Several hundred occurrences of chloride-bearing salt deposits have been proposed in terrains within the southern highlands of Mars on the basis ofThermal Emission Imaging System and Thermal Emission Spectrometerinfrared observations. The spectral identification of chloride salts by remote sensing is challenging because they are transparent over much of the thermal infrared portion of the spectrum. Further ambiguity arises from the diverse geologic settings in which the putative chloride-bearing materials are found. In order to better constrain the composition of these unique compositional units, we perform a global survey of these materials in the Near-Infrared (NIR) domain with theObservatoire pour la Minéralogie, l'Eau, les Glaces et l'Activité(OMEGA) imaging spectrometer. The spectral signatures of the deposits are consistent with – although not specific of – chlorides. We do not observe olivine to be associated with the deposits, which confirms that sulfides are an unlikely alternative candidate. Our systematic search reveals the global lack of association with hydrated minerals (phyllosilicates, sulfates, hydrated silica) except for a few deposits (noteworthy in northwestern Terra Sirenum) where a small fraction of chloride material overlaps Fe/Mg-rich clay-bearing terrains. Even in these locations, the morphology and crosscutting relationships of the deposits suggest two separate episodes of mineralization, first phyllosilicates then chlorides, followed by subsequent formation of sulfates. Our study shows that local groundwater upwelling seems to be the most frequent source for the water involved in the formation of chloride, rather than surface runoff.
 Chloride salts are among the secondary minerals that were probably produced by aqueous alteration of the Martian surface. Evidence for their presence has first come from studies of Martian meteorites [e.g., Bridges et al., 2001] and from in situ analyses of the Martian soils [Ming et al., 2008]. Recently, potential deposits of chloride salts have been mapped from orbital observations [Osterloo et al., 2008, 2010] using Decorrelation Stretch images (DCS) from the Thermal Emission Imaging System (THEMIS) and infrared spectral data (7–43 μm) from the Thermal Emission Spectrometer(TES). These potential chloride-bearing materials form km-sized deposits and are widespread across the southern highlands of Mars (seeFigure 1) [Osterloo et al., 2010]. A total ∼640 deposits has been identified with formation times spanning over the first billion years of Mars history based on geologic units association and crater counting technique [Osterloo et al., 2010]. These deposits could result from precipitation in ponds and are therefore potential witnesses of the paleo-environmental aqueous conditions of Mars [e.g.,Kendall and Harwood, 1996].
 The identification of chloride-bearing materials in THEMIS and TES data is based on their particular thermal infrared features that are the results of an underestimation of the derived temperature in the process of converting radiance to emissivity [e.g.,Bandfield, 2009]. The cause of the incorrect temperature derivation has been associated to the chloride salts relatively uncommon propriety of a maximum emissivity less then unity in THEMIS and TES thermal infrared spectral range [Osterloo et al., 2008]. As Osterloo et al.  pointed out, chloride salts and more generally halides are the best candidates to explain these spectral characteristics, but are not the unique compounds that could account for the non unity emissivity. Sulfides could be alternative candidates, which are proposed to form on Mars by ultramafic lava flows or by hydrothermal deposits [Burns and Fisher, 1990]. The conditions leading to the formation of such materials are different from those of chloride salts. However, recent laboratory measurements have shown that sulfide-basalt mixtures should exhibit spectral evidence for oxidation which has not yet been observed within these deposits [Jensen and Glotch, 2011].
 Most of the aqueously altered minerals found on Mars have been detected by their near-infrared (NIR, 1–3μm) spectral characteristics in reflectance spectra [e.g., Poulet et al., 2005; Gendrin et al., 2005; Milliken et al., 2008; Murchie et al., 2009]. Specifically, absorption features due to water of hydration near 1.9 μm and metal-OH vibrations in the 2.2–2.5μm range enable the identification and mapping of hydrated minerals [Clark et al., 1990]. Neither (anhydrous) chlorides nor (unoxidized) sulfides exhibit absorption band, thus they cannot be identified solely on the basis of NIR [Jensen and Glotch, 2011; Hanley et al., 2011]. So far, analyses of NIR data from Compact Reconnaissance Imaging Spectrometer for Mars(CRISM) signatures of a few putative chloride sites have revealed a lack of absorption bands. Studying NIR spectral properties of putative chloride-bearing materials (hereafter quoted PCs) deposits is, however, relevant for the following reasons: 1) sulfate salts and clays may form in the vicinity of PCs due to the expected presence of liquid water during chloride formation [Baldridge et al., 2009], 2) sulfides may show typical spectral signatures if oxidized [Jensen and Glotch, 2011; Burns and Fisher, 1993], 3) ultramafic minerals such as olivine are expected to be found in association with sulfides [Osterloo et al., 2010]. Detecting hydrated minerals would exclude sulfide as a main component of the material and would help to constrain the geochemical conditions. On the other hand, the presence of ultramafic minerals such as olivine would support sulfides as a likely bulk component.
