An enduring enigma of Neoproterozoic earth history is the intimate association of glacial diamictites with typical warm-water carbonates. Among the many hypothesized explanations for this paleoclimatic dichotomy are reduced solar luminosity, CO2 drawdown, high orbital obliquity, true polar wander, and reinterpretation of the diamictites as mega-impact ejecta. The Otavi carbonate platform bordering the Congo craton in Namibia contains two discrete intervals of diamictite and associated glaciomarine deposits, sandwiched by thick carbonates from which we have obtained detailed carbon isotopic records. Most of the pre-glacial carbonates are strongly enriched in ¹³C, signifying high proportional burial of organic carbon and consequent drawdown of atmospheric CO2. Directly beneath the younger glacial interval is a regionally extensive regressive parasequence, in which d¹³C values descend monotonically from +7‰ to -3‰. Above the glacial interval, d¹³C values continue their descent stratigraphically upward to a nadir of -6‰, before rising gradually back to positive values over a stratigraphic thickness of 400 m. A more condensed negative excursion is observed above the older glacial interval but its pre-glacial record is erosionally truncated. The magnitude and timing of the isotopic variations, in particular the newly-observed deflection beneath the younger diamictite, permits critical assessment of existing hypotheses. We conclude that the Neoproterozoic world was devastated by global glaciations, which resulted in protracted collapse of surface-ocean productivity.top
Louis Agassiz' (1840) apocalyptic vision of ice ages so severe that continents were glaciated in the tropics and organic activity was stilled on land and sea was overwrought for the Quaternary, but what about the extraordinary events of the late Neoproterozoic (750-543 Ma)? Brian Harland (1964) first drew attention to the global distribution of glacial deposits of this age and their often close association (Fig. 1B) with presumed warm-water carbonates (Fairchild, 1993). Two or more discrete glacial intervals occur in many sections, thus the extent of any individual glaciation depends on correlation. However, there is robust paleomagnetic evidence for equatorial glaciomarine sedimentation at ~600 Ma in South Australia (Schmidt and Williams, 1995; Sohl, 1997) and at ~720 Ma in northwest Canada (Park, 1997). Quaternary ice sheets never went below 40°N along any coast.
Spanning the late Neoproterozoic ice ages are huge positive-to-negative carbon isotopic shifts, unique in the past two billion years (Fig 2). Carbonates become very strongly enriched in ¹³C prior to each ice age and then descend to extremely depleted values typically found in lithologically distinctive transgressive carbonate units, which "cap" the glacial deposits and may extend far beyond them, aiding regional-scale correlation (Knoll et al., 1986; Kaufman and Knoll, 1995; Kennedy, 1996). Such large shifts signify a combination of changes in (1) fractional organic to total carbon burial rates (Scholle and Arthur, 1980), (2) biological productivity in the surface ocean (Broecker, 1982), (3) vertical circulation rate for the whole ocean (Brass et al., 1982; Knoll et al., 1996), (4) isotopic composition of carbon sources (Beck et al., 1995; Dickens et al., 1997), and (5) isotopic fractionation related to carbonate ion concentration (Spero et al., 1997). All are linked to global climate change. Here we illustrate and discuss the stratigraphic relations with new high-resolution data spanning two (Sturtian, 750-700 Ma) glacial intervals on a carbonate-dominated platform margin exposed in northern Namibia. Such observations gain sway only in the arena of ideas, so we begin with a partial review of current hypotheses.top
A welter of hypotheses contend to shed light on the Neoproterozoic climate puzzle. Stars in their main evolutionary sequence become more luminous as their helium cores grow more massive: our sun would have had only about 93% of its present luminosity at 700 Ma. However, the absence of middle Proterozoic (0.8-2.2 Ga) glacial deposits (Fig. 2) shows that other factors must be involved. Higher concentrations of greenhouse gases, notably CO2, may have offset the dimmed early sun most of the time, but this dependence could have left the earth susceptible to severe chilling in event of an unbalanced CO2 drawdown (Kasting, 1992). Coupled energy-balance models suggest that such scenarios are sensitive to the distribution of land masses, but that sub-zero temperatures at the equator were possible with a 570-Ma paleogeography (Dalziel, 1992) and reduced solar luminosity if CO2 concentration fell to 100 ppm, or about 28% of the present level (Crowley and Baum, 1993). Considering qualitatively albedo effects and other feedbacks, Kirschvink (1992a) suggested that the earth alternated between greenhouse and icehouse climate states, the oceans occasionally freezing over almost completely-a true "snowball earth"-with profound implications for ocean circulation and biological activity.
