We present a large-scale spring hypothesis for the formation of various enigmatic light-toned deposits (LTDs) on Mars. Layered to massive LTDs occur extensively in Valles Marineris, chaotic terrains, and several large craters, in particular, those located in Arabia Terra. Most of these deposits are not easily explained with either a single process or multiple ones, either in combination or occurring sequentially. Spring deposits can have a very wide range of internal facies and exhibit complex architectural variations. We propose the concept of large-scale spring deposits for explaining LTDs on Mars. Stable volcano-tectonic settings, such as the ones typical on Mars, are compatible with a large-scale, long-term, multistage formation of spring deposits. The large-scale spring deposit model can explain the formation of LTDs with a common process, although active in different times and locations, compatible with coeval local or regional processes and deposits, such as volcaniclastic ones. LTDs, if formed as spring deposits derived from subsurface fluids, could potentially offer favorable conditions both to life and to the fossilization of past life forms.
 A number of interpretations have been proposed for Interior Layered Deposits (ILDs) in Valles Marineris [Lucchitta et al., 1992]: the proposed formation mechanisms in the literature include sedimentary genesis [Malin and Edgett, 2000], volcanic origin [Chapman and Tanaka, 2001], salt diapirism [Milliken et al., 2007], or mechanisms related to eolian processes or pyroclastic events [Peterson, 1982]. Despite the various hypotheses proposed during last decades, the process of formation of ILDs is still debated and theories about the extensive hydrated sulfate-rich outcrops in Valles Marineris ILDs [Gendrin et al., 2005] are being put forward. In some cases mineralogical variations have been documented [Murchie et al., 2007] within the layered sequences, e.g. in Candor Chasma.
 Also, an origin as evaporite deposits, responsible for the formation of outflow channels has been proposed [Montgomery and Gillespie, 2005]: this theory would require that the ILDs are older than the Valles Marineris canyon system. In this paper the discussed light-toned deposits are designated LTDs, including the ones in Valles Marineris. However, because of the custom in the literature [Lucchitta et al., 1992], those particular deposits are also mentioned as ILDs in the text.
 In this study we investigate these features in more detail by testing the spring deposit hypothesis [Rossi et al., 2006, 2007a] for these layered deposits, discussing the geological and geometrical characteristics of a number of LTDs. Although each LTD has its unique features in terms of scale, stacking pattern, and composition, a significant set of characteristics are shared by LTDs described here. Their common characteristics led us to propose a generic scenario for their formation.
 Spring deposits on Earth exhibit an extremely wide range of features, morphologies and facies associations [Ford and Pedley, 1996; Pentecost, 2005] and variable size, morphology and internal structure. The typical size for terrestrial spring deposits ranges from several meters to several hundred meters, up to few kilometers, while the local relief can usually top several tens of meters.
 This extremely high variability makes complicated the use of environmental or morphological classifications [Pentecost, 2005]. Among several classification schemes, in situ (autochthonous) and clastic (allochthonous) can be distinguished [Pentecost and Viles, 1994]. The first group includes morphologies such as mounds, fissure ridges, paludal, cascade, dam and different sorts of cemented crusts. The allochthonous ones comprise alluvial bars and lake or valley fills, thus subject to more extensive transport.
 The range of actual morphologies that can occur on terrestrial spring deposits is extremely wide (Figure 2): spring mounds on their own can occur in a variety of different types, according to local conditions, including regular mounds, terraced mounds (Figures 2a, 2b, and 2c), almost flat subhorizontal ones, with or without a hosted lake; in addition spring mounds can be characterized by parasite submounds, positioned eccentrically with respect to the main structure. The possible combination of many end member morphologies can make the resulting assemblage of different depositional events, over a certain time span, rather complex.
 The local topography can also affect the final shape of a spring mound: deposits forming on a slope would have a preferential progradation direction downslope, with an increased complexity in the geometry of the mound (Figure 2a).
 Some spring deposit morphologies, such as fissure ridges (Figure 2d) are formed along structurally controlled lineaments, such as faults, fractures or joints [Pentecost, 2005] and the resulting spring deposits can be aligned, although possibly at discrete locations along the alignments, because of local variation of the local fracture/faulting geometry.
 Notably terrestrial spring deposits can show strong and rapid changes of facies and geometry, both at the micro-, meso- (Figures 2e and 2f) and macro-scale [Guo and Riding, 1998, 1999; Pentecost, 2005], often because of the rapidly changing hydrodynamic and chemical conditions over short distances from the source areas.
 All those characteristics in nature can occur combined in a variety of ways, determining a large and complex set of final morphologies, facies association and extents of the spring deposits.