Osterloo et al.  did not find any PC in hydrated areas detected by OMEGA. However, this comparison was based on OMEGA data derived from the nominal mission (one Martian year) only [Poulet et al., 2007]. A few PCs analyzed with CRISM NIR orbital data have revealed the co-occurrence of Fe/Mg phyllosilicates [Murchie et al., 2009; Wray et al., 2009]. The mineralogical context of the terrains indicates that chloride salts are the main component of the material, and precipitation in ponds as the favored formation process [Glotch et al., 2010; Davila et al., 2011; Wray et al., 2011]. CRISM high spatial resolution observations have a limited spatial coverage (10 × 10 km images, 15–19 m/pixel) [Murchie et al., 2007], therefore most PC sites and their surroundings have not been observed and their nature remains mainly unconstrained in this wavelength range. Here we use the NIR global coverage observations provided by the Observatoire pour la Minéralogie, l'Eau, la Glace et l'Activité (OMEGA), [Bibring et al., 2004; Ody et al., 2012] on board Mars Express (MEx) to constrain the spatial mineral assemblage of these deposits and their surroundings. We then perform a detailed mineralogical and morphological study of PC sites where hydrated minerals are also detected. We finally discuss the composition and possible formation processes of PCs.
2. Data and Methods
2.1. OMEGA Mineralogical Data Analysis
 Since 2004, OMEGA/MEx has been acquiring hyperspectral images of the surface of Mars, thus providing visible to NIR spectral properties of the surface. At these wavelengths diagnostic spectral features of minerals are detectable and allow for the mapping of compositional units at the surface [e.g., Griffes et al., 2007; Loizeau et al., 2007; Mangold et al., 2007; Poulet et al., 2007, 2008; Williams et al., 2010; Wiseman et al., 2010]. The OMEGA instrument measures reflected sunlight from the surface between 0.4 to 5.1 μm. Here we focused on wavelengths of the C channel (from 1 to 2.7 μm), where key spectral features of mafic, ultramafic and hydrated minerals are present [Bibring et al., 2005]. In order to recover surface components from the measurements, radiance spectra were corrected for solar irradiance and atmospheric absorptions using the OMEGA standard reduction scheme of Langevin et al. . These corrections provide I/Fcos(i) (with i the incidence angle) reflectance factor spectra at a spectral binning of 13 nm at the beginning of the mission (128 spectral elements). Due to detector aging, some spectral channels were then lost.
 We conducted a systematic study of the spectral properties of the PCs for the entire OMEGA data set (more than 3 Martian years). The last NIR C-channel observation was obtained by OMEGA in July 2010 [Ody et al., 2012]. The data set corresponds to about 8600 cubes, namely ∼700 millions spectra. Given the small size of the PCs (an averaged extent of ∼24 km2 although with a wide range of variation [Osterloo et al., 2010]), we selected OMEGA data cubes with high spatial resolution (<2.5 km/pixel), low noise and of the nadir-pointing mode. Periods of dust storms and observations with clouds, which partially masked surface spectral properties, were identified by visual inspection and excluded. These filtering processes allowed the analysis of 77% of the ∼640 PCs mapped byOsterloo et al.  (Figure 1).
 We implemented an automated procedure to select a window size of ∼60 km centered on each PC identified by Osterloo et al. . We then performed the following survey:
 (i) Visual inspection of the reflectance spectra of the PC unit and its surroundings. Attention was given to noise effects, instrumental artifacts, presence and strength of the pyroxene band, overall slope and mean reflectance value. These spectra gave preliminary insights into the spectral properties of the surface material and the data quality.
 (ii) Building of spectral maps, based on spectral parameter. Each spectral parameter measures the strength of a characteristic mineral absorption band. Absorption bands of pyroxenes, hydrated silicates and monohydrated sulfates were calculated with the band depth method (Table 1, Figure 2, and section 2.3). In order to reduce false detections, a threshold defined by Poulet et al.  and Griffes et al.  was used for each band depth parameter (Table 1). Finally, the boundaries of each PCs provided by Osterloo et al.  were outlined on the different parameter maps.
