> 2100 km 3 ) collapse feature that formed within Morella Crater. The walls of the collapse structure expose ≥1 km-thick olivine-rich volcanic oor deposits. Though Ganges Cavus is signi cantly larger and deeper than Halema uma u in Hawaii, the morphology, structure, slope and depth/diameter ratio (~ 0.15) is the approximately the same in both structures.
The east wall of Morella Crater is breached at the boundary of the ~ 180 km-long Elaver Vallis out ow channel, which terminates further eastward at the rim of Ganges Chasma. The mid-point of the channel(s) is approximately 250 m higher elevation than the channel origin or terminus, and the higher elevation terrain in the channel is composed of chaos blocks. The chaos might represent previously icerich terrain that was disrupted by the extreme volcanism in and around Morella Crater. I propose that late stage aqueous activity occurred, mobilizing volcanic materials and confusing evidence for aqueous versus volcanic processes. In this scenario, the volume of water involved would have been signi cantly less than what has been previously proposed, and huge volumes of water are not required because water is not the primary agent of erosion.
Most of the channel deposits appear to be buried within the oor of Ganges Chasm, but it is notable that fan deposits at the terminus of the out ow channel show olivine enrichment, but this might also re ect younger volcanic material deposited onto the fan deposits. In any case, Ganges Chasma is the sink of the channel deposits and the oor of the chasma contains abundant olivine.
Previous authors have argued that the collapse of Ganges Cavus and the formation of Elaver Vallis can all be explained through the catastrophic release of pressurized groundwater. The aqueous model for the origins of Ganges Cavus and Elaver Vallis rely on the release of huge volumes of groundwater (> 2000 km 3 ) to form an ice-dammed lake, which failed, resulting in a jokulhaup that carved the channel. Any aqueous model is unable to explain the origin, recharge mechanism or hydraulic pressures required for such a model to operate in this region.
Though there are no known analogs for catastrophic release of groundwater on any planet, there is signi cant context to interpret nearly all features of this system in terms of volcanic processes. Ganges Cavus is a caldera formed by collapse of the magma chamber due to extrusion of lava, subsurface migration of magma or likely, both. Volcanic deposits within Morella are demonstrably forsteritic olivinerich and seemingly low silica. The lavas could reasonably be interpreted as komatiitic or picritic and would have been high temperature, low viscosity lavas. It is plausible that the lava lake breached the eastern wall of Morella Crater and that the low-viscosity lava carved Elaver Vallis, at least in large part.
The chaos deposits east of Morella might represent terrain that was previously ice-rich, but which became disrupted during thermal erosion of the substrate. In this way, a late stage of aqueous activity, albeit of much lower volumes and character than what has been proposed, might have occurred, potentially owing from the mid-point of the channel both eastward into Ganges Chasma and westward into Morella Crater.
Given that Mars is a cold, volatile-rich planet that has experienced sustained volcanic activity for billions of years, it is inescapable that magma or lava, groundwater, and surface or near-surface ice have affected each other through time. It would not be wise to frame the geological puzzle of the origins and evolution of out ow channels as having resulted entirely due to volcanism or aqueous activity. In the case of Ganges Cavus, Morella Crater and Elaver Vallis, the geology is clearly volcanic from source to sink. That is not to say that aqueous processes played no role at all in shaping the morphology of this area, but the conclusion is that volcanism was the dominant process driving surface collapse, deposition of (lava) lake deposits, channel erosion and resurfacing.

Methods
The geology of Morella Crater was studied primarily using remote sensing data from the following missions  21 . This is a unique dataset available within JMARS amounting to a global Digital Elevation Model (DEM) informed by both MOLA and HRSC, at 200 m/pixel resolution. Geomorphology was evaluated using visible and thermal infrared images. Thermal infrared data from the MO Thermal Emission Imaging System (THEMIS) available in a global 100 m/pixel mosaic within JMARS provide a mesoscale base map 22 . Visible images used here include high resolution data available from HRSC at 10-20 m/pixel, CTX at ~ 6 m/pixel and HiRISE at ~ 0.25 m/pixel. Data were obtained in radiometrically corrected and geometrically projected format and ingested into a geographic information system (GIS) for analysis and interpretation. Visible images were draped over digital topography in order to create 3D views of bedding and surface morphology.
