Internal Structure and Early Thermal Evolution of Mars from Mars Global Surveyor Topography and Gravity

See allHide authors and affiliations

Science  10 Mar 2000:
Vol. 287, Issue 5459, pp. 1788-1793
DOI: 10.1126/science.287.5459.1788


Topography and gravity measured by the Mars Global Surveyor have enabled determination of the global crust and upper mantle structure of Mars. The planet displays two distinct crustal zones that do not correlate globally with the geologic dichotomy: a region of crust that thins progressively from south to north and encompasses much of the southern highlands and Tharsis province and a region of approximately uniform crustal thickness that includes the northern lowlands and Arabia Terra. The strength of the lithosphere beneath the ancient southern highlands suggests that the northern hemisphere was a locus of high heat flow early in martian history. The thickness of the elastic lithosphere increases with time of loading in the northern plains and Tharsis. The northern lowlands contain structures interpreted as large buried channels that are consistent with northward transport of water and sediment to the lowlands before the end of northern hemisphere resurfacing.

The interior structure of Mars contains information on bulk composition, differentiation and melting history, and mantle dynamics, all of which bear on the planet's thermal evolution. Global models of martian topography from the Mars Orbiter Laser Altimeter (MOLA) (1) and gravity from Doppler tracking (2) obtained during the mapping mission of the Mars Global Surveyor (MGS) spacecraft (3) have now achieved the coverage and quality required for the first reliable determination of Mars's crustal and upper mantle structure.

We analyzed updated models of topography (4) (Fig. 1A) and free-air gravity (5) (Fig. 1B) expanded to spherical harmonic degree and order 60, corresponding to a spatial resolution of ∼178 km. Global correlations of the fields are shown in Web figure 1 (6). To investigate the martian internal mass distribution, we calculated Bouguer gravity anomalies, for which the gravitational attraction of surface topography is subtracted from the free-air gravity field to reveal the subsurface mass distribution. Bouguer gravity anomalies have been calculated for Mars, but the uncertainties limited geophysical interpretation (7).

Figure 1

Global maps of (A) topography, (B) free-air gravity, and (C) crustal thickness (8) of Mars (Mercator projection). On all panels, the Tharsis province is centered near the equator in the longitude range 220° to 300°E and contains the east-west–trending Valles Marineris canyon system and the major volcanic shields Olympus Mons (18°N, 225°E), Alba Patera (42°N, 252°E), Ascraeus Mons (12°N, 248°E), Pavonis Mons (0°, 247°E), and Arsia Mons (9°S, 239°E). The Arabia Terra region is centered at 10°N, 10°E, the Elysium rise is at 30°N, 150°E, the Tempe Terra region lies at 40°N, 290°E, the Syria Planum region is centered at 25°S, 270°E, and the Terra Cimmeria region is centered at 60°S, 180°E. Major impact basins include Hellas (45°S, 70°E), Argyre (50°S, 320°E), Isidis (12°N, 88°E), and Utopia (45°N, 110°E). The hemispheric dichotomy boundary is shown as a red line in (C), solid where distinctively expressed and dashed where estimated. This analysis uses an areocentric coordinate convention with east longitude positive. One degree of latitude on Mars equals ∼59 km.

The simplest interpretation of the Bouguer gravity is that subsurface mass variations are a consequence of variations in the thickness of a crust of uniform density. This interpretation provides a first-order indication of Mars's internal structure but does not account for density variations within the crust or upper mantle. A model based on this premise provides an approximate lower limit on mean crustal thickness and volume (8). In this model, a crust of density 2900 kg m−3 has an average thickness of 50 km and represents 4.4% of Mars's volume (Fig. 1C). The minimum thickness is 3 km beneath the Isidis basin, and the maximum thickness of 92 km occurs in the Syria Planum region of Thaumasia.

