Biological diversification linked to environmental stabilization following the Sturtian Snowball glaciation

The body fossil and biomarker records hint at an increase in biotic complexity between the two Cryogenian Snowball Earth episodes (ca. 661 million to ≤650 million years ago). Oxygen and nutrient availability can promote biotic complexity, but nutrient (particularly phosphorus) and redox dynamics across this interval remain poorly understood. Here, we present high-resolution paleoredox and phosphorus phase association data from multiple globally distributed drill core records through the non-glacial interval. These data are first correlated regionally by litho- and chemostratigraphy, and then calibrated within a series of global chronostratigraphic frameworks. The combined data show that regional differences in postglacial redox stabilization were partly controlled by the intensity of phosphorus recycling from marine sediments. The apparent increase in biotic complexity followed a global transition to more stable and less reducing conditions in shallow to mid-depth marine environments and occurred within a tolerable climatic window during progressive cooling after post-Snowball super-greenhouse conditions.

Model B serves to shorten the timing of recovery from the initial post-Sturtian anoxic episode, but does not result in any notably different pattern or timescale of geochemical response relative to that shown between the geochemical data calibration of Model A and the FOAM climate model output.
Given the more reasonable deglaciation age of ca. 661 Ma and negligible difference in geochemical calibration between models A and B, we consider all alternative models relative to Model A (Fig.   S1B).

Model C (Non-glacial duration ca. 15.5 Myr)
Model C assumes an age of ca. 661 Ma for Sturtian deglaciation, but extends the duration of the non-glacial interval to 15.5 Myr, with Marinoan re-glaciation occurring at ca. 645.5 Ma (Fig. S1C).
Model C results in crystallization of the dated Thorndike submember zircons approximately coincident with the Taishir anomaly (Cn3), rather than the Keele Peak as suggested by (13). This remains possible when considering that the Thorndike age is conservatively interpreted to represent a maximum depositional age (M.D.A.; 19), and in recognition of the complexity of preserved δ 13 Ccarb within the regional depositional environment of the Thorndike submember (13). Given these associated uncertainties, we consider models A and C to be equally plausible based on integrated consideration of all available radiometric data.
Whilst the duration of the non-glacial interval is 3.5 Myr longer in Model C than Model A, the calibration of all geochemical and biotic data remain internally consistent. The result is that the timing of stabilization of geochemical trends (e.g., equilibration of Fe-speciation data) in inner shelf to slope environments appear to align perfectly with the timing of modelled climate-carbon steady state ca. 4.5-5 Myr post-Sturtian deglaciation (Fig. S1A, S1C). In Model C, this timing also aligns with the maximum possible age for the first appearance of green algal and putative sponge steranes ( Fig. S1C).

Model D (Non-glacial duration ca. 21 Myr)
Model D also accepts an age of ca. 661 Ma for Sturtian deglaciation, but extends the non-glacial interval to occupy 21 Myr (Fig. S1D), which is close to the maximum possible duration based a zircon U-Pb age of 639.29 ± 0. 26 Ma from an ash interbedded with the Marinoan-age Ghaub diamictite (21). This age model strongly suggests that the dated horizon in the Thorndike submember is younger than the maximum depositional age from zircon U-Pb geochronology (Fig. S1D). This age model also extends the minimum age of the basal black shale interval of the Datangpo Fm to ca.  Table S1). (E) Key radiometric ages considered in construction of each age model (17,19,21). D.A. sections (17).
In light of the caveats, we currently consider this maximum non-glacial duration less likely than models A-C. However, even in Model D, the interval of geochemical re-equilibration following Sturtian deglaciation remains largely within the 4.5-5 Myr interval of decreasing global temperature and increasing O2 saturation state of surface seawater, and preceding the attainment of climatecarbon steady state (Fig. S1D).