A completely different approach is to redefine the problem from global glaciation to preferential low-latitude glaciation, consistent with paleomagnetic data. Williams (1975, 1993) proposed that the earth's axial tilt (the obliquity of the ecliptic) exceeded 54° until the end of the Proterozoic, after which it stabilized at much lower values (23.5° today). Accordingly, the earth's climatic zonation would have been reversed in the Proterozoic, meaning lower insolation at low-latitudes than at the poles. On the other hand, the seasonal cycle would have been greatly amplified, which would not have been favorable for glaciation. The association of carbonates and glacial deposits is not explained by this model, nor is the middle Proterozoic glacial hiatus-once low obliquities were established, high obliquity should not recur (Laskar et al., 1993). The model has the merit of a simple falsifying test, a high-latitude glaciation. None are known.
Another approach that avoids global glaciation, but which could explain rapid transitions from normal low-latitude to high-latitude conditions, is inertial-interchange true polar wander (IITPW). True polar wander refers to a rotation of the entire solid shell (crust and mantle) of the earth relative to the spin axis, in this case related to an interchange of the major and intermediate axes of the earth's non-hydrostatic moment of inertia tensor. The requisite condition arises if the geoid figure closely approximates a prolate ellipsoid: the prolate axis will stabilize in the equatorial plane to conserve angular momentum (Goldreich and Toomre, 1963), but if the other two axes are nearly equal, small changes in mass distribution may cause the globe to rotate 90° around the prolate axis in a few million years (Deutsch, 1963; Fisher, 1974). As a result, continents located near the poles of the prolate axis will undergo large rotations and continents located far from the axis will migrate through 90° of latitude at velocities far exceeding those due to plate tectonics. IITPW has been invoked in the interpretation of Cambrian paleomagnetic data (Kirschvink, 1992b; Kirschvink et al., 1997) and, if the inferred prolate geoid was inherited (viz. Anderson, 1982) from the former supercontinent Rodinia (1.05-0.72 Ga), then IITPW may have occurred several times in the late Neoproterozoic (Evans, 1997). This model attributes the carbonate-glacial association to rapid changes in paleolatitude: it does not explain low paleomagnetic inclinations obtained from the glacial deposits themselves (Schmidt and Williams, 1995; Park, 1997; Sohl, 1997). No changes in global climate are required in the model but they would surely occur as ocean circulation patterns change in response to migration of the earth's rotation axis. Continents moving in or out of the polar regions would experience sea level falls or rises, respectively, because of their migration with respect to the earth's rotational bulge (Mound and Mitrovica, 1998): the sea level changes would mimic those of glacio-eustasy.
Recently, Rampino (1994) has attempted to derail the entire debate by suggesting that the purported glacial deposits are actually impact ejecta. His argument is that ballistic debris-flows resulting from large impacts exhibit many features conventionally identified with glacial and glaciomarine deposits, including faceted and striated clasts and grooved bedrock pavements. Rampino (1994) does not discuss C isotopic anomalies but negative excursions do accompany impact-induced mass extinctions (Hsü and McKenzie, 1985; Zachos et al., 1989). In the absence of terrestrial biomass burning (Ivany and Salawitch, 1993), cessation of surface ocean productivity (Strangelove ocean) would cause rapid (<10³ yrs) loss of the ocean's isotopic gradient (~2‰ in the present ocean), meaning that d¹³C in the surface ocean would fall to whole-ocean values (Kump, 1991). Only if productivity ceased for hundreds of thousands of years would d¹³C approach the ocean input value of -5‰ (Kump, 1991). Barring a rapid succession of large impacts, the isotopic excursions should strictly post-date the diamictites, providing a test for the impact hypothesis. The new data from Namibia may serve to critically test this and other contending hypotheses.top
In late Neoproterozoic time, the Congo craton was a low-lying platform the size of the conterminous United States. It was blanketed by carbonates and shales containing regionally mappable diamictite horizons of glacial origin (Hambrey and Harland, 1981). Two discrete glaciogenic intervals are contained in carbonates of the Otavi Group (Fig. 1A)(Hoffmann and Prave, 1996), which drape the southern promontory of the craton in northern Namibia (Fig. 3). We consider the Otavi Group to be entirely pre-Vendian in age (Fig. 2), based on chemo-, chrono- and biostratigraphic arguments. Paleomagnetic data from the eastern part of the Congo craton (Meert et al., 1995; Meert and Van der Voo, 1996) imply that the Otavi Group was at ~12°S paleolatitude at 743 ± 30 Ma and ~39°S at 547 ± 4 Ma (compare with age constraints in Fig. 1A).