 (iii) Potential detections of hydrated minerals were then verified manually by spectral ratioing. The spectral ratio technique divides a spectrum of a potential hydrated spot (Region Of Interest, ROI) by a nearby spectrum (e.g., located in a non-hydrated region). The ratio highlights the spectral features of the hydrated minerals. This technique was also applied to the PCs. Given the small size of these deposits that often cover only a few OMEGA pixels, such a technique emphasizes the signature especially when there is a spatial mixture of PC with other materials. Where possible, binning of OMEGA pixels was performed to increase the signal-to-noise ratio (SNR). The retrieved ratio was compared to spectral libraries by visual inspection [Clark et al., 2007] (RELAB - Reflectance Experiment Laboratory - Brown University).
2.2. Image Data Sets
 The ROIs detected with the method described above are then investigated for their geologic context. This is performed with a geographic co-registration of key minerals summary maps with THEMIS IR night images at 100 m/pixel [Christensen et al., 2004] and various visible images (described below). This step is carried out in a Geographic Information System (ESRIs ArcGIS) software package often with a manual co-registration of the maps. The identification of PCs within the outlined boundaries is based on visible image characteristics described byOsterloo et al. : lighter toned than the surrounding terrain and often light blue in High Resolution Imaging Science Experiment (HiRISE) false color images. THEMIS IR night images are used as a proxy for thermal inertia and thus are helpful in the identification of dust free outcrops: e.g., high brightness temperature in THEMIS IR night imagery would suggest relatively high thermal inertia [e.g., Fergason et al., 2006].
 Visible images are provided by several cameras. High Resolution Stereo Camera (HRSC) panchromatic nadir images provide almost complete coverage of the sites with resolution between 10 and 50 m/pixel [Neukum et al., 2004; Jaumann et al., 2007]. The Context Camera (CTX) provides images with a resolution of 6 m/pixel [Malin et al., 2007], HiRISE image data have a maximum resolution of 0.25 m/pixel [McEwen et al., 2007]. Note that detailed morphological study was not performed for all ROIs because of unavailable HiRISE coverage. Finally, we used topographic data from Mars Orbiter Laser Altimeter (MOLA) [Smith et al., 2001] for investigating PC sites elevations and for the identification of basins and channels.
2.3. Spectral Parameters
 The pyroxenes band depth is calculated using a broad absorption band at 2 μm resulting from the mixture of low-calcium pyroxenes with a broad band centered at ∼1.9μm, and high-calcium pyroxenes with a broad band centered at ∼2.3μm (Table 1) [Poulet et al., 2007]. Hydrated phases such as most phyllosilicates and polyhydrated sulfates are detected based on a narrow absorption band centered at 1.9 μm due to vibrational overtones and combinations of H2O in the mineral lattice [Clark et al., 1990; Cloutis et al., 2006]. Other diagnostic features occur in the 2.0–2.6 μm wavelength range due to OH stretch and metal-OH bend for phyllosilicates and S-O overtones as well as H-O-H/O-H combinations and overtones for sulfates [Bishop et al., 1994, 2008; Cloutis et al., 2006]. These features are used to characterize the nature of phyllosilicates and sulfates. Monohydrated sulfates (e.g., kieserite) have the characteristic absorption band of hydrated silicates at 1.9 μm shifted toward 2.12 μm [Gendrin et al., 2005; Arvidson et al., 2005; Cloutis et al., 2006]. A band depth is calculated at this wavelength to ensure identification of this additional sulfate mineral class [Griffes et al., 2007]. It should be noted that water ice and carbon dioxide ice also show an absorption band near 1.93 μm [e.g., Langevin et al., 2007]. Since they are seasonally present on the surface in the southern and northern mid- to high- latitudes [e.g.,Vincendon et al., 2010], they are mapped using specific absorption bands (Table 1) in order to avoid false detections of hydrated silicates.
 The search for ultramafic associate minerals (olivine) is based on the global distribution of the olivine rich-terrains derived with the entire OMEGA data set (A. Ody et al., Global investigation of olivine on Mars: Insights into mantle and crust composition, submitted toJournal of Geophysical Research, 2012): planetary scale maps of olivine-rich terrains are compared with the PCs mapped byOsterloo et al. . When a close association was found, the detections of the olivine-bearing materials have been validated with the spectral ratio technique.
3.1. Associated Mineralogy and Distribution
 We find that spectral signatures of PCs mostly correspond to typical, widespread OMEGA spectra dominated by pyroxene and/or dust. In this section, we analyze the spectral signatures of PCs and adjacent terrains to derive an associated mineralogy, and to constrain the relationships among these association at the planetary scale.