 The role of groundwater in Martian geological history and its hydrological cycle is being increasingly recognized in recent years. Groundwater upwelling and evaporation may have played an important role in the deposition and subsequent alteration of rocks and soils at Meridiani Planum [Andrews-Hanna et al., 2007; Squyres et al., 2004]. Also Valles Marineris shows evidence of groundwater activity during its evolution [Okubo and McEwen, 2007; Treiman, 2008]. There is various evidence pointing toward diagenetic activity on Mars, on the basis of the analogy of Martian “blueberries” found in Terra Meridiani [Squyres et al., 2004] to concretions on Earth [Chan et al., 2004].
 We propose a common formation scenario for some light-toned deposits (LTDs) in Valles Marineris, neighboring chaotic terrains and crater bulges, interpreting the large, thick outcrops of layered material as complex assemblage of spring mounds of varied age.
2.1. Global Distribution, Setting, and Common Features
 LTDs in Valles Marineris, chaotic terrains and crater bulges share a relative low topographic setting, from the floor of several chasmata to the low portions of chaotic terrains, or on crater floors (Figure 1); similar deposits in Meridiani Planum are also present, in flat, plateau-like areas. Globally, most thick bulge deposits in craters are also at moderate to low absolute elevation. Several common features can be traced among these thick, architecturally and structurally complex deposits: the range of thickness is variable, being greatest in Valles Marineris, with the maximum thickness exceeding 6–7 kilometers. Thicknesses are strongly correlated with the local relief differences (Figure 3), with lesser thicknesses in Capri Chasma, and greater thicknesses in Hebes and Ganges chasmata (Figures 3 and 4) [Lucchitta et al., 1994]. Quasi-tabular (Aram) or mound-like (Iani) deposits have a variable visible/apparent thickness, with a maximum of few hundreds of meters (Figure 5). Crater bulges, regardless the size of the host crater, are up to 2.5–3 km thick (Figure 6), in most cases not exceeding 1.5 km. The level of erosion is varied for both crater bulges and the impact crater structures, suggesting a wide range of formation ages, although the degree of induration could also affect the observed level of erosion.
 Deposits in Valles Marineris [Quantin et al., 2006] range in elevation of the base from about −4000 to −1500 m (with respect to MOLA (Mars Orbiter Laser Altimeter) aeroid-referenced topography) (Figure 3). The floors of crater bulges, concentrated in Arabia Terra and surroundings (Figure 1), have a slightly wider elevation distribution, with minimum elevations as low as −4000 m, and maximum values not exceeding 0 m.
 Moreover, a significant number of these mound-shaped deposits occur in closed depressions, such as Hebes Mensa and most deposits in crater bulges, which are surrounded by high plateaus. In only a few cases the bounding plateaus or rims are partially eroded (e.g., Gale crater northern rim).
2.2. Setting-Dependent Features
 The LTDs in Valles Marineris, being the thickest and most extensive ones among those treated in the present work, are also the ones with the largest morphological and facies variability. A wide range of forms is present, ranging from thick, massive deposits, to finely layered sequences (Figures 4a–4d), with very different possible subenvironments present within the same chasma (e.g., Candor, Melas, etc.). Extensive unconformities have been documented within the LTD in Valles Marineris (e.g., Candor Mensa) by various authors [Fueten et al., 2006; Mangold et al., 2007]. At regional scale, some of the deposits in Valles Marineris, such as those in Ophir and Candor Chasma and, slightly less, in Capri Chasma show a streamlined surface geometry in plan view, suggesting the action of fluid during/after their emplacement. At a multikilometer scale, LTDs in Valles Marineris show a tabular to mound-like morphology. Both vertical and horizontal stacking pattern variations are visible at HRSC (High Resolution Stereo Camera) [Jaumann et al., 2007; Neukum et al., 2004a] scale (10–20 m/pixel). Both albedo and architectural variability is still high at MOC (Mars Orbiter Camera)/HiRISE (High Resolution Imaging Science Experiment) scale in certain areas within LTDs (Figure 4).
 Valles Marineris ILDs, although sharing several common features, locally show local morphological and apparent facies variation over a very short spatial distance, e.g., quasi-horizontal finely bedded sequences bounded by unconformities.
 Chaotic terrains host relatively small (in terms of volume) LTD patches or extended cover, e.g., in Aram Chaos. Their thickness to extent ratio is among the lowest. They show a certain variability of surface texture, being often affected by eolian erosion, with possible yardangs [Sowe et al., 2007]. At a scale of tens of meters they tend to show flat, subtabular morphologies, although even the flattest looking deposits in Aram Chaos have a slightly bulged expression [Glotch and Christensen, 2005; Massé et al., 2007].