The Otavi Group is exposed in fold and fault belts along the southern and western borders of the cratonic promontory (Hedberg, 1979; Frets, 1969). The contractional deformation dates from the amalgamation of west Gondwanaland at 0.6-0.5 Ga (Stanistreet et al., 1991). The southern belt coincides with the edge of the Otavi platform, which is flanked by correlative slope and deep-sea fan deposits draped over extended Congo crust (Henry et al., 1990). A long-lived basement high, the Huab ridge (Porada et al., 1983), separates the platform margin from an intra-shelf basin to the north. The ridge was intermittently active, shedding clastic wedges bilaterally into the flanking carbonates during early and middle Otavi times. The parautochthonous fringe of the western belt provides a transverse profile of the intra-shelf basin and the Huab ridge, which along with the western part of the southern belt has been the locus of our work (Fig. 3). Regional subsidence patterns suggest that successive rifting of the western and southern margins ceased in early and middle Otavi times, respectively.
The area is advantageous because thick carbonates, representing a wide range of primary facies, are in direct contact with the glacial deposits. Therefore, detailed isotopic records of the comings and goings of the ice ages can be obtained, provided that the contacts do not represent significant hiatuses. Superb exposure permits sequence boundaries (subaerial hiatuses) to be systematically mapped out on the dissected escarpment bordering the Namib Desert. Our work indicates that there is no significant hiatus between the glacial deposits and their overlying "cap carbonate" sequences, and that a strong isotopic harbinger is preserved in otherwise unexceptional carbonates beneath the Ghaub diamictites (Fig. 1A). We recognize no primary carbonates within either glacial interval. On account of their age, the Otavi carbonates lack skeletal components and are not bioturbated. Consequently, lamination and bedforms are finely preserved and form the basis for differentiating primary lithofacies (Fig. 5). "Rhythmites" are parallel-laminated micrites, dolomicrites or marlstones deposited below storm wave base. "Ribbon rocks" are wavy and partly current-rippled carbonate siltstones deposited above storm wave base. "Stromatolites" are colonial columnar microbialites forming sublittoral shoal complexes (excepting those in the Rasthof cap carbonate which appear to have formed in deep water). "Grainstones" are crossbedded intraclastic and oolitic carbonate sandstones deposited in the surf zone and in tidal channels. "Microbialaminites" are undulose stromatolites with tepee structures, tepee breccias and flat-pebble conglomerates indicative of tidal-flat deposition. Sequence boundaries are mappable subaerial exposure surfaces and are typically brecciated, silicified and/or ferruginized.top
Both the Chuos and Ghaub glaciations were preceded by carbonate ramps or platforms that built repeatedly to sea level. The Chuos glaciation succeeded the Ombombo Subgroup, a bilateral pair of distally-steepened ramps on which cherty dolomites and subordinate limestones were deposited. The ramps flanked the Huab ridge, which shed tongues of basement-derived clastics and reworked dolomite-chert breccias into the middle and upper Ombombo Subgroup (Fig. 1A). North of the ridge, the Ombombo strata were tilted 1.5 degrees basinward prior to the Chuos glaciation, resulting in their progressive truncation toward the ridge (Fig. 5). Paleovalleys, with up to 180 m of local relief, occur directly beneath the glacial diamictites. Thus, a significant hiatus exists between the Ombombo carbonates and the Chuos glaciation in the vicinity of the Huab ridge.
