Interior layered deposits (ILDs) within western Candor Chasma were studied by mapping lithologies, measuring layer attitudes and comparing the stratigraphy of two adjacent mounds. Layering tends to dip in the same direction as the local topographic slope, although at different angles. Regionally consistent attitudes do exist, suggesting postdepositional block rotations. The stratigraphy of two adjacent mounds correlates, but the thicknesses of the units differ. Most layered material appears to have been deposited conformably, with one late major unconformity. Several fault populations are identified and correlate well with regional faults associated with the formation of Valles Marineris. The data suggest that here the ILDs predate the faulting and may be early basin fill. According to our model for ILD formation, ILDs are deposited syntectonically during early basin collapse. Subsequent subsidence of surrounding areas, accompanied by little or no sedimentation, left the early deposits as individual mounds, remnants of the former subbasins. Stratigraphic differences between mounds resulted from different subsidence rates of subbasins. A significant change in depositional environment, from depositional to a near cessation of deposition and the onset of a major erosion event, possibly coincided with the opening of the main Valles Marineris troughs. We further suggest that groundwater played an important role in the formation of sulfates. The youngest unit identified, not including surficial deposits, is likely the result of a posttectonic, regionally limited volcanic event. If basin collapse continued following the cessation of deposition, this model can also account for mounds within closed basins, such as Hebes.
 The chasmata themselves are thought to have developed as a two-stage process [Lucchitta et al., 1994; Schultz, 1998]. Lucchitta and Bertolini  and Lucchitta et al.  proposed that ancestral basins (closed trough depressions) with irregular outlines formed prior to the opening of Valles Marineris -related linear troughs. In particular, for the present study area, they suggested that the southern portion of Candor Chasma formed as an early ancestral basin, in which ILD were deposited later and at least partially within a deep lacustrine environment. Faulting associated with the later Valles Marineris opening [Schultz, 1998] connected the basins and created the current pattern of interconnected depressions.
 Mapping of these ILDs throughout Valles Marineris has largely been limited to identifying their surface properties, with little information about their geometry and attitude. Several studies have included measurements of bedding attitudes [Lucchitta, 2004; Beyer and McEwen, 2005; Fueten et al., 2006; Hauber et al., 2006; Zegers et al., 2006; Gaddis et al., 2006], showing that the layering tends to dip shallowly (typically less than 20°) in the direction of the topographic slope, suggesting a drape origin. Recent measurements [Fueten et al., 2006] revealed a more complicated pattern in west Candor. While down-slope dips are common, some areas show a consistent regional dip that is not related to topography. A major unconformity is present, indicating significant gaps in the depositional record.
 Here, we expand on the work of Fueten et al.  and develop a model for the formation of these ILDs. We propose that ILDs formed during the formation of ancestral basins and that a major unconformity can be correlated with the onset of Valles Marineris-related extension.
2. Geological Setting
 Within the chasma, the valley floor is a complex series of mounds and depressions (Figure 1), composed of various layered and massive deposits, as well as more recent surficial debris (dunes, landslides and crater ejecta). Dominating are layered deposits which show a range of layer thicknesses, albedos and erosional characteristics, indicating differing compositions and degrees of competency [e.g., Komatsu et al., 1993, 2004; Malin and Edgett, 2000] divide the deposits into three main types: light- to intermediate-toned “layered” units, light- to intermediate-toned “massive” units, and “thin mesas”, which consist of dark- to intermediate-toned capping units. Light- to intermediate-toned units differ in their erosional characteristics. The layered units tend to exhibit stair-stepped or cliff-bench structures indicating competency differences within the layering.
 Commonly, the ILDs are thought to be younger in age than the formation of the chasmata themselves (lower Hesperian; e.g., Tanaka ). For this study the area is subdivided into a number of tectonostratigraphic domains based on topography, stratigraphy and structure (Figure 2). These domains, labeled A–F for simplicity, tend to be bounded by linear depressions.
 The imaging spectrometer OMEGA [e.g., Bibring et al., 2006] has found spectral signatures of pyroxenes, sulfates and iron oxides within the study area. A detailed analysis of the distribution of these materials throughout west Candor Chasma is given by Mangold et al. . Within the study area kieserite is located along the northern slope of the hill within the domain F, throughout much of domains C and E, and along the northern boundary of the domain B. The kieserite cannot be correlated with obvious stratigraphic units and occurs at various elevations, even on top of mounds. Chasma wall rock to the south and the domain D are pyroxene-rich regions. Pyroxene-rich signatures also occur along the southern slope of the domain F and in isolated parts of the domain C. Only one minor occurrence of iron oxides is present within the domain E.
