The ∼150 km wide Holden crater lies in an area characterized by high density of valley networks implying conditions conducive to forming of water-related environments. We undertook geological mapping and a stratigraphic survey in order to probe the evolution of water-related landforms and their paleoenvironmental implications. Our investigations lead us to propose that the Holden area was subjected to a “wet” lacustrine phase of Hesperian age and an “icy” phase during the Amazonian. Deltaic, coastal, and lacustrine environments occurred during the “wet” phase, some displaying a cyclic depositional pattern presumably related to autogenic processes. Water was delivered to the basin by the Uzboi Vallis and by surface runoff channels from a series of drainage basins along the crater walls. Fan delta geometries and coastal onlap enabled estimation of major water levels. Two levels of major stand of the water have been recognized, possibly reflecting allogenic controls. Geologic units related to this “wet” lacustrine phase were subsequently eroded by glacial abrasion and plucking and were disconformably overlain by glacial deposits of Amazonian age, defining an “icy” phase. These features are consistent with a warm-based glacier entering the Holden crater through the wide Uzboi Vallis to form a proglacial lake in the central part of the crater. Changes in sedimentary units reflect changes of depositional environments probably connected with climatic variation.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Holden crater is a ∼150 km complex crater of Noachian age [Scott and Tanaka, 1986] breached to the south by Uzboi Vallis. The rim is still well preserved (Figure 2) and notably asymmetric topographically, the western rim being 2000 m higher than the eastern one (see altimetric profile in Figure 2). Remnants of the central peak are still visible in the centre of the crater.
 In order to identify all possible water-related deposits and landforms, we undertook geological mapping in a GIS (Geographic Information System) environment; this permitted multiple coregistered data sets to be used for maximizing the scientific return [Ori et al., 2004]. Data used in this study include MOC (Mars Orbiter Camera) narrow-angle and wide-angle images, THEMIS (Thermal Emission Imaging System) VIS and IR (both nighttime and daytime) images (Figure 1b). These data were processed and prepared for GIS ingestion using the USGS-ISIS (Integrated Software for Imagers and Spectrometers) system. We used MOLA (Mars Orbiter Laser Altimeter) topographic data as both PEDR (Precision Experiment Data Record) [Smith et al., 1999] profiles and grids through in-house gridding using GMT (Generic Mapping Tools) [Wessel and Smith, 1998]. Spatial resolution of gridded MOLA data was set at 1/128° per pixel.
 Single PEDR profiles have been used locally for detailed topographic analyses using the USGS-ISIS program “mocmola” together with selected MOC narrow-angle frames.
 Topography and image data were integrated into a GIS archive using a Mercator projection, and a Mars ellipsoid with the length of major and minor axes at 3376.19 and 3337.62 km, respectively. Area and length measurements were performed with equidistant and equal-area projections, respectively, e.g., for crater counting.
 In order to elucidate the geological history of this area, we mapped several sedimentary units discriminating their relative age on the base of stratigraphic relationships. Geomorphic features were used to constrain interpretation of the paleoenvironmental significance of the geological units. Availability of high-resolution images and altimetric data allowed us to undertake detailed geological mapping. We have used terrestrial methodology for stratigraphy and sedimentology to reconstruct the overall depositional history of this intriguing basin.
 Crater counting was conducted on selected units of the Holden crater. Due to the small area of most mapped units, we grouped ones interpreted as coeval. The minimum crater diameter considered for crater counting is 300 m. The results were plotted with the Mars isochrones from Hartmann  and Hartmann and Neukum .
2. Description of Geologic Units and Inferences
 In this section we describe the morphologic and textural characteristics allowing discrimination of geologic units and morphologic features. Geologic units, inferred from stratigraphic relationships, are described in stratigraphic order, older to younger, and their paleoenvironmental significance is specified. The resultant geological map is shown in Figure 3.
 This unit predates the putative sedimentary materials, constituting the older terrains on which water-related surface-modifying processes acted, eroding and depositing sediments. The Substratum crops out along the outer and inner rim, and on the central peak (Figure 3). It consists of intermediate-toned massive, texturally smooth and uniform materials. The Substratum is probably a reworked mixture of old impact breccia, lava flows, and pyroclastic material [Scott and Tanaka, 1986].
