The Mangala Valles system of channels heads at one of the graben that comprise Memnonia Fossae, extends northward nearly 1000 km across Noachian and Hesperian highlands, and terminates at basins contiguous with Amazonis Planitia. The Mangala Valles system has previously been interpreted to have formed through one or more catastrophic aqueous floods on the basis of similarities between the characteristics of its channel landforms and those of terrestrial systems, including the Channeled Scabland of Washington. Although aqueous mechanisms for formation of Mangala Valles are broadly congruous with known characteristics of the channel system, an alternative volcanic hypothesis for formation of the system appears to be worthy of consideration on the basis of (1) its consistency with the volcanotectonic nature of the system and (2) commonalities between the basic nature of the system and that of large volcanic channels of the inner solar system. Estimates based on thermal considerations suggest that formation of Mangala Valles could conceivably have taken place through eruption of a lava volume of ∼2 × 105 km3, or roughly the total volume of the terrestrial Columbia River Basalt Group. The volcanic hypothesis for formation of Mangala Valles, and a hybrid hypothesis involving formation of Mangala Valles through aqueous processes followed by or in concert with substantial modification of the system by volcanic erosion and deposition, appears viable and worthy of future consideration.
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.
 Partly on the basis of spatial variation in crater counts, the Mangala Valles system has been previously hypothesized to have formed through multiple flow events, some or all of which may have been correlated in time with late-Hesperian or Hesperian-Amazonian episodes of volcanism and tectonism in the region [Tanaka and Chapman, 1990]. Regardless of the exact timing of the formation of Mangala Valles, the system is generally interpreted to have formed by catastrophic aqueous floods over a combined period of days to thousands of days [e.g., see Ghatan et al., 2005].
 Hypotheses previously proposed to account for the existence and characteristics of the Mangala Valles drainage system rely upon aqueous mechanisms of formation that involve the sudden release of large volumes of water from the subsurface through fractures associated with the source graben. These proposed mechanisms have substantial potential to account for the basic nature of the Mangala Valles system. However, it is noteworthy that there are no satisfactory analog processes, terrestrial or otherwise, for formation of large systems of incised channels by the catastrophic release of groundwater from fractures. Although proposed aqueous mechanisms for formation of Mangala Valles are congruous with known characteristics of the channel system, an alternative volcanic hypothesis for formation of the system appears worthy of consideration on the basis of its apparent consistency with landforms interpreted as volcanic at Mangala Valles and on other solar system bodies. Terrestrial, lunar, Venusian, and Martian landforms collectively suggest that nonaqueous volcanic processes may be sufficient to account for the existence and nature of the Mangala Valles system, without the need to invoke the action of past catastrophic aqueous flooding.
 This paper outlines a hypothesis for formation of the channels of the Mangala Valles system through nonaqueous volcanic processes. A summary of the geological setting of Mangala Valles is presented, followed by a review of the rationale behind previous aqueous interpretations of the channel system. An alternative volcanic hypothesis for formation of the channel system is outlined and discussed, and quantifications of basic aspects of this hypothesis are presented.
 Dune-like ridges located within tens of kilometers of the main source graben of Mangala Valles have been previously interpreted to have formed through explosive eruptions related to phreatomagmatic processes that preceded large outpourings of groundwater hypothesized to have formed the channels of Mangala Valles [Wilson and Head, 2004]. More prominent lobate-margined deposits located adjacent to the main source graben (Figure 4) have been previously interpreted to have formed by the outward flow of ice lobes formed by the freezing of water emitted from the graben [Head et al., 2004]. The span of the source graben at its intersection with a highland ridge to the east of the head of Mangala Valles (Figure 1: R; Figure 5a) is defined in part by numerous collapse pits considered suggestive of subsidence at a subgraben dike [Ghatan et al., 2005]. Extensive plains to the east of these collapse pits are interpreted to be composed of volcanic units that bury the source graben of Mangala Valles, and partly consist of flows erupted from the graben itself [e.g., Craddock and Greeley, 1994; see also Wilson and Head, 2002] (Figure 5b).
 Extensive smooth plains to the south of the graben (Figure 1: P) have been previously interpreted to have formed through aeolian or fluvial processes [Craddock and Greeley, 1994], although more recent research suggests that these plains are likely of volcanic origin, on the basis of the distribution of similar plains in the vicinity of circum-Tharsis ridges, their association with Tharsis volcanic flows, the manner in which older topography is embayed by the plains deposits, and the presence of features such as wrinkle ridges and apparent flow fronts [Ghatan et al., 2005]. Similar volcanic interpretations have been made for other ridged plains deposits located in the Mangala Valles region [e.g., Chapman and Tanaka, 1993], and a general regional association exists between volcanic plains and the Tharsis-radial graben system [Wilson and Head, 2002]. The plains to the south of the head of Mangala Valles (Figure 1: P) are interpreted to have formed volcanically in conjunction with the volcanic resurfacing of Noachian terrain to the north, prior to the events that formed the channels of Mangala Valles [Ghatan et al., 2005].