 We have divided the NIR characteristics of PCs into four mineralogical associations: “pyroxene,” “featureless,” “Fe/Mg phyllosilicates” and “olivine” (Figure 3). Around 60% of the sites are characterized by a signature of pyroxene covering the PCs and extending in its surrounding. These sites exhibit NIR reflectance values lower than 0.2. The spectral signature of pyroxene varies significantly from site to site and can be explained by variations in pyroxene abundance, grain size, composition or dust coverage. The spectral ratio technique applied to these sites shows a spectrum with a red slope in the 1–2.5 μm range and lacking absorption bands (Figure 3b). These characteristics of the PC have been also reported in CRISM studies by Murchie et al.  and Glotch et al. . We should note here that the uncertainty on the shape of the derived ratio is great, depending notably on the choice of the surrounding spectrum used in the ratio.
 The second most common (30%) associated mineralogy is spectrally featureless (Figures 3c and 3d) with an albedo in the 0.2–0.25 range. These values fall between the low albedo values of the mafic-rich highlands and the high albedo values (>0.30) of the dusty northern plains. It should be noted that such spectrally featureless regions are note unique on Mars, since they are also observed in highland areas devoided of PCs. Finally, the spectral ratio applied to these sites also reveals a lack of absorptions and a red slope similar to the pyroxene associated sites (Figure 3d).
 Around 9% of the surveyed sites indicate the presence of hydrated spot(s) within the 60 × 60 km2 size window. The spectral ratios reveal that all the spots are characterized by absorption bands at 1.93 μm and ∼2.3 μm with an additional 1.4 μm narrow band for most of the spots. The observed spectral signatures are thus consistent with Fe/Mg-rich hydrated phyllosilicates: Fe/Mg smectites, vermiculite and smectite-chlorite mixed-layer clays provide the best matches (compareFigures 3g–3h with spectra e and f of Figure 2). The PCs spectra associated with the Fe/Mg-rich hydrated phyllosilicates show similarities with the pyroxene and/or the featureless deposits (Figure 3e). The spectral ratio always reveals the characteristics common to the two above associations: absence of absorption bands and red slope (Figure 3f). There was no spectral evidence for hydrated sulfates based on the 1.93 μm and the 2.12 μm criteria.
 Comparisons between the locations of PCs and olivine spots revealed that both minerals occur together in a spatially limited area (<30 km) for a few sites only. In most cases, they are found at large distances (hundreds of km).
Figure 4shows the spatial distribution of the four associated mineralogies over the MOLA map and OMEGA-derived dust coverage map fromOdy et al. . At a global scale the pyroxene association is widespread on the southern low albedo, dust free (yellow areas of Figure 4, bottom) and pyroxene-bearing highlands [Poulet et al., 2007]. Conversely, the spatial distribution of the featureless deposits is uneven and correlated with regions exhibiting intermediate values of the OMEGA-derived dust coverage. These spots are rarely found in pyroxene-rich regions. High albedo (>0.30 at 1.1μm) is globally due to the nanophase oxides rich dust layer [Poulet et al., 2007], and it is possible that the observed sites are partially covered by dust [Ody et al., 2012]. The Fe/Mg phyllosilicate association occurs over a wide range of longitudes and latitudes, but is mainly concentrated in Terra Sirenum (30°S, 190°E) and in Margaritifer Terra (5°S, 345°E). The few spots associated with olivine are scattered over the highlands in low albedo, dust free areas.
3.2. Geomorphologic Contexts
3.2.1. Featureless and Pyroxene Associated Mineralogy
 To characterize the context of the featureless deposits, we use albedo and pyroxene maps combined with high spatial resolution imagery. The albedo map indicates intermediate values associated with the PC deposits, while the pyroxene map indicates small values. This reflects low pyroxene content and intermediate dust coverage of the surface. One of these sites is shown in Figure 5 where the PC corresponds to brighter material with no pyroxene signature. At high spatial resolution, the PC corresponds to a stratigraphic low and bluish unit. It is surrounded by a brown smooth thin mantle with pyroxene signature identified in its eastern part. The small size of the deposits is at the limit of the OMEGA spatial resolution. Hence, the OMEGA signature is likely a spatial mixture of PC and of mantle material. Another example is shown in Figure 6. The small PCs are hardly distinguishable from the surrounding terrain and the entire region appears to be covered by a fine-grained layer. A weak pyroxene signature is also present at this site. This suggests that these sites are partially covered by dust and low content pyroxene-bearing mantle, which is possibly masking the composition of the substrate.
