Dark polygons associated with fans and spots appear during the spring on the southern seasonal cap. The basal sublimation of the translucent cap and the venting of the CO2 gas are responsible for their formation, as previously proposed for the spots and fans. Dark polygons appear when dark material emerges from elongated vents, whereas spots and fans form from point sources. A class of erosive features (etched polygons) is associated with depressions a few meters to tens of meters in diameters connected to a network of radiating troughs (“spiders”). Spiders are shaped by the scouring action of the confined gas converging toward point sources, whereas the etched polygons result from the forced migration of the CO2 gas over longer distances. The minimum age of the spiders is 104 years. They result from one of the most efficient erosive processes on Mars, displacing 2 orders of magnitude more dust per year than a typical dust storm or than all the dust devils during the same time period. In the north, parts of the seasonal cap are translucent between Ls = 355° and Ls = 60° and are associated with spots, fans, dark polygons, and possibly spiders, suggesting that the basal sublimation and venting of the cap triggers a subice gas and dust flow that is modifying the morphology of the surface layer. However, perennial features are extremely uncommon on the north regolith, indicating that the conditions for their formation or conservation are not met. The reduced basal energy budget of the north cap compared to the south and the shorter seasonal life time of the north translucent ice may explain the relative scarcity of features in the north. The polar layered deposits contain the stratigraphic record of climatic changes and catastrophic events. Both polar deposits may have been locally disrupted by the seasonal subice gas flow and the stratigraphic record may have been partially lost.
 Every winter, a layer of CO2 frost condenses from the atmosphere and covers the polar regions of Mars, as the surface temperature drops to the vapor equilibrium temperature (∼148 K) and forms the seasonal caps. Leighton and Murray  predicted that the seasonal and perennial caps were composed of CO2 ice and act as a buffer for the atmosphere. Orbital data from the Mariner 4, Mariner 9 and Viking spacecrafts, along with terrestrial observations, confirmed the presence of CO2 frost in the north and south polar regions [Herr and Pimentel, 1969; Kieffer, 1970; Neugebauer et al., 1971; Larson and Fink, 1972]. The Thermal Emission Spectrometer (TES) onboard the Mars Global Surveyor (MGS) spacecraft has characterized the density, the recession date as defined by the “Cap Recession Observations indicating that the CO2 has Ultimately Sublimated” (CROCUS date), and the thermal and visual properties of the polar caps [Kieffer et al., 2000; Kieffer and Titus, 2001]. The Observatoire pour la Minéralogie, l'Eau, les Glaces, et l'Activité (OMEGA) onboard Mars Express has acquired repeated spectral data of the polar ices and dust in the visible and near infrared and mapped their distribution [Bibring et al., 2004; Langevin et al., 2006]. It showed that in places, the CO2 ice displays the signature of coarse CO2 crystals on the seasonal cap and the spectral absorption features of dust [Langevin et al., 2006]. The MGS Mars Orbiter Laser Altimeter has provided an estimate of the thickness of the seasonal caps: ∼1–2 m of CO2 frost [Smith et al., 2001]. The Thermal Emission Imaging System (THEMIS) onboard the Mars Odyssey spacecraft has acquired excellent repeated visual and thermal observations critical for understanding the polar processes, most of which occur during the spring [Kieffer et al., 2006; Kieffer, 2007]. The MGS Mars Orbiter Camera (MOC) and Mars Reconnaissance Orbiter High Resolution Imaging Science Experiment (HiRISE) have returned high-resolution images of the polar regions unveiling numerous exotic features (e.g., spots, fans, “fried eggs” and spiders [Kieffer, 2003; Piqueux et al., 2003; Kieffer et al., 2006; Hansen et al., 2007]) with no equivalent on Earth.
 The gas venting model of the south seasonal cap developed by Kieffer , Kieffer et al. , and Kieffer  predicts the formation of these features and requires the erosion of the surface of the polar layered deposits (spiders) and the redeposition through atmospheric settling of dust to sand in the vicinity of the spiders (fans and spots). The process leading to the formation of the spiders (i.e., the confined flow of pressurized gas) may have disrupted the south polar layered deposits' (SPLD) stratigraphic record with important consequences for future studies and missions.
 It is important to evaluate the effects of the erosive subice gas flow on the degradation of the climatic record in order to better constrain the stratigraphic resolution that can be expected from these units for future studies and missions. In this paper, we present new observations of seasonal features in the south, update the model of the venting of the cap proposed by Kieffer [2003, 2007], Piqueux et al. , and Kieffer et al. , test this model on the north seasonal cap and show that the stratigraphic record of the NPLD may have been partially lost or disrupted.
 The spots are often associated with fans (Figure 2). Fans are usually darker than the surrounding material [Malin and Edgett, 2001] although bright fans are also observed [Kieffer et al., 2006]. Fans appear slightly later in the spring than the spots. Each fan originates from a single source but their tails can be oriented in multiple directions. They are usually tens to hundreds of meters long and 10–30° in angular size [Piqueux et al., 2003; Kieffer et al., 2006]. In a given region, all of the fans point in similar directions. The fans emanate from the central depression of complex, radial troughs (spiders). Fans have been proposed to be formed by wind-driven near-surface transport of sand-sized particles carried to the surface of the ice [Piqueux et al., 2003; Kieffer et al., 2006: Kieffer, 2007]. A large proportion of the fans occur where the CO2 seasonal cap appears dark from orbit, in the so-called “cryptic region” [Piqueux et al., 2003].