2.2. Proximal Layered Deposits (Sed 1 Unit)
 Proximal layered deposits (Sed 1 Unit) crop out in the southern and outermost part of the crater, as scattered patches in the terminal part of Uzboi Vallis, and just outside the terminal part of Uzboi Vallis (Figures 3 and 4). They consist of smooth bright and dark toned interlayered deposits seen only with MOC narrow-angle images (Figure 5a). They were recognized in the Holden crater and interpreted as sedimentary rocks by Malin and Edgett .
 Bright layers appear to be generally thicker than darker ones. The bright deposits in Figure 5a seem to change upward gradually into darker ones and then, suddenly, are bright again. This pattern is repeated suggesting cyclic organization of the depositional pattern. A MOC-MOLA M2302006 profile passing through this outcrop has average thickness of about 8.5 m per cycle (Figures 5b and 5c).
 Rock debris generated by Sed 1 bright layers accumulated at the foot of cliffs very close to the source area and, in places, may have filled interdune depressions (Figure 6), but did not develop eolian dune fields. This suggests the grain size was too coarse for eolian transport. The upper limit of grain size for forming dunes on Earth is about granule size (<3 mm diameter) [Greeley and Iversen, 1985; Greeley et al., 1992]. Because debris deposits do not occur at the toe of the slopes, Sed 1 Unit dark layers probably consisted of finer clasts, possibly removed by wind. We infer from the above that the bright layers of Sed 1 Unit consist of coarser (at least granule-size) material with the darker layers finer, mainly sand but possibly also silt. In this way, superposed fining and upward-thinning sequences could be inferred (see fining upward in Figure 5a).
 Bright-to-dark alternating layers could be connected with changes in sedimentary input and water discharge. In this scenario, bright layers would correspond to coarse materials delivered to the basin during intervals of higher discharge, dark layers to periods of lower discharge and possibly widespread decantative processes. Alternately, these layers could reflect internal dynamics of the sedimentary environments: the result of lateral shifting of differing subenvironments.
 The geometry of most Sed 1 Unit outcrops characterized by MOLA data accords with horizontal bedding. Low-dipping Sed 1 Unit strata terminate against the older crater wall Substratum unit in an onlap relationship (Figures 5 and 6) indicating paleotopography and gradual infilling of the basin. Sed 1 Unit thickness can be estimated using MOC-MOLA profiles M0202300, M0302733 and M2302006 coupled with stratigraphic mapping. The measured values range from a few meters to at least 140 m, but the thickness could be higher because the basal contact is often not readily identifiable.
2.3. Fan Deltas
 Along the inner part of the crater walls, especially in the south, a very well defined complex drainage system developed in response to extensive fluvial erosion (Figure 3). Some of these systems extend over areas ranging from around 20 to 325 km2. At the toe of these basins, alluvial fans and fan deltas can be discriminated (Figure 7), formed by reduced transport energy, related mainly to decreased topographic gradient. A spectacular fan delta has been imaged in the northeastern crater, next to the Holden crater (Figure 3) [Malin and Edgett, 2003; Moore et al., 2003]. Fan deltas are the most compelling indicators of the presence of former standing bodies of water on Mars.
 In order to discriminate possible fan deltas, fan-like features have been studied using MOLA data. In the area imaged in Figure 7, where two superposed fan-like structures occur, the topographic profile from crater wall to crater floor shows a surface dipping slightly toward the crater floor, becoming more inclined toward the crater center. An elevation drop occurs between the proximal low dipping and distal high dipping part of the fan (Figures 7a and 7b). Continuing toward the crater center, another elevation drop occurs (Figures 7a and 7b). These features may have originated when fluvial-dominated processes began to interact with putative wave-dominated processes related to the presence of a standing body of water. This implies that the fan-like features observed in the study area may have acted as fan deltas and might also indicate the Holden crater to have been a lake during their emplacement. The gently dipping parts may have been the subaerial portion of the fan delta, with the more inclined part equating with the submerged delta front. This pattern accords with a multiphase evolution of the putative lacustrine system with two hypothesized stands corresponding to topographic steps of the two fans. No definitive elements have been found to establish the relative age of these two hypothesized stands. Nevertheless, incised valleys cutting Sed 1 Unit strata in the topographically higher fan delta suggest this fan was formed before the topographically lowest one (Figure 7c). A drop in water table may have caused erosion (incised valleys) in the exposed part of the fan delta and formation of a prograding more distal fan delta (i.e., a forced regression scenario) [Posamentier et al., 1992] (Figure 7c).
