Transient fertilization of a post-Sturtian Snowball ocean margin with dissolved phosphate by clay minerals

Marine sedimentary rocks deposited across the Neoproterozoic Cryogenian Snowball interval, ~720-635 million years ago, suggest that post-Snowball fertilization of shallow continental margin seawater with phosphorus accelerated marine primary productivity, ocean-atmosphere oxygenation, and ultimately the rise of animals. However, the mechanisms that sourced and delivered bioavailable phosphate from land to the ocean are not fully understood. Here we demonstrate a causal relationship between clay mineral production by the melting Sturtian Snowball ice sheets and a short-lived increase in seawater phosphate bioavailability by at least 20-fold and oxygenation of an immediate post-Sturtian Snowball ocean margin. Bulk primary sediment inputs and inferred dissolved seawater phosphate dynamics point to a relatively low marine phosphate inventory that limited marine primary productivity and seawater oxygenation before the Sturtian glaciation, and again in the later stages of the succeeding interglacial greenhouse interval.


Sedimentation rates and the iron-based redox proxy
The best estimate is that the BDF succession up to the top of member 3, formed within 250 Kyr following the end of the Sturtian snowball glaciation, according to the following logic: (1) In the original Snowball Earth hypothesis, cap carbonates were thought to have formed within 10 Kyr, based on instantaneous melting models.However, subsequent magnetic reversals within cap carbonates pointed to longer timeframes, consistent with post-glacial sediment starvation because of seawater level rise 15 .There is a transition from coarse-grained sediments to being very fine grained up the studied sequence, consistent with a rise in seawater level and reduction in detrital loading after the Sturtian glaciation.
(2) Globally, δ 13 Ccarbonate values generally recover from -5‰ towards more positive excursions following the transition from the Sturtian icehouse into the greenhouse interval 5,16 .Similarly, published δ 13 Ccarbonate values for the immediate Bonahaven dolomite formation (BDF) succession upwards to member 3, as are our data, switch towards a more positive δ 13 Ccarbonate values immediately above the PATF.
(3) The abrupt shift of δ 13 Ccarbonate to up to +10‰ in member 4 in the BDF is far more extreme than any reported worldwide, leading some workers to equate this to the "Keele Peak" in Canada.However, the occurrence of member 4 halfway up the Sturtian post-Snowball interval towards the Marinoan Snowball Earth poses a problem, particularly because member 4 is estimated to be older than the Keele Peak.Although no explanation for the anomalous positive δ 13 Ccarbonates value was given, Fairchild interpreted member 4 in a PhD thesis as a supratidal dolomitic deposit.Therefore, the top of member 3 in the BDF lies within the bounds of the timeframe when the negative δ 13 Ccarbonates values were yet to consistently rise towards the strong positive values recorded in member 4.
(4) In the absence of radiometric dates for the BDF, we use unique cyclostratigraphic constraints in the widely accepted classic Sturtian Snowball Earth succession in Svalbard 4 , to estimate sedimentation rates for the BDF.Here, the δ 13 Ccarbonate values > -5‰ are consistently recorded up to ~10 m in the immediate post-glacial section which terminates in a shale facies succession.Within this succession, 0.5 m thick precessional cycles estimated to represent 18 Kyr in duration are linked to 6-8 Myr interglacials 17 .To a first approximation, the 10 m thick succession above which δ 13 Ccarbonate switches to ≥ -5‰ values, formed over an estimated 200-250 Kyr interval, is related to the time of deposition of the 300 m thick BDF unit up to member 3, based on stratigraphical correlation and similar and consistent ≤-5‰ δ 13 Ccarbonate values.This suggests an estimated 30 times faster sedimentation rate for the BDF compared to the Svalbard succession, accounted for by syndepositional faulting and basin subsidence associated with the Dalradian Supergroup.This has implications for the accuracy of the iron-based redox proxy because sedimentation rates of up to 0.0077 m yr -1 are reported to falsify its sensitivity and accuracy [18][19][20] .The above calculations suggest a maximum sedimentation rate of 0.0015 m yr -1 for the BDF up to member 3, which is 4× slower than the rate expected to strongly interfere with the accuracy of the iron-based redox proxy [18][19][20] .
Moreover, evidence suggests that depositional rates during the formation of the BDF sequence, switched to a slower pace because of global seawater transgression slowing down sediment supply to the basin.Hence if anything, the iron-based proxy is expected to become more accurate in their prediction of bottom seawater oxygenation at this time, as false oxygenation trends can be produced during rapid sediment accumulation [18][19][20] .The signal remains consistently anoxic during the deposition of the 1.1 km thick PATF succession, whose sedimentation rates were likely much faster than for the BDF.Moreover, throughout the sequence, the redox signal appears unaffected by carbonate concentration; staying consistently anoxic or oxic, depending on the interval, regardless of whether carbonate concentrations varied from near 0.0 wt.% to 50 wt.%.Ccarbonates and δ 18 Ocarbonates is interpreted to suggest limited diagenetic alteration of the primary δ 13 Ccarbonates signal 1,2 .Broadly, the data are consistent with primary C signatures that have been correlated world-wide to the Sturtian Snowball deposits [3][4][5][6][7] .

Fig. S4 |Fig. S5 |
Fig. S3 | (a) Fe/Ti and Al/(Al+Fe+Mn) cross plots showing a mainly detrital origin for the sampled sequence.Red arrow=hydrothermal origin (cut off minimum threshold at 100≥Fe/Ti>1000).Blue arrow shows continuous gradation towards mixed chemical+detrital to pure chemical sediments 9 .(b) Relationship between TiO2 and Zr, showing that the analyzed rocks have a strong detrital origin.

Fe 2 Fig. S6 | 5 (
Fig. S6 | Relationship between facies Ca, P, Corg and δ 13 Corg content.(a) CaO versus P2O5 distribution along sequence stratigraphy.(b) CaO versus P2O5 scatter plot.(c) Scatter plot for representative data available for P2O5 and Corg (d) Scatter plot for representative data available for P2O5 and δ 13 Corg.PAT1 and PAT2 are samples collected at Port Askaig.PAT3 and PAT5 represent sample LL-16-a at the top of the LLF that conformably underlines the Port Askaig tillites.

P 2 OFig. S8 |Fig. S9 |Fig. S11 |
Fig. S7 | Box and whisker plots showing P distribution in various mineral phases relative to bulk P concentrations (P 2 O 5 ) in low P2O5 pre-Snowball Lossit Limestone Formation and immediate post-Snowball high P2O5 and late post-snowball low P2O5 intervals.Bulk unreactive P was calculated by deducting the concentrations of highly reactive iron oxide-bound P and sheet silicate (phyllosilicate) associated P from total bulk measurements for corresponding samples.(a) Total unreactive P in ppb.(b) % unreactive P in bulk.(c) % Fe oxide-bound P in bulk.(d) % sheet silicate P in bulk.Centre line = median value; whiskers = minimum and maximum values; box limits = lower and upper quartiles.

Table S2 |
Semi-quantitative bulk XRD mineralogical composition for representative samples measured in weight percent (wt.%).PAT1 and PAT2 are samples collected at Port Askaig.PAT3 represents sample LL-16-a at the top of the LLF that conformably underline the Port Askaig tillites.nd=not detected.

Table S3 |
Trace and rare earth elements (ppm) and major elements (wt.%) data for representative samples across the sampled section.LD, lower than detection limit.PATF, Port Askaig Tillite Formation.

Table S4 |
Bulk iron isotope data for sampled cross section.