Cryogenian age model limitations
Given ongoing uncertainties in paleogeographic reconstructions, and variability in regional and global δ 13 Ccarb associated with possible local effects and variable diagenetic regimes, Sturtian deglaciation and Marinoan reglaciation are conservatively considered to have been globally synchronous at the resolution of each age model. Globally synchronous Sturtian deglaciation is consistent with current radiometric constraints and associated uncertainties (Fig. 1G) (16)(17)(18), and is also consistent with a geologically instantaneous global deglaciation predicted from climate model simulations under immediately post-glacial super-greenhouse conditions (78).
With the exception of the timing of Sturtian deglaciation and the associated Rasthof anomaly, and the age of positive δ 13 Ccarb values associated with the Keele Peak (but see above), the non-glacial chemostratigraphic age model presented herein suffers from a dearth of radiometric ages to anchor the durations and absolute ages of individual δ 13 Ccarb excursions (Fig. 1B, G). As such, only relative trends in δ 13 Ccarb can presently be used for correlation. As noted in the main text, between-section variability in the magnitude of δ 13 Ccarb during each δ 13 Ccarb excursion of the Cryogenian non-glacial is well established, especially across the Taishir (Cn3) and Keele Peak (Cn4) anomalies. This variability in magnitude may reflect differences in the fidelity of preservation of seawater δ 13 C associated with fluid-buffered versus sediment-buffered diagenetic regimes (12,13,23), which leads to some ambiguity in the correlation of minor δ 13 Ccarb trends that may be present within the Cn2 and Cn4 intervals, and the discrimination between the Cn3 versus Cn5 excursions. The latter case is especially problematic when considering successions of mixed lithology that do not host both excursions; perhaps most clearly demonstrated by the negative δ 13 Ccarb anomaly recorded in the lower Ice Brook Formation in some sections of the Wernecke Mountains (Yukon) (10). Specifically, (Tindelpina Shale Member), which gradually shallows-upwards and is conformably overlain by the shallow marine oolitic grainstone of the Brighton Limestone (33). To the east, shale of the upper Tapley Hill Formation interfingers with equivalent marine carbonates of the Balcanoona Formation, which in turn shallow to iron-rich peritidal dolomite and dolomitic shale of the Angepena Formation (7,83,84). In SCYW-79-1A, the Brighton Limestone is overlain by the Whyalla sandstone, which has been interpreted as a periglacial aeolian sandstone that was deposited on the Stuart Shelf equivalent to the Elatina Formation diamictite of the Marinoan cryochron to the east (70).
The Aralka and Tapley Hill formations contain a characteristically depauperate Cryogenian nonglacial microfossil assemblage dominated by leiosphaerid acritarchs, alongside prokaryotic Synsphaeridium, simple filamentous forms, and rare occurrences of ornamented and double vesicle acritarchs (85). The associated assemblage is similar to microfossil assemblages from the correlative Macdonaldryggen Member of the Elbobreen Formation, Svalbard, and the lower Datangpo
Iron-rich sandstones and banded iron formation deposits of the Fulu Formation are variably reported as overlying the Gucheng Formation (93), comprising the Liangjiehe and Gucheng members (90), or underlying the Gucheng Formation diamictites (92). However, litho-and sequence stratigraphic constraints support a syn-Sturtian position for the Fulu Formation, equivalent to the Xieshuihe and Liangjiehe formations, and conformably underlying the Gucheng Formation (34,91,92).
A Sturtian age for the lower diamictite-bearing deposits of the Nanhua Basin is confirmed by a zircon U-Pb SHRIMP age of 691.9 ± 8.0 Ma from tuffaceous siltstone of the Xieshuihe Formation, and a zircon U-Pb CA-ID-TIMS age of 658.8 ± 1. 49 Ma from a tuff layer at the top of the Tiesi'ao Formation (91,94). More recent zircon U-Pb CA-ID-TIMS ages of 660.98 ± 0. 74 (20).
Here we present data from two drill core profiles that transcend continuous siliciclastic deposits of the Datangpo and Xiangmeng formations (Fig. S2B). Core ZK102 was recovered in the Daotuo area near Tongren City of eastern Guizhou Province and records outer shelf-slope deposits of the Datangpo Fm. Core ZK3603 was recovered in eastern Hunan Province and records deeper, slopebasin deposits of the Xiangmeng Fm (Fig. S2B). In both cores, the contacts between the Datangpo/Xiangmeng formations and the underlying and overlying diamictites appear to be conformable (14).
The lithostratigraphy of drill core ZK102 closely resembles that described for neighbouring drill core ZK105 (Daotuo section of 11), and comprises a basal transgressive unit of ~10 m interbedded Mn-rich carbonate and black shale that conformably overlies diamictite of the Tiesi'ao Formation, followed by ~30 m organic-rich black shale, ~160 m grey shale, and ~60 m grey siltstone. The boundary between the upper Datangpo Formation and overlying lower Nantuo Formation in the Daotuo area is locally conformable in drill core and outcrop sections (96). The lithostratigraphy of core ZK3603 has been subdivided into 4 units; 0.4 m interbedded Mn-carbonate and organic-rich black shale, 82 m black shale, 1.1 m grey-black Mn-bearing limestone, and 5.5 m calcareous and pyritic grey-black shale (14). The microfossils assemblage of the lower Datangpo Formation is dominated by cyanobacteria and a variety of small vesicle acritarchs including (but not limited to) Protosphaeridium, Synsphaeridium, Trachysphaeridium and Leiosphaeridia (86,97).