 Sed 1 Unit crops out similarly in both fan deltas (Figure 7a), implying that at least part of it may have been deposited in a deltaic depositional environment; the strata in both fan deltas are subhorizontal without clinoforms, even where MOC NA coverage is relatively good. The delta front scarp has been estimated to dip 3°–5° on the basis of MOLA data. This does not accord with a Gilbert-type delta wherein the front scarp is commonly around 20° [Nemec, 1990]. We speculate that this depositional architecture may resemble terrestrial shallow water fan deltas [Galloway, 1976; Wescott and Ethridge, 1990]. On Earth such shelf-type fan deltas encroach onto low-gradient shelves with very shallow water depths near river mouths (e.g., Copper River). These systems do not develop clinoforms, but display a gently inclined delta front [Wescott and Ethridge, 1990]. It is difficult to compare these systems because, on Earth, wave- and especially tide-related processes affect the delta front, frequently leading to formation of barrier islands [Galloway, 1976], not observed on our Martian counterparts.
2.4. Distal Layered Deposits (Sed 2 Unit)
 The distal layered deposits, Sed 2 Unit, are dark-toned and layered, with apparently smooth surface texture; this unit is apparently uniform throughout the basin (Figure 8). Sed 2 Unit crops out in the central part of the crater, approximately in the area surrounded by the inner rim (Figure 3). Onlap on the inner rim and the central peak has been discriminated (Figure 9), suggesting infilling of the basin. It developed in the topographically lowest part of the crater, which was also a zone of accumulation of recent eolian sediments. The eolian cover is extensive inhibiting precise characterization of Sed 2 Unit (Figure 8). Because its base never crops out, its total thickness cannot be precisely estimated; geological sections suggest a minimum thickness of about 300 m.
 Several flat platform-like features bound the inner crater wall in its eastern sector (Figure 3); they occur at two elevations, roughly −1950 and −2060 m below the Mars datum. Their geometry is partly masked by talus deposits, but they can be traced for 7 to 20 km in length and 0.5 to 1.2 km in width. According to MOLA profiles, they dip at a low angle (<2°) toward the center of the crater. Their surface appears smooth at MOC NA and THEMIS VIS (∼18 m/pixel) and IR (∼100 m/pixel) resolution; the Substratum rocks are mantled by a thin layer of dark consolidated dust (Mantling Unit) or by talus. Such morphology could be related to either wave-cut or depositional lacustrine terraces or to collapse of the impact structure [Melosh, 1989] (see discussion by Ori et al. [2000a]). In the latter case, the platforms would be fault-bounded and rimward dipping, whereas on the MOLA profile they appear clearly to dip toward the crater floor. No evidence consistent with the possible presence of faults has been found.
 We conclude that terrace formation was not due to collapse of the impact structure rim, but is another line of evidence for presence of a standing body of water inside the basin. In essence, the two orders of terraces match the two water levels inferred from topography of the fan deltas (Figure 7).
 Terraces could be produced by erosional or depositional processes related to wind/wave action [Bradley, 1958: Bradley and Griggs, 1976]. No layered deposits have been discriminated on top of the uppermost terraces, thus implying that depositional processes were negligible at those elevations. Higher order terraces (−1950 m) should thus reflect mainly erosional processes, as with wave-cut terraces. On the contrary, the lower order of terraces (−2060 m) is characterized by the presence of Sed 1 Unit layered deposits on their top. In this case, either depositional processes played an important role in terrace formation or Sed 1 Unit strata were deposited during a postulated phase of higher water stand (coeval with the higher order of terraces) and then eroded, forming what we interpret as wave-cut terraces.
2.6. Smooth Deposits (Sed 3 Unit)
 Smooth deposits (Sed 3 Unit) are dark deposits occurring in the southern and eastern portions of the outermost part of the crater covering Sed 1 Unit (Figures 3, 4, 5, and 6). The boundary between units Sed 1 and Sed 3 is clearly unconformable (Figure 5) suggesting that erosional episodes occurred between the periods of deposition. Bedding is very faint. Because eolian dunes from reworked Sed 3 Unit sediments are often recognizable on top of that unit, we infer the grain size of the latter to be mostly sand and/or silt. Sed 3 Unit thickness up to 50 m has been measured by MOC-MOLA profiles. The way this unit mantles Sed 1 Unit layered deposits is consistent with eolian deposition.