 Their overall appearance is rather mound-like, regardless the amount of erosion. Remarkable light-toned deposits are also present, e.g., in Iani Chaos, as two main distinct bodies, a few hundred kilometers distant from each other (Figures 5a and 5b). In Aram Chaos, both the thickness and elongation of the light-toned deposits show evidence of structural control [Oosthoek et al., 2007]. Also, geological and geomorphological mapping [Glotch and Christensen, 2005] in Aram Chaos showed evidence of tilting and bulging of the LTDs within Aram [Glotch and Christensen, 2005]. Possible structural control (doming/bulging) on LTDs in Valles Marineris is also emerging from attitude analysis over large areas where those deposits crop out [Fueten et al., 2006].
 It is clear at all scales, from HRSC to HiRISE, that LTDs in chaotic terrains are onlapping on knobs and mesas (Figures 5a and 5c), thus postdating chaotic terrain formation [Sowe et al., 2007].
 Crater Bulges appear as pristine to heavily eroded tabular to mound-like structures located in a number of craters with diameters above several tens of kilometers.
 Deposits in crater bulges have the widest range of latitudinal and longitudinal distribution, although most of them are concentrated in Arabia Terra (Figure 1). Their elevation distribution is less wide, being usually at moderate elevations (∼−2000 ± 2000 m) above MOLA datum.
 All LTDs in crater bulges mentioned and analyzed here share a lack of signs of relevant lacustrine activity in their respective basin. This includes evidence of terraces, deltas, fan deltas [Pondrelli et al., 2005] and significant drainage basins associated with the crater itself.
 Also, there is no or little evidence of fluvial activity in the immediate surroundings of the craters hosting bulges and within their rim. A few non-eroded crater bulges appear to be mantled, with low exposure of underlying material. High-resolution observations, such as with HiRISE images over Gale Crater (Figure 6c) show finely layered bright material resembling sedimentary rocks. LTDs in craters locally show evidence of structural control in their deposition/emplacement [Bridges et al., 2007].
2.3. Volumes, Composition, and Internal Geometry
 Volume determinations show a wide range of values: the largest ones belonging to thicker, low-lying deposits in Valles Marineris. The volume of most ILDs in Valles Marineris is of the order of 104 – 105 km3 (although smaller ones are present), with a total volume of 105 – 106 km3 in the Valles Marineris interior [Lucchitta et al., 1994]. This is roughly one order of magnitude lower than the total eroded or removed volume: i.e., the missing volume ideally filling the canyon system to the top of the surrounding plateau; it is not implied that this missing volume has ever been filled with deposits, we simply compare the differences in magnitude between LTDs and maximum potential available volume [Lucchitta et al., 1994].
 The following volume estimates have been conducted using Geographic Information System-based (GIS) tools in order to calculate thicknesses and volumes from digital raster topography data. The volumes have been calculated assuming a base level of the deposits derived from interpolation of available bedrock outcrops at the bottom of the layered sequences [Oosthoek et al., 2007]; uncertainties in the measurements are lower for LTDs located in rather flat areas (e.g., Gale crater floor) and higher for LTDs in more rugged terrains, such as Iani Chaos (characterized by small-scale mesas and knobs).
 LTDs in chaotic terrains such as Aram or Iani are much smaller, starting from few tens to a few hundred km3, e.g., above 30–40 km3 (uncertainty because of unknown basal topography) in Iani southernmost LTDs, as measured from the HRSC Digital Terrain Model (DTM) [Gwinner et al., 2005; Jaumann et al., 2007; Neukum et al., 2004a]. Volumes in crater bulges vary from a few tens of cubic kilometers in small bulges, up to 103–104 km3 in large ones. For Gale and Henry craters total volumes of about 3000 km3 and of about 6000 km3, respectively, have been measured. Within the bulge, its upper, steeper and apparently clinostratified portion (Figure 6) contains approximately 700 km3, as derived from both MOLA and HRSC DTMs.
 Quantitative measurements of bedding attitude in Valles Marineris ILDs show outward dipping layers [Fueten et al., 2006; Gaddis et al., 2006; Hauber et al., 2006], consistent with a mound-like structure. Several deposits show central (crater bulges) or elongated symmetry (Hebes and Gangis Mensae). In particular, Hebes Mensa is E–W oriented, showing consistency with some of the regional tectonic trends, e.g., [Schultz, 1998].