 The channels of Mangala Valles are incised into the uplands of Memnonia, extending from the source graben northward to the dichotomy boundary. The primary channel network of Mangala Valles extends directly northward from the source graben, whereas a major branch of the system, Labou Vallis, reaches the dichotomy boundary by a northwestward path [e.g., Chapman and Tanaka, 1993] (Figure 1: M and L). The downstream gradient of most of the Mangala Valles system is less than one degree, steepening only in the vicinity of the dichotomy boundary (Figure 2) [see also, e.g., Zimbelman et al., 1992, 1994; Ghatan et al., 2005]. Differences in crater-count ages within the Mangala Valles system have previously been used to suggest that multiple flow events formed its channels [Tanaka and Chapman, 1990], but it has been recently noted that measured channel gradients are not necessarily consistent with such age determinations [Ghatan et al., 2005]. It is also possible that differences in crater statistics between reaches of Mangala Valles instead reflect the sampling difficulties expected of regions of relatively small areal extent [Zimbelman et al., 1994].
 Longitudinal grooves discontinuously extend along the floors of many channels of Mangala Valles (Figure 8) [see also Chapman and Tanaka, 1993], and have been previously interpreted at other Martian outflow channels as having formed through the action of catastrophic flooding or mud and debris flows [e.g., Baker and Milton, 1974; Thompson, 1979]; other hypothesized mechanisms have included aeolian processes [e.g., Sharp and Malin, 1975] or glacial processes [e.g., Lucchitta et al., 1981; Lucchitta, 2001]. Some landforms that appear at relatively coarse image resolutions as longitudinal grooves and ridges are recognizable at higher resolutions instead as dune fields and yardangs that in some cases deviate notably from channel paths (e.g., MOC image M0806100). Similar aeolian landforms are also found in upland regions surrounding parts of the Mangala Valles system (e.g., MOC images E1003072 and E1301768).
 Terraces are common along the margins of the channels of Mangala Valles [Chapman and Tanaka, 1993] (Figure 8), and in places several tens of stacked terraces are evident [Ghatan et al., 2005]; such terraces have been interpreted as having formed by erosion of layered volcanic units deposited prior to or during formation of the channel system, or having formed as fluvial depositional terraces during the waning stages of aqueous flooding [e.g., see Chapman and Tanaka, 1993; Ghatan et al., 2005]. Narrow leveed subchannels are present along some reaches of Mangala Valles, and a long (100 km) example has been interpreted as a rille that likely acted as a conduit for fluid volcanic flow (Figure 9) [Chapman and Tanaka, 1993].
 Plains deposits, some of which have lobate margins or pitted surfaces, are present at the mouths of Mangala Valles channels at the dichotomy boundary. These deposits have been interpreted as possible mudflow deposits and ice-rich deposits, with more distal plains units interpreted as volcanic on the basis of the presence of wrinkle ridges and lobate margins [e.g., Chapman and Tanaka, 1993; see also Scott and Tanaka, 1986]. Deposits of the Medusae Fossae Formation extend across the terminal basins of Mangala Valles, and are interpreted as possible airfall deposits of volcanic origin [e.g., Bradley et al., 2002; Hynek et al., 2003; see also Chapman and Tanaka, 1993].
 Dune and megaripple fields (e.g., MOC images M0300429 and S0500890) overlay portions of Mangala Valles, with coverage greatest within several hundred kilometers of the dichotomy boundary. Much of the Mangala Valles system and surrounding region is mantled by other aeolian materials, as is indicated by the widespread presence of wind streaks, dark talus streaks, and dust-devil tracks.
3. Evidence in Support of Aqueous Mechanisms of Channel Formation at Mangala Valles
 At the Channeled Scabland, there are numerous subfluvially formed erosional transverse ledges that interrupt the floors of deeply incised inner channels and that acted as recessional cataracts to paleoflow [Bretz, 1928]. These ledges may be generally described as horseshoe-shaped headcuts with scour-formed closed depressions at the bases of each cataract alcove [Baker, 1978b]. The Dry Falls cataract, one of the largest such features in the Channeled Scabland, is ∼5.5 km across and 120 m high [e.g., Bretz, 1923; Baker, 1978b]. Although confident identification of this class of landform at Mangala Valles is difficult, inner channels that head at scarps oriented transversely to directions of past channel flow are relatively common features along much of the system's length (Figure 13) [see also Baker and Milton, 1974].