2.7. Erosional Features
 The walls and floor of the Holden crater display complex erosional geometries where the large U-shaped Uzboi Vallis debouches into the crater. The area is characterized by erosional grooved features up to a few km in length and a few hundred meters width (Figures 10 and 11a), but are usually smaller by an order of magnitude; they are either linear, bend slightly (Figures 10 and 11a) or are sinuous (Figure 10). As the valley enters the crater, Sed 1 Unit bedrock has a smooth surface with streamlined forms. Ridges and troughs are elongated in the direction of the valley and become fan-like inside the crater; they are a few kilometers long and about a kilometer large (Figure 10). Striations are present at places (Figure 11c). From SW to NE, they display a smooth profile rising gently and then descending, suddenly breaking into a steep scarp (Figure 10). On top of these features or inside the grooves may be smaller-scale “stoss and lee” morphologies, with gentle slopes passing into steep, jagged edges can be recognized (Figures 10, 11a and 11b). Striations with smooth and bright texture are visible on the stoss side (Figures 10 and 11b). The transition between the stoss and the lee side is irregular and fragmented (Figures 10 and 11b); in some cases, boulders occur on the lee side (Figure 11b).
 Wind-related processes form grooved terrains consisting of juxtaposed ridges (yardangs) and troughs. In eolian systems these features are linear or slightly inclined, whereas in the study area some grooves are clearly sinuous. Ridges and troughs, moreover, are fan-like inside the crater (Figure 10). The smaller-scale “stoss and lee” morphologies accord with eolian dunes, but the fragmented transition between the stoss and the lee side (Figure 10a) supports erosional processes rather than depositional ones. As a consequence, it appears unlikely that wind-related processes were responsible for formation of these structures.
 Catastrophic flows can be effective erosion agents producing, among others, longitudinal grooving, scouring and streamlining of residual terrain [Baker et al., 1992]. They nevertheless do not usually produce very sinuous grooves like some of those observed in the study area (Figure 10). Stoss and lee morphologies inside the channels recall recessional headcuts (Figure 11a), a feature consistent with catastrophic flows. Stoss and lee morphologies found on top of the streamlined ridges (Figure 10a) could represent primary tractive depositional structures. These features in fact have been interpreted as sand waves [Parker and Grant, 2001; Grant and Parker, 2002]. The wavelength of these “stoss and lee” features is between roughly 150 and 300 m. As a consequence, MOC NA resolution appears accurate enough to detect possible cross-stratification. Nevertheless, no evidence of cross-stratification has been found in the several MOC NA images of this area; strata below the lee and stoss morphology appear uniformly horizontal or faintly inclined. Moreover, as already stressed, the transition between the stoss and the lee side is fragmented (Figures 10 and 11b), suggesting erosion rather than deposition. These elements do not entirely fit with a catastrophic flow origin. The presence of gently streamlined ridges and grooves may be generated by glacial erosion. Grooves developed by subglacial erosion have been widely documented on Earth (Figure 11a) and can develop a very sinuous shape. The prominent scarps bounding the frontal part of the ridges recall rock drumlins (Figure 10), as do extensive striations on the smooth surfaces of the stoss sides (Figures 10 and 11b) and the jagged and disrupted transition between the stoss and lee sides. These are consistent with possible origin in a subglacial setting. The stoss side in this scenario would display striations caused by abrasion as well as a very smooth texture due to polishing of the bedrock. Plucking would occur on the steeply dipping lee side (Figures 10a, 10b, 11a, and 11b).
 We suggest that the glacial-related erosional hypothesis is the most consistent with our data. Earthly counterparts of these structures develop beneath warm-based glaciers [Benn and Evans, 1998; Siegert, 2001]. Abrasion can take place also beneath cold-based glaciers, but the resulting erosional features are not as uniform and consistent as those formed on bedrock abraded by warm-based sliding ice [Atkins et al., 2002].