 During the southern spring, the seasonal cap's albedo remains locally lower than the rest of the cap (<0.37 versus >0.40), whereas its temperature is stable near the CO2 ice temperature, i.e., ∼150 K [Kieffer, 2000; Kieffer et al., 2000; Piqueux et al., 2003]. These terrains of low-albedo seasonal ice were termed “cryptic” because of the contradiction between the different observations: they grow dark but remained cold, giving them the visual properties of defrosted soil and the thermal characteristics of CO2 ice. The process leading to the formation of the dark looking ice has been first proposed by Kieffer  and expanded upon by Kieffer , Kieffer et al. , and Kieffer . TES spectra have shown that the CO2 ice that appears dark from orbit is composed of larger crystals than elsewhere on the cap, and that it likely takes the form of a slab [Kieffer et al., 2000, 2006]. The slab appears dark due to the underlying low-albedo material visible through it and the dark dust settled at the surface. Because this behavior has now been modeled [Kieffer, 2007], the dark and cold seasonal cap should no longer be called “cryptic”; we propose the term “translucent seasonal cap” (TSC). Piqueux et al.  have mapped the distribution of the south TSC (STSC) winter/spring/summer cycle. It extends from the pole to −65°N at all longitudes. It shrinks during the spring and stabilizes between Ls = 200° and Ls = 240° where it is centered near −80°N and 180°E. The dark-looking CO2 ice disappears as the seasonal cap retreats.
 Here we present a series of observations of polygonal features on the south seasonal cap. Networks of polygons marked by dark outlines are observed in the regions of active sublimation, early in the southern spring. Seasonal polygons have four or five ∼50 to ∼80 m long dark edges (Figures 3–5). Most of these features are incomplete, missing one or several edges. The polygons' edges vary from relatively thin and straight, to large and diffuse. Where they are diffuse, the edges are not of uniform tone but usually form a gradient from very dark (near the edge, Figures 3 and 4) to gray a few tens of meters away in the general direction similar to nearby fans. This suggests that both types of features on the seasonal cap are formed by a similar process. Using Fenton and Richardson  wind field maps generated by a global atmospheric circulation model, we have verified that the orientation of the fans is consistent with the regional prevailing wind direction. Some variability occurs, which we attribute to the low resolution of the map by Fenton and Richardson  near the poles and local topographic effects on the wind field. Fans or spots are always associated with these polygons. Between the fans, the seasonal cap is usually darker than the pristine ice visible elsewhere outside the low albedo cap, whereas it stays bright between the spots. Hereafter, we call similar features visible on the cap “polygons on the seasonal cap.” Rarely, bright material (presumably CO2 frost) originates from one of the edges of these polygons (Figure 5). These bright features are ∼50 m in length, which is smaller than typical well developed bright fans (hundreds of meters).
 Three examples of polygons on the seasonal cap are presented in Figures 3–5. Figure 3 shows a portion of the seasonal cap at Ls = 199° when the CO2 ice sublimes. The subliming seasonal cap (right) is covered with 600 m by 200 m wide fans, and displays a gray surface that contrasts with the bright pristine ice (left). The eruption of sand-sized particles that fall near the vent forms the fans, and erupted dust-sized particles that are dispersed both laterally and further downwind can explain the gray tone of the cap. The transition between the fan-rich area and the clean cap is clearly delimited by the presence of a network of seasonal polygons whose dark boundaries are sometimes near or below the resolution of the image in width (5.8 m per pixel) and sometimes wide and diffuse (up to 30 m). Occasionally, the fans are not as well developed but the polygonal network is. This is illustrated in Figure 4, where the polygons are associated with short fans and spots within a region covered with a thin, dusty, darker layer (right). The brighter region (left) appears to be a region that has not experienced the eruption of sand and dust onto the surface. Figure 5 shows a region of active sublimation consistent with CO2 gas emerging from linear vents, forming a dark subpolygonal network. Along with the dark dust settled on the downwind side of the cracks, bright CO2 ice associated with dark fans originating from the vents is also visible.
2.2. Perennial Features
 Once the seasonal cap has entirely disappeared, a variety of surface textures can be observed in the locations where the seasonal features formed during the phase of sublimation of the CO2 ice. These features are (1) spiders [Piqueux et al., 2003] and (2) networks of polygonal troughs. Spiders are features visible on the defrosted SPLD. The spiders are depressions a few meters deep and typically ∼50 m across connected to a network of radiating troughs cut in the south layered material (Figure 6). In some cases, the spiders form a network of depressions rather than a well defined central depression with radial merging branches. Piqueux et al.  have mapped the distribution of these depressions which occur primarily, but not uniquely, where the ice is translucent at ∼Ls = 210°. Spiders are ∼150 to ∼1000 m in diameter including the numerous merging and branching troughs radiating from the center depressions. Occasionally, away from the center depression of the spiders, these troughs expand and cover large areas (tens of square kilometers near −87°N, 90°E), forming networks on the terrains that are translucent during the winter and spring. Similar to the branches of the spiders, these somewhat polygonal troughs are not linear and are not always well defined, with diffuse edges, giving them a characteristic appearance clearly different from other classified polygons [van Gasselt et al., 2003; Mangold et al., 2004; Mangold, 2005].