 Two typical geomorphologic contexts of the pyroxene association are exemplified on Figures 7 and 8. Pyroxene spectral maps show an extensive signature over PCs as well as on neighboring areas. The pyroxene signature comes from the basaltic composition of the uppermost material, which is widespread in the southern highlands [Bandfield, 2002; Poulet et al., 2007]. In Figure 7, high spatial resolution imagery reveal that dark dunes partly overlay bright toned deposits (interpreted to be the PCs), thus indicating the dark material to be mobile. The pyroxene signature found at the PC therefore probably comes from this dark eolian material. In Figure 8, the context image reveals dark streaks indicative of eolian activity. At higher resolution eolian material may be inferred by the smoothness of the area. The PCs appear as bluish patchy materials covering a small fraction of the layout mapped by Osterloo et al. . The spectral ratio that emphasizes subtle spectral differences in a mixture removes the (presumably) pyroxene signature and provides insight into the composition of the bright material (presumably PC) (Figure 3b). Except for a red spectral slope (in contrast to the flat infrared continuum slope characteristic of dusty material), no clear signature is present in the spectral ratio. These spectral properties are consistent with a significant content of a high-albedo, anhydrous phase such as a chloride salt [Jensen and Glotch, 2011].
3.2.2. Olivine Association
 The olivine detections correspond to a wide range of settings, but occur within well-defined geomorphologic units.Figure 9ashows that olivine-bearing materials correspond to remnant hills. An extensive smooth plain embays the hills indicating that the olivine-rich hills are outcrops of stratigraphically lower material. Because PCs are only found within or superposing the plains, they do not have a clear genetic relationship with the olivine-bearing material and are probably of younger age. InFigure 9bthe olivine-bearing material corresponds to a smooth plain that covers the floor of heavily degraded impact crater. By contrast, PCs are found in a rugged unit away (>20 km) from the smooth crater floor. We thus exclude a clear genetic and stratigraphic relationship between the two materials.
3.2.3. Phyllosilicates Associations
 The signatures of Fe/Mg phyllosilicates are found in two different settings. (1) Localized, relatively high thermal inertia eroded terrains. These terrains are found at several to tens of km away from PCs and are referred to as “regional phyllosilicates” in section 188.8.131.52. They have not been previously reported nor investigated. (2) Knobby, relatively high thermal inertia terrains. These terrains are only found in spatially limited areas and are juxtaposed to the PCs. They are presented as “juxtaposed phyllosilicates” in section 184.108.40.206. Three of them have been previously reported and one site has been previously investigated.
220.127.116.11. Regional Phyllosilicates
 In Margaritifer Terra, in elevated sites within Thaumasia Planum, and in other scattered locations, PCs are found within 60 km from spots of Fe/Mg phyllosilicates. At a regional scale, the common characteristic of these spots is their low stratigraphic position (Figures 10 and 11). Their surface has an eroded and rough appearance at the CTX resolution. In one case they are associated with impact ejecta and therefore are likely excavated material (Figures 11c1 and 11c2). These characteristics are common for outcrops of Fe/Mg phyllosilicate-bearing crust, and hundreds of similar terrains have been reported [Murchie et al., 2009; J. Carter et al., Hydrous minerals on Mars as seen by the CRISM and OMEGA imaging spectrometers: 1. Updated global view, submitted to Journal of Geophysical Research, 2012]. PCs are usually located at >10 km apart from these Fe/Mg phyllosilicates-bearing terrains and are distinct, stratigraphically higher and superficial deposits. Since Fe/Mg phyllosilicate spots are fairly common at the surface and presumably within the subsurface, we consider these detections of PCs and Fe/Mg phyllosilicates in the same scene as coincidental.
18.104.22.168. Juxtaposed Phyllosilicates
 Six terrains of juxtaposed Fe/Mg phyllosilicates are clustered in northwestern Terra Sirenum (NW TS hereafter), between 180 and 210°E and 20–35°S (Figure 12), three of which correspond to previously identified spots of Fe/Mg phyllosilicates. They occur on shallow intercrater plains that have been mapped as cratered and subdued cratered unit (Npl1 and Npl2) by Scott and Tanaka . Visual investigations show that the terrains are located at different elevations and that the surrounding topographic settings do not suggest the presence of basins. Visual inspection indicates that the terrains are also away from dendritic >∼1 km wide valley networks mapped by Hynek et al. . The Fe/Mg phyllosilicate-bearing material forms a few tens of km wide and 10–20 m thick deposits with high thermal inertia and knobby surfaces (Figures 13, 14 and 15). By contrast, PCs form a discontinuous, thinner layer with a smooth surface and forming inverted channels in places (Figure 13f and 14). The inverted channels are 10s of m wide and several km long.