2.8. Blocky Deposits (Sed 4 Unit)
 Where the Uzboi Vallis debouches in the Holden crater, a unit with peculiar characteristics has been recognized. Unlike the adjacent Sed 1 and Sed 2 Units, this unit is limited to a few patches which, in aggregate form a sort of tongue inside the crater (Figures 4 and 11c). These patches have a clearer tone than Sed 2 and Sed 3 Units, but are not as bright as Sed 1 Unit; they appear very rough. They consist of extremely disorganized, poorly sorted boulders of various dimensions dispersed in a finer matrix. Some faint layering is slightly inclined toward the centre of the crater. This disorganized and poorly sorted texture might accord with an origin connected with mass wasting, but this hypothesis is not supported by any morphologic features, nor can a possible source area be specified. The same textural patterns are typical of till deposits (Figure 11c). We favor the latter genetic hypothesis. It is consistent with the morphology of Sed 4 Unit and would explain the absence of an obvious source area.
 The fact that Sed 4 Unit crops out along a tongue bounding possible glacial-related erosional features (Figures 4 and 11c) accords with emplacement as till, resembling frontal moraines, by a glacier (Figure 11c), specifically a warm-based glacier [Benn and Evans, 1998; Siegert, 2001] (see paragraph 2.7).
2.9. Alluvial Fans
 Fan-like features have been recognized at the base of the crater walls, mainly in the southern part of the Holden crater (Figure 3). The MOLA topographic profiles obtained along these structures have a constant topographic dip (Figure 12), suggesting that they are alluvial fans and that, even if they needed water to form, there was no standing body of water in this portion of the basin at the time of formation; in short, deposition occurred in a subaerial environment. In most cases, several alluvial fans are superimposed (Figure 12) indicating persistence of conditions favoring formation of alluvial fans.
 The alluvial fans consist of uniform dark-toned deposits (Figure 13) without discernible layering. Lobate features characterize the surface of these structures, suggesting that late stage deposition occurred, at least partly, by viscous mass flow (Figure 12). The use of MOC-MOLA profile M0202300 supplemented by geologic map data (Figure 4) enable verification that the alluvial fans lay on top of Sed 1 Unit and at least partly on Sed 3 Unit, demonstrating that the alluvial fans are younger than these units (Figure 13). The alluvial fans have a steep scarp close to the Uzboi Vallis (Figures 6 and 12); this reflects an erosional episode, presumably related to a flood episode from Uzboi Vallis.
 Processes associated with mass wasting appear to be dominant and superimposed on alluvial fan deposits in the NW sector of the inner crater wall. Creep and translational slides along bedding planes seem to be the most common processes.
2.10. Mantling and Dune Fields
 During all periods of Martian geological history in which water and ice were absent or limited on the planet surface, winds were the main surface-modifying process, even if features like gullies are active even now in the eastern portion of the crater walls (Figure 3). Stony deflation zones as well as wind streaks are widespread, but also buttes and yardangs confirm the importance of eolian erosion. Eolian landforms include fossil and recent dunes with wavelength up to 1.2 km. Recent or even contemporaneous eolian dust and dunes are present all over the crater, but are extensive in the central topographically deepest part of the crater inside the inner rim (Figure 3). Extensive sinuous and barchan dunes with wavelength sometimes exceeding a km (Figure 8), as well as wind streak directions, indicate an approximately N-S wind direction. Unconsolidated dust is dominant in the deepest parts of the crater, protected from wind action. Small craters (<1 km) tend to be filled with sediments, thus preventing crater counting for the Sed 3 Unit lying below.
 Materials we interpret to be indurated dust crop out discontinuously throughout the studied area. This material consists of very dark-toned smooth rocks that appear to mantle younger units. Along the crater rim and outside of the crater, mantling occurs in patches corresponding to flat areas, for example on wave cut terraces (Figure 3). Inside the crater, this unit crops out in association with Sed 2 Unit; they are often difficult to map separately. No dunes have been observed in these mantling deposits; their textural characteristics are consistent with consolidated eolian dust. These indurate dusts could correspond to the ice-rich mantling deposits discussed by Mustard et al.  and Head et al. .
2.11. Large-Scale Polygons and Tectonics
 Two kinds of fractures occur in the study area, one apparently involving only superficial sedimentary deposits located mainly in the southern part of the crater, the other deforming also the Substratum units (Figure 3). The first system consists of large-scale polygons, with fractures up to 80 km long (Figure 14). The mechanism of formation of these features has been debated for many years [Carr and Schaber, 1977; Morris and Underwood, 1978; Pechmann, 1980; Lucchitta, 1983; McGill, 1986; McGill and Hills, 1992; Lane and Christensen, 2000]. Desiccation could possibly represent the last phase of the “wet” history of the Holden crater, with progressive transition to arid conditions. On the other hand, these fractures may correspond to faults resulting from down-dip increasing in the southern part of the crater where we suggest the sedimentation rate was greatest for the crater, due to sedimentary input from the Uzboi Vallis.