 Pre-flood tributaries to channels deeply excavated from earlier stream valleys in the Channeled Scabland form hanging valleys with floors up to hundreds of meters above the floors of excavated channels [e.g., Baker, 1978b]. At Mangala Valles, there are numerous examples of tributary channels stranded above more deeply incised channels (e.g., THEMIS VIS image V10429001). Some of the most prominent such examples include subchannel segments that terminate abruptly at the peripheries of large residual landforms within the main channel system itself (e.g., Figure 7).
 At the Channeled Scabland, distinctive scour depressions are found along the floors of numerous conduits, and appear to have formed by the action of vortices around and in the lee of bedrock projections and large boulders [e.g., Baker, 1978a, 1978b]. Channeled Scabland scour features and associated protuberances together define a characteristic butte-and-basin topography, with a total relief of about 30–100 m; the dimensions of scours are as great as kilometers in length and tens of meters in depth [Baker, 1978b]. The pitted surfaces of some channel units of Mangala Valles have been identified as having possibly formed through scour by floodwaters [Sharp and Malin, 1975; Zimbelman et al., 1992], and some areas of irregular topography at Mangala Valles are candidate examples of butte-and-basin topography [Baker and Milton, 1974].
 Depositional landforms at the Channeled Scabland include longitudinal bars, eddy bars, and giant current ripples. Longitudinal bars, depositional features with long dimensions oriented parallel to flow direction and thicknesses of tens of meters, are found in relatively uniform Channeled Scabland reaches that lack abrupt changes in width [e.g., Bretz et al., 1956; Baker, 1978b]. Such bars, which include pendant bars, are present in the Channeled Scabland downstream from bedrock projections and other flow obstacles. Scabland eddy bars, poorly sorted sediment accumulations tens of meters thick and deposited in former environments of high turbulence, block the mouths of drainages entering the Cheny-Palouse scabland tract, and are associated with slackwater deposits that extend further up the tributary valleys [e.g., Bretz, 1925, 1928; Baker, 1978b; Benito and O'Connor, 2003]. Giant current ripples, composed of parallel ridges and swales with relief of up to ∼15 m, are distributed in a number of scabland channels [Bretz et al., 1956; Baker, 1973, 1978b].
 At Mangala Valles, identification of landforms clearly formed through aggradation involving fluvial or diluvial processes has not yet been made, although there are many examples of channel features at the lee of obstacles to flow, and that conceivably could be depositional landforms (Figure 6) [see also, e.g., Milton, 1973; Tanaka and Chapman, 1990; Zimbelman et al., 1992; Ghatan et al., 2005]. Confident discrimination between erosional and depositional landforms in these cases has been problematic [e.g., Baker and Kochel, 1978a], and recent work suggests that most or all of the streamlined forms at Mangala Valles are erosional [Ghatan et al., 2005]; Burr  has identified candidate depositional landforms associated with Athabasca Valles, a Martian channel system that heads at Cerberus Fossae [e.g., Head et al., 2003; Manga, 2004]. Giant current ripples cannot yet be confidently identified at Mangala Valles, in part because candidate alluvial ripples cannot be clearly discriminated from aeolian landforms that can have similar appearances and dimensions (although high-resolution photoclinometry and other techniques have potential in this regard [e.g., Burr et al., 2004]).
4. Volcanic Processes as Alternative Mechanisms of Channel Formation at Mangala Valles
 The aqueous-flood hypothesis for incision of the Mangala Valles channel system has been considered compelling on the basis of the strong and seemingly unique correspondence between the landforms of this system and those of terrestrial drainage systems such as the Channeled Scabland. The aqueous-flood hypothesis can account for the basic nature of the Mangala Valles system, and has as a result been appropriately used as a foundation for recent studies of the system. However, the conspicuously volcanotectonic nature of the Mangala Valles system, and shared morphological and contextual characteristics with large and apparently volcanic channels of the Moon and Venus, suggest that an alternative volcanic mechanism of formation should be developed for Mangala Valles, for consideration alongside the aqueous-flood mechanism.