Figure 7 is a view of this type of polygonal terrain, which we termed “etched polygons.” They differ from any other type of ground pattern by the extremely irregular shape of their wide edges, which are rugged troughs, connected to smaller converging elongated depressions. In some cases, the resolution is high enough to estimate a minimum depth of ∼4 m for these depressions using shadow measurements. In addition to the topographic troughs delimiting each polygon, a diffuse, darker material surrounds the edges of the etched polygons. When polygons and spiders are adjacent, they often share one or several troughs. This is illustrated in Figure 8 which was acquired early enough in the season such that the seasonal polygons have started to form, but late enough such that the topography of the spider is clearly uncovered. The troughs forming the “legs” of the spiders become shallower as the distance from the central depression increases. Away from the vent of the spiders, little or no topography is visible and the spiders legs merge with the polygonal network. The close relationship between the polygon networks and the legs of the spiders (Figure 8) and between the fans and the polygons (Figures 3–5) provides the foundation for a unified model of formation.
Kieffer [2000, 2003] proposed a preliminary model (later refined by Piqueux et al. , Kieffer et al. , and Kieffer ) for the formation of the spiders associated with the fans and the STSC. In this model, the seasonal cap is dusty and opaque to the solar radiation before ∼Ls = 190°. During the spring (from Ls = 180° to ∼Ls = 240°), the dust trapped in the CO2 frost is removed [Kieffer et al., 2006; Kieffer, 2007]. At some time, either during the ice formation or during this dust and sand cleaning process, the cap forms an impermeable slab. At this stage, about three quarters of the solar rays reach the top of the dark regolith [Kieffer, 2007], which is in contact with the base of the seasonal cap. As the substrate warms up, the overlying CO2 ice is not stable and sublimes, forming a layer of pressurized gas trapped between the substrate and the slab. The gas pressure may be ∼10 times that of the atmosphere (4000 Pa) [Kieffer, 2007]. Wherever the cap is mechanically too weak to bear the increasing gas pressure, cracks form. The pressure gradient between the atmosphere and the confined gas will trigger a rapid convergence of the gas toward the lower pressures, i.e., under any crack in the slab connected to the atmosphere. The mechanical action of the flowing gas scours the surface and entrains the products of this erosion (dust and sand-sized particles). The gas loaded with dust converges toward the cracks of the seasonal cap and erupts in jets that form the dark spots and fans on the surface of the ice. The erosion associated with the gas flow forms the spiders, which are systems of channels that converge toward the vents.
3.2. Dark Polygons Formation
 In the classic jet model [Kieffer et al., 2006; Kieffer, 2007], point vents are formed, which produce spots and fans on the seasonal cap and radial channels (spiders) on the underlying substrate. Because of the cap being brighter (i.e., cleaner) between spots than between fans, we propose that the dust is not ejected as high in those places where spots form compared to locations where fans are observed and that the finer material is not dispersed as efficiently. The wind velocity may also be higher where fans are observed. However, our observations suggest that the slab may not always be punctured, forcing the gas to migrate laterally toward lower pressures, confined between the cap and the substrate and eroding the substrate along its path. The repetition of this process over many years may lead to the formation of a network of troughs used by the gas as a preferential path during its migration. We suggest that this network evolves into a field of etched polygons where the troughs are densely intertwined and dispersed with occasional spiders. In addition, the moving gas erodes the overlying CO2 cap during its migration resulting in lines of weakness that produce vertical planar vents evolving into elongated depressions at the surface. The sand and dust that erupt from these linear vents form the dark polygons on the seasonal cap (Figures 1–5 and 8).
 Laboratory experiments indicate that CO2 ice is mechanically weaker than H2O ice or bedrock [Clark and Mullin, 1976; Durham et al., 1999]. The surface layer of the PLD is most likely composed of a thin layer i.e., a few millimeters [Titus et al., 2003] to a few centimeters [Mellon and Phillips, 2001] of unconsolidated material on a harder H2O ice cemented substrate [Mitrofanov et al., 2002; Boynton et al., 2002; Feldman et al., 2004; Bandfield, 2007] although undetectable significant variability is possible. The apparent depth of the spiders seems to exceed that of the unconsolidated surface layer, and the walls of the spiders may be steeper than the angle of repose of unconsolidated material (see C. J. Hansen et al.'s (Seasonal erosion of Mars' southern high latitude terrain, submitted to Geophysical Research Letters, 2008) Figure 2c, which is an anaglyph of a spider where the central depression is ∼1.8 m deep and where the angle of the walls near the center depression approaches or exceeds 25°, possibly larger than the angle of repose of uncemented regolith), indicating that the scouring power of the flowing gas could be sufficient to erode the underlying H2O hardened layer and that spiders are maintained at least partially in consolidated material. Also, it is possible that the sublimation of the ice near the eroded surface is faster than the gas-powered erosion rate, and that erosion only takes place in soft, unconsolidated material. Additional work should focus on HiRISE anaglyphs of spiders to determine if the angle of the walls consistently exceeds that of the angle of repose expected for uncemented regolith. The strength of the substrate (cemented or not by water ice) is not known, but if water ice is present, it may approach or exceed that of CO2 ice. In any case, the abrasion is not limited to the regolith during the migration of the gas; the moving fluid loaded with sand and dust likely scours the overlying seasonal cap as well, forming a network of troughs and channels similar to the one observed on the regolith.
