TILTING AT SNOWBALLS Paul F. Hoffman and Adam C. Maloof Department of Earth & Planetary Sciences Harvard University, Cambridge, MA, USA 02138 January 30, 2001 ABSTRACT Ten arguments were recently presented in The Australian Geologist [1] against the ‘snowball Earth’ scenario for equatorial glaciation, a scenario that explains many previously enigmatic features of Neoproterozoic Earth history such as post-glacial cap carbonates, large shifts in stable isotope ratios of seawater proxies, and extensive oxide-facies iron formations. We show that the counter arguments are based on incomplete understanding of the ‘snowball Earth’ scenario, in particular the continuously changing conditions that must occur in response to rising atmospheric CO2 levels during the glacial period. There is no conflict with geological observations when this and other factors are considered. We also suggest that evidence central to the theory of high orbital obliquity for Proterozoic low-latitude glaciation may be open to alternative interpretation. Introduction None has done more to focus attention on the Proterozoic glacial record than George Williams and Phillip Schmidt, and to this end we welcome their ten-point discussion [1] of the ‘snowball Earth’ hypothesis. However, many of their points derive from a false or incomplete portrayal of the hypothesis, and their “rebuttal” of it is therefore less “formidable” than they imagine. That aspects of the snowball Earth scenario are misunderstood is not surprising. The hypothesis itself [2,3,4] is comparatively recent and its literature is terse. An ice-covered globe and its ultra-greenhouse aftermath invoke very different worlds from any familiar one and much remains to be learned about how such worlds would work. Even the parents of the hypothesis don’t know exactly how the child will turn out. In defense of the snowball Earth scenario as best we understand it, we respond to each of Williams and Schmidt’s [1] ten points in order, using their headings (with question marks added or deleted). We also suggest a possible alternative explanation, independent of seasonality, for the occurrence of polygonal sand wedges near the equator. That observation and its conventional seasonal interpretation lies at the heart of the hypothesis preferred by Williams and Schmidt [1,5], a high obliquity of the ecliptic, or ‘Big Tilt’ for short. 2 1. Vigorous hydrological cycle shows no global freeze-up? When we described a snowball Earth as having “continental ice cover that was thin and patchy because of the virtual elimination of the hydrologic cycle” [3], we were referring to the situation immediately after the ice-albedo runaway. From then on, the frigid climate will slowly ameliorate due to the steady increase in atmospheric CO2 (from volcanic outgassing in the absence of sinks for carbon). As temperatures rise (most rapidly early on because of non-linear CO2-greenhouse forcing [6]), the ablation of sea ice in the tropics and subtropics will feed a weak hydrological cycle [7]. Wally Broecker and Richard Alley pointed out to us that even the very low rates of snow accumulation at higher elevations (due to the lapse rate) and higher latitudes that would result from such a hydrological cycle should, over the course of millions of years, result in dynamic, wet-based glaciers. Ablation rates are high where air currents descend, as in the Antarctic dry valleys (~30 cm yr-1 [8]). An equal accumulation rate would produce 3,000 m of ice in 10,000 years. Given that a Proterozoic ‘snowball Earth’ would last for millions of years [6,3], dynamic glaciers would have existed in coastal areas with strong sunlight and significant topography in all but the initial stages of a snowball Earth. This is consistent with the presence of Proterozoic glacial deposits, locally >1000 m thick, with faceted and striated clasts of extrabasinal derivation [9]. 2. No extreme fall of sea level rebuts global glaciation? If every continent in the world were covered by ice sheets of comparable mean thickness to those in Greenland and Antarctica, then the absolute fall in sea level after isostatic adjustment would be on the order of 900 m. (Isostasy would be maintained assuming the ice caps grew slowly.) George Williams has repeatedly pointed to the fact that no evidence for extreme sea- level fall has been observed geologically. This is a false test because it is not the absolute but the relative sea-level fall that leaves its mark geologically. The late Pleistocene rise in sea level can be tracked because we have ocean islands and continental margins that were never depressed by ice caps. But if all the continents were burdened by ice the thickness of Greenland or Antarctica, then their mean isostatic depression would be on the order of 750 m, offsetting much of the glacio-eustatic fall in sea level. In detail, the relative sea-level change at continental margins would vary spatially on the length-scale of lithospheric