 A stratigraphic relationship can be inferred between juxtaposed Fe/Mg phyllosilicate and PCs. PCs embay and overlay Fe/Mg phyllosilicate-bearing materials, which is in favor of a temporally uncorrelated formation. The two most obvious examples that support two distinct episodes of mineralization are shown inFigure 13. In Figures 13a–13c, the PC appears to blanket the underlying topography of Fe/Mg phyllosilicates-bearing knobs. This setting suggests that the PCs constitutes a very thin (few meters thick at most) layer. InFigures 13d–13fthe PC layer is seen to overlap the Fe/Mg phyllosilicate-bearing knobs. From these examples, the PC formation must have occurred subsequently to the formation of Fe/Mg phyllosilicates. InFigure 14, PC is seen to surround a Fe/Mg phyllosilicate-bearing mesa (Figure 14b2) implying a formation during two distinct episodes.
 A deposit of juxtaposed Fe/Mg phyllosilicates is also observed outside NW TS (Figure 15) near the Knobel impact crater, at the boundary between the highlands and the lowlands (140°E–7°S). The material is found in a local topographic low at the mouth of a valley system and at the source of a northeast trending valley. In addition, there are several linear structures that could be inverted channels indicative of a possible role of surface water runoff. Fe/Mg phyllosilicates materials are scattered in this region and associated to the most eroded terrains. The largest clay-bearing unit is spatially correlated with eroded deposits covered by a dark cap unit (Figure 15c). The PC appears as light-toned patches partially covered by darker units and are found near scattered Fe/Mg phyllosilicate-bearing deposits. The curvilinear boundary of the PC is interpreted to be the result of wind erosion and removal of the overlying dark unit and not due to deposition [Osterloo et al., 2010]. Because of the strong eroded appearance of both the PC and the Fe/Mg phyllosilicate-bearing terrains, no clear relationship can be established, while the close vicinity in this topographic depression does not exclude a possible genetic relationship between the two minerals. The observation of a valley system near these materials strongly suggests hydrologic activity in the form of surface water runoff [e.g.,Howard et al., 2005]. Formation and remobilization of secondary minerals in fluvial systems and open-basin lakes has already been observed on Mars [e.g.,Milliken and Bish, 2010; Erkeling et al., 2012; Goudge et al., 2012]. This site, however, presents a unique mineral assemblage with PCs and Fe/Mg phyllosilicates.
4.1. Composition of PCs
 One of the ambiguities regarding the PC composition is that their NIR characteristics are similar to those of sulfide-bearing materials. However, the discrimination between chlorides and sulfides can be inferred by characterizing the co-occurrence of primary or secondary minerals. For sulfides, these can be olivine formed in a volcanic system (i.e., lava flow) or oxidation products of sulfides, such as sulfates or iron oxides, formed in a hydrothermal system. Based on our systematic survey we find no evidence of any association between the above minerals and the PCs. Therefore the proposed formation of sulfide related to volcanic or hydrothermal activities can be excluded by our study. Our results are in agreement with the previous investigations ofOsterloo et al. , which concluded that this formation processes was less likely due to the absence of lava flows, volcanic constructs or heat sources.
 The most common spectral NIR features observed on the deposits result from pyroxene-bearing or dust-bearing mantles likely of eolian origin. High spatial resolution images show that the PCs are patchy and discontinuous, thus suggesting spatial mixtures at the spatial resolution of OMEGA. These effects could mask the real composition of the PCs. However, the spectral ratio technique, which partly reduces the spectral effect of spatial mixing, indicates that the spectral signature of these PCs is characterized by a red slope that lacks absorption features. This spectral signature is consistent with an anhydrous halite composition [Jensen and Glotch, 2011].