 The fluid that formed the Mangala Valles system must have incised its anastamosing channels (Figure 7) and carved or otherwise formed features such as the elongate landforms located in the lee of obstacles (Figure 6). The natural fluid to consider for formation of such features, on the basis of terrestrial analogs, is water. However, although volcanic processes on the Earth are normally constructive in nature, theoretical and field-based studies have indicated that substantial amounts of incision and scour can result from the flow of molten rock through the processes of thermal and mechanical erosion (see discussion below). The possibility of formation of Martian outflow channels by the flow of lava has been suggested previously [Schonfeld, 1976, 1977, 1979; Cutts et al., 1978; Leverington, 2004; see also Carr, 1974b] and this possibility is consistent with the basic nature of large lunar and Venusian channels of apparent volcanic origin.
 Large and complex Venusian channels include a prominent system located northeast of Ozza Mons (Figure 16) [Komatsu et al., 1993] and Kallistos Vallis, a 1200-km-long outflow channel at Astkhik Planum [e.g., Baker et al., 1992; see also Leverington, 2004, Figure 14]. Systems such as these share a wide range of characteristics with Mangala Valles, including channels that appear to have been incised into country rock, streamlined residual landforms, complexly anastamosing reaches, channels of variable width, exposure of channel floors through partial or complete drainage of late-stage fluid flows, and channel lineations oriented along directions of paleoflow. The highest-resolution Magellan radar images have pixel dimensions of ∼100 × 100 m (resampled to ∼75 × 75 m for the full-resolution radar map, or FMAP, databases), rendering confident identification of bedforms (e.g., longitudinal grooves) on this spatial scale difficult. The coarseness of Magellan topographic databases of Venus, with resampled Global Topographic Data Record (GTDR) data characterized by pixel sizes of ∼5 × 5 km, severely inhibits meaningful quantification of the relief of channel bedforms and of channel depths. Magellan stereoimage data sets offer potential for generation of topographic databases of improved spatial resolution, but processing of these data sets can be challenging [e.g., Howington-Krause et al., 2006], and these data sets are not available for systems such as Kallistos Vallis. Importantly, although the attributes of numerous Venusian channels are suggestive of fluid erosion [e.g., Baker et al., 1992; Komatsu et al., 1993; Komatsu and Baker, 1994; Baker and Komatsu, 1999], the lack of high-resolution topographic data for Venus currently places limits on the confidence with which conclusions can be made in this regard.
 A volcanic interpretation of the Mangala Valles channel system is consistent with the basic properties and formation mechanisms of large volcanic channels of the inner solar system. This interpretation also appears to be consistent with the nature of Mangala Valles landforms previously interpreted as having formed through aqueous processes. The lobate-margined and festooned flow deposits proximal to the main source graben of Mangala Valles, previously interpreted as glacier-like landforms produced through phreatomagmatic processes and related outpourings of groundwater [Wilson and Head, 2004; Head et al., 2004], may alternatively be interpreted as ridged volcanic flows erupted from within the graben structure or channeled through the graben from Daedalia Planum. Such an interpretation is supported by the presence of similarly ridged and festooned volcanic flow units erupted from the same source graben at the volcanic plains of adjacent Daedalia Planum (Figure 17a), as well as at, e.g., the region east of the northern mouths of Mangala Valles (Figure 17b) (a volcanic interpretation is also consistent with the basic nature of volcanic deposits located elsewhere on Mars, Moon, and Earth [e.g., Mollard and Janes, 1984; Leverington and Maxwell, 2004; Greeley et al., 2005]). A volcanic interpretation is compatible with a wide range of climate conditions, whereas the predicted instability of surficial and near-surface water ice at latitudes within ∼30° of the equator [e.g., Farmer and Doms, 1979] does not favor the presence of glaciers in this region.