 Several unconformities have also been documented, qualitatively and quantitatively [Fueten et al., 2006]. Even horizontal or subhorizontal layers, such as those in the deposits of Hebes Mensa, are difficult to trace over long distances, showing more or less low lateral continuity. The Gale crater bulge [Rossi et al., 2007b] is characterized by internal unconformities, the largest and most evident being the one separating the horizontally bedded crescent-shaped basal mound and the steeper, less cratered upper mound (Figure 6). Apparent facies variations are visible in most deposits, both laterally and vertically (Figure 6). The bedding dip and direction and the attitude of the contacts between different subunits have been observed and measured [Rossi et al., 2007b] using high-resolution topography and coregistered ortho-rectified imagery, in order to observe the 3D location of the interfaces (bedding, unconformities, etc.) and their respective relationships. In this way angular unconformities could be easily spotted and mapped.
 Gale crater central mound is locally higher than the current rim, which is differentially eroded (mostly the northern side), but the highest point actually corresponds to what is probably the central peak of this 150 km wide crater; the northern rim, coinciding roughly with the Martian dichotomy boundary, is located in an area which appears affected by degradation, collapses and incipient chaotic terrain formation [Watters et al., 2007].
 Dating based on crater size-frequency analysis (Figures 7 and 8) has been performed using a variety of data sets (Figure 7) on various examples among the light-toned deposits presented in this study, covering a wide age span (Figure 8). Nevertheless, all dated deposit surfaces have a relatively young exposure age, being in general younger than 0.5 billion years (Gyr).
 Ages as young as 10 Myr have been determined for Hebes Chasma (Figure 4b) and these very young ages are therefore likely to be erosional ones: this age should be carefully evaluated and is among the less reliable among those presented here. The ILDs in Juventae Chasma have been dated using both HRSC and MOC data [Neukum et al., 2007] providing ages as young as about 400 Myr.
 LTDs in chaotic terrains such as Iani or Aureum Chaos appear moderately eroded, with low crater densities. Measured ages on the slightly bulged LTD in Aram Chaos provide values of around 300 million years (Myr) and 70 Myr (Figure 8) in its eastern portion (Figure 7). Also, a resurfacing event around 100 Myr seems likely, as visible from the characteristic kink in the distribution of the crater size-frequency measurements (Figure 8).
 All young ages derived from crater size-frequency measurements may indicate the timing of deposition of the uppermost LTD layer. In most cases it is rather more likely that either erosional effects have lead to a general underestimation of true ages or that deposition of, e.g., wind-blown material have masked true ages which are consequently supposed to be significantly younger. A decision on which process is more likely cannot be taken.
 Iani Chaos is characterized by two main distinct bodies of light-toned deposits, both embedding chaos knobs: a northern outcrop (Figure 5a) and a southern one (Figures 5b and 7). Dating of both deposits provides consistent ages, around 10 Myr, in terms of the last resurfacing event (Figure 8). These isolated, small, easily erodible deposits within Iani Chaos show remarkably comparable ages, although they are separated by almost 200 km.
 LTDs in Gale and Crommelin Crater bulges have been dated using crater counting (Figures 7 and 8). Crater size-frequency analysis for Gale crater central bulge was carried out using HRSC, CTX (Context Camera) and MOC data (Figure 7). The Gale basal tabular unit (Figures 6 and 7) has a measured age of about 100 Myr; the younger age limit obtained for the uppermost part of Gale mound is around 5 Myr. This apparent range is consistent with the geological and geomorphological observations on the variably eroded bulge (Figure 6). This age is not necessarily a depositional one, but the age range could provide an estimate of the time interval of activity during the formation of the bulge itself.
 Therefore two main age groups emerge from our dating on 3 different classes of deposits (LTDs in Valles Marineris, chaos deposits, crater bulges): an older set of depositional/resurfacing events at a few hundred Myr (e.g., Aram, Crommelin) and a younger one close to 10–20 Myr (Iani, Hebes, Gale, etc.).
 In general, the ages cover a certain time span, but most of them are relatively young. Some, e.g., in Gale Crater, are surprisingly recent, suggesting some kind of geological activity in the very recent geological past [Neukum et al., 2004b, 2007]. This activity could be likely related to eolian erosion, but it would not necessarily be the only active process/agent, in the case of relatively young late-stage depositional activity.
 The presence of different geometries, stacking patterns and multiple unconformities co-existing often in the same chasma or crater suggest variations in the sedimentary evolution (Figures 4, 5, and 6). Spatial and temporal variations in spring mounds could lead to vertical and lateral variations in sedimentary environments, facies and internal geometries [Pentecost, 2005]. A complex internal architecture, developed during multiple formation stages, is common on spring mounds on Earth [Martin-Algarra et al., 2003]: Terrestrial spring deposits do show the co-existence of several different sedimentary subenvironments within the same assemblage of spring deposits [Guo and Riding, 1999; Martin-Algarra et al., 2003; Pentecost, 2005].