The major features in the free-air gravity field (Fig. 1B) include a broad but geometrically complex anomaly associated with the Tharsis province, positive gravity anomalies over large volcanoes, localized mass concentrations centered on large impact basins, and mass deficits (negative anomalies) associated with the Valles Marineris canyon system. Aside from these structures, the gravitational field is relatively smooth (9). As a consequence, the crustal thickness map, to first order, correlates with the principal features of surface topography (Fig. 1A). The primary topographic feature on Mars is the global south-to-north slope (10, 11). Cross sections of crustal structure (Fig. 2) indicate a progressive thinning of the crust from high southern latitudes toward the north. The northward slope and corresponding decrease in crustal thickness continue through the Tharsis province (Fig. 2B), but the crustal structure changes to a zone of approximately uniform thickness (∼35 to 40 km, except for buried basins) under the northern lowlands and the Arabia Terra region of the southern highlands (Fig. 2A). On the basis of their similarity in crustal structure, we infer that Arabia Terra is related to the northern lowlands and may represent exposed basement of the resurfaced northern plains.

Figure 2

Circum-Mars profiles of crustal thickness along longitude lines of (A) 0° to 180°E and (B) 70° to 250°E. Light gray represents crust, and dark gray represents mantle. In the figures, the south pole is at both ends of the plot, the north pole is at the center, and the lower longitude profiles (0°E and 70°E) are on the left sides of the plots. Apparent crustal thickening beneath the north and south polar regions is an artifact of the assumption that layered terrains and ice caps are composed of material with the same density as the crust rather than less dense ice plus dust. The arrows in (A) show the location of the hemispheric dichotomy boundary. The vertical exaggeration is 30:1.

The transition between the major crustal provinces at Mars's hemispheric dichotomy boundary separates the older, heavily cratered southern hemisphere highlands from the relatively smooth, resurfaced, and younger northern hemisphere lowlands (12). The boundary zone (Fig. 1A) has a width of ∼700 km and is characterized by complex geology (12). In many areas, the dichotomy boundary also exhibits a regional elevation change, sometimes described as a scarp, with relief of up to 4 km over distances of 300 to 1300 km (13). In the longitude range 110° to 190°E, the transition from the northward-thinning to uniform-thickness regions of crustal structure coincides with the dichotomy boundary (Figs. 1C, 2, and 3). Most of the dichotomy boundary, however, does not correspond to the crustal thickness transition, and it is apparent that the geologic expression of the dichotomy boundary is not a fundamental feature of Mars's internal structure.

Figure 3

Polar stereographic projection of martian crustal thickness. The figure encompasses latitudes from 20°S to 90°N. The contours are at intervals of 5 km.

One hypothesis for the formation of the northern lowlands invoked a large impact or impacts in the northern hemisphere (14,15). By this hypothesis, the scarp-defining part of the boundary between the hemispheres is interpreted as remnant basin rim material (15). Impact origin hypotheses for the northern lowlands predict a global association of crustal thickness and the dichotomy boundary, as is seen in lunar basins (16); topography revealing the shape or shapes of the basins; and the presence of gravity signatures representing basins of appropriate size and distribution to explain the northern lowlands. Three specific features of the results from gravity and topography are inconsistent with an impact hypothesis. First, the global distribution of inferred crustal thickness does not correlate with the dichotomy boundary (Figs. 1C and 3). This result indicates that any crustal thickness signature of an impact origin for the northern lowlands, if it once existed, has since been erased by subsequent processes. Second, the topography (11) and crustal thickness (Fig. 3) make clear the complex outline of the northern basin (11), and impact events have not been otherwise observed to produce such structures. Third, there is no gravity signature or signatures of the appropriate size and distribution to account for the northern lowlands by the formation and subsequent modification of large impact structures (9). An unfilled early basin would likely be compensated, whereas a filled basin, more consistent with the resurfacing age of the northern lowlands, would be expected to have a positive mass anomaly. Although the Utopia basin fits this description and other smaller positive gravity anomalies can be found in the northern plains (9), this pattern is not seen for the northern lowlands as a whole. An impact origin of the northern lowlands would require that the impact depression is fully compensated and lacks a substantial thickness of fill (that would otherwise produce a mass anomaly); the impact or impacts must have been highly oblique to yield a complex depression; and the event or events must have occurred so early in martian history (i.e., earlier than the time of formation of the Utopia basin) that any crustal thickness signature has been modified by subsequent processes. We view this combination of requirements as unlikely.