The situation differs for the Ghaub glaciation. The preceding upper Abenab Subgroup comprises two sequences, the Gruis and Ombaatjie formations of Hoffmann and Prave (1996), which Prave (1996) separates by an unconformity. The older Gruis Formation was deposited on bilateral ramps flanking the Huab ridge. Proximal tongues of basement-derived clastics pass upward and outward into pale peritidal dolomites with conspicuous tepee horizons. The southern ramp was truncated by a major syndepositional fault (Rockeys fault, Fig. 3), south of which the sequence is represented by a low-stand fan of deepwater clastics and redeposited carbonates. The younger Ombaatjie Formation is a stack of five or more shoaling-upward parasequences (Fig. 5), noteably composed of black limestones and variably dolomitic grainstones. Parasequence correlation, including chemostratigraphic mapping of the topmost parasequence (Fig. 6), indicates that the surface representing the Ghaub glaciation has remarkably little relief on the platform north of the Huab ridge (<20 m relief over a north-south distance of 70 km). On the southern foreslope, however, the Abenab Subgroup is deeply dissected beneath the thickest sections of Ghaub diamictite (Fig. 3).top
The glacial deposits are characterized by diamictites, sheets of unstratified wackestone carrying a variety of outsize matrix-supported clasts. The ice sheets that fed the diamictites crossed extensive carbonate platforms: basement rocks were encountered only on the Huab ridge and a similar ridge in the Baynes Mountains northwest of the intra-shelf basin. The area of basement exposure contracted between Chuos and Ghaub time (Fig. 3) and the overall composition of the respective diamictites changes accordingly. The Ghaub diamictites contain limestone and dolomite debris, much of it recognizeably derived from the Ombaatjie Formation. Clast and matrix compositions covary but, within each formation, separate diamictite sheets differ in composition, depending on local bedrock incision.
The Chuos diamictites are associated with fluvial outwash facies and were probably deposited from land-based glaciers. The Chuos Formation as a whole is characteristically ferruginous but iron mobilization during orogenesis obscures its origin. The Ghaub Formation consists of glaciomarine slope deposits south of the platform margin, which first developed after the Chuos glaciation. Both glaciations are well represented in the intra-shelf basin axis, but the Ghaub Formation becomes thin and discontinuous toward the Huab ridge (Fig. 3). The relics of Ghaub diamictite occur consistently above the isotopically distinct parasequence at the top of the Ombaatjie Formation and below the basal "tube" stromatolite of the Maieberg cap-carbonate sequence (Fig. 6). This relation is critical to the timing of the isotopic shift with respect to the glaciation.
The criteria for a glacial origin challenged by Rampino (1994), striated and faceted clasts and grooved bedrock pavements, are rarely observed in carbonates because surfaces are karstic. Aside from stratigraphic relations, our best evidence for glacial transport is the widespread occurrence of dropstones with pierced and draped laminations (Fig. 1C). These are extremely abundant at the top of a thick succession of massive Ghaub diamictites and graded subaqueous debris-flows on the southern foreslope.
Both glacial intervals have highly complex and variable internal stratigraphies, but their external contacts are remarkably similar and consistent. On the platform, both their basal contacts show intense brecciation of the underlying carbonates, variably silicified and ferruginized, typical of subaerial exposure. Both their upper contacts are knife-sharp and lack lag deposits or any evidence of subaerial exposure or hiatus. Where the glacial deposits are subaerial, the upper contact is a smooth flooding surface; where they are subaqueous, there is simply an abrupt cessation of ice-rafted material.top
The two "cap-carbonate" sequences-the Rasthof and Maieberg formations of Hoffmann and Prave (1996)-are single depositional sequences of distinct stratigraphic position, lithology and thickness. They directly overlie glacial deposits or their equivalent sequence boundaries (Fig. 3). They contain highly unusual lithofacies (Fig. 4), and their average thickness, 200-300 m, is an order of magnitude greater than that between any other successive exposure surfaces in the platform facies of the Otavi Group (Fig. 5). Thus, anomalous accommodation space must be created for them. Post-glacial "cap dolostones" elsewhere are relatively thin (Fairchild, 1993; Kennedy, 1996), but they may only represent transgressive systems tracts, not complete depositional sequences (Mitchum et al., 1977). On the platform, where the depositional surface remained close to sea level both before the glaciation and after the cap carbonate sequence was deposited, the accumulated sediment must equal the sum of the tectonic subsidence, net glacial erosion and the effect of sediment loading. Compaction effects are negligible in the Otavi Group carbonates and net erosion during the Ghaub glaciation was only 10-20 m. Assuming a tectonic subsidence rate of ~15 m/m.y., appropriate for early stages of thermal subsidence of moderately stretched lithosphere and equivalent to a carbonate accumulation rate of ~50 m/m.y. (Bond et al., 1989), the time needed to create sufficient accommodation space for the Maieberg cap-carbonate sequence would have been 4-5 m.y.