 The stable volcano-tectonic settings on Mars could potentially provide long-lived fluid supply to the surface, especially in Valles Marineris: this could justify, possibly together with the lower Martian gravity, compared to Earth, much larger spring deposit buildups. The still very large scale difference between spring deposits on Earth and the candidate spring deposits on Mars discussed here might be more easily explained given the longer time span available for their progressive formation, with multiple discrete events in the same area over long time. No spring deposit on Earth has had as much time (and endogenic power) available as on Mars: in this perspective the lifetime of Tharsis volcanic activity, exceeding 3.5 billion years (Gyr.) [Neukum et al., 2004b], would provide a reasonable engine for driving groundwater over large distances [Andrews-Hanna et al., 2007] for a long time. Regional slopes would also provide an important control on groundwater flow [Andrews-Hanna et al., 2007] in the surroundings of Arabia Terra and Meridiani Planum: this may also explain the concentration of LTDs in Arabia Terra [Allen and Oehler, 2008; Oehler and Allen, 2008].
 The extent of this Earth-Mars spring deposit scaling is not well constrained and should be further investigated. Moreover, a few of the (less extensive) LTDs are not very close to major tectonic provinces, although they could be have been affected by more local effects. Among crater bulges, Gale is located in a more important tectonic region, being positioned on the dichotomy boundary.
 Smaller, thinner deposits such as LTDs in chaotic terrains (e.g., Iani) are apparently less affected by tectonic deformation: they could relate to supplies of subsurface fluids active for shorter periods of time, possibly deriving from smaller and/or more local sources.
 The age estimates of these deposits vary considerably (Figure 8) in the literature, from Noachian to Hesperian [Malin and Edgett, 2000; Scott and Tanaka, 1986] and in some cases Upper Hesperian to Amazonian [Lucchitta et al., 1992] (Figure 8). Moreover, because of the relatively high erodibility of LTDs, the crater counting age could record portions of the craters embedded in the deposits during their formation, which might have been in multiple stages, as we suggest in the present work. Therefore the actual age of the uppermost portions of some LTDs (e.g., Gale upper mound, Figure 7d) might be much younger than that derived from all embedded craters.
 Our new age measurements show relatively (locally very) young ages for LTDs in different geological settings. Although these new ages should be taken with some care, they definitely show that the geological activity, and possibly not only the erosion, might be sensibly younger than Noachian and possibly Hesperian as indicated by the absence of large craters.
 Absolute measurements ages show some correlation among a few of the LTDs described in the present work. Among all dated deposits, some clustering of absolute ages appears at a few million years and a few hundreds of millions of years. Interestingly, they share ages of the same order of magnitude with recent volcanic, fluvial and glacial activity on Mars [Neukum et al., 2004b, 2007].
 Alleged Noachian crater bulges [Malin and Edgett, 2000] in Arabia Terra, similarly to what is proposed for Meridiani Planum soils [Andrews-Hanna et al., 2007] could be possibly related to earlier Tharsis activity, while Chaotic Terrains and Valles Marineris could be linked to later stages (Figure 9). Hematite bearing LTDs in Aram Chaos are considered to be substantially younger than those in Meridiani Planum [Glotch and Christensen, 2005].
 New extensive dating would be needed for constraining the age of several crater bulges, since most currently available age estimates are based on Viking-era data or are poorly constrained: our first attempts, presented here, showed very young ages that do not necessarily reflect the true ages of formation [Malin and Edgett, 2000], however further focused dating efforts would provide better constraints for the formation of LTDs in crater bulges.
 Links could be drawn between multiple depositional events visible in LTDs, both in Valles Marineris and large bulges, such as Gale, and multistage Tharsis activity [Andrews-Hanna et al., 2007; Neukum et al., 2004b]. Indeed, Tharsis activity dates back to the Noachian and developed during Hesperian and Amazonian [Scott and Tanaka, 1986], but the volcanic, and likely hydrothermal and enhanced hydrologic activity, has persisted possibly up to recent times [Neukum et al., 2004b, 2007].
 Hydrated minerals have been detected in various areas of Mars [Bibring et al., 2006], and mono and poly-hydrated sulfates are widespread in Valles Marineris [Gendrin et al., 2005]. A complex association of sulfates and hematite in various segments of the canyon system, such as Ophir and Candor chasmata is present [Weitz et al., 2008], similarly, although with differences possibly because of the different erosion history and physiographic setting [Weitz et al., 2008], to what has been observed by various authors in Aram Chaos [Glotch and Christensen, 2005] and Meridiani Planum [Hynek, 2004].