Detailed methods
Drill core material was trimmed of any visible surficial alteration or veining and samples with macroscopic euhedral pyrite were not used. Samples were crushed and powdered using a tungsten carbide piston and agate disc mill. Unless specified otherwise, chemical extractions and analyses were undertaken at the Cohen Laboratories, School of Earth and Environment, University of Leeds.

Elemental analysis
Bulk digestions were performed on 50-80 mg of sample powder using HNO3-HF-HClO4 at ~70°C for 12 h, followed by addition of H3BO3 to prevent the formation of Al complexes. Major element (Al, Ca, Fe, K, Mg, Mn, Na, P) concentrations were measured using inductively coupled plasma optical emission spectrometry (ICP-OES, Thermo Fisher iCAP 7400), and trace element (Mo, Re, U) concentrations were measured using inductively coupled plasma mass spectrometry (ICP-MS, Thermo Fisher iCAPQc). Total digestions of a standard material (SBC-1, United States Geological Survey) yielded values within the certified range for all elements analyzed (<3.4%).

TOC and organic C isotopes
Samples were fully decarbonated via 3-4 12 h HCl (10% vol/vol) dissolutions, repeatedly washed with 18MΩ H2O to remove all residual acid (pH >4), centrifuged and dried prior to analysis. Carbon concentrations were measured using a LECO carbon-sulfur analyzer, with replicate analyses yielding a precision of ±0.09 wt% (2σ). The organic carbon fractions of samples from core ZK102 were analyzed for organic C isotopes (δ 13 Corg) by elemental analyzer isotope ratio mass spectrometry (EA-IRMS) at Iso-Analytical Ltd. All data are reported relative to the Vienna Pee Dee Belemnite standard, with replicate analyses yielding a precision of ±0.14‰ (1σ). Organic carbon isotopic ratios in samples from the BR05-DD01 core were measured by EA-IRMS in IsoLab at the University of Washington. Replicate analyses yielded a precision of ±0.03% (1σ) for TOC and ±0.13‰ (1σ) for δ 13 Corg.

Fe speciation
Fe speciation was performed after the established methodology of Poulton and Canfield (71) (98). A pre-leach in boiling 50% (vol/vol) HCl confirmed that no acid volatile sulfides were present.
All analyzed samples have total Fe concentrations (FeT) >0.5 wt%, including 5 carbonate-rich samples from the base of the Datangpo Formation in core ZK102. In order to ensure that the sequential leach quantitatively extracted each operationally-defined Fe phase, a recently-certified Fe speciation reference material (WHIT) was run alongside each batch (99). The results are in agreement with certified values, with mean values of Fecarb = 0.599 wt%, Feox = 0.069 wt%, and Femag = 0.107 wt%, and relative standard deviations (RSDs) of <5% (n=17).

Pyrite sulfur isotopes
The isotopic composition of pyrite sulfur (δ 34 Spy) was determined on Ag2S produced through sample reaction with boiling chromous chloride. Samples from core ZK102 were analyzed for δ 34

P phase association
Selected samples were subjected to a separate sequential P extraction following the modified SEDEX methodology (75) adjusted for use on ancient sediments (Table S5; 76). This method isolates operationally-defined P pools including P bound in Fe(oxyhydr)oxide minerals (PFe), organic matter (Porg), authigenic carbonate fluorapatite, biogenic apatite and CaCO3 (Pauth), and detrital apatite (Pdet). Reactive P (Preac) equates to the sum of PFe, Pauth and Porg (36). The concentrations of P in Porg, Pauth and Pdet leachates were measured spectrophotometrically using the molybdate-blue method on a Spectonic GENESYS 6 at a wavelength of 880 nm, whereas P concentrations in the PFe leachates (including PFe1, PFe2 and Pmag) were measured by ICP-OES (76). A mean P recovery of 89% of PTot (as measured by ICP-OES after total digestion) was achieved by the sequential extraction protocol. Replicate analyses of the Fe speciation standard (WHIT, n=9) gave a relative standard deviation of <7% for all extraction steps, with the exception of PFe (18%) due to very low concentrations of PFe (near detection).