 An estimation of the stress within the slab induced by the flexure of the cap forced by the gas pressure is provided by Portyankina  with a stress-strain elastic model. She concluded that the stress in the cap (over 105 Pa) exceeds that of the putative yield stress of CO2 ice before Ls = 175°, which is consistent with the observations by Piqueux et al. . The mode of rupture is not discussed, but it is plausible that networks of linear fractures can result in the seasonal cap from the rupture as well as point vents.
 The equilibrium temperature of CO2 ice at the surface is ∼148 K but increases by at least 10 K at the base of the seasonal cap due to the increase of the gas pressure [Kieffer, 2007]. A mechanical analysis of the stress in the subliming seasonal cap is not within the scope of this paper, but differential thermal expansion of the ice should result from this temperature gradient (coefficient of thermal expansion for solid CO2 is α = ∼2 × 10−4 K−1 from Gibbons and Klein ), increasing the tensile stress within the cap and eventually causing it to fracture.
 In addition, the flow of warm pressurized gas (∼162 K) may also thermally erode the seasonal cap where its equilibrium temperature is cooler, i.e., where the gas pressure is lower, closer to the surface, during the phase of vertical eruption of the CO2. This is consistent with the observation of white fans composed of CO2 frost formed by the adiabatic decompression of pressurized warm gas at the surface [Titus et al., 2007]. The formation of the cracks leading to the seasonal polygons on the cap may therefore be due to a combination of (1) the stress induced by the flexure of the cap, (2) the physical scouring by moving gas loaded with dust, (3) the thermal erosion by warmer, compressed gas, and (4) the fracture initiation and propagation by thermal stress.
 This updated venting model is consistent with the chronological observations of the bright and pristine seasonal cap (albedo > 0.40), the appearance of the darker STSC (albedo < 0.37), the formation of the dark polygons, spots and fans, their distribution on the STSC, the observation of the spiders and etched polygons on the SPLD, and the occurrence of dark material (spots, fans) near CO2 ice temperature [Kieffer et al., 2006]. It must be noted that the spatial resolution of TES (3 km) may not be sufficient to detect smaller isolated patches of translucent CO2 ice elsewhere near the poles. This may explain the occasional occurrence of spots, fans or spiders outside the mapped limits of the southern dark and cold CO2 ice.
 We conclude that the polygons on the cap are a class of seasonal features associated with the basal sublimation and the venting of the seasonal cap similar to the spots and the fans. Figure 9 illustrates the schematics of the model developed by Kieffer  updated in this paper, and emphasizes the spatial and temporal relationship between the fans, the polygons on the cap, the spiders and the etched polygons. It is consistent with all the observations made by Kieffer [2000, 2003], Piqueux et al. , Kieffer et al. , and Kieffer .
 A small fraction of the polygonal ground patterns mapped by Mangold  are formed on the seasonal cap. The “Large polygons around the Polar Cap” (LPC) class of features in particular (40–250 m polygons, characterized by dark outlines) includes images near the south pole occurring well before the seasonal cap has been removed. The CROCUS date is Ls = 320° right next to the perennial cap [Piqueux and Christensen, 2008], whereas the LPC polygons usually occur at Ls ∼ 240° when the venting is the most active and the seasonal features are already well developed. LPC polygons are not ground patterns but seasonal features on the cap and we propose that they result from the venting of the cap. Mangold [2005, p. 338] noted that “it is not clear why cracks appear dark” but our model predicts that vents coated with dark particulate material and surrounded by erupted sand on the CO2 frost should be present at this time of the year on the seasonal cap (Figures 3–5 and 8).
 The dark changing polygonal pattern observed by van Gasselt et al.  in the troughs of the south perennial cap may be another example of seasonal polygons resulting from the venting process. In the summer, when the cap has entirely sublimed, MOC images show that the seasonal polygonal patterns cease to exist. These observations are also consistent with the updated venting model presented here. The region of study of van Gasselt et al.  is located on terrains where water ice is either exposed or near the surface [Bibring et al., 2004; Piqueux et al., 2008], perhaps preventing the substrate from being efficiently scoured (no spiders or etched polygons) and the moving gas from displacing as much dust as other polygonal patterns do elsewhere (no spots and fans).
3.3. Age of the Spiders
 We evaluate the time required to form a spider. Kieffer et al.  observed that (1) ∼30% of the surface of the STSC is covered with spots, (2) the albedo of a spot is 0.21, and (3) the surface temperature of the spots is ∼3–5 K hotter than the pristine CO2 ice. At ∼Ls = 230° and −86°N where these measurements have been made (“Manhattan Island” [Kieffer et al., 2006]), our thermal modeling of the cap (using Kieffer's [2007, Tables 1 and 2] parameters) indicates that the spots have to be made of a (1) 0.5 mm thick layer of 2 μm dust, (2) 1 mm of 10 μm dust, or (3) 5 mm of 150 μm sand, that is in thermal contact with the cap. Kieffer et al.  found that a layer of ∼1 mm is appropriate to account for the thermal data (no indication of the grain size). Kieffer  calculated that the shape of the fans is consistent with ∼100 μm particles settling to the surface. The STSC at Ls = 210° is ∼3 × 106 km2. If 30% is covered with a 1 mm layer of dust, 3 × 1010 m3 or 4 × 1013 kg of dust is involved in the cycle in the south every year. For comparison, this is equivalent to 200 times the mass of dust lifted in the global dust devil activity on Mars per year [Whelley and Greeley, 2008], or 100 times the mass of dust in a global dust storm [Martin, 1995]. For this reason, the venting of the seasonal cap is a major geomorphological agent displacing a tremendous amount of material and shaping the southern polar surface.