The Tharsis province is characterized by regional variability of crustal thickness superposed on the south-to-north crustal thickness trend. The Solis Planum (southern Tharsis) and Tempe Terra (northeast Tharsis) regions represent zones of thicker crust than their surroundings, although the former area is underlain by a crustal root whereas the latter is a topographic dome without a root. The spatial variability of crustal thickness in Tharsis supports the notion of volcanism as a contributor to the elevation of the province (17). The Valles Marineris canyon system (Fig. 1C and Web figure 2A) (6) displays crustal thinning beneath the trough axis, which may be a consequence of lithospheric stretching caused by extensional stresses associated with the Tharsis rise (18).

Derived crustal thicknesses reveal differences in style associated with volcanic constructs. The ancient Alba Patera structure that dominates northern Tharsis (Fig. 2B) is a zone of thicker crust and displays a broad morphologic similarity to the Elysium rise (Web figure 2B) (6). In contrast, younger shields such as Olympus (Fig. 1C) and Elysium (Web figure 2B) (6) Montes, as well as the Tharsis Montes, display apparent crustal thinning (8). The differences in crustal structure among these volcanic centers may reflect some combination of variations in magma composition, eruptive properties, and lithospheric structure through time.

On Mars, as for the moon (16), all major impact basins including Hellas, Argyre, Utopia, and Isidis exhibit varying degrees of crustal thinning. Thinned crust beneath martian basins may be the consequence of excavation and mantle rebound associated with the impact process (16). The large Hellas and Utopia structures are surrounded by annuli of thickened crust (Web figure 2, C and D) (6), as are major basins on the moon (16) and the Chicxulub structure on Earth (19).

Average crustal thicknesses of 100 to 250 km, larger than that in Fig. 1C, have been suggested on the basis of the geochemistry of martian meteorites (20). These suggestions raise the question as to whether or not a much thicker crust is as plausible as our 50-km mean model. To investigate this possibility, we estimated the sustainability over time of the long-wavelength component of the crust. A viscoelastic relaxation model (21) indicates that a pole-to-pole (spherical harmonic degree 1) crustal thickness anomaly in a crust with mean thickness of 50 km could be maintained for >108years, the approximate time scale of high early heat flow indicated by thermal models, provided that the time-averaged viscosity of the lower crust is >1020 Pa·s. Larger crustal thicknesses require a still higher viscosity in the lower crust. The 100-km lower limit of crustal thickness suggested by martian meteorite chemistry would require a value of viscosity for the lower crust, 1022Pa·s, that exceeds the present-day value for Earth's upper mantle. We consider such a scenario unlikely, particularly because our estimation represents a time average that must factor in the thermal state of an early, warmer martian interior. On the basis of thermal history models (22, 23), it is conceivable that early in martian history the lower crust could have been characterized by a viscosity as low as 1015 Pa·s. Because the minimum crustal thickness is only ∼3 km in the current model, a thinner crust would require a higher crustal density in the northern hemisphere than assumed here. The crustal relaxation model indicates that relatively rapid cooling of a thin crust is required to maintain the degree-1 component of crustal structure and favors a mean crustal thickness of ≤50 km.

Joint analysis of gravity and topography permits the determination of the effective elastic thickness (T e) of the lithosphere (i.e., the outer shell of long-term strength) on a regional basis for most of the planet. To first order,T e can be regarded as the depth to an isotherm (∼650°C) beneath which the martian interior is too weak to support stresses over geologically long (∼108 year) intervals. We applied spatio-spectral localization and multitaper coherence methods and compared observed values with those predicted by models for loading of an elastic lithosphere (24). Results indicate variability in lithosphere structure among major physiographic provinces (Fig. 4). For example, the southern highlands, including the Hellas impact basin, indicate a best fitT e ranging from 0 to 20 km, whereas the Alba Patera volcano displays T e = 50 km. Olympus Mons and the other Tharsis volcanoes are characterized byT e >100 km. Within the southern highlands, the Terra Cimmeria region of prominent linear magnetic anomalies revealed by the MGS magnetometer experiment (25,26) shows T e = 20 km. This value is comparable to the 30-km depth extent of coherent magnetization [i.e., the depth of the Curie temperature at the time of magnetization (27)] estimated from observed magnetic anomaly magnitudes (25, 26) for specific magnetizations comparable to that of Earth's upper oceanic crust (27), the primary contributor to marine magnetic anomalies.