4.2. Formation Processes of PCs
Osterloo et al.  proposed the following processes for the formation of PCs on Mars: (i) precipitation in ponds from surface runoff or groundwater discharge, (ii) hydrothermal brine via impact or volcanic activity and/or (iii) efflorescence via fumarolic or atmospheric interactions. Osterloo et al.  noted that efflorescence is unlikely to have been a dominant formation mechanism, and hydrothermal activity is less likely because of the lack of ancillary minerals. For one deposit (Figure 13a), efflorescence could be the formation mechanism that best explains the observed thin and mantling nature of the material. However, for most deposits, deposition via precipitation from a ponded evaporating brine derived from groundwater upwelling and/or surface runoff appears to be the most likely formation scenario and thus may be indicative of a past aqueous environment [e.g., Kendall and Harwood, 1996]. For such a scenario one would expect also the formation of a set of associated secondary minerals, such as carbonates and sulfates. These are commonly found in terrestrial playa evaporates [e.g., Eugster and Hardie, 1978; Baldridge et al., 2004; Sgavetti et al., 2009]. While carbonates have not been detected by OMEGA so far [Jouglet et al., 2007], they have been remotely observed as excavated materials only with CRISM higher spatial resolution data [Ehlmann et al., 2008; Michalski et al., 2012]. It has been suggested that the rarity of carbonates on Mars could be explained by the absence of an Earth-like carbonate cycle during early Mars [Ehlmann et al., 2008; McLennan and Grotzinger, 2008], and/or later dissolution under exposure to acid conditions [Chevrier et al., 2007; Fernández-Remolar et al., 2011]. On the other hand, hydrated sulfates have been identified on Mars in extensive and diverse deposits by OMEGA and CRISM [e.g., Gendrin et al., 2005; Langevin et al., 2005; Mangold et al., 2008; Murchie et al., 2009]. In addition, sulfur is present in anomalously high concentration in the Martian soils compared to Earth [e.g., Clark et al., 1976; Rieder et al., 1997; Brückner et al., 2008]. Despite hydrated sulfate deposits abundances, our systematic survey has revealed the lack of hydrated sulfates near PCs. This absence of correlation between sulfates and PCs at the global scale might provide insights into the formation processes. If the geochemical conditions at the time of chloride formation were similar to those of sulfates, one would expect dissolution of sulfur regardless of its source, and thus also the deposition of sulfate salts in association with PCs. Since sulfate is not observed, possible explanations are: (i) sulfur was not present in sufficient concentrations in the Martian surface/subsurface at the time of PC formation, or (ii) fractionation of a SO3-Cl brine has occurred [Moore and Bullock, 1999] resulting in halide/sulfate deposits at different locations [e.g., Dickinson and Rosen, 2003]. Fractionation is known to have occurred on Mars at smaller scale, for example at Meridiani Planum, where observations by the Mars Exploration Rover Opportunity have revealed differences in Cl versus SO3 concentrations [Rao et al., 2009]. Enrichment in Cl, which in part it may be due to the high solubility of Cl and aqueous transport [Karunatillake et al., 2009], has been observed across the surface of Mars by the Gamma Ray Spectrometer on board Mars Global Surveyor.The separation could be favored in a low temperature environment where only a chlorine-rich brine could have remained fluid, as in the Antartic Dry Valleys [Marion, 1997]. The brine could have then percolated through the sub-surface and subsequently froze, resulting in a residue of chloride salts. The establishment of a stratigraphy of PCs and sulfates based on morphologic relationship would allow further assessment of the formation processes. Such stratigraphy, however, could not be inferred because the two mineral deposits are not found in a common location. Overall, the global absence of sulfates associated with PCs and their different location apart from each other suggest that these two weathered products were likely formed during two distinct periods under distinct physicochemical conditions, with the formation of sulfates postdating the formation of chloride.
 The only hydrated phases found in close association with a few PC sites are phyllosilicates (juxtaposed Fe/Mg phyllosilicates association), which could reinforce an aqueous-related formation process, although this association is scarce compared to the widespread occurrence of both Fe/Mg clays and PCs in the highlands. The stratigraphy between the PCs and Fe/Mg clays is based on morphologic relationship and indicates that PCs have formed after Fe/Mg clay-rich deposits. This relationship and the different locations between PCs and sulfates might suggest three separate episodes of mineralization, first Fe/Mg phyllosilicates then PCs, followed by the subsequent formation of sulfate deposits.
 In the next section we discuss in more details juxtaposed Fe/Mg phyllosilicates occurring in NW TS, for which the geomorphologic context yield evidence of the formation process.