3. Methodology and Data Sets
 The primary tool used for understanding the structure and tectonics of this region is measurement of the layering attitudes, done using the software package Orion. Fueten et al.  discuss the methodology in detail.
 The data used for this study are HRSC stereo images [Jaumann et al., 2007] from Mars Express orbit 2116. The nadir panchromatic orthorectified image was processed to 12.5 m/pixel resolution. Stereo images were used to calculate the elevations to a resolution of 50 m/pixel [Gwinner et al., 2005; Scholten et al., 2005]. Orion uses bilinear interpolation to compute the intermediate elevations needed to match the nadir image resolution. Subsequently, a CTX [Malin et al., 2007] image mosaic for the area was constructed and registered to the HRSC DTM. In addition to HRSC images, we used topographic data from the Mars Orbiter Laser Altimeter (MOLA) [Zuber et al., 1992] and image data from the Mars Orbiter Camera (MOC) [Malin et al., 1992], as well as from the High Resolution Imaging Science Experiment (HiRISE; McEwen et al. ).
 Layer attitudes were calculated by sampling the elevations along visible contacts and computing the best-fit plane through the three-dimensional set of points. The average sample spacing was 50 m or more, often 100–150 m, equal to or greater than the DTM resolution. The measured end-to-end trace lengths were typically from 600 m to 11 km, with an average length of 2.6 km. We conclude that the measurements are largely independent and little biased by the procedure used to produce the DTM. Even where the higher-resolution CTX images were used, these restrictions were maintained.
 For all data presented here, the maximum deviation of the point furthest from the calculated plane is less than 1.5% of the plane's trace length. For all planes with dips greater than 2°, the dip error is less than the dip of the plane, with an average dip error of 1.7 degrees. Layer measurements are presented in Figure 3. Measurements for massive and layered units, which are described in detail later, are differentiated for clarity.
4. Lithologies and Layer Attitudes
 This portion of west Candor Chasma had most recently been mapped by Lucchitta  who suggested that ILDs in this area are of Hesperian or Amazonian age. The deepest part of the chasma is in the northeastern part of our area (domain D) where layered material is covered by deposits which appear to be darker. The layering attitude could be measured in a minor depression and it indicates that the layering is nearly horizontal. We focus primarily on the lithologies that make up the topographically elevated mounds and subdivide the rock units based on their bedding and competency characteristics, adopting the terminology proposed by Malin and Edgett . For the purpose of our work, the lithology can broadly be classified into massive units and layered units.
4.1. Massive Units (MU)
 The best exposed outcrops (Figure 4A) of units that appear massive at HRSC image resolution (10–20 m/pixel), but display fine layering at MOC (few meters per pixel, Figures 4B and 4C) and HiRISE (25–32 cm/pixel, Figure 4D) resolutions, respectively, occur in domains B, C and E. They are resistant to erosion, normally display spur-and-gully erosional features and have a steeper topographic profile than layered units. They form a ridged topography, with ridges often parallel to strike where it can be measured. In HRSC images, layering thickness is on the order of tens of meters and layers appear to occur in packages several hundred meters in thickness. Where the outcrop permits, layer attitudes can be measured with great constancy over distances of several kilometers. In domains B and C, massive units are conformably overlain by layered units (Figure 4B). The exposures of massive units do not coincide with the sulfates and high-calcium pyroxenes detected by the OMEGA spectrometer [Mangold et al., 2008]; their composition is therefore unknown.
 There are five distinctive outcroppings of massive units which form mesas. One, which had been tentatively described by Fueten et al.  as a basement unit, is unfortunately not completely covered by the 2116 HRSC image and DTM. Elevation traces across the four mesas completely covered by the DTM were used to estimate their elevation above background topography. Although a sloping background topography and debris aprons make an exact thickness measurement difficult, we suggest that an estimate of 800–900 m is reasonable, as illustrated in Figure 5A. For the fifth mesa, not entirely covered by the image, approximately 900 m is our best estimate for its thickness (Figure 5E).