Figure 4

Effective elastic thicknessT e and corresponding heat flowq obtained from lithospheric thickness inversions (24, 30). Shown for comparison (dashed line) is the depth to the Curie temperature indicated from analysis of MGS magnetic data in Terra Cimmeria (26,27).

In the determination of lithosphere thickness, the flat northern plains of Mars (28) represent a special case because deposition and erosion have smoothed topography to such an extent that inversions based on isostatic response methods yield incorrect results (29) unless bottom loading and erosion are also considered. To obtain a reliable estimate of T e for the northern lowlands, we used a forward model for the lithospheric response to filling of the Utopia basin (30). For basin-fill density comparable to the crustal density assumed in the model of Fig. 1C (volcanic fill), or for fill densities appropriate to sedimentary loading of a locally denser crust (30), the model yields a lithospheric thickness of ∼100 km in the vicinity of Utopia.

If lithospheric thickness is interpreted as a proxy for time since crustal formation, then the progressively greater lithosphere thicknesses from the southern highlands to the northern plains to Tharsis would be in conflict with the younger surface ages in the north indicated by crater densities and stratigraphic relations (31). The lithospheric thickness results, however, are explained if, as on Earth (32), elastic thickness reflects the thermal state of the martian lithosphere at the time of loading. The southern highlands are oldest, indicative of crustal stabilization and loading earliest in martian history, followed by the northern lowlands, with an elastic thickness that reflects the thermal state during the period of northern hemisphere resurfacing. For instance, for a conductively cooled lithosphere originally emplaced at temperatures close to that of the convecting mantle, theT e value derived for Utopia suggests that filling of the topographic depression occurred ∼500 million years subsequent to crustal formation. The Tharsis province displays the thickest lithosphere, implying the most recent loading, consistent with young surface ages (31). The collective results provide constraints on the thermal state of Mars at various stages in its evolution.

Estimates of heat flow q at the time of loading derived from estimates of elastic lithosphere thickness and considerations of temperature-dependent flow laws for crust and mantle material under a range of representative strain rates (Fig. 4 and Web table 1) (6,33) may be compared with the predictions of thermal history models. Isoviscous models (22) characterized by a lid with a constant basal temperature produce the slowest cooling during early martian evolution. In contrast, models with strongly temperature-dependent viscosity described by scaling laws suggesting that Mars convected in a stagnant lid regime (23) show the most rapid falloff in early heat flow.

The fundamental differences in lithospheric structure between the southern highlands, northern lowlands, and Tharsis rise provide quantitative constraints on the cooling of Mars over time. The lithospheric thicknesses determined for large areas of the southern lithosphere indicate a heat flow on early Mars that is considerably less than the global averages suggested by most thermal evolution models. If this discrepancy is correct, then a disproportionally large fraction of the heat loss must have been localized in another part of Mars, in a region or regions where lithosphere thicknesses at the earliest times cannot be determined because values in such regions reflect the thermal state at times of subsequent geological activity. High heat flow, distributed approximately uniformly over a broad region, would have cooled the mantle and led to lower melt production and to a lower crustal production rate (i.e., thinner crust) later in martian history (34). On the basis of crustal structure and relative surface age (31), the most plausible locus of high heat flow on early Mars is the northern lowlands.

It has been previously suggested by Sleep (35) that the earliest phase of martian history was one of crustal formation by a martian analog of plate tectonics and the formation of the northern hemisphere crust during the final stages of that era. The identification of linear magnetic anomalies over a part of the southern hemisphere, suggesting alternating polarity (or at least discontinuous magnetization amplitude), has been taken as support for a crustal spreading origin for the southern highland crust during a time when a strong global magnetic field underwent dipole polarity reversals (26). The cessation of an early plate tectonic heat flow regime may have resulted in an episode of mantle heating that suppressed core heat loss and led to the demise of the core dynamo (36).