 The sulfate-bearing outcrops within the LTDs in Meridiani Planum show evidence of aqueous processes [Grotzinger et al., 2005, 2006]. It has been suggested that the original sedimentary environment for the formation of these LTDs could have been a playa/sebkha [Grotzinger et al., 2006].
 In principle, putative playas in Meridiani Planum would be possibly benefit from groundwater availability [Andrews-Hanna et al., 2007] and both rover and orbiter data analysis show that a rise of the water table could have been responsible at least for the alteration of the volcanic bedrock [Arvidson et al., 2006]. In addition, a fumarole-driven hydrothermal alteration of volcanic and volcaniclastic material in Meridiani Planum has been invoked in order to explain the diagenetic features observed by the MER rover Opportunity [McCollom and Hynek, 2005]. The “playa” interpretation for the deposits in Meridiani Planum, possibly associated with groundwater movement and evaporitic pumping is in principle compatible with spring activity and the formation of spring mounds [Komatsu et al., 2007]. Therefore Meridiani playa-like LTDs and Crommelin mound-like LTDs could be the two faces of the same medal.
 Variations on the hydrological, physical and chemical regime through time could justify variations in composition and mineralogy [Murchie et al., 2007] along the stratigraphical sequence of LTDs (Figure 9). Moreover, the possible detection of mafic minerals interbedded with sulfates in Valles Marineris ILD [Murchie et al., 2007], for example in Hesperian to Amazonian units [Lucchitta, 1999] could be related to the presence of tephra trapped inside the LTDs produced by Tharsis volcanic activity (Figure 9), as suggested in the case of Hebes Chasma [Peterson, 1982].
 Estimates of maximum hydraulic heads for Valles Marineris are compatible with the present maximum topographic relief of ILDs, since none is higher than immediately surrounding plateaus (Figure 3). Those theoretical maximum values would require the latest spring deposits to be formed at the top of the LTDs in the canyon system and to have the presence of confined aquifers, with the possibility of a temporary water fill of the basins. Lateral accretion rather than vertical over multiple generations of spring mounds would require smaller hydraulic heads and local aquifer confinement, which is likely to be achieved with cementation within the mounds themselves.
 Gale crater appears to be an exception, but its position at the very edge of the crustal dichotomy boundary, where extensive erosion, degradation and chaos formation has occurred [Watters et al., 2007], could justify the significant erosion and reduction of relief of its rim.
 The thickness and base level of ILDs show a strong correlation with both plateau elevation and location (Figure 3), with inferred hydraulic heads diminishing eastward (e.g., Capri Mensa is less thick than Candor, Ophir Mensae) in the Valles Marineris canyon system [Lucchitta et al., 1994]. This correlation would strongly suggest the role of regional and local hydrogeology in the emplacement and shaping of ILDs (Figure 3). A certain amount of erosion of LTDs should also be taken into account, depending on the setting: likely minor in a completely closed basin (e.g., Hebes Chasma), but eolian erosion (not easily constrained) on these apparently not very resistant materials should also be taken into account. The topographic setting of Valles Marineris with respect to the Tharsis bulge (Figure 3) is consistent with the presence of (at least in the past) aquifers beneath the Tharsis bulge and in the high plateaus surrounding the canyon region [Harrison and Grimm, 2007].
 The possibility of such a large volume of spring deposits would require the presence of a significant amount of water in the subsurface. The presence of such water reservoirs has been proposed for the ancient pre-Tharsis development of the Valles Marineris area [Dohm et al., 2001] (Figures 9b and 9c). Although terrestrial spring deposits are mostly related to meteoric waters [Pentecost, 2005], Martian ones could be more related to juvenile sources. The extremely long lived Tharsis super-plume could have led to a significant supply of water from deep mantle sources [Farmer, 1996].
 In the case of Valles Marineris ILDs, the formation of spring deposits would generally postdate the canyon formation itself, consistently with an emplacement of ILDs within an already formed canyon [Chapman and Tanaka, 2001; Lucchitta et al., 1992].
 The fluted or streamlined shape of various LTDs in Valles Mariners, mainly Ophir, Candor and Capri Mensae, indicates at least a morphological record of fluid flow and erosion, possibly related to large-scale outflow and/or smaller-scale spring activity. If the flow were syn-post Valles Marineris formation, this would put some constraint on the relative formation/modification age for these deposits. Ganges Mensa, which is located in a semi-open chasma, appears much less streamlined. More obviously, Hebes Mensa does not show major signs of unidirectional flow modifications, being in a closed basin.
 The scenario of spring deposits formed after the emersion of subsurface fluids would be consistent with a pseudokarst-like mechanism [Otvos, 1976], between Tharsis rise and the plateaus surrounding Valles Marineris, compatibly with subsurface cavernous systems as proposed by some authors [Rodriguez et al., 2003].