 The vertical thickness estimates presented in Figure 5 need to be corrected for the measured dip of the units to obtain true stratigraphic thicknesses which are less than the vertical estimate. For MU1 to MU4 with dip values of approximately 15° the difference between an estimated 800m vertical height and the corresponding stratigraphic thickness is less than 25 m and thus within the range of estimated error for arriving at the original vertical measurement. However, attitude measurement on MU5 indicate steep dips of 22° and 33° to the northeast, representing among the steepest measurements in the area. These values correspond reasonably well to steep measurements of 29.7° and 41.6° reported by Fueten et al. , made using a lower resolution DTM. Using a dip value of 30° for MU5 indicates that an estimated vertical thickness of 900 m corresponds to a stratigraphic thickness of 780 m.
 On the basis of their common appearance, thickness and lack of interbedded finely layered units within them, we suggest that these free-standing MU mesas are the remnants of a single, competent member. The best estimates for the thickness of this unit are between 750–800 m. Because of their location within the stratigraphic column, as will be discussed later, these freestanding mesas will be referred to as the Upper MU member.
 A distinctive darker resistant cap unit (CU, identified as AHire by Lucchitta ) forms the approximately 50 m thick cap on the central mound of domain F and has been shown to unconformably overlie layered material [Lucchitta, 2004; Fueten et al., 2006]. It is preserved primarily along a south-west-facing slope with a sharp break at the ridge of the mound and appears to have limited regional extent. CTX imagery (P01_001443_1740_XN_06S074W) draped over the HRSC DTM shows that the steep slope contains fine layering which is truncated at a shallow angle by the cap rock (Figures 6A and 6B). Spectrographic measurements using the Mars Express OMEGA instrument shows a strong pyroxene signature [Mangold et al., 2008], indicating that this unit is most likely a basaltic unit, either a flow or a pyroclastic deposit that has become well lithified. This unit dips 6–8° to the southwest (Figure 3), compatible with the attitude of its exposed top surface. This fact likely indicates that, here, this unit has been little eroded since its formation.
4.2. Layered Units (LU)
 Layered units occur throughout the area. Using HiRISE image TRA_000836_1740, Okubo and McEwen  demonstrated that sub-surface fluid flow was responsible for the alteration of fractures present within a layered sequence of ILD material with a kieserite signature. This outcrop is within our domain C.
 Layered units are characterized by alternating dark and light layering on the meter scale, and are less resistant to erosion than the massive units. Layered units often show a stair-stepping or terracing erosion surface, as described by Malin and Edgett , suggesting alternating high and low erosional competencies. In domains B and C, packages of layered units alternate conformably with massive units (e.g., Figure 4B).
 Domain A appears to be entirely composed of layered units. Attitudes near the southern end where they could be measured indicate 5°–15° dips toward the south to southwest.
 Along the southern edge of domain B where the strata are well exposed in cross section, Layered units appear in three packages, separated by massive units. Here, at low elevations, attitudes dip consistently approximately 5°–10° toward the SE. These attitudes are compatible with those of massive units in the same locations. At higher elevations toward the northwest portion of domain B, the dips become shallower and show more variability in direction, which tend to mimic local topographic slope. The determination of strike becomes more sensitive to the placement of data points at shallow attitudes, which explains some of the variability in strike. At the northwest end of domain B, attitudes gradually change to a more easterly dip.
Fueten et al.  suggested that an elliptical feature (Figure 6A) within the layered unit in the north part of domain F was consistent with a gentle fold, suggested to be draping of a layered unit over a local block. These attitude measurements are confirmed and show further that the layering becomes more easterly in dip on the east side of the mound.
4.3. Layering Attitude Summary
 With some notable exceptions discussed below, the attitude measurements (Figure 3) are regionally consistent within each topographically defined domain, but vary between domains. Dip directions are to the south in domain A, to the east in domain E, to the northeast in domain C, and nearly horizontal within the domain D. Within domain B, layer attitudes change from southerly trends on the western margin to easterly trends in the eastern portion of this domain. Most dip values are in the range of 5°–10°, while dips in some localities such as the domain F and mesa MLU5 may reach 20°–30°. In most cases, dip direction parallels the topographic slope, while the dips themselves are generally either steeper or shallower than the slope. Only in a few places is the topographic slope parallel to the layering. As can be expected from conformable units, most measurements of massive and layered units in close geographic proximity are consistent.