 Tectonic structures discriminated in the studied area are partly related to the cratering process; they are distributed radially and concentrically throughout the crater. Two systems are observed, oriented NNE-SSW and WNW-ESE, respectively (Figure 3). These systems are subparallel and possibly related to structures controlling the first stage of Valles Marineris formation [Tanaka et al., 1991].
2.12. Drainage Basins
 The overall drainage pattern is centripetal with rivers and streams flowing down the crater walls toward the crater floor. Drainage basins show two different kinds of pattern: (1) dendritic drainage systems, characterizing most of the drainage basins (Figure 3), and (2) treelike drainage patterns, usually characterizing areas of uniform, erodible rocks, possibly Substratum rocks subjected to impact shock. In a few cases, parallel drainage patterns have reflected mainly impact-related radial fractures.
2.13. Stratigraphic Relationships
 The impact event that generated the Holden crater and formed the Substratum has been dated to the Noachian by Scott and Tanaka . A geological section through the Holden crater (Figure 15) summarizes the geographic distribution and stratigraphic relationships of the mapped units. These reflect a complex depositional architecture: differing depositional and erosional processes producing different stratigraphic sequences located in various areas of the basin.
 As shown in the previous sections, in the southern, outermost part of the crater, layered Sed 1 Unit (mainly of deltaic origin) was deposited unconformably on top of the Substratum. In the central part of the crater, Sed 2 Unit unconformably covers the Substratum, and is presumably coeval with Sed 1 Unit. We tried to estimate the age of these units using crater counting. We considered the areas occupied by Sed 1 Unit, fan deltas, and depositional terraces, interpreted as coeval. Due to the recent extensive eolian deposition that covers Sed 2 Unit, this unit has not been counted. The small area of the units greatly hampers the results and overly confidence should not be placed on the derived ages. The small crater population of Sed 1 Unit is perhaps underestimated because small craters could be buried beneath the sedimentary infill that covers a significant part of the area. An attempt to determine absolute ages results in a Hesperian age for these units (Figure 16). Our results are consistent with ages provided by Scott and Tanaka , who dated the fluvio-lacustrine systems of the entire Margaritifer Sinus as Upper Hesperian.
 Sed 1 Unit is disconformably covered by Sed 3 Unit and Sed 4 Unit. Close to the crater walls, Sed 1 Unit is covered by several alluvial fans clearly postdating its deposition. These fans were later subjected to erosion by the latest flood event recorded in the Holden crater.
 We tried to date the alluvial fans, assuming they were coeval or formed during a geologically small interval. In this case, also, the small area of the units greatly hampers the results. Nevertheless, they do not have a population of relatively medium-sized craters, thus suggesting a younger age than Sed 1 Unit. This is consistent with the observed stratigraphic relationships. To determine the absolute ages gives a problematic result that the alluvial fans span the Amazonian, as shown in the Figure 16. Eolian dunes and dust, the youngest unit mapped, stay on top of all these units.
 Geologic analysis of the Holden area revealed a complex sedimentary evolution possibly related to climate changes after the Noachian impact that generated the Holden crater. The various sedimentary deposits and morphologies have been analyzed in order to infer the processes that contributed to their formation. Stratigraphic disconformities and different phases of fan-formation allow discrimination of major erosional, or at least “non depositional”, episodes. On the basis of our interpretation of the processes forming the different units, we speculate on the nature of the depositional systems present through time in the study area, and interpret their environmental significance on the basis of stratigraphy and sedimentology. A space-time diagram representing the stratigraphic relationships of the sedimentary bodies and their evolution through time is shown in Figure 17. The space-time diagram (x axis is space, y axis is time) is commonly used in stratigraphic analysis to better convey the temporal and spatial relationships among geological units. With this tool, the different geological events, both erosional and depositional, can be outlined.