Thermophysical properties were evaluated using THEMIS and MGS Thermal Emission Spectrometer (TES) data Golombek et al., 2005). The TES data were previously processed into a global thermal inertia dataset, though at relatively coarse spatial resolution (8 pixels per degree). Despite the coarse spatial resolution, these data provide a stable and reliable measure of thermal inertia (J m − 2 K − 1 s − 1/2 ). Areas with high THEMIS TI translate to surfaces with coarser grains, rocky materials, better indurated materials or some combination of all three of these scenarios.
Surface mineralogy was investigated using multiple near infrared, short-wave infrared and mid-infrared datasets. THEMIS daytime IR radiance images were processed using a decorrelation stretch method (DCS), which emphasizes spectral radiance differences attributable to compositional variation. Spectral interpretations of bulk surface mineralogy were further constrained using basic applications of TES atmospherically adjusted emissivity spectra. Near infrared data from the MEx Observatoire pour la Minéralogie, l'Eau, les Glaces et l'Activité (OMEGA) and the MRO Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) were processed to correct for instrument effects and minimize atmospheric absorptions using a standard data analysis pipeline 14 . Both datasets rely on the use of an atmospheric transmission spectrum scaled by elevation (path length) with atmospheric effects estimated for a given Martian L s (day of Martian year). As a result, (due to the estimated rather than measured atmospheric properties) there are usually some residual atmospheric effects even in the corrected spectra. These can be minimized by using spectral ratios, which divide a spectrum of interest (numerator) by a spectrum from the same scene over a spectrally unremarkable terrain of comparable albedo (denominator) (e.g 25 . This technique has the advantage of emphasizing unique spectral features related to a certain surface of interest (usually mineralogical absorptions of interest), but it has the disadvantage of converting the spectra from units of I/F (which is comparable to radiance) into unit-less spectra, comparable in shape to laboratory spectra but not comparable in absolute units.
We analysed all CRISM images targeted within Morella crater. Our analyses focused primarily on CRISM observations corresponding to: Full Resolution Targeted (FRT) images (18/pixel), with some Half Resolution Long (HRL) images (36 m/pixel) and Half Resolution Short (HRS) images. Some of the images used in this work were acquired by CRISM after 2012, when the one of the cryo-coolers failed. As a result of this normal system decay, data collected after that date contain more noise than data collected before the cooler failed. Images collected in the newer observing mode are called Full Resolution Short (FRS). CRISM I/F images were also converted to spectral summary products. These data products were created using the CAT_ENVI software package using combinations of spectral ratios tuned for sensitivity to various minerals or mineral groups 26 . Such maps are useful for evaluating the likely presence of a particular mineral and for mapping relative signal strength corresponding to the unique features associated with that mineral, but these maps do not correspond to actual mineral abundances and they require veri cation in order to validate mineral occurrences.
Geologic mapping was carried out using THEMIS daytime IR and CTX image data as the base. Geological units were de ned based on their geomorphological expression, response to erosion, texture, mineralogy and relative age relationships. Elevation data were used to delineate chaos and channel boundaries. Figure 1 MOLA elevation data draped over THEMIS daytime IR show the relationship among Morella Crater, Elaver Vallis, and Ganges Cavus (a), all located to the south of Ganges Chasma. An inset of Ganges Cavus shown in "b" contains a different elevation scale, illustrating the depth and steep walls of the collapse feature. The topographic pro le from N to S located in "a" is illustrated in "c." Figure 2 HRSC/MOLA blended elevation data overlaid on THEMIS daytime IR are shown in "a." A geologic map of the Elaver Vallis area is shown in "b."    Figure 4). A pro le of TI data on the wall (c) shows differences values of ~350 corresponding to recessive (labelled "r") units and ~450-475 corresponding to cliff-forming units (labelled "cf").  MOLA-HRSC topography draped over THEMIS daytime IR (a) show the western part of Morella Crater. Note terraces present in the wall with at benches that are ~200 m higher than the oor. THEMIS DCS data (b) show in purple the occurrence of olivine in materials draping these benches. THEMIS thermal inertia data (c) show that the terrace materials are relatively high inertia (600-700), corresponding to indurated or rocky deposits. CTX visible data (d) show the light-toned olivine-bearing unit.  A schematic cross section shows as simpli ed version of Morella Crater, ooded by lava from a deep magma source. In "a," the topography of the crater is accurate, except that Ganges Cavus is omitted to show the pre-cavus con guration. Panel "b" shows the actual north-south topographic pro le with interpreted geology. Note the outcropping of crater-ll lava in the cavus wall. Withdrawal of magma, likely driven in part by the movement of magma toward Ganges Chasma to the north resulted in collapse of the Page 24/24 magma chamber. Collapse was controlled in part of the existence of impact-related structural discontinuities.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.