 Layer attitudes in the domain F show the greatest variability. The capping unit CU dips 6–8° to the southwest (Figure 3), which differs from the attitudes of the underlying units, but is compatible with the attitude of its exposed top surface. The layered units below this cap have some of the steepest dips measured in this study and dip approximately 20°–25° to the northeast. Within the linear depression forming the western boundary of the domain C, some layered units dip to the southwest, in contrast to all surrounding measurements.
 Layered and massive units are present in domain B and C, and appear to be conformable with each other. To be able to compare units in both blocks and determine the relative stratigraphic position of all units, we attempt to develop the generalized stratigraphy of the recognized units by taking the upper and lower contacts of packages of layered and massive units from the topographic profiles (Figure 7). Only a few locations allow for an adequate cross section of the stratigraphic column. Figure 7 shows the attempt to measure and correlate the stratigraphy. Cross sections AB, CD and EF show the apparent dip. Because the profile EF in domain B is parallel to strike direction, it appears horizontal in cross section. Because of uncertainties in locating contacts, all thickness values are rounded to 50 m increments.
 Section AB and CD are approximately parallel to the dip direction in the domain C. Section AB includes MU1, which forms the highest stratigraphic level within this domain, while section CD provides the best detailed information for the lower stratigraphy. The simplest stratigraphy, compatible with the topographic profile for this region, is constructed by fitting layers of 5° (section CD) and 7° (section AB) dip, which agrees well with the direct dip measurements of 5°–10°.
 Section EF in domain B is composed entirely of alternating layered and massive units. A change in layer attitudes, from southeasterly to easterly across this domain makes the construction of a simple stratigraphic column more complex. A measurement of the central layered unit yields a thickness of 350 m, while the upper layered unit has a minimum thickness of approximately 200 m. The east dipping massive unit (Figure 7, hatched area on map inset), which has a thickness of approximately 200 m near location F, is stratigraphically above the layered unit if projected according to its attitude. This agrees with observations within the image that the massive unit overlies the layered unit. However, as the topography of domain B descends beyond location F along section line EF, the elevation of this massive unit is at the same elevation as the more westerly layered units. As outlined above, we suggest that the massive unit attitude reflects a regional change within this domain and place the massive unit stratigraphically on top of the layered unit, in accordance with observation and structural projection.
 To obtain the best match between the sections and allow for the simplest correlation of the massive units outlined in the map area, two stratigraphic sections are correlated along the upper contact of the uppermost layered units in both sections. Hence we suggest that the capping massive unit in domain B can be correlated with the massive unit below the free standing MU mesa in domain C.
 A comparison of the two stratigraphic sections indicates significant differences between the two domains. The lowest massive unit within domain C, for which the thickness can be determined accurately, is 50% thicker than the thickest equivalent unit within domain B. The lowest-most exposed unit in domain B is a layered unit.
 Even though the unconformable cap unit CU does not outcrop along the sections, it is illustrated in the composite section in its presumed position. As erosion has removed much of the upper MU member and CU is clearly unconformably emplaced, we tentatively suggest that the event that caused the erosion of the upper MU member is the same as that represented by the unconformity in domain F.
5. Faulting and Lineaments
 Numerous studies have documented the existence of faults within VM [Schultz, 1995, 2000; Mège and Masson, 1996; 2000Schultz and Lin, 2001; Peulvast et al., 2001] and ample evidence exists for multiple fault sets within this area. We suggest that linear depressions and planar scarps seen in this area are regionally significant faults. In most cases this inference is supported by observations made at smaller scales. As outlined below, most structural features can be grouped into two sets, each appearing to follow a regional trend. These two dominant trends are approximately parallel to the south and west walls of west Candor Chasma (Figure 1). In the absence of geochronological data or clear cross-cutting relationships no clear age relationships between the features can be established. Features of similar orientation, corresponding to the two regional trends, have been grouped together for discussion purposes although they may have different ages.
 Arguably the most significant structural trend in the area is oriented NW–SE, at approximately 120° which is parallel to the south wall of the chasma and the overall Valles Marineris trend (Figure 1). This trend also parallels the major linear topographic low within western Candor Chasma. Within the study area, topographic depressions of that attitude form the boundaries of domains A, B, C, D (Figures 8 and 9) . The edge of the elevated topography within the area is marked by a sharp boundary (C–D boundary) trending 120°.