 The oldest sedimentary units are represented by layered sequences (Sed 1 and Sed 2 Units). Much evidence suggests that Sed 1 Unit was deposited in deltaic and coastal settings. Since the maximum elevation recorded by coastal onlap corresponds to the maximum elevation of water level, we analyzed the distribution and elevation of the coastal onlaps in the Holden basin. Two elevations of the coastal onlap, marked by Sed 3 Unit deposits, have been discriminate (Figures 5 and 18). According to these coastal onlaps, water level elevation would have reached −1962 m and −2066 m below the Mars datum (Figure 18). It is pointed out that they should not be considered absolute values because of errors related to altimeter resolution, data elaboration and, above all, due to the erosion recorded between deposition of Sed 1 Unit and Sed 3 Unit. Nevertheless, these water levels can help to estimate the magnitude of the lacustrine system. These values have been compared with the ones obtained from altimetric profiles of the fan deltas. The topographic drop between proximal low dipping and distal high dipping parts of the fans (Figure 7) indicates the water elevation during formation of the fans. Erosion prevents obtaining precise values, but the inferred water surface elevations obtained with the two different observations are similar (Figures 7 and 18). This implies that these two elevations probably correspond to two major stands of the water table in the Holden crater. These elevations also correspond roughly to the terrace elevations recognized in the eastern part of the basin.
 Putative cyclic depositional patterns recognized in Sed 1 Unit (Figure 5) could be related to autogenic processes in the depositional environment rather than to allogenic-driven processes. The autogenic processes are more likely controlled by avulsions of the distributary channels.
 Contemporaneous with deposition of Sed 1 Unit in the outermost Holden crater (Figure 17), conditions were favorable for lacustrine deposition in the deeper central parts of the basin. The characteristics of Sed 2 Unit as well as its distribution in the middle of the crater are consistent with deposition in a lacustrine environment. The transition between the Sed 2 and Sed 1 units occurs roughly around the inner rim (Figures 3 and 15). During phases of low discharge and surface runoff due to precipitation, playa deposition could have occurred in the central part of the basin.
 Along the crater wall, immediately west of the Uzboi Vallis mouth, is a series of shallow channels developed in an ∼6 km long valley (Figure 19). The overall channel width is large compared to channel depth and they cut the outer rim (Figure 19). This morphology, resembling outflow channels, requires abundance of water and high velocity of flow; it is not consistent with a simple drainage system fed by rain. Because these channels originate at the top of the outer rim, their formation implies a source of water in an area outside the Holden crater and the Uzboi Vallis. According to Parker , Grant and Parker [2001, 2002], and Parker et al. , Uzboi Vallis flowed in the Holden area prior to formation of the Holden crater. The crater rim should have been an obstacle that was impossible to breach, even if there had been sufficient fracturing to allow water infiltration and favor erosion. Though these processes allowed water to flow freely inside the Holden crater, we suggest formation of an ephemeral lake just outside the crater border. This scenario is consistent with the presence of layered Sed 1 Unit deposits in these areas (Figures 3 and 19).
 On the base of our interpretation of the processes of formation and the sedimentary environments of the recognized units and morphologies, we speculate that the Sed 1 and Sed 2 Units were deposited during a lacustrine phase in the Hesperian during at least two major phases (Figures 7, 17, and 18). The presence of extended drainage systems developing large fan deltas (and not directly fed by Uzboi Vallis) underlines the significance of surface runoff from precipitation [Craddock and Howard, 2002; Hynek and Phyllips, 2003].
 The disconformity between the Sed 1 and Sed 3 units indicates that erosional processes occurred in the interval between deposition of those units (Figures 5 and 17); the erosion appears to have been mainly eolian, as suggested by residual pavements (produced by wind deflation) and possible yardangs and buttes.
 The erosional and depositional features described in sections 2.7 and 2.8 (Figures 10 and 11) in a confined area close to the Uzboi Vallis mouth display characteristics that appear consistent with glacier-related processes. They are consistent with a glacier tongue having entered the crater through the Uzboi Vallis, a frontal moraine marking the position in which ice was most extensive. Abrasion-related features such as striations, polished surfaces and grooves, as well as the presence of till deposits, suggest formation of these features beneath and in front of a warm-based glacier [Benn and Evans, 1998; Siegert, 2001]. The distribution of the separate patches of Sed 4 Unit is also consistent with the presence of water melting from a retreating glacier and subsequent opening of a passage through the moraine. Under these conditions, a proglacial lake could have existed in the central part of the basin. If so, Sed 2 Unit deposition might have continued during this glacial stage as well (Figure 17). Sed 3 Unit (Figure 5) lies on top of an erosional surface produced by glacial abrasion and was, locally, subjected to similar erosional processes. This suggests that Sed 3 Unit may have been coeval with the glacial erosion and deposition, possibly connected with wind transportation (i.e., loess).