flexure that must accommodate the shift in water load from the oceanic to the continental areas. In addition, any tectonic subsidence (or uplift) that occurred during a prolonged snowball event would change the net relative sea-level fall. The bottom line is that the fall in relative sea level to be expected in a snowball Earth is probably in the range of 0-300 m, consistent with observations [10]. More extreme falls are still possible locally (Fig. 1). The deepest stratigraphic incisement reportedly associated with Neoproterozoic glaciation is located not far outside Adelaide and ironically was discounted as being too extreme (~500 m) to be eustatic in origin [11]. Regardless of the truth in this example, the idea that global glaciation should leave evidence of extreme falls in relative sea level is untrue. 3. Tidal rhythmites show an open ocean? We are baffled as to why the presence of tidal rhythmites demands an open ocean [1]. Tidal forces act on ice as they do on water. A global sea-ice shell with a mean thickness of only 1 km 3 (and only ~10 m at the equator [7]) would offer negligible resistance to flexure of tidal amplitudes (meters) and wavelengths (20,000 km at the equator). Cracks would form at ice grounding lines (E.J. Gaidos, personal communication, 2000). Tidal channels and deltas might develop, for example, where the underside of a floating ice shelf hugs the crest of an abandoned ‘surge moraine’ (see below)[12]. In any event, patches of open water at low latitudes were assumed to exist in the original snowball Earth hypothesis [2] and the Adelaidean coast could well have opened before the main ocean near the end of a snowball event. Aeolian sandstones of Marinoan age on the Stuart shelf show unimodal offshore-directed paleowinds (from present northwest) [13]. If these winds were descending from an inland ice plateau, adiabatic heating would make them strongly ablative at lower elevations, causing the coastal sea ice to thin and locally disappear. Open coastal waters would lead to increased precipitation and more dynamic glaciers. 4. Wave action indicates an unfrozen sea? Williams [14] has described submarine slumps with superimposed storm-wave structures in the Marinoan glacial succession in South Australia. Storm-wave activity requires open water. The unanswered question is at what point in the glacial cycle did the storm-wave structures form? The entire Marinoan succession in the area described is less than 150 m thick and could have been deposited towards the end of the glacial period. Areas of open water must surely have developed at low latitudes around the time of deglaciation. According to the snowball hypothesis, high atmospheric CO2 and reverse ice-albedo feedback would drive deglaciation violently. Strong warming in low-albedo areas of open water and steep temperature gradients at ice margins would make for very stormy weather. Slumps and storm-wave structures are to be expected near the end of a snowball event, and are commonly observed in Neoproterozoic post- glacial ‘cap’ carbonates. 5. Strongly seasonal climate confutes a frozen-over Earth? The existence at low latitude of periglacial sand wedges and varves (if truly seasonal) is the observation at the core of the Big Tilt hypothesis. High orbital obliquity is the simplest way to enhance seasonality close to the equator [15]. The existence of sand-wedge polygons in Neoproterozoic glacial successions of unknown paleolatitude [16] presents no problem, but those on display in the Cattle Grid open pit in South Australia [17] formed at ~8˚ latitude [18] (or ~16˚ if the geomagnetic field had a 25% octupole component). The wedges are far too deep (many 2.5 m) to result from diurnal temperature fluctuations. Does the snowball hypothesis provide for any alternative explanation? Glaciers that surge periodically could cause large, multi-annual, temperature fluctuations in periglacial soils (Fig. 2). Many glaciers in cold, arid regions are polythermal (‘subpolar’). Their thin, cold-based extremities undergo ablation, but their upper parts become thick and wet-based because of net accumulation [19]. This is an inherently unstable condition—the glacier’s toes are frozen to the bed and create a barrier to the flow of incoming ice from the accumulation zone. A positive feedback results where the surface gradient of the glacier increases, leading to increased precipitation rates in the accumulation zone. As the accumulation zone thickens, more and more subglacial meltwater is produced that cannot drain through the frozen ablation zone. Subglacial water pressures build and the surface gradient grows until the toe-hold releases and the glacier 4 surges. During surges, ice evacuates the upper reaches of the glacier and flows down into a proglacial repository bounded by a ‘surge moraine’. Surges last for 1-3 years, during which time the ice line (or the ice grounding line of an ice shelf) may advance tens of kilometers [19].