Although the crustal structure within the zone of prominent magnetic lineations is not distinctive from other parts of the southern highlands, it is the northern lowlands region proposed as the locus of most recent plate tectonic activity (35) that displays a distinctive crustal structure. The 40-km crustal thickness contour corresponds closely to Sleep's proposed limits of late-stage mobile plates (35). On Earth, passive upwelling of warm mantle material produces a crust consistency about 6 km thick. A 40-km-thick northern lowlands crust could be the product of crustal spreading if the mantle potential temperature were somewhat greater than that typical of modern mid-ocean ridges (34). This argument, and the comparison of heat flow estimates (Fig. 4) with thermal history models, are consistent with early martian plate tectonics, although the hypothesis is far from conclusively demonstrated.

The northern hemisphere lowlands of Mars, the flattest known large-scale surface in the solar system (28), were shaped by the resurfacing of heavily cratered terrain. The gravity field shows mascon anomalies that represent evidence for impact basins buried beneath the fill, of which Utopia is the largest (9). Geological reasoning [e.g., (12, 31)] has favored volcanism as the principal mechanism of resurfacing of the northern plains. Our model for the gravity anomaly over the Utopia basin is permissive of either volcanic or sedimentary fill (30). In other parts of the northern lowlands, however, there is clear evidence that sediments contributed substantially to the resurfacing.

Outflow channel termini in the Chryse region have been shown to occupy a nearly constant geopotential elevation (37) that corresponds to a geological boundary [Contact 2 in (38)] proposed on the basis of Viking Orbiter image interpretation to represent the former shoreline of a large standing body of water. A prominent negative gravity lineation over Valles Marineris reflects a mass deficit of the trough depression (Web figure 2A) (6). This lineation follows the outflow channels first east and then north into Chryse beyond Contact 2 and into the northern lowlands, bifurcating toward the northwest and, more prominently, the northeast, and in each case following the highland-lowland boundary. The northwest branch of the gravity anomaly intensifies and broadens at the mouth of Kasei Vallis (25°N, 310°E). It then extends northward of 75°N and includes a major branch to the east (39). Within the northern plains, these negative gravity anomalies lack topographic expression (Fig. 5B), indicating that the source of the mass deficit is in the subsurface.

Figure 5

High-resolution gravity (43) draped over a shaded relief map of part of the northern hemisphere showing evidence for buried channels: (A) observed free-air gravity and (B) predicted gravity due to attraction of surface topography. The fact that (A) and (B) differ indicates that the gravity lows in the northern lowlands shown in (A) are due to subsurface mass deficits. The plot is a Hammer projection from 30°S to 90°N latitude, 270° to 360°E longitude. North is at the top of each panel. Valles Marineris is the prominent feature at the lower left.

Unless the martian crust has thickened in quasi-linear zones that emanate from outflow channels, these mass deficiencies must correspond to regions within the crust that are of lower density than their surroundings. Although the linear gravity anomalies do follow, in part, the hemispheric dichotomy boundary, they do not follow transitions in crustal structure (Fig. 1C) and thus are unlikely to be tectonic in origin. We interpret these features as the extension of outflow channels that are now buried beneath the northern plains. The hypothesis that the negative gravity lineations are due to buried channels is testable in terms of such parameters as the volume of material removed, channel bottom gradient, depositional basins for outflow, and the trade-off between infill density and channel depth.

By terrestrial analogy (40), the geometric characteristics and flow properties of these structures are consistent with either subaerial or subaqueous flow. The lineations are about 200 km in width and thousands of kilometers in length. If these features mark buried channels, then their scale indicates that they would have transported large quantities of water to mid to high northern latitudes in Mars's ancient past. Such channels also would have carried a large sediment load that plausibly contributed to the resurfacing of the northern lowlands. The gravity signature implies an average subsurface relief of 1.5 to 4.5 km if the channels are sediment-filled or about 1 to 3 km if the structures contain an ice-sediment mix (39). If large lakes or an ocean ever developed in the northern hemisphere of Mars (37, 38), then the geometric characteristics of the buried channels imply that bodies of water could have accumulated rapidly. In addition, these features indicate that substantial amounts of water and sediment were transported to the northern lowlands before the cessation of resurfacing.

  • * To whom correspondence should be addressed. E-mail: zuber{at}

  • Present address: Lunar and Planetary Institute, Houston, TX 77058. USA.


View Abstract

Navigate This Article