 Surface exposure of the deposits of the terminal basins of Mangala Valles is poor due to cover by materials of the Medusae Fossae Formation [e.g., Scott and Tanaka, 1986; Bradley et al., 2002; Hynek et al., 2003], as well as by dune fields and wind-scoured cover (e.g., MOC image M0001747), but exposed units are discontinuously found along the dichotomy boundary (Figure 18) as well as north of the Medusae Fossae Formation [e.g., Scott and Tanaka, 1986]. Exposed terminal deposits characterized by quasi-circular depressions and cuspate margins (Figure 18a and Figure 18c (top)), and similar deposits found within some channels of Mangala Valles (Figure 8), have previously been interpreted as degrading water-ice-rich flood deposits [Zimbelman et al., 1992; Levy and Head, 2005]. However, deposits with comparable features have been emplaced at volcanic plains located, e.g., at the source graben of the Mangala Valles system at Daedalia Planum (Figure 19a), within the Arsia Mons caldera (Figure 19c), and on the eastern flanks of Pavonis Mons (Figure 19d). Furthermore, parts of the same Mangala Valles terminal deposits have a ridged and platy nature (e.g., Figure 18c (bottom)) that is matched at nearby volcanic units of Daedalia Planum (e.g., Figure 19b). The pitted and cuspate-margined units near the mouths of Mangala Valles channels are overlain in places by ridged and lobate-margined flows derived from the east and of likely volcanic nature (Figure 17b), but the partly buried units do not show visible signs of thermal disturbance related either to the heat of flows during emplacement or to the upward shift in melting isotherm that might have been associated with the thickening of surface cover over ice-rich units. A volcanic origin for the cuspate and pitted deposits at Mangala Valles is considered consistent with (1) characterization by surface textures matched at volcanic plains in the region; (2) association of these materials with volcanic flows in a range of environments far removed from outflow channels; (3) a lack of response of materials to changes in thermal conditions related to burial by volcanic flows; and (4) the presence of these deposits at equatorial latitudes where water ice is unstable [e.g., Farmer and Doms, 1979]. A volcanic origin for a substantial volume of materials beyond the mouths of Mangala Valles is required of the volcanic hypothesis, since although many large volcanic channel systems on the Moon and Venus lack lava deltas at their mouths, extensive volcanic deposits are always present at the terminal basins of these systems [e.g., Leverington, 2004]. The characteristics of the units at the northern mouths of the Mangala Valles system appear to be consistent with this requirement, as is the ridged and lobate-margined nature of units exposed north of the Medusae Fossae Formation [e.g., Scott and Tanaka, 1986].
 Prominent inverted-channel features found in a segment of the northern Mangala Valles system (Figure 12) have characteristics consistent with those expected of lava flows [e.g., Chapman and Tanaka, 1993], but these resistant materials need not have been emplaced in a channel system carved by water. Many of the inverted features only partly fill the main Mangala Valles reaches within which they are nested (even accounting for subsequent erosional recession of channel rims), implying that host materials of the flows were channel-fill materials of relatively low resistance. The surface elevations of inverted channel-fill deposits generally match the elevations of corresponding rims of the main Mangala Valles channels, suggesting that host materials once filled the main channels here, rendering a diluvial origin for these host materials unlikely. Accumulation of ash or aeolian deposits would be consistent with the nature of infilling of these channel reaches, and is consistent with the once-greater sediment cover of the region implied by the abundance of aeolian deposits and yardang fields found across the region, extending south from the Medusae Fossae Formation. Other inverted channels in the region (e.g., features at the nearby volcanic plains of Daedalia Planum, V14859003, and a leveed channel in Noachian uplands adjacent to Mangala Valles, V16594001) also imply possible formation during confinement by now-eroded host materials of nonalluvial nature.
 Aspects of the nature of the flow of lava are yet to be fully described in a quantitative manner [e.g., Mouginis-Mark et al., 1992; Fagents and Greeley, 2001] and a comprehensive theory of both thermal and mechanical erosion by lava has not yet emerged. The viability of volcanic mechanisms for formation of the Mangala Valles system is therefore, at present, best illustrated by the existence and geomorphological characteristics of large lunar and Venusian channels interpreted as volcanic [e.g., Greeley, 1971a, 1971b; Swann et al., 1972; Carr, 1974b; Strain and El-Baz, 1977; Wilhelms, 1987; Komatsu and Baker, 1994; Baker et al., 1992, 1997; Komatsu et al., 1993]. However, a crude estimate of the volume of lava that might have been involved in the hypothesized volcanic formation of Mangala Valles can be calculated on the basis of the minimum volume of material removed during channel formation. The main Mangala Valles system (Figure 20) has a total surface area of 43,000 km2. The form of the channel system is complex, with typical channel depths ranging from ∼100 to as much as ∼700 m (Figure 20). Assuming a representative channel depth of 200 m (a depth of 100 m is assumed by Tanaka and Chapman ), the minimum volume of material removed from the main Mangala Valles system is estimated to be about 8600 km3, a volume that is larger than the 3000 km3 estimate of Tanaka and Chapman  and the 5700 km3 estimate implied by Hanna and Phillips , but smaller than the 13,000–20,000 km3 estimate of Ghatan et al.  (which also considers a region east of the main channels of Mangala Valles; see Figure 20). A volume greater than this 8600 km3 estimate could have been removed under the volcanic hypothesis for formation of Mangala Valles, on the basis of the tendency of lava to emplace solid fill materials within channel subbasins toward the terminations of individual eruption events, thus potentially leading to the removal of this additional material during subsequent flow events.