The Rasthof sequence is composed of dark to medium grey dolomite and limestone (0.03-0.3 wt%C), contrasted with pale pink limestone and dolomite (0.001-0.02 wt%C) in the Maieberg sequence. The complete Rasthof sequence in the intra-shelf basin consists of essentially three stratigraphic units (Fig. 6). At the base are <20 m of finely and smoothly laminated micrite, showing m-scale dolomite-limestone rhythms and increasing southward in thickness due to the addition of cm-scale calcareous turbidites. Transgressive deposits are virtually absent: the laminated unit lies directly on a knife-sharp flooding surface atop the Chuos glacial deposits or the subglacial erosion surface where glacial deposits are absent. The top of the flat-laminated unit is marked by a profligate development of irregular domal stromatolites, the peculiar geometries of which suggest fault-propagation folds, verging in all directions and developed contemporaneously with aggradation. This suggests that the stromatolites resulted from lateral expansion. Up-section, the stromatolite unit contains numerous reversibly coiled roll-ups (Fig. 4A), which formed while the sediment was cohesive (microbially-bound?) but flexible (lightly mineralized). Also present are irregular synsedimentary collapse-breccias, composed of microbialaminite blocks and void-filling thrombolite. The stromatolitic unit, >100 m thick, is devoid of desiccation structures and current bedforms. At its top, the lamination fades away and passes cryptically into regressive pale grey grainstones, which coarsen upward to a sequence boundary marked by tepees and chert-breccias. The entire Rasthof sequence is erosionally truncated over the Huab ridge by the Gruis Formation (Fig. 3). Where it reappears south of the ridge, it undergoes remarkable facies changes possibly related to crustal flexure in advance of faulting at the platform margin (Gawthorpe et al., 1997).
The transgressive systems tract of the Maieberg sequence over the platform is an unusual "tube" stromatolite (Fig.4B), reminiscent of the post-glacial Noonday dolomite in eastern California (Wright et al., 1978). This pale dolomite unit consists of microbialaminite and long slender columnar stromatolites, between which the "tubes" formed as sediment- and cement-filled synoptic depressions. The "tube" stromatolite thickens over the Huab ridge and develops into a mounded reef complex at the platform margin (Fig. 3). Talus blocks at the base of the mounds are infilled by m-scale silica fans, pseudomorphic after aragonitic sea-floor cement. The "tube" stromatolite is absent on the southern foreslope: there the basal Maieberg consists of pale dolomite rhythmite, sheeted with early isopachous carbonate cement. On the platform, the stromatolite is overlain transgressively by marly rhythmite, followed by a thick regressive sequence of pink limestone rhythmite, pale dolomite rhythmite, and dolomite grainstone that coarsens upward to an exposure surface made prominent by downward-penetrating box-work chert. Directly above the sequence boundary is a multitude of m-scale parasequences full of tepee structures. Preliminary isotopic correlations indicate that the thick shallow-water Elandshoek Formation on the platform (Fig. 5) is equivalent to deepwater rhythmites and slump breccias on the southern foreslope.top
The carbon-isotopic composition of Phanerozoic marine carbonates is complicated by skeletal components precipitated out of isotopic equilibrium with sea water, by terrestrial runoff of waters isotopically fractionated by land plants, and by pervasive bioturbation. Proterozoic carbonates are not plagued with these biological problems. The primary Otavi Group carbonates were pure lime muds, silts, sands and microbial micrites, most of which underwent early fabric-retentive dolomitization rendering them relatively impermeable. Oxygen-isotopic compositions and elemental ratios sensitive to diagenesis indicate minimal alteration of most samples, and even the brecciated and ferruginized samples deliberately collected at sequence boundaries seldom have anomalous carbon-isotopic values (for details of analytical procedures used in screening for diagenetic alteration, see Kaufman et al., 1991; Kaufman and Knoll, 1995). Most important, the carbon-isotopic trends for closely-spaced samples are stratigraphically coherent and regionally reproducible.