 Let us consider a single spot and the associated spider. Assuming that a spot is a circular layer ∼50 m in diameter and 1 mm in thickness, each one is made of 2 m3 of articulated material. A typical spider center depression is 50 m in diameter [Piqueux et al., 2003] although sizes range from a few meters (Hansen et al., submitted manuscript, 2008) to over 150 m. Let us assume that it is 5 m deep (consistent with the 4 m depth of the troughs calculated on the etched polygons, Figure 7). For comparison, Hansen et al. (submitted manuscript, 2008) find a depth of 1.8 m based on the analysis of an anaglyph of a shallow spider (Hansen et al., submitted manuscript, 2008, Figure 2c). The volume of a spider is 1 × 104 m3 (assuming that the branches of the spiders represent a negligible volume). If all the material forming the spot directly originates from the erosion of the substrate, it takes 104 years to scour a depression having the volume of a spider. This assumes that all of the dust erupted on the cap is the product of the erosion of the substrate during the spider formation process. It is more likely that a significant fraction of the dust involved in the formation of a spot is not the result of the abrasion of the subsurface but originates from a preexisting layer of mobile particulate material that is regenerated every summer, brought by winds and atmospheric sedimentation. In addition, much of the material lifted each year settles back onto the surface each summer. Finally, using TES and THEMIS IR data, Titus et al.  have shown that the terrains near the southern cap are composed of a water ice table under a thin layer of very fine unconsolidated dust. This mobile dust layer may represent a large fraction of the material forming a fan and therefore the time calculated here to form a spider is a lower limit, the spiders being many times or orders of magnitudes older than 104 years.
 Because the spiders take over 104 years to form, the locations in which the seasonal CO2 ice is translucent and impermeable must have been similar to those of today over tens of thousands of Martian years. The present surface of the south PLD is 10 to 100 Ma old [Plaut et al., 1988; Herkenhoff and Plaut, 2000; Koutnik et al., 2002] which would be sufficient time for hundreds of generations of spiders to develop. The fact that such an old surface is not covered with fossil spiders indicates that the distribution of the spiders in the past must have been very similar to what it is today (assuming that little erosion erased possible features). In other words, spiders have probably always been confined to small regions and may not have been prevalent in the past.
 The age of the spiders implies that they have likely experienced several cycles of precession (5 × 104 years) and possibly obliquity (1.2 × 105 years). The erosive power of the confined gas is controlled by its velocity, which is in turn a function of the sublimation rate of the cap [Kieffer, 2007, equation (2) and (3)] as well as other parameters (vent diameter, intervents distance). At other epochs, the solar insolation near the poles and the seasonal cap characteristics were different than what is observed now and presumably led to various seasonal cap sublimation rates. The solar input varies with changes of the Martian orbital parameters. At low obliquity, the Sun does not rise more than ∼15° above the horizon and the caps are assumed to be thicker than they are presently [Toon et al., 1980]; both effects reduce significantly the amount of energy reaching the top of the substrate (also see Kieffer  for the transmission of light in a pure CO2 slab as a function of the thickness). As a result, the basal sublimation of the cap is expected to be less efficient at low obliquity than what is observed today and the formation of spiders may not have been possible. In addition, the perennial caps are expected to be more widespread during periods of low obliquity, and spiders cannot form on high-albedo surfaces. At high obliquity, when the Sun is ∼35° above the horizon at noon, the situation is reversed and spiders are more likely to form. Further, the perihelion is presently reached during the southern summer when the seasonal cap retreats, which seems to be a favorable configuration to form spiders because the amount of solar energy available is maximal at the right time. When the aphelion occurs during the southern summer, the southern cap receives ∼20% less energy and the sublimation rates are expected to be smaller.
 The calculation of the age of the spiders assumes a constant rate of erosion and therefore conditions of insolation similar to those observed today. Yet, the present southern perihelion is most favorable to the formation of southern spiders and the obliquity is currently increasing [Ward, 1992], suggesting that the erosion rates have been most likely lower in the past for extended periods of time. Further quantifications are difficult without more detailed knowledge of the mechanical parameters of the substrate but the age we have calculated is a lower estimate and it is possible that spiders do not form at all epochs (e.g., low obliquity, aphelion during the summer).
4. North: Observations and Interpretation
 The model of the venting of the cap is based on observations and modeling applied to the south seasonal ice. Little or no work has been invested in analyzing data available on the north seasonal cap despite the well documented exotic defrosting features visible on the sand dunes [Malin and Edgett, 2001]. The basal sublimation of the cap is a major erosive agent in the south and it is important to better characterize the extent of its effect on the northern stratigraphic record.