 On a finer scale, seen in HiRISE images, this boundary is marked by a series of parallel joints (Figure 8B) and fracture sets of that attitude are visible within domain D (Figure 8C). Fractures (Figure 8D) and a significant offset (Figure 8E) parallel to the linear depression that marks the boundary between domains B and C are visible within HIRISE images.
 Within the southern portion of the area, two planar scarps (Figure 9B) occur within massive units along the SE edge of domain B. A NW–SE trending fault is visible within HiRISE image PSP_001641_1735 (Figure 8C). Fractures trending approximately 100°–110° are also visible within CTX image P01_001443_1740_XN_06S074W on the east side of domain F (Figure 9D). Here close-spaced fractures occur at the bottom of the mound high in the stratigraphic section of the finely layered unit.
 The second prominent set of structural features trends SW–NE (∼40°–070°) and is thus approximately perpendicular to the first set. Many regional-scale features show this trend, including the walls forming the blunt end of western Candor Chasma (Figure 1). Linear topographic depressions (Figure 10) oriented at 040° mark the northwestern and southeastern boundary of domain B. Fractures of that attitude can be identified within HiRISE images at several locations, (Figures 10C to 10E), including near several 040°-trending linear breaks visible within the lower MU (Figure 10B). In the southwestern portion of the area, fractures considered to be part of this set, though trending closer to 060°, are present in HiRISE images PSP_002129_1735 and PSP_001641_1735 (Figure 11).
 Two prominent planar scarps, which trend approximately E–W, are oblique to these regional trends. Both dip approximately 30° to the north and are in close proximity to each other (Figures 12A, 12C, and 12D) within what appears to be the same competent massive unit. A clear truncation of finely layered units, parallel to the base of the scarp (Figure 12B), is visible near the northern scarp. We suggest that these scarps are large east–west trending faults.
 Any complete model for the geological history of this area must account for all observations presented here, the primary ones being:
 1. Layer attitudes primarily follow the regional topographic slope.
 2. The stratigraphic sections between adjacent blocks differ. There is no exposed equivalent within domain B for the thick lower massive unit found in domain C.
 3. The sedimentary units alternate between competent massive and less competent layered units, which itself is composed of alternating thinner beds of competent and incompetent material. No significant unconformity can be identified within this LU/MU package.
 4. A significant erosional event affected at least the upper MU member. This event required the removal of at least 750–800 m of stratigraphic section, leaving isolated mesas behind.
 5. Individual blocks of layered deposits appear to be bounded by faults, with differential displacements having taken place between these smaller blocks.
 6. Mounds of layered deposits show signs of internal deformation. Fractures visible within HiRISE images are present at all stratigraphic levels. Their attitude appears to correlate well with regionally significant linear depressions and scarps.
 The observations above clearly indicate that the formation of interior layered deposits cannot be a late-stage or postdeformation event. Valles Marineris most likely did not develop in a single tectonic episode. Rather, ancestral basins formed some time prior to the opening of rift-like linear chasms [Lucchitta et al., 1994; Schultz, 1998; Peulvast et al., 2001]. In the case of Candor Chasma, Lucchitta et al.  suggested that the southern portion opened as an ancestral basin and that if the deposits formed in a lacustrine setting, a water depth of several kilometers was necessary. For this reason, they suggested that a volcanic origin is more likely for interior layered deposits. We agree that they are most likely associated with the ancestral basin formation, but suggest that the deposits were emplaced syntectonically during the basin formation, rather than after it. At any time, the depositional surface need not have been very far below the regional elevation of the surrounding plateau.
 The subsiding floor of an ancestral basin would have been enclosed by a number of bounding faults which would have accommodated the downward displacement of the basin floor. Given the size of the areas occupied by deposits, the minimum basin size would have been on the order of hundreds to thousands of square kilometers. A slab of crust of that size would not likely have maintained its integrity while being displaced downward several kilometers. It is much more likely that the crust of the basin fractured into a number of smaller segments, which we suggest to be on the order of several square kilometers in area. The subsidence itself would then have taken place with differential displacements between these smaller blocks, some subsiding faster or further than others. This implies that the subsiding floor would have developed an irregular topography during the subsidence.