Secular variations in d¹³C for carbonates through the entire Otavi Group are summarized in a composite section from the intra-shelf basin north of the Huab ridge (Fig. 5), and details of the isotopic trends directly above and below the glacial intervals are shown in Figure 6. The thick shallow-water carbonates of the Ombombo Subgroup are strongly and uniformly enriched in ¹³C (>+5‰), consistent with pre-glacial Neoproterozoic carbonates world-wide (Kaufman et al., 1997). An influx of siliciclastics (from the Huab ridge) and erosional truncation at the top of the Ombombo Subgroup precludes isotopic study of the youngest pre-glacial sediments in the area of the composite section (Fig. 5). Stratigraphically higher Ombombo Subgroup carbonates occur 100 km to the north in the axial zone of the intra-shelf basin, where the Chuos and Ghaub diamictites are 1005 and 115 m thick, respectively, but these deeper water sections have yet to be studied in detail. Following the Chuos glaciation, the basal Rasthof carbonates are strongly depleted in ¹³C, beginning near -4‰ and rising steadily to ~0‰ near the top of the flat-laminated member. At the base of the overlying stromatolitic member, d¹³C values rise abruptly and ultimately stabilize near +5‰ for >150 m through the remainder of the Rasthof Formation. The shift from negative to positive values is widely correlated with the onset of deep-water stromatolite development in the intra-shelf basin.
Prior to the Ghaub glaciation, d¹³C values through most of the Ombaatjie Formation hover between +5 and +9‰ (Fig. 5), exemplifying the familiar pre-glacial enrichment in ¹³C. However, the ultimate parasequence, directly beneath the Maieberg cap carbonate or remnants of the Ghaub diamictite, displays a pronounced downward trend from +5‰ at the base to -3‰ at the top (Fig. 6), reaching -5‰ in some sections. The isotopic trend to more negative values with decreasing water depth is observed in the ultimate pre-glacial parasequence in many sections up to 50 km apart. The trend is opposite to the depth-dependent d¹³C gradient of the ocean and presumably represents a strong secular trend in advance of the Ghaub glaciation. There is little change in d¹³C values across the glacial interval: the basal Maieberg "tube" stromatolite begins near -3‰ and declines gradually up-section to a nadir of -6‰ near the inferred maximum flooding surface. While d¹³C values gradually rise above this level, they remain negative through the sequence boundary at the top of the Maieberg Formation. In fact, the crossover to positive d¹³C values does not occur for at least another 150 meters, comprised of repetitious peritidal parasequences with ubiquitous subaerial exposure surfaces (Fig. 5). Although the top of the Tsumeb Subgroup is truncated by the Mulden Group in our study area, a strong positive d¹³C excursion to values approaching +10‰ is preserved in the Hüttenberg Formation (Fig. 1A), 350 km to the east, and is possibly a harbinger of the first Vendian glaciation (ca 650-590 Ma) (Hoffmann, 1989; Kaufman et al., 1991; Germs, 1995).top
What do the observations tell us about the models? The persistent enrichment in ¹³C through ~400 m thickness of platformal carbonates beneath both glacial intervals (Fig. 5) is consistent with very high proportional rates of organic to total carbon burial, sustained for >8 million years assuming a rapid accumulation rate of 50 m/m.y. The consequent drawdown of atmospheric CO2 must have made the earth more susceptible to severe glaciation by reducing greenhouse warming (Kaufman et al., 1997). High organic carbon burial may have been occasioned by the opening of the Pacific and Iapetus ocean basins, attended by the creation of new continental-margin repositories for organic carbon (Des Marais et al, 1992; Young, 1995; Torsvik et al., 1995). However, the ages of glaciation and ocean opening are still too poorly known for this suggestion to be tested.
The continuous plunge in d¹³C from +7‰ to -3‰ in the topmost parasequence beneath the Ghaub glaciation (Fig. 6) is an important new observation. Bounded by exposure surfaces and deposited mainly above wave base, this 20-m-thick parasequence should represent >400 k.y. assuming ongoing thermal subsidence. It has been traced for over 50 km along strike and lacks structural or sedimentological evidence for accelerated tectonic subsidence. Far higher subsidence rates could be achieved by glacial loading, but that would require deposition of pure carbonate within ~100 km of the advancing ice front. The isotopic trend implies that the proportion of organic to total carbon increased in the sources or decreased in the sinks. Release of methane is an effective way of lowering oceanic d¹³C, but the duration of the observed isotopic shift is perhaps 50 times greater than the latest Paleocene shift which has been attributed to methane release (Dickens et al., 1997). Moreover, protracted release of methane, a strong greenhouse gas, would be an unlikely prelude to glaciation. Erosion of organic carbon on continental shelves exposed by glacio-eustatic sea-level fall could have lowered the oceanic d¹³C composition, provided that the organic matter was oxidized and not merely redeposited (Beck et al., 1995). However, a large sea-level fall would have exposed the Otavi platform, precluding deposition of the parasequence containing the isotopic shift we seek to explain.