 Thus, summarizing, we propose a scenario for the large-scale spring deposit hypothesis on Mars: Valles Marineris LTDs are interpreted as an association of large-scale spring mounds (in a broad sense) formed over varying durations with discrete steps of activity, possibly related to periods of increased volcanic activity. This activity, likely partially or totally contemporaneous to the formation of LTDs would have been recorded within the sedimentary record, possibly as tephra (Figure 9a). The amount of erosion which occurred after the formation and emplacement of the LTDs is not easily constrained (ideally indicated by the thick dashed line in Figure 9a). Especially in the case of Valles Marineris, the close presence of Tharsis and its long-lasting activity and possibly large aquifers beneath the plateaus [Dohm et al., 2007] would offer a reservoir of subsurface fluids (Figure 9b), which could have supported: (1) subsurface erosion and mass transfer, producing collapses and formation/enlargement of cavi/chasmata [Rodriguez et al., 2003] and (2) provided dissolved sulfates to be deposited within the depressions/collapsed areas as spring deposits formed over multiple stages (Figure 9c). An association of collapses and tectonic deformation during initial and later stages of Valles Marineris development [Schultz, 1998] is also possible: such a complex combination of processes would provide even more pathways for subsurface fluids to reach the surface [Treiman, 2008]. The diagram in Figure 9 summarize the large-scale spring deposit scenario, including the structure and interplay between ILDs formation and the Tharsis volcaniclastic contribution (Figure 9a), the relationship between the Tharsis province, Valles Marineris and smaller cavi and chasmata (Figure 9b), and the possible extrapolation through time of the scenario presented here (Figure 9c). The possible long-term role of tectonic extension in Valles Marineris is also highlighted (Figure 9c).
 Large, coalescing spring-mounds, with local variations of facies, environment and geometry, active in different times as a response to increased activity, linked to Tharsis volcanism [Andrews-Hanna et al., 2007; Neukum et al., 2004b] and related to its products, would explain the geological variability of LTDs in Valles Marineris and more peripheral deposits in chaotic terrains. Smaller-scale widespread activity, not only related to Tharsis, could be linked to the peculiar mounds in large craters, mostly in Arabia. The need for long-term strict Earth-like conditions [Baker et al., 1991] would not be a too stringent requirement in order to explain the deposition of LTDs in Valles Marineris and chaotic terrains as spring deposits: still, liquid water should be stable, but the climate would not have to be as warm and wet as that needed to develop and sustain a full hydrologic cycle. In particular, ILDs in Valles Marineris, which date back to Hesperian or Amazonian [Lucchitta et al., 1992; Lucchitta, 1999], were deposited during possibly unfavorable climatic and environmental conditions [Bibring et al., 2006], which would be too harsh to support the development of lacustrine basins. In this respect, some of the advantages and the attractive features of the subice volcano hypothesis [Chapman and Tanaka, 2001] are shared by the large-scale spring deposits scenario presented here, with the advantage of not requiring a large stable presence of ice occupying the canyons and chasmata.
 The present and future robotic missions to Mars could offer the possibility of testing, up to a certain extent, the large-scale spring deposit hypothesis presented here.
 Not all the specific characteristics of spring deposits on Earth could be used to uniquely identify their Martian counterparts, including micro- or meso-facies, in the absence of lander/rover close-up data over candidate spring deposits. A combination of features that could possibly be of help in testing the spring deposit hypothesis for LTDs include those listed below. In brackets the spatial scale of the feature is indicated, classified into mega- (at the scale of the entire deposit: as visible from medium-high resolution, wide-swath satellite imagery, e.g., HRSC) macro- (at the scale of several km to few hundreds of meters, e.g., as visible on high-resolution images, e.g., MOC, CTX), meso- (at the scale of hundreds to few meters, e.g., HiRISE), micro- (at the scale of the hand specimen/micro-imager, e.g., with imaging instruments on rovers):
 1. Sedimentary-like appearance of the deposits at regional, local and outcrop scale (mega-/macro-/meso-/micro-).
 2. General mound-like morphology (mega-/macro-/meso-).
 3. Coexistence of different textures (e.g., on remotely sensed data), facies, morphologies over short distances (macro-/meso-).
 4. Local compositional homogeneity or systematic/periodical compositional variations along the stratigraphical sequences. (macro-/meso-/micro-).
 5. Evidence of fault or fracture-controlled deposition of the sedimentary-like layers (macro-/meso-).
 6. Presence of possible source areas within the extent of the deposit, for autochthonous spring deposits (macro-/meso-).
 7. Directional variation of thickness/facies over short distances ascribable to discrete sources (macro-/meso-).
 8. Correlation between LTD extent/elongation and local/regional structural trends (mega-/macro-).
 9. Morphology and internal structure (from both orbital images and outcrop observations, e.g., with rovers) suggesting draping, accretion or progradation over bedrock or other LTDs (macro-/meso-/micro-).