 It is also reasonable to assume that deposition of material within a juvenile ancestral basin occurred as soon as a topographic depression had developed, especially as Mars is thought to have been more active during the Hesperian period. This assumption implies that deposits filled the basin as the floor subsided, eliminating the need for several kilometers of water depth. There is no direct evidence that the basins were, in fact, filled with water, either continuously or intermittently. The observed erosion event, as discussed below, is more likely the result of fluid erosion. However, that does not imply that all deposition within the basins took place within a lacustrine environment. Certainly, crater ejecta, ashfall and spring deposits are possible contributors and cannot be discounted.
 The observation that the lower portion of the stratigraphic sequences in domains B and C do not match in thickness is directly compatible with this model of basin formation. If domain C subsided faster during early basin formation, it may well have accumulated a thicker lower massive unit than the adjacent domain B.
 We suggest then that the individual mounds of layered deposits formed as small basins, or more likely small subbasins within a larger basin, which subsided and were filled during subsidence. The observed dips may be partially the result of draping over basement topography as the basins developed. However, differential basin subsidence and tilting could also have produced minor rotation and tilting of the basin blocks, and thus account for some of the present dips.
 Our observations also put some constraints on the mode of formation of the sediments. Two key factors need to be included in any sedimentary model used to explain the ILDs: the presence of fine layered material with contrasting competencies and the alternation of thick packages of layered and massive units. These features point to systematic changes of source or depositional conditions, with two periods of cyclicity, short periods during deposition of the fine layering and a superimposed long period producing the alternation of layered and massive units. Nedell et al.  used the short-period cyclicity to argue for lacustrine deposition; however, the interbedded thick competent units are more difficult to explain in this way.
 A further point to make here is that the sequence of deposits, alternating between massive and layered units, is similar in the two areas where we can determine the stratigraphic sections, except that the units are thicker in one (domain C) than the other (domain B). This observation implies that the source of sediment is the same for both areas. If, as we suggest, domain C subsided faster than domain B, then either domain C received a greater rate of sediment influx or the deposition surface was very close to the topographic level of the surrounding plateau. Without knowing the sediment source, we cannot choose between these two possibilities.
 The opening of the rift-like portion of Valles Marineris must have been accompanied by faulting and subsidence [Schultz, 1998]. Faulting associated with this tectonism would have used pre-existing crustal-scale fractures, where possible. A composite tectonic map containing the major proposed faults with cross section is presented in Figure 13. We suggest that the planar scarp along the boundary between domain C and D most likely represents fault motion related to Valles Marineris opening. Any layered deposit north of this boundary was displaced downward. The fact that the floor has very little topographic expression suggests that such buried layered deposits did not have much topographic expression or that the floor deposits are quite thick. Hence we suggest that the major lineaments discussed above are basin faults, with some being reactivated during Valles Marineris opening. Since the scarp along the northeast edge of domain C is up to 600 m in height, the vertical component of displacement during Valles Marineris tectonism may be quite significant on reactivated faults, perhaps on the order of several kilometers. Tectonism-related displacement of fault blocks would also have led to rotation. However, because of the limited amount of extension actually produced during Valles Marineris opening [Mège and Masson, 1996; Schultz, 1995], the amount of rotation would also have been limited.
 Two regions within the area provide some evidence for rotation. With the exception of MU5 which appears to dip 20°–30°, most layer dips observed in the study area are less than 20°. If, as we suggest, MU5 is located on a fault active during Valles Marineris opening, the high dip angle of MU5 is most likely the result of fault-related rotation during this late stage.
 We also suggest that the shallow northeast dip observed within domain C is partially the result of a rotation or tilt toward the northeast (a clockwise rotation about a horizontal axis oriented approximately at 300°). Our proposed mode of deposition can account for different stratigraphic sequences for different subbasins, but is unlikely to account for the formation of the observed narrow linear depressions, as they would be filled easily. We showed earlier that the 120° linear depression bounding domains B and C is fault-related. A minor rotation of domain C along a 300°-trending horizontal axis, i.e., a tilt to the northeast, could result in the opening of a tensile region in the area of the depression (Figure 13). Such a proposed rotation would require oblique or transform motion along bounding faults, motion that in the simplest case would be approximately perpendicular to the rotation axis. The two scarps trending approximately east–west could serve that purpose (Figure 13), though we have no direct evidence for an oblique motion.