Detailed paleoredox and P assessment
The importance of a multi-proxy dataset in the characterization of regional paleoredox We note recent challenges to the Fe speciation technique, and interpretations of paleoredox data based on Fe speciation alone (100). Indeed, there are several caveats that must be considered carefully when evaluating Fe speciation data, which are well described in the literature but were ignored in the analysis of Pasquier et al. (100).
Secondly, rapid deposition (e.g., turbidites) may mask anoxic FeHR enrichment and lead to muted FeHR/FeT (102). However, no turbidites were analyzed in this study, and up-core changes in depositional rate are evaluated and discussed relative to changes in all available paleoredox data. In particular, whilst each of our cores show sedimentological evidence for shallowing (and likely corresponding increases in depositional rate), the redox shifts do not always correspond directly with recorded lithological shifts. For example, the redox shift in BR05-DD01 occurs within a monotonous dark grey siltstone interval (Fig. 2), and dominantly occurs across a transition from black to grey shale in ZK102 (Fig. 4). In the latter case, this color change does not correspond with a decrease in sediment grain size, but more likely represents enhanced organic carbon content of the basal unit of black shale (Fig. 4A). Diagenetic models also suggest that muted Moauth would be expected in intervals of enhanced sedimentation rate (41). However, our deepest (and likely slowest deposited) core (ZK3603), exhibits low Moauth throughout the non-glacial interval despite occasionally elevated Fepy/FeHR up to 0.77 (e.g., Fig. 4N). If increasing sedimentation rates throughout the non-glacial were solely responsible for the observed decrease in (e.g.) Moauth in shelf-slope environments, then we may expect elevated Moauth to persist throughout deeper environments where sedimentation rates remained low and pore waters were occasionally sulfidic (e.g., ZK3603).
Thirdly, the interpretation of paleoredox conditions from Fe speciation systematics can potentially be affected by the degree of chemical weathering (103). In some modern subtropical environments, high riverine concentrations of FeHR supplied by intense chemical weathering can bypass the inner shore zone of Fe (oxyhydr)oxide trapping (e.g., 104), and lead to FeHR/FeT values in marine sediments that are enriched relative to the normal oxic threshold of 0.38 (103). It is thought that the initiation of Sturtian deglaciation was a consequence of super-greenhouse conditions that followed a multi-million-year build-up of atmospheric CO2 during the Sturtian cryochron (1,105,106). Previous studies utilizing the chemical index of alteration (CIA) proxy have suggested a high degree of chemical weathering in the Cryogenian non-glacial Nanhua Basin (107,108). However, the pronounced and unidirectional global trends in TOC, FeHR/FeT, Fepy/FeHR, and redox sensitive trace metal enrichments (from high to low values) are distinct from the persistently elevated CIA values observed throughout the non-glacial Nanhua succession, and so were more likely driven by changes to depositional redox conditions (107). This interpretation is further reinforced by a corresponding up-core increase in diameter and decrease in abundance of framboidal pyrite morphology (e.g., 40,109), which are controlled by depositional redox conditions but are independent of chemical weathering-derived FeHR.
Fe speciation has been extensively calibrated in both modern depositional environments and using ancient rocks (39,73,101,103,104,110,111). The ancient rocks used in calibration have by definition undergone the diagenetic transformations that Pasquier et al. (100) claim compromise the utility of the proxy for paleoredox interpretation. As emphasized in the literature, Fe speciation data are most appropriately used in combination with other indicators of water column paleoredox (e.g., redox sensitive trace element concentrations) and all pertinent information to constrain changes in the depositional rate and depositional environment (e.g., 112,113). This is the approach taken here, which provides a particularly robust reconstruction of paleoredox conditions based on an integrated consideration of independent multiproxy redox data. Below, we assess all available proxy data to build a conservative and integrative framework for the characterization of depositional redox conditions in each studied core.

Regional variability in the marine inventory of redox sensitive trace elements
Extensive study of elemental concentrations (including redox sensitive and detrital unreactive elements) in soils from across the continental United States found notable heterogeneity in trace element concentrations and ratios traditionally used to infer enrichments relative to detrital contribution. As such, the determination of trace element enrichments relative to crustal cutoff values has been considered to oversimplify intra-and inter-regional variability (114). Specifically, detrital fluxes of redox sensitive trace elements may regionally fall well below the crustal average, thereby effectively concealing paleoredox-related authigenic enrichments (see below) (114). As such, we consider authigenic trace element concentrations (auth) where possible, but also consider trace element enrichment factors (EF) in cases where the calculation of authigenic concentrations obscure trends in the dataset (e.g., U, see below). When discussing trace element enrichment factors, all values and trends are considered relative to local oxic baseline values for each dataset (see below).   to zero, we consider both Xauth concentrations and trends in XEF for each element throughout the nonglacial interval. As emphasized above, anoxic elemental enrichments are considered relative to oxic baseline values for each dataset in the main text and following discussion. We further note that none of the analyzed cores show any systematic trend in Al concentrations throughout the non-glacial interval that may affect trends observed in normalized trace metal enrichments. Figures S4 and S5 show all core intervals colored by their interpreted depositional redox conditions based on a combination of Fe speciation data and observed degrees of enrichment in the redox sensitive elements Mo, U and Re. These three redox sensitive elements have differing reduction potentials and/or differing mechanisms for sedimentary enrichment, which, when considered in combination, permit a more detailed assessment of depositional redox conditions (e.g.,