 For the purposes of order-of-magnitude estimations, the initial temperature of a magma of basaltic composition can be taken as 1450 K [e.g., Kauahikaua et al., 1998; Burgi et al., 2002] (although higher temperatures are possible), and initial cold-surface temperatures of Martian terrain can be taken as 220 K [e.g., Carr, 1996]. Substantial changes to the rheological properties of basaltic lava can take place at temperatures below ∼1400 K [e.g., Keszthelyi and Self, 1998], and thus the bulk temperature of basalt involved in hypothesized volcanic flows is not likely to have dropped substantially below this temperature during the main phases of hypothesized channel incision. Making the simplifying assumption that the densities and specific heat capacities of the solid and liquid phases of basalt are equal, the ratio of the mass of cold regolith to the mass of lava that would allow lava temperatures to be maintained at ∼1400 K can be estimated on the basis of conservation of energy to be ∼0.04. Assuming a minimum volume of removed material to be 8600 km3, this figure implies a minimum total channel-forming lava volume of ∼2 × 105 km3. This volume estimate accounts for that amount of lava required to raise regolith to and above melting temperatures exclusively through thermal mechanisms, and does not take into account the energy required for phase change of heated surface materials. Assuming a density of basalt of 3000 kg/m3 [e.g., Hulme, 1973, 1982] and a latent heat of fusion for basalt of 3.3 × 105 J/kg [Yoder, 1976], an additional ∼8.5 × 1021 J would be required to melt the full 8600 km3 of basalt bedrock at the melting temperature of ∼1275 K. Assuming a total erupted lava volume of 2 × 105 km3 and a specific heat of mafic lava of 1200 J/kg K, this value represents an additional drop in melt temperature of ∼11.8 K and thus, as expected [e.g., Hulme, 1982], is a relatively minor influence on the simplified scenario given above. The loss of energy associated with this drop in temperature is ultimately returned to the system as these melted materials re-crystallize, and the crystallization of an additional 2 × 105 km3 of erupted basalt results in the release of an additional ∼2 × 1023 J (energy sufficient to raise the temperature of this volume by ∼280 K). The large influx of energy resulting from crystallization of erupted lava is only partly offset by expected cooling along the channel system (cooling of ∼0.1 K/km [Keszthelyi et al., 2006] results in cooling of 100 K along a 1000 km system) and by expected cooling at flow fronts and the source vent (amounting to an approximate total of tens of K [e.g., Keszthelyi et al., 2006]). Note that the bulk density of removed regolith is unlikely to have been equal to that of nonvesicular basalt (vesicular or particulate basalt can have much lower densities that can approach ∼1500 kg/m3) [e.g., Yan et al., 2001], and it is likely that less than 40% of consolidated regolith would have needed to have undergone a phase change for channel erosion to have taken place [e.g., Hulme, 1973], and thus the energy required to warm and melt regolith is likely overestimated above for the volume estimate of 8600 km3.
 The timing of an event of total volume 2 × 105 km3 is extremely difficult to meaningfully estimate. Assuming an incision rate of 4 × 10−6 m/s (previously estimated for a smaller lunar channel located in the Marius Hills [see Hulme, 1973]), a total event time of 580 days is predicted, implying an average eruption rate of 4 × 106 m3/s (a total event time of 2000 days is predicted if a rate of incision of 1.16 × 10−6 m/s, measured at the small Hawaiian channel discussed above [Kauahikaua et al., 1998], is used). Adding to the substantial uncertainties involved in these estimates is the lack of consideration of such factors as (1) the action of mechanical processes of erosion [e.g., Kauahikaua et al., 1998, 2002; Ferlito and Siewert, 2006] and (2) the influence of expected high levels of flow turbulence [e.g., Howard et al., 1972] on thermal erosion.
 The above estimates suggest a total volume of erupted lava of 2 × 105 km3, an average eruption rate of 4 × 106 m3/s, and a combined event duration of 580 days. These values are crudely derived minimums that could conceivably underestimate the nature of hypothesized volcanic events by multiple orders of magnitude (though these values may also overestimate aspects of volcanic events, should, e.g., erupted magma temperatures have been notably greater than 1450 K, or should processes of mechanical erosion have been significant). Formation of the Mangala Valles system by graben-fed flows, or by Daedalia Planum flows channeled through a segment of the source graben, might be speculated to have taken place in concert with the latter stages of formation of Arsia Mons. Such a scenario could have involved episodic development of the system over geologically extended periods of time that are not reflected in the duration estimate.