The negative d¹³C shift might also be related to decreased biological productivity in the surface ocean, as observed at times of mass extinction (Holser et al., 1989; Magaritz, 1989). Complete cessation of the ocean's biological "pump" causes geologically instantaneous isotopic homogenization of the ocean, which would be manifest as a negative shift from about +2.5‰ to +1.0‰ in surface waters of the modern ocean (Kump, 1991). Logan et al. (1995) postulate an increased isotopic gradient for the Proterozoic ocean, which would magnify the isotopic shift, but our comparisons of deep-water precipitates with contemporaneous shallow-water carbonates has so far failed to confirm this postulate. If primary productivity is extinguished for over 105 years, the ocean's d¹³C composition would decline more dramatically and approach the riverine input value (Kump, 1991), which is estimated to be -3.8‰ at present (Delaney and Boyle, 1988). This scenario may have occurred at the Permo-Triassic boundary but not at other Phanerozoic mass extinction events (Magaritz, 1989), and is compatible with the isotopic trends straddling the Neoproterozoic glaciations (Fig. 6). Extinguishing primary productivity in the surface ocean for >105 years is difficult but could presumably result if most of the world ocean was covered by sea ice (Kirschvink, 1992a). The younger of the two biological catastrophes could not have resulted from a mega-impact at the time of the Ghaub diamictite, as proposed by Rampino (1994) because the carbon-isotopic shift largely predates the diamictite.
What was the time-scale of post-glacial biotic recovery? As argued earlier, subsidence analysis suggests that the interval from the onset of glaciation to the top of the cap-carbonate sequences should represent >4 m.y. However, we do not know how that time is partitioned between the glacial interval and the cap-carbonate sequence. This is because of the competing effects of sea-level variation due to changing ice volumes (glacio-eustasy) and vertical ground motion due to changing ice loads (glacio-isostasy). Theoretically, the duration of cap-carbonate deposition could be extremely short, limited only by their sedimentation rates. The lower 150 m of the Rasthof Formation are well-laminated (flat-laminated and stromatolitic), with an average lamina thickness of ~0.3 mm (Fig. 4A). If the laminae are annual layers, then the total laminated interval represents ~500 k.y. and the basal 15 m having negative d¹³C values would represent ~50 k.y. In contrast, the most depleted values in the Maieberg cap-carbonate sequence occur ~50 m above the base and negative but rising values continue for ~450 m, including >150 m of lower Elandshoek peritidal deposits representing >3 m.y. based on inferred subsidence rates.top
Continuous carbonate sections of the Otavi platform and their attendant carbon-isotopic profiles provide critical insights to the Neoproterozoic climatic dichotomy. The postulate that low-latitude glaciation was favored by high orbital obliquity (Williams, 1993), meaning that the tropics were colder than the poles, fails to account for the presence of thick primary carbonates, including dolomites and aragonitic precipitates, in direct contact with glacial diamictites. The supposition that the diamictites are ballistic ejecta blankets resulting from large impacts (Rampino, 1994) is inconsistent with the timing of the negative isotopic deflection, which occurs in advance of the Ghuab diamictite. Comparable isotopic shifts accompany Neoproterozoic diamictites the world over, so the association can hardly be fortuitous (Kaufman et al., 1997). The hypothesis that certain regions migrated rapidly between low and high latitudes as a consequence of intertial-interchange true polar wander (Kirschvink et al., 1997) does not explain the temporal association of the isotopic shifts with glacial deposits, nor do we find transitional mid-latitude facies between the diamictites and their respective cap-carbonate sequences as predicted by this hypothesis.
On the other hand, the sheer magnitude of the carbon-isotopic anomalies (Fig. 2) testifies that biogeochemical carbon cycling is the heart of the problem. The long-lived strong enrichments in ¹³C indicate high proportional burial of organic carbon, consistent with late Neoproterozoic fragmentation of the supercontinent Rodinia and consequent increase in continental margins (Hoffman, 1991), which are the major repositories for organic carbon (Hedges and Keil, 1995). A drawdown of atmospheric CO2 and reduction in greenhouse capacity followed inevitably, setting the stage for glaciation. The plunge in d¹³C values accompanying Neoproterozoic glaciations have been attributed to invigorated thermohaline circulation and consequent upwelling of bicarbonate- and CO2-charged waters, releasing CO2 to the atmosphere and driving the precipitation of isotopically-depleted bicarbonate as limestone in highly oversaturated surface waters (Knoll et al., 1996; Kaufman et al., 1997). However, the timing of the isotopic plunge in advance of the Ghaub glaciation creates problems for this model, inasmuch as the release of CO2 would raise the atmospheric greenhouse capacity before the Otavi platform was glaciated. The model is also problematic as an explanation for isotopic trends in post-glacial cap carbonates because meltwater injection should retard thermohaline circulation during deglaciation.