119).
The removal rates of Mo, U, and Re to anoxic sediments are controlled by water column and pore water chemistry (116,120). Authigenic enrichments in U and Re occur in sediments deposited under anoxic water column conditions, in addition to more minor enrichments in anoxic sediment pore waters underlying weakly oxygenated bottom waters (120)(121)(122)(123)(124). Both U and Re enrichments primarily occur via diffusion from the water column into reducing sediments across the sedimentwater interface (126)(127)(128). U and Re thus behave similarly to changes in depositional redox state, however Re has a higher sensitivity to weakly reducing conditions than U, and is also efficiently sequestered in the sediment under dysoxic conditions characterized by O2 penetration depths of <1 cm below the sediment-water interface (119)(120)(121)128). Enhanced U and Re enrichments can take place under anoxic conditions characterized by low dissolved H2S, whereas high dissolved H2S is a prerequisite for high authigenic Mo enrichments (129). Mild authigenic Mo enrichments may also occur in sulfidic porewaters beneath weakly oxic to anoxic ferruginous water column conditions, and the magnitude of enrichment scales with sulfide availability (120)(121)(122). Importantly, widespread anoxic (and for Mo, euxinic) water column conditions will necessarily result in substantial redox sensitive trace element depletion in the global ocean or restricted basin, leading to muted authigenic enrichments (130,131).
We characterize changes in water column and pore water redox conditions throughout each core based on an integrated assessment of Fe speciation and redox sensitive trace element systematics (Table S2).    (2) indicated by green stars.
Redox data presented stratigraphically. See main text figures 2-4 for keys to lithostratigraphy.

Long-term trends in paleoredox data and the composition of weathered source material
Long-term intra-basinal trends in elemental data may simply reflect long-term changes to weathering style or the source of weathered material within a catchment (114). This may be partially tractable in studies that resolve consistent trends between multiple sections from different depositional basins (114). However, the Cryogenian non-glacial interval may represent a specific case where all catchment areas and regional depositional environments were influenced by broadly similar long-term trends in both weathering regimes and depositional rate (albeit with variability in magnitudes controlled by differences in paleolatitude and regional tectonics). If long-term trends in trace metal enrichments are primarily related to long-term shifts in source composition, this may be reflected in corresponding trends in the concentrations of detrital elements used for normalization.
However, as noted above, none of the analyzed cores show notable systematic unidirectional trends in Al concentration throughout the non-glacial interval. The correlation of trends observed between proxies that rely upon detrital normalization (e.g., UEF) and those that are independent of detrital normalization (Fe-speciation) therefore strongly supports a common driver that is most parsimoniously related to changing depositional paleoredox conditions, which is further supported by the observed correlation between interpreted paleoredox conditions and P-speciation data (e.g., Figs.

Critical assessment of P speciation data
One possible complication when interpreting P speciation data is the potential for late stage diagenetic transformation of carbonate fluorapatite (Pauth) to more crystalline P-bearing phases (76,132). The resulting crystalline phases would be operationally extracted as Pdet, thereby reducing the  substantially lower relative to all other cores, which is an expected consequence of distance from the detrital P source throughout deposition (Fig. S6).
We note that there is no significant correlation between Pauth and Pdet in any of our cores (Fig.   S7A). However, we also note that, with the exception of Wallara-1, there is no significant positive correlation between Pdet and Al (Fig. S7B). If high values of TOC/Preac observed in some euxinic/  Maintenance of high C/Preac (mean = 353) in our deepest core (ZK3603), supports a degree of bioavailable P recycling from sulfidic sediment pore waters in some anoxic basinal environments throughout the non-glacial interval, even if C/Porg of primary biomass remained elevated relative to the Redfield ratio.   Table S5. Analytical Data 2: Total P concentrations (P Tot ), and P speciation. Depth (m) indicates sample position relative to drill core datum. BD = below detection; ND = not determined.