 Though enormous relative to the volumes of large historical terrestrial volcanic eruptions (a volume of ∼15 km3 was erupted during the 1783–1784 fissure eruption at Laki, Iceland [Thordarson and Self, 1993]), the estimated Mangala Valles eruption volume of 2 × 105 km3 roughly corresponds to the total volume of 1.7 × 105 km3 of the terrestrial Columbia River Basalt Group [Tolan et al., 1989]; the largest individual flows of the Columbia Plateau flood basalts had volumes that in some cases exceeded 1000 km3 [Reidel, 1998]. An estimated eruption volume at Mangala Valles of 2 × 105 km3 is equivalent to ∼3–8% of the total estimated volume of lunar mare basalts if average mare thicknesses are taken to be between 1.0 and 0.4 km, respectively [DeHon, 1974; Wilhelms, 1984], and represents ∼13% of the volume of the main Arsia Mons edifice on Mars [Smith et al., 2001]. The terminal basins of the Mangala Valles system extend from Lucus Planum northward to Amazonis Plantitia and Arcadia Planitia. Exposed units in these regions have characteristics consistent with volcanic origins (see discussion above), and flow of 2 × 105 km3 of lava into this broad region would result in total deposit thicknesses averaging 50 to 400 m for surface areas of 4 × 106 to 5 × 105 km2, respectively.
 The source of Mangala Valles is a component of the Tharsis-radial graben system interpreted as having formed in response to plume-related dike intrusion, with individual dikes predicted to have widths of ∼0.4 km, heights of ∼100 km, and lengths of up to 3000 km [Wilson and Head, 2002]. A dike with these dimensions will have a volume of ∼1.2 × 105 km3, which is of the same order of magnitude as the magma volume estimated above; the graben at the head of the Mangala Valles system is ∼1500 km long, corresponding to an estimated dike volume of ∼6 × 104 km3.
 The estimated effusion rate of 4 × 106 m3/s is far greater than the tens of cubic meters per second typical of modern terrestrial eruptions involving a wide range of lava compositions [e.g., Hon et al., 1994; Calvari et al., 2002; Kauahikaua et al., 2002], as well as the rates of less than 400 m3/s estimated for large historical eruptions at the Hawaiian Islands [e.g., Rowland and Walker, 1990]. A maximum effusion rate of ∼8.7 × 103 m3/s has been estimated for the Laki fissure eruption of 1783–1784 [Thordarson and Self, 1993]. However, the geological record suggests that past high-effusion volcanic events have taken place on the Earth, with eruption rates of ∼1 × 104 to 1 × 106 m3/s regarded as possible for individual flows of the Columbia River Basalt Group [Reidel, 1998]. The largest known effusive eruption of modern times appears to have occurred on Jupiter's moon, Io, at Pillan Patera, where peak effusion rates associated with the 1997 eruption of 31 km3 of lava over a 100-day period are estimated to have exceeded 1 × 104 m3/s [e.g., Davies et al., 2006]. Extreme rates of lava effusion have been cited as necessary to account for the sizes of large basalt flows on the Moon, with a rate of 1.3 × 106 m3/s estimated for emplacement of a single 400-km-long flow erupted from a fissure vent at Mare Imbrium [Schaber, 1973; Schaber et al., 1976]. Extrapolation of terrestrial and lunar effusion estimates suggests that rates up to and perhaps in excess of 1 × 106 m3/s should be associated with formation of flows with lengths of many hundreds of kilometers [Walker, 1973; Schaber et al., 1976]. Consistent with this perspective, rates of effusion equal to or exceeding 1 × 106 m3/s have been estimated for some Martian lava flows [Baloga and Pieri, 1986; Cattermole, 1987; Keszthelyi et al., 2006] and rates of up to ∼1 × 108 m3/s have been estimated for large individual volcanic flows on both the Moon and Mars [Zimbelman, 1998]. Rates of lava discharge of between 1 × 105 and 1 × 106 m3/s have been estimated for large flow fields on Venus [e.g., Lancaster et al., 1995], and a rate of up to 5 × 107 m3/s has been estimated for formation of Venusian outflow channel Kallistos Vallis [Baker et al., 1997].
 Formation of the Mangala Valles system by catastrophic aqueous floods is consistent with similarities in the nature of landforms of this system and those of systems including the Channeled Scabland. However, a volcanic interpretation of the Mangala Valles system represents a viable (but unproven) alternative hypothesis on the basis of (1) the location of the head of the system at a large structural feature known to have acted as a source for effusive volcanic eruptions; (2) the distribution of landforms with volcanic affinities along and beyond the Mangala Valles system, suggesting past conveyance and pooling of large volcanic flows by the system; and (3) the potential for lunar and Venusian landforms to act as both morphological and process-based analogs for formation and evolution of the system. Indeed, the large lunar and Venusian channels developed through voluminous discharge from the subsurface, and, on this basis alone, might be argued to be more complete analogs to Mangala Valles than terrestrial systems formed in association with ice-dammed lakes. Rudimentary estimates of lava volume and effusion rates associated with hypothesized mechanisms of formation of Mangala Valles are consistent with those made previously for large volcanic landforms of the solar system.