 Alluvial fans occur widely at the toe of drainage basins eroding the crater walls. They are very well preserved in the southern part of the crater, but in the northern part are partly covered by loose material. Stratigraphic relations confirm the alluvial fans are younger than Sed 1 Unit (Figure 13) and, as a consequence, younger than the lacustrine phase of the Holden crater's geological history. Alluvial fans represent, together with their drainage basins, a phase in which water was still abundant, but not enough to fill the crater, at least up to the areas proximal to the outer rim. The only water-rich phase recognized in the Holden crater, after the lacustrine episode, is the glacial one. Even if no direct stratigraphic relationships have been found, we speculate that alluvial fans were coeval with the glacial phase. The erosion observed at places at the toe of some alluvial fans (Figure 12) might have been related to the flooding following melting of the ice. If it is assumed that the alluvial fans were coeval with the glacial phase, we can try to estimate the age of the latter by dating the alluvial fan surfaces. Using crater counting, we estimated a putative Amazonian age for the alluvial fan and, as a consequence, for the glacial phase (Figure 16). The Holden crater glacial episode might be quasi-contemporaneous with the Amazonian cold-based mountain glacier deposits recognized on the western flank of Arsia Mons [Head and Marchant, 2003].
 The complex geological history of the Holden crater has been interpreted by means of detailed geologic and stratigraphic analyses. We identified two putative stages in evolution of the Holden basin: a “wet” phase with lacustrine deposition and an “icy” phase with glaciers and, possibly, a proglacial lake (Figure 20). These two stages are marked by a large stratigraphic hiatus where mainly erosion occurred. On the basis of stratigraphic relationships and the crater counting dating on units with high areal extent, we constrained the “wet” period to the Hesperian and the “icy” phase in the Amazonian.
 The “wet” phase consists of a lacustrine depositional system with related subenvironments and structures (Sed 1 Unit, Sed 2 Unit, Fan-deltas, Terraces). These deposits represent deltaic and possibly coastal depositional environments laterally passing to fully lacustrine conditions. Deltaic successions show a cyclic depositional pattern reflecting internal dynamics and discontinuities in the water discharge. Water was delivered to the basin via the large Uzboi Vallis, but also by some surface runoff channels formed after episodes of rain; from these originated a series of drainage basins along the crater walls. At least two major stands of the water table during this phase have been documented by means of fan deltas and coastal onlap, thus proving that water was supplied with relative continuity to the basin for a long time. Incised valleys cutting the topographically upper fan delta allow addressing the relative age of these stands, with the upper one, roughly located at −1962 m, older than the lower one, around −2066 m below the Mars datum (Figure 20). Control on this fluctuation was probably allogenetic. The “wet” phase (Holden Lake) is tentatively dated to have occurred during the Upper Hesperian. The end of this “wet” phase for the Holden crater is marked by erosion. During a not estimable time interval, arid conditions prevailed with widespread eolian erosional and depositional activity, and without evidence for significant water supply.
 Eolian deposits and older deltaic-lacustrine sedimentary rocks were later eroded by abrasion and plucking by glaciers during the “icy” phase (Figure 20). A glacial tongue entered the basin through the U-shaped Uzboi Vallis, eroding by abrasion and plucking processes and depositing a frontal moraine (Sed 4 Unit); this marks the maximum extension of ice. Erosional as well as sedimentary patterns are consistent with deposition beneath and in front of a warm-based glacier. A subglacial lake is thought to have formed in the center and deepest part of the crater during this phase. Drainage systems along the crater walls, and associated possible debris-flow-dominated alluvial fans at the toe of the slope were probably coeval with this event.
 The complex environmental evolution discriminated in the Holden crater area helps chronicle the climatic changes that occurred during Martian history.
 We are grateful to our colleagues Gian Gabriele Ori and Goro Komatsu for fruitful discussions, to John Talent, Maquarie University, for polishing our English, and to anonymous reviewers for suggestions that greatly improved the manuscript. Our research was funded by the Italian Space Agency and the Italian Ministry of Universities and Research.