4.3. Northwestern Terra Sirenum
 The quantity and diversity of hydrated minerals in NW TS are witnesses to a complex aqueous history (Figure 12). The most complex mixture of secondary minerals are found in impact craters (Columbus, Cross and Dejnev impact craters), while scattered exposures of Al-phyllosilicates and one isolated mound with opaline silica have been reported in intercrater plains [Wray et al., 2011]. Fe/Mg smectites have been reported in several basins (e.g., Gorgonum Basin) belonging to the broad Eridania Basin [Noe Dobrea et al., 2008; Annex et al., 2011]. Fe/Mg phyllosilicate/PCs associations were previously reported at five sites using CRISM data, and geologic context investigation were performed for the largest PC site at 154°W–33°S (marked as “intercrater sedimentary basin” in Figure 12) [Murchie et al., 2009; Glotch et al., 2010; Davila et al., 2011; Wray et al., 2011]. Based on observations of the above site, previous studies presented the co-occurrence of the two minerals as the result of transport and deposition of detrital material as well as chemical precipitation [Murchie et al., 2009]. Such a deposition may have occurred in a possible paleobasin [Davila et al., 2011], or as a result of evaporation from groundwater discharge [Glotch et al., 2010]. We have provided geologic context investigation for each of the previously identified sites and three newly identified sites, providing the regional extent of the co-occurring minerals in NW TS. In three of eight independent sites, three show inverted channels consisting of PC (Figures 14 and 13f and Davila et al. ) and may indicate a past surface runoff. Since the size of these morphologies is relatively small, 10s of meters wide channels, the possible surface runoff events would have been relatively small, i.e., of a local scale. In addition, the non-correlation between the sites and the mouth areas of the dendritic valleys, where concentration and evaporation would be expected, does not support major surface runoff events as proposed byMurchie et al. . Several observations are instead in favor of groundwater upwelling: the relatively sparse valley networks, the presence of likely sapping channels and the possible presence of springs [Swayze et al., 2008; Wray et al., 2011; Capitan and Van De Wiel, 2011]. Furthermore, it is important to note that all the juxtaposed phyllosilicate/PC deposits are isolated and scattered throughout the intercrater plains. They are not in relationship with the other complex deposits identified in this region, such as hydrated sulfates or Al-rich phyllosilicates. This is in favor of a local groundwater source for the fluid. In addition, geologic settings do not suggest the presence of basins, differing from the findings in the previously investigated site, discussed byDavila et al. . Our observations are thus more consistent with groundwater discharge [Glotch et al., 2010].
 The juxtaposed Fe/Mg phyllosilicates show a rare knobby morphology, which has been found in only one other region on Mars in a deposit inside the Schiaparelli impact crater [Wiseman et al., 2009, Figure 6]. In the latter, the morphology is related to sulfate/clay-bearing deposits that may be explained by groundwater upwelling and evaporation [Wiseman et al., 2009]. In NW TS, a fluid could have been enriched in chlorine through chemical weathering or atmospheric transport [Bao et al., 2008], and subsequent evaporation or freezing/sublimation could have resulted in the deposition of chloride salts.
 The NIR spectrum of PCs exhibits shallow features indicative of a basaltic dust eolian mantle, partially covering the PCs phases, which are free of absorption bands and display a red slope. These properties match well with NIR laboratory spectra of anhydrous chloride mixtures obtained by Jensen and Glotch . A chloride composition for materials identified by Osterloo et al. is thus consistent with NIR observations. Our study also shows that areas adjacent to PCs lack ultramafic related minerals (olivine) as well as sulfates. This result confirms that sulfide as an alternative candidate for the material's bulk composition is unlikely, and furthermore excludes PCs formation by hydrothermalism. The most likely scenario for PCs formation, instead, could be precipitation from ponds either from surface runoff or groundwater discharge. Closely juxtaposed deposits of PCs and hydrated minerals (Fe/Mg clays only) are rare and essentially found in NW TS. The morphology of the two-mineral assemblages, however, suggests formation during two distinct wet periods. The assemblages have a scattered distribution in the region, are not related to valley networks, and do not occur in basins. This setting suggests a local groundwater source for the fluid. In addition, the morphology of the Fe/Mg clays associated with PCs is distinct from other Martian Fe/Mg clays-bearing materials, further suggesting particular formation. Outside from NW TS, near Knobel impact crater (132°E–6°S), the same assemblage is observed within a valley system that suggests involvement of water runoff. Most of the PC sites, however, occur without any associated hydrated minerals, thus differing from terrestrial evaporitic environments where a set of hydrated minerals and anhydrous salts is observed. Other environments where fractionation would lead to a chlorine dominated fluid may be better analogs for Martian chlorides. In Antartic Dry Valleys, for example, ponds almost uniquely dominated by chloride salts are observed. In such regions, studies have shown that chloride deposits can form both at the surface and in the subsurface [Marchant and Head, 2007] thus the geologic settings where chloride may be found are diversified and do not necessarily require a basin. Interestingly, in Antarctica Fe/Mg clays are observed as an associate mineral in chloride salts rich areas [Campbell and Claridge, 2009]. Terrestrial analog studies in cold and dry environment, as the one in Dry Valleys [Fastook et al., 2012], could provide insight into environmental conditions required and responsible for the formation of the PCs observed in NW TS.
 The authors would like to thank Mikki M. Osterloo and an anonymous reviewer for useful comments and suggestions.