 The volcanic hypothesis for formation of the Mangala Valles channel system is not encumbered by present or past instability of liquid water at the surface of Mars, whereas this issue has the potential to limit the viability of some aqueous scenarios [e.g., see Carr, 1996; Hoffman, 2000]. For example, the mechanism of near-surface dike emplacement for release of water from a cryosphere-confined aquifer could require periods of water release that are on the order of several million years [McKenzie and Nimmo, 1999], lengths of time that are not consistent with stability considerations for low latitudes [e.g., Farmer and Doms, 1979; Rossbacher and Judson, 1981; Clifford and Hillel, 1983], nor with the absence of large basins that could have acted as long-term surface reservoirs for later catastrophic floods. Uncertainty remains regarding the manner in which vast aquifers could have accumulated in the southern highlands of Mars, perched several kilometers above the floor of the northern lowlands in highly permeable and regionally interconnected regolith units [see, e.g., Clifford and Parker, 2001]. Deliberation also continues regarding the capacity of such aquifers to have catastrophically supplied massive volumes of water to the surface at rates required for aqueous outflow-channel formation [e.g., Head et al., 2003; Manga, 2004].
 The aqueous and volcanic hypotheses each represent end-members of a spectrum of possible processes that could have played key roles in the development of Mangala Valles, with the aqueous hypothesis involving simple volcanic resurfacing of a channel system otherwise formed primarily by catastrophic aqueous flows, and with the volcanic hypothesis involving channel formation by lava flow. A hybrid hypothesis would involve formation of the Mangala Valles channels through incision by aqueous flows, followed by or in concert with substantial modification of the system through erosion by volcanic flows. Other processes (e.g., glacial or CO2-related) cannot be dismissed out of hand, either as end-member hypotheses or as processes of secondary importance in channel evolution. Certainly, aeolian and mass-wasting processes have greatly influenced the landscape of the Mangala Valles region since the time of channel formation.
 A volcanic interpretation of Mangala Valles relies upon assumptions that (1) present volcanic interpretations of large lunar and Venusian channels are correct and (2) similarities in the characteristics of Martian, lunar, and Venusian channels reflect similarities regarding processes of formation. It is possible that neither assumption is valid. Progress regarding assessment of the end-member volcanic hypothesis should follow both from continued investigation of Mangala Valles and from further numerical modeling of large volcanic channel systems; aspects of such modeling efforts would be facilitated by acquisition of high-resolution topographic data sets for the Moon and Venus. Importantly, the end-member volcanic hypothesis discussed in this paper is testable. For example, identification of an aggradational origin (involving alluvial or diluvial processes) for large streamlined islands of the system would validate these features as true braid bars rather than simple erosional residuals, and would negate the end-member volcanic hypothesis for formation of Mangala Valles.
 Similarities in the nature of the channel landforms of Mangala Valles and those of aqueous flood systems have suggested possible commonalities in mechanisms of channel formation. However, a volcanic origin for the Mangala Valles system also appears viable on the basis of the following factors: (1) the system heads at a structural feature that acted as a source of past effusive volcanic eruptions; (2) the nature of surface materials at Mangala Valles is suggestive of past flow of large volumes of lava through the channel system to its terminal basins; and (3) large lunar and Venusian channel systems have the potential to act as both morphological and process-based analogs for formation and evolution of the system. Crude estimates suggest that formation of the channel system through thermal mechanisms of lava incision could have taken place through eruption of ∼2 × 105 km3 of lava, a volume that under some conditions would involve an average eruption rate of ∼4 × 106 m3/s over a total event duration of almost 600 days (though likely taking place through multiple events over geological timescales). A smaller volume of lava would be required if mechanical processes of erosion were also substantially involved. The volcanic hypothesis for formation of Mangala Valles, and a hybrid hypothesis involving formation of Mangala Valles through aqueous processes followed by or in concert with substantial modification of the system by volcanic erosion and deposition, appears worthy of future consideration.
 The helpful reviews of Laszlo Keszthelyi, Devon Burr, and an anonymous reader are appreciated. The author acknowledges the use of Mars Orbiter Camera images processed by Malin Space Science Systems and Thermal Emission Imaging System images processed by Arizona State University.