4.1. North TSC Mapping
Kieffer and Titus  mention the presence of a northern dark region on the seasonal cap between Ls = 357° and Ls = 60° characterized by an albedo as low as of 0.21 and a CO2 ice temperature of 152 K. Titus et al.  attribute spectra of the cap acquired by TES to the presence of a transparent slab of CO2 ice. This dark and cold region is an equivalent of the STSC, indicating that regions of the north seasonal cap are translucent during the spring and locally coated with dark dust. This northern dark ice is visible between 135°E and 225°E and latitudes 80°N to 85°N, mostly on the polar erg [Kieffer and Titus, 2001]. The TES albedo of the seasonal cap between Ls = 345° and Ls = 60° is shown on Figure 10 and confirms the finding of Kiefer and Titus . Before Ls = 350°, the seasonal cap's albedo is fairly uniform with a typical albedo of 0.40. Near Ls = 350°, a fraction of the cap darkens (albedo < 0.20) near 75°N 300°E for a short period of time (e.g., Ls = 5°–10°). This darkening is observed during the 3 consecutive years of TES observations. The translucent ice (albedo of 0.25–0.30) first appears between Ls = 355° and Ls = 360° with a few dark patches near 75°N 300°E. Between Ls = 0° and Ls = 5°, the translucent ice occupies a considerably larger surface area (e.g., ∼8 × 105 km2) encircling the NPLD. The darkest part of the north TSC (NTSC) (e.g., albedo ∼0.25) shrinks to half this surface area at Ls = 10° and almost disappears at Ls = 15°. However, the rest of the cap brightens to an albedo of ∼0.45. The ring shaped NTSC brightens to albedos near 0.30–0.35 and shrinks continuously until Ls = 40°. At this time, the cap locally reaches albedos of 0.60 or more, consistent with Paige et al.  measurements. After Ls = 45°, most of the NTSC retreats and displays an albedo of 0.35. After Ls = 60°, the NTSC is not visible anymore and most of the seasonal cap has sublimed away. Kieffer and Titus  mapped the CROCUS date at the vicinity of the H2O perennial cap slightly after Ls = 80°.
 It is possible that the NTSC expands farther toward the higher latitudes, covering the NPLD and the perennial cap but albedo data alone are not sufficient to confirm this hypothesis. The H2O cap has an albedo > 0.5 that is also close to CO2 frost albedo. Therefore, translucent ice would be bright above the north seasonal cap. Through the spring long, the outline of the perennial cap is clearly brighter than the rest of the seasonal cap (near 0.6, Figure 10), indicating that the NTSC is not ring-shaped as suggested by Figure 10 but a subcontinuous translucent region covering the terrains from 75°N to the pole.
 The existence of the NTSC indicates that the seasonal cap is translucent with the underlying substrate being visible through the slab and/or that dark material is deposited on the cap. Spectral mapping of the surface material of the cap can help to differentiate locally between these two processes [Langevin et al., 2006]. In the south, the STSC is associated with the basal sublimation and venting of the cap. If the same model is valid for the north, seasonal (spots, fans, dark polygons) and perennial features should be detected and their distribution should be correlated with the NTSC.
4.2. Active Venting Features
 Where dunes are present, dark spots often form at the crest as they commonly do in the South [e.g., Malin and Edgett, 2001, Figure 73] (Figures 11, 12, and 13a). Some spots are subcircular with a slight elongation in the possible prevailing wind direction (Figure 13a); others are elongated toward the direction of the greatest slope, suggesting that the dark material has been deposited on the CO2 ice after it has been brought to the surface of the cap (Figure 12). Next to the polygons on the cap, very small spots are often aligned in groups, comprising an incomplete polygonal network that is consistent with the venting model, with spots and seasonal polygons forming a continuum of vent landforms. In both polar regions, the fans have similar characteristics (shape, size, albedo contrast with the surrounding terrains) but the fields of the northern fans are sometimes less dense, with a interfan distances easily exceeding 300 m (Figure 13b) versus 100 m in the south [Kieffer, 2007]. They are also commonly adjacent to fields of dark spots. Bright fans, relatively rare in the south, have not been observed on the north seasonal cap.
 We have mapped the distribution of the spots and fans between 87°N and 70°N with all the available MOC high-resolution images acquired between Ls = 0° and Ls = 70° when most features are visible on the seasonal cap (Figure 14a). Over 2300 MOC images have been scrutinized for features (Figure 14c for the data distribution) and 688 images displaying spots or fans have been found. Very few features are mapped on the H2O perennial cap, between the pole and 80°N despite the extensive MOC coverage. This is consistent with the venting model because the substrate has to be dark (the albedo of the cap is 0.5) and the surface material mobile (unlikely on the surface of the perennial cap) to form spots and fans.
 Other observations confirm that the dark spots on the sand dunes are the source of the material that sediments on the cap and darkens it further. Figure 15 shows a group of a dozen of isolated dunes covered with dark spots. Kilometers-long dark streaks originating from the defrosting dunes are visible and oriented toward the prevailing wind direction deduced from the shape of the dunes and the model by Fenton and Richardson . This observation, similar to that presented in the south in Figure 3, confirms that the low albedo of the seasonal cap in the TSC is due to the contribution of the dark substrate visible trough the translucent ice and the particulate material erupted and dispersed at the surface of the cap.
 Almost all the seasonal features are mapped between 82°N and 75°N. This distribution matches the location of the NTSC (Figure 14) and the polar erg. This observation is compatible with the cap basal sublimation model which requires dark looking ice and mobile particulate material. Whether the surface layer of the north polar erg is consolidated is not known, but dune movement has been detected [Bourke et al., 2008], suggesting that at least some surfaces are made of loose material. 88 images showing spots or fans have been found between 30°E and 330°E. This area corresponds to the region where the NTSC is not well defined in Figure 14. This region is dark at other times of the spring (Ls = 350°−355° and Ls = 0°−5°, Figure 10) which confirms that almost all the seasonal features are seen on the NTSC. The distribution of spots and fans presented in Figure 14a is biased by the heterogeneous MOC data coverage (Figure 14c). However, Figure 14a is still a good indicator of the correlation between the location of the NTSC and the seasonal features because 84 images are available between 73°N and 70°N, outside the NTSC, and less than 10 display seasonal features (∼12%), whereas 1640 images are available between 80°N and 75°N, on the NTSC, and 443 show spots or fans (27%). Similarly, over 300 images are available above the perennial cap and none show features (we have excluded the features seen in the spiraling troughs and chasmas, which are also darker than the rest of the seasonal cap).