6.3. Unconformities and Erosion
 No evidence was found for any significant unconformities within the LU/MU package. While it cannot be stated that deposition throughout the package was continuous, the continuity of layers and the lack of erosional horizons that cut layers argue against significant erosional events during the time of LU/MU deposition. If the free standing MU mesas represent the remnants of a continuous unit, a minimum estimate of its regional extent would be between 600 and 800 km2. To create the free standing MU mesas would require the removal of most of this 800 m thick regionally extensive unit. Such a massive erosional event is clearly different from the deposition-dominated period preceding it. We propose that this event marks a significant change in the environment of Candor Chasma and possibly Valles Marineris as a whole.
 A significant angular unconformity exists between cap unit CU and the underlying material. CU appears to be of limited extent, its boundaries cannot be correlated with any tectonic features and there is no evidence that it has been affected by any deformation. Hence we suggest that it was deposited very late in the geological history, possibly after tectonic activity had ceased.
6.4. Timing of Events
 In the absence of crater counts or other geochronological data, we can only determine relative timing and sequence of events. We propose that ILD deposition occurred simultaneously with ancestral basin formation and most likely continued throughout basin formation. Currently elevated mounds of layered deposits correlate with individual subbasins, each of which followed its own subsidence path. The fact that free-standing MU5 appears to be rotated by a Valles Marineris opening-related fault suggests that deposition of the LU/MU package was complete prior to Valles Marineris related tectonic activity.
 The significant erosion event following deposition of the LU/MU package represents a major change in the local environment. Given the severity of the event, it is tempting to place it at the beginning of Valles Marineris opening. Opening of chasmata would provide passages for large amounts of erosion products and possibly the mechanisms for rapid erosion. A more open system of chasmata would also make it difficult to deposit significant thicknesses of material locally on elevated locations. Since at least one chasma, Hebes, retained its closed basin form while still containing a significant isolated ILD, a simple increase in erosion cannot account for the observations. The sedimentary conditions must also have changed. An intriguing possibility is that water disappeared as a possible transport/deposition medium at that time, perhaps because Valles Marineris opening created a new connected network of passageways that drained away the water from the surface of the Tharsis crust. Whatever is the cause, it appears that subsidence continued in Valles Marineris with little or no deposition other than the products of wall rock erosion and collapse and wind transport. The idea of chasma formation with and without deposition was first proposed by Lucchitta et al. , but not developed further.
 The unconformable cap unit CU is the last ILD unit to be emplaced and appears to have been affected only by erosion. We suggest that it postdates the main phase of Valles Marineris tectonism and may not be part of the ILD, in the strictest sense.
6.5. Sulfate Occurrences
 As pointed out above, the appearance of kieserite cannot be correlated with a particular lithology or stratigraphic level. Kieserite is not pervasive throughout Candor Chasma [Mangold et al., 2008] and, while the condition for the formation of the detected kieserite is under debate, its lack of widespread abundance suggests that the conditions are localized. We have suggested that LU and MU units from domain B and C formed under similar conditions during basins subsidence, but kieserite is common within the domain C, but mostly absent from the domain B. Hence we suggest that it is unlikely that kieserite is a primary mineral formed during deposition.
 If Valles Marineris opening did open a drainage network as suggested above, it is difficult to envisage standing water to the elevation at which kieserite is found in an interconnected chasma network. This constraint makes a groundwater origin for the kieserite units more likely. The observation of Okubo and McEwen  of fluid alteration along fractures in domain C demonstrates the presence of groundwater. We suggest then that kieserite may be the result of alteration or may have formed as spring deposits, as suggested by Rossi et al. . An alteration scenario dominated by groundwater also seems to be very plausible on the basis of recent global hydrologic modeling [Andrews-Hanna et al., 2007].