Alternatively, the burial of organic carbon may have lowered concentrations of CO2 in the atmosphere and the surface ocean to the point of limiting photosynthesis (hypocapnia), while land-and sea-ice cover continued to expand through albedo feedback (Kirschvink, 1992a). The plunge in d¹³C values may record collapsing productivity, due to the combined effects of hypocapnia and climatic degradation (Kaufman et al., 1997), after which the ice would finally come to envelope the low-latitude Otavi platform. Release from the icy tomb would eventually come as atmospheric CO2 concentrations were restored by volcanic emissions. The aftermath of such an ecological catastrophe is best recorded isotopically in the Maieberg cap-carbonate sequence (Fig. 6), for which the closest Phanerozoic analogue is the Permo-Triassic mass extinction (Magaritz, 1989). Whether or not Agassiz' (1840) vision of panglacial catastrophes is valid for the Neoproterozoic, Namibia teaches us that Agassiz was correct when he said, "If you learn about Nature in books, when you go out of doors you cannot find Her."top
This work was funded by the Geological Survey of Namibia, University of Victoria, National Science and Engineering Research Council of Canada, Harvard University, and the National Science Foundation Earth Science and Earth Systems History programs. Key field observations were made by Guowei Hu, Adam Maloof, Tony Prave and Gaddy Soffer. Sam Bowring is thanked for permission to quote the age of the tuff in the Ombombo Subgroup. We gratefully acknowledge valuable discussions with Sam Bowring, Dave Evans, John Grotzinger, Charlie Hoffmann, Bill Holser, Andy Knoll, Mike McElroy, Paul Myrow, Gerry Ross, Dan Schrag and Jim Walker.top
Figure 1. (A) Generalized stratigraphy of Neoproterozoic cover on the Congo craton in northern Namibia, showing geochronological constaints. (B) Galen Halverson points to the sharp contact between Ghuab diamictite and Maieberg cap carbonate. Below him is a graded debris flow, behind him is a laminite with numerous dropstones, and above him are deepwater rhythmites with undulations due to isopachous cement sheets. All three units are composed entirely of dolomite. (C) Ice-rafted dolomite dropstone pierces underlying laminae in glaciomarine Ghaub dolomicrite. 2 cm coin for scale.
Figure 2. Secular variation in d¹³C (PDB) values of marine carbonates from 2.5 Ga to present. Triangles indicate times of extensive glaciation. Note extreme late Neoproterozoic isotopic variability and correlation at that time of negative excusions with glaciations. Bar indicates inferred Otavi Group age span.
Figure 3. Regional correlation of depositional sequences on the western Otavi platform and its southern foreslope. Inset map shows locations of measured sections and Huab ridge.
Figure 4. Unusual lithologies in post-glacial cap carbonates. (A) Microbial roll-ups in middle Rasthof Formation (2 cm coin). (B) Vertical "tube" stromatolite in basal Maieberg Formation (10 cm scale divisions).
Figure 5. Composite section through entire Otavi Group north of Huab ridge, showing variations in d¹³C (PDB). Note negative excursions in lower Rasthof cap carbonate (above Chuos diamictite), uppermost Ombaatjie parasequence (below Ghaub diamictite) and Maieberg cap carbonate (above Ghaub diamictite). Note lack of correlation of d¹³C values with depth-dependent lithofacies.
Figure 6. Detailed d¹³C variations bracketing Chuos diamictite (A) and bracketing Ghaub diamictite in two sections (B and C) located 12.3 km apart. Note negative trend in uppermost Ombaatjie parasequence beneath Ghaub diamictite and its equivalent erosion surface. Inset photo shows sequence boundaries (arrows) between (Ab) uppermost Ombaatjie parasequence with negative d¹³C values, (Tg) remnant of Ghaub diamictite and (Tm) basal Maieberg "tube" stromatolite.