 Although dark polygons on the seasonal cap do not appear to be as common in the north as they are in the south, many instances (i.e., 65 MOC images) have been detected between Ls = 0° and Ls = 70°, when the sublimation is most active. The morphological characteristics of these polygons are very similar to their southern counterparts (both straight and diffuse edges, sometimes elongated toward one consistent direction, with varying widths, occasional missing edges) although they may be slightly smaller (40–60 m in size). In Figures 11 and 12, the polygons in the north are ∼50 m in size. In both cases, the polygons on the seasonal cap are accompanied by a few dark spots. The locations in which dark polygons are visible are shown on Figure 14b. The majority of the dark polygons are mapped on the dark NTSC which is expected based on their proposed formation model, but several (14 images) are mapped elsewhere on the polar erg. It is possible that these dark polygons form on small patches of TSC that are under Figure 14b resolution (e.g., 1 pixel per ∼15 km). Similar observations of spiders mapped outside the region where the ice is translucent during the summer are reported by Piqueux et al.  on the SPLD. The correlation between the NTSC and the dark polygons is consistent with the basal sublimation model.
 In addition to the polygons, spots and fans, we have identified on the CO2 ice-free surface a total of 3 groups of dark, dendritic patterns with thin branches converging and merging with larger branches located near 85.4°N 180.5°E (red dot on Figure 14b). The dark branches vary in width from 10 to 40 m. In all three cases, the resolution of the available image is low (5.1 m per pixel) and it is not known if the visibility of these features is enhanced by shadows due to their own topography or albedo, but the numerous branches merging toward a larger segment is strongly suggestive of the spiders observed in the south polar region. In the case of the features seen on Figure 16, the overall diameter of the dendritic pattern is ∼500 m, which is twice the size of the other two groups observed. These objects display the size of spiders observed in the south. The albedo of the NTSC where they are located is ∼0.3, which is brighter than the typical albedo of the STSC where southern spiders are most common (e.g., albedo ∼0.2). If they are formed by the scouring action of confined CO2 gas, these spider-looking (araneiform (Hansen et al., submitted manuscript, 2008)) objects would have been expected on the terrains where the NTSC has an albedo of 0.2 (Figure 10, Ls = 5°–10°). A few dark polygonal networks are found nearby (Figure 14b) and it is possible that these araneiform features are on a patch of dark NTSC that is under Figure 14 resolution.
 The low-albedo seasonal ice indicates that parts of the cap may be translucent and that the south cap basal sublimation model applies to the north cap as well. The observation of spots, fans and dark polygons further suggests that similar processes take place under both seasonal caps and their distribution is consistent with the model. It is also possible that spiders form on the NPLD, under the NTSC.
5. North and South Differences
 Fewer seasonal and perennial features are observed in the north compared to the south suggesting that the conditions for their formation are not ubiquitous. Let us review the requirements for the generation of seasonal and perennial features and discuss the possible differences between both polar regions.
 The TSC must be sufficiently transparent and the substrate must be dark enough to absorb a significant fraction of the solar energy in order to maintain the basal sublimation rate (see Kieffer  for the relationship between substrate albedo, sublimation rate, and gas pressure). A comparison of Figure 10 and Piqueux et al.'s Figure 4 of the albedo of the south seasonal cap (similar color scale) indicates that the NTSC is slightly brighter than its southern counterpart, whereas the north polar erg is much darker than the SPLD (Δalbedo = 0.15 [Paige et al., 1994; Paige and Keegan, 1994]). To account for the similar albedo despite the difference in substrate albedo, we propose that the NTSC is less transparent than the STSC. However, a similar fraction of solar energy hits the surface of the PLD since they both have a similar albedo. Therefore the regolith albedo difference does not explain the north and south differences.
 The seasonal cap has to be translucent for many consecutive years to form perennial features. Viking mapped but did not recognize the STSC in 1976 [Kieffer, 1979; Kieffer et al., 2000]. Since the beginning of the MGS TES mapping era (1999), the north and south TSC have been observed repeatedly with little variability, suggesting that the occurrence of translucent seasonal ice is common in the current epoch. Several instruments orbiting Mars (THEMIS, OMEGA and CRISM) will continue to observe both TSCs for springs to come, helping to monitor their long-term evolution.
 The seasonal cap has to be translucent long enough during the spring to form perennial features. In the south, the darkest (e.g., albedo < 0.25) STSC is observed before Ls = 180° to Ls = 250°, when the seasonal cap is mostly removed [Piqueux et al., 2003]. This corresponds to a 70° Ls period of time when the basal sublimation and venting are active. For comparison, the darkest (albedo = 0.25) NTSC is not clearly visible from before Ls = 0° and is mostly gone after Ls = 15°. From Ls = 15° to Ls = 60°, the NTSC is still detected but it is much brighter (albedo ∼0.35). As a result, the spike of energy reaching the NPLD is reduced to a period of 15°Ls (70°Ls in the south).