6.6. A Model of ILD Formation
 As ILDs are common throughout Valles Marineris, a generalized model of ILD formation should provide a hypothesis which can be tested in other locations. According to our model, layered deposits initially form as basin fill during the formation of ancestral basins (Figure 14). Basin floor collapse proceeded by differential displacement of kilometer-sized fault blocks. Basal layered material covered the fault blocks by passive draping and hence may have been deposited at a small angle, rather than horizontally. The mode of deposition is unknown, but either aqueous or aerial deposition are feasible. This model of gradual basin floor collapse does not require deep standing water at any time during the collapse. Indeed, if deposition rate is close to subsidence rate there is no room for significant water depth. Differential displacement of subbasins also resulted in differences in the stratigraphy between sub-basins. Deeper subbasins may have accumulated a thicker sequence of deposited material, as demonstrated in domain C above (Figures 14B and 14C). Even though fine layered units alternate with more massive layered units, the depositional environment throughout basin formation appears not to have been interrupted by significant periods of erosion, as no significant unconformities can be identified.
 We postulate that this deposition environment changed significantly during Valles Marineris opening. Valles Marineris related faulting most likely used existing zones of weakness where possible and led to vertical displacement of fault blocks (Figure 14D). In west Candor Chasma it also appears to correlate with a significant erosion event. If basins were filled by water, opening a drainage network by faulting could indeed lead to erosion and transport of significant amounts of ILD material. In a closed chasm, such as Hebes, any water, probably not a large amount at any time, may well have drained out through fractures, though it may have lacked the erosive power of the open chasm.
 While it is tempting to attribute post-Valles Marineris faulting to Valles Marineris opening–related processes sensu stricto, it is not certain that basin collapse did not continue as well. If, for example, Hebes basin continued to collapse, growing in an outward fashion following the end of significant deposition, it would attain its current topography. The elevated ILD marks the site of the original ancestral basin which collapsed early and was filled. Parts of the basin with little or no ILD material formed after the environment favorable for deposition ceased to exist. Small basins devoid of layered material or basins with asymmetrically distributed deposits might be explained this way.
 While processes such as subaerial deposition could certainly have continued during the erosional event and Valles Marineris opening, the present erosional surface is composed largely of the eroded MU/LU units. If a significant thickness of material had been deposited on top of the eroded MU/LU units, then these too would have to have been eroded subsequently. This would require a second erosion event capable of removing significant amounts of stratigraphy. In the absence of evidence for such a secondary erosion event, we suggest that the opening of Valles Marineris coincided with the end of significant deposition. There is evidence of relatively minor deposition following that opening. On the basis of its limited extent and stratigraphic location, we suggest that a late localized minor volcanic event may be responsible for the deposition of cap unit CU.
 The model of ILD formation as outlined above can be tested by examining the structure and stratigraphy of other layered deposits. One would expect draped layering, evidence of faulting and a significant change in the depositional environment, possibly associated with a major erosional event in open chasmata.
 It is proposed that mounds of interior layered deposits within west Candor Chasma originated as syntectonic basin fill during ancestral basin collapse. Differences in stratigraphy between adjacent mounds result from differential collapse of subbasins. The depositional environment, which produced in excess of 1.5 km of apparently nearly conformable layered material, changed during the Valles Marineris opening. The change in environment is marked by a cessation of deposition and by a major erosional event. Spectral signatures of kieserite occur in several locations, not directly related to specific stratigraphic layers, suggesting that groundwater may have played an important role in the formation of kieserite. Groundwater entering the ancestral basins soon after their initial subsidence, perhaps forming shallow ponds or playas, may have altered some, but not all layered deposits. The stratigraphically highest unit CU was most likely emplaced very late as the result of local volcanic activity. The model presented accounts for the observations in the study area. In addition it can also account for the present day geometries of ILDs in closed basins such as Hebes.
 This study demonstrates the importance of including both structural and image analysis in unraveling the complex history of this part of Mars. Knowledge of the layering attitudes and stratigraphy is crucial to understanding the sequence of events and can lead to testable models for the development of similar ILD deposits.
 We thank the HRSC Experiment teams at DLR Berlin and Freie Universitaet Berlin and the Mars Express Project teams at ESTEC and ESOC for their successful planning and acquisition of data, as well as for making the processed data available to the HRSC team. We also want to thank the MOC, MOLA, and HiRISE teams for making their data available. Two constructive thorough reviews by Baerbel Lucchitta and Ross Beyer helped to substantially improve the manuscript. This project was partially funded by an NSERC discovery grant to F. Fueten. Pangaea Scientific thanks Paul Budkewitsch and Canada Centre for Remote Sensing for support of ORION under contract NRCan-01-0102. We thank M. Lozon for preparing the illustrations.