 10. Evidence for non-volcanic origin (e.g., composition compatible with sedimentary deposits, as derived from imaging spectrometers on board orbiters and rovers).
 A few of the features listed above could be detected and studied with remote sensing instruments, either medium to high-resolution imagery and high-spatial (and spectral)-resolution images and spectrometers; some others could be more reliably addressed with detailed observation at the meso- and micro-scale, which would make current and future landing missions more suitable.
 Single processes could not easily produce such a large morphological, geometrical, textural and compositional variability in the resulting LTDs. Alternative mechanisms able to produce sedimentary deposits and morphologies somehow similar to those produced by spring activity include: volcanism [Chapman and Tanaka, 2001] or fluvio-lacustrine processes [Pondrelli et al., 2005]. Locally facies of lacustrine or spring origin could be confused, but lacustrine ones would have a much wider lateral continuity and they would occur within a traceable basin. Of course spring-fed deposits can include lacustrine or paludal facies as well [Pentecost, 2005], but in the context of local sources rather than with a basin-wide occurrence. A point of discrimination, as mentioned above, would then be the stronger vertical and lateral facies variation in any of the LTDs, if formed as a spring mound or as the result of the coalescence of several individual mounds over a large period of time.
 Therefore the key for testing the spring-deposit hypothesis on Martian LTDs is the co-occurrence of several features (like those listed above) at multiple scales, from both orbiter and rover platforms, using multiple data sets (such an approach is indeed useful for most geological problems).
 Although the task of uniquely identifying spring deposits within sedimentary material on Mars is challenging and not easily solvable by robotic missions orbiting or landing, in particular future high-mobility rover missions, such as Mars Science Laboratory (MSL) or ExoMars, could provide both the detailed morphological, geological, mineralogical information together with the local/regional outline needed to constrain the nature of eventually investigated LTDs (if any of the several LTDs mentioned here are indeed selected as landing sites).
 In the meantime, further work on comparing the stratigraphical, compositional, geometrical and structural features of Martian LTDs with terrestrial spring deposits is needed in order to better constrain the combination of key features to be matched.
 We propose a large-scale spring mound origin for various light-toned deposits on Mars. Given the wide range of possible architectural styles, facies, morphologies and surface composition possibly resulting from spring mounds, this concept can explain in a comprehensive way several Martian light-toned deposits. Although precipitation from hydrothermal waters has been proposed for certain LTDs, for the hematite associated with them [Catling and Moore, 2003; Chan et al., 2004; Glotch and Christensen, 2005; Weitz et al., 2008] and for several deposits in craters, the scenario proposed here explains the actual buildup of major light-toned deposits on Mars.
 The spring deposit scenario particularly suits the materials with strong analogies with sedimentary rocks, but not easily fitting with lacustrine or volcanic generic interpretations, unlike clearly fluvio-lacustrine basins such as Holden [Pondrelli et al., 2005]. If light-toned deposits, unevenly distributed on the surface of Mars, are indeed, even just in part, spring deposits; this would have several implications for the geological evolution of Mars and the variety of its sedimentary environments and their possible potential for life. Also, spring deposits locally as young as Amazonian [Lucchitta, 1999] would have offered a suitable setting for life to sustain and fossilize even in a very “dry” scenario for Martian post-Noachian hydrological evolution [Bibring et al., 2006], being a possibly protective shelter even during harsh surface and atmospheric conditions. Testing of the large-scale spring deposit scenario for LTDs on Mars with future landing missions, could offer valuable new information on Mars' still enigmatic geological features.
 We thank the HRSC Experiment Teams at DLR Berlin and Freie Universität Berlin as well as 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 acknowledge the effort of the HRSC Co-Investigator Team members and their associates who have contributed to this investigation in the preparatory phase and in scientific discussions within the team. We thank MOLA, MOC, THEMIS, CTX, and HiRISE respective teams for making data available to the public through PDS. Thanks for insightful comments and discussion to Nicolas Mangold, Vincent van Hinsberg, and Paul Mason. We greatly thank Aileen Yingst and an anonymous reviewer for their thorough and constructive comments and suggestions that provided valuable information and substantially improved the manuscript. Our warmest thanks to Greg Michael for reading the final manuscript and providing precious comments. Many thanks to Mihaela Glamoclija and Gian Gabriele Ori for providing useful pictures. This work was carried out during a research fellowship at the European Space Agency (ESA).