 In addition, the north cap is translucent for 80°Ls after the beginning of summer, when the solar radiation is maximal, whereas the STSC is clearly visible starting 50°Ls from the southern summer (Ls = 220°), indicating that the amount of solar energy reaching the substrate integrated over the entire spring is significantly larger in the south than in the north.
 Because of the date of the perihelion (Ls = 336°) and the eccentricity of the Martian orbit, the solar heating is reduced by 20% in the north compared to the south during both springs. As a result, the rates of basal sublimation must be lower in the north, leading to a lower pressure gradient with the atmosphere and slower transport of gas, dust and sand.
Kieffer  has shown that the subice gas velocity is not only a function of the sublimation rate but also of the vents' spacing. Because the rapid migration of the gas under the slab is responsible for the scouring of the substrate, it is critical to determine if the velocities are lower in the north to explain the scarcity of features. In the south, the typical vent spacing distance is 100 m [Kieffer, 2007]. In the north, this distance seems to be highly variable. In Figures 11 and 13a, the interspots distance is ∼100 m, whereas it is smaller in Figures 13b and 15. In Figure 12, the dark spots are usually separated by over 200 m. Our observations show that the intervent distances are variable in the north but roughly similar to the southern ones. As a result, the spacing of the vents does not seem to be a factor controlling a preferential formation of spiders and etched polygons in the south.
 The slab of ice has to be impermeable to allow the gas pressure to build up under the cap and form geysers. The permeability may depend on the thickness of the frost deposit (along with the grain size and impurity concentrations which influence the strength of the bulk material). Hypothetically, a difference in cap thickness may result in a difference in mechanical strength. A thicker cap may be more difficult to puncture or crack, allowing the gas to migrate over larger distances and the gas pressure to build up. Conflicting data regarding the relative thicknesses of the caps have been published. On the basis of repeated MOLA observations of the Martian high latitudes, the maximum thickness of the northern seasonal cap (1.5 ± 0.25 m) is larger than the maximum thickness of the southern seasonal cap (0.9 ± 0.30 m) [Smith et al., 2001]. However, gamma ray data [Feldman et al., 2003; Litvak et al., 2004; Kelly et al., 2006] and global circulation models [Smith et al., 1999] suggest a greater seasonal cap thickness in the south. No conclusions can be drawn.
 The substrate must be friable enough to permit the formation of troughs and the eruption of dust and sand, but competent enough to conserve the troughs and spiders over time. Piqueux et al.  have shown that the physical nature of the substrate is a key factor for the formation of the spiders in the southern polar region. In particular, all the occurrences in the south have been found on the SPLD or on the surrounding mantled terrains [Piqueux et al., 2003]. The STSC expands locally on the cratered highlands where no spiders have been observed [Piqueux et al., 2003] suggesting that spiders are shallow landforms that require a somewhat erodable substrate to form. The north polar erg is populated with sand dunes, part of which might be made of loose material [Edgett and Christensen, 1991; Bourke et al., 2008]. Unconsolidated sand may offer a porous medium in which ice is readily sublimated from the pores, releasing sand and dust that can be easily transported. Therefore, the polar erg is a favorable place to form seasonal features on the cap (consistent with the observations). However, loose dust and sand may not be able to maintain perennial features. In the south, the walls of the depressions forming the core and the legs of the spiders are steeper than the angle of repose of particulate material, confirming that spiders and etched polygons have been carved in indurated materials. The hypothetical softer nature of the north surface may prevent the conservation of well defined perennial features and may result in difficult to detect shallow and smooth depressions.
 The reasons for the scarcity of perennial features near the north pole are not clear. The differential of solar energy reaching the bottom of the north seasonal cap versus the south cap may be an important contributing factor along with the mechanical characteristics of the north surface layer, which may be unable to maintain steep spider-like depressions.
 1. During the southern spring, where the seasonal ice is translucent and impermeable, the basal sublimation and rapid flow of the gas is responsible for the eruption of sand and dust at the surface, scoured from the regolith. The dust forms spots, fans, and dark polygons.
 2. Dark polygons form when the cap is fractured and where the gas escapes from elongated vents, whereas spots and fans form when the cap is punctured in small vents.
 3. Networks of etched polygons scoured in the substrate are a class of perennial landform resulting from the venting of the cap. Etched polygons form when the gas migrates laterally, forming linear troughs, whereas spiders form from gas that converges toward a point source.
 4. The minimum age of the spiders is 104 years old.
 5. The scouring of the SPLD by the confined gas corresponds to one of the most efficient erosion agents on Mars. It displaces orders of magnitude more material than global dust storms and the annual dust devil activity.
 6. We have mapped the location and seasonal variability of the north translucent cap and shown that it correlates with the distribution of spots, fans, dark polygons, and possibly spiders. These observations are consistent with the model of the basal sublimation and venting of the seasonal cap developed for the south. The surface layer of the NPLD is swept and gardened every spring by confined moving gas and particulated material.
 7. Kieffer et al.  proposed that the repetition of the process over long periods of time may have affected the conservation of part of the stratigraphic record in the south. Our work suggests that not only the southern high latitudes may have been affected but also the northern.
 This paper has greatly benefited from S. Milkovich and an anonymous reviewer's comments. This work was supported by the NASA THEMIS Mars Odyssey Project.