Lithium isotope evidence for enhanced weathering and erosion during the Paleocene-Eocene Thermal Maximum

Description


Samples
Here we analysed both marine carbonates and detrital shales from multiple localities covering different ocean basins to determine the lithium (Li) isotope response to the PETM.
One shale section is from the borehole Store Norske Spitsbergen Kulkompani (SNSK) BH9/05, located on the eastern flank of the Paleogene Central Basin of Spitsbergen, Svalbard (17,40,88). The Central Basin began to form as a foreland basin adjacent to the West Spitsbergen fold-and-thrust belt from around 61.8 Ma (89). The subsidence marks the beginning of sustained compression between Greenland and Svalbard that resulted in continuous deposition into the late Eocene and is associated with the Eurekan deformation (90,91). During the Paleocene-Eocene transition, the study area was a marine shelf facies dominated by laminated shales (92). The Svalbard section is anomalous for PETM sections as high deposition rates precede and post-date the PETM, although deposition during the PETM CIE still shows elevated sedimentation rates (93).
The other shale section is from Fur Island, Denmark. In the Paleogene this locality was part of the Norwegian-Danish Basin, a marginal extension of the epicontinental North with maximum values similar to modern core-top carbonates ( Fig. S4) (7). There is no obvious trend in Li/Ca ratios across the PETM for any of the sections (Fig. S2), but a trend would not necessarily be expected, given regional temperature and salinity effects (6,104), as well as changing seawater Li concentrations. Similar stability was found for Li/Ca ratios from Ocean Anoxic Events 1a and 2 (42,45). Mn/Ca ratios generally exhibit a slight peak at the PETM, although absolute values vary between cores, and such variability likely reflects changing redox conditions during this time period (105). There is no correlation between Mn/Ca and d 7 Li (Fig. S3), indicating no effect from Mn-oxyhydroxides. On the note of oxyhydroxides, Site 1210 has also been investigated for Cr isotopes, which would be controlled by such minerals if strongly present (106). In all cases, Al/Ca ratios are low, indicating that the effect of silicate dissolution during sample leaching is insignificant (Table   S1). All Al/Ca values are below the cutoff used in prior reconstructions of seawater Li isotope ratios from bulk carbonates (42), while most are also beneath the lower cutoff recently suggested in a study specifically examining at the effects of diagenesis on Li isotope ratios in carbonates (29). Where Rb/Ca ratios are available (Table S1)      In contrast, the duration at Site 1210 is 64 kyr using the age model of (87), or 105 kyr using the age model of (85). The oceanic residence time of Li is on the order of a million years (108), and therefore the oceans are isotopically homogenous at any given time. This finding therefore raises concerns about the completeness of the age model for Site 1210, in particular for the recovery phase of the CIE. Overall, this concern is supported by the dynamic box models of Li discussed below (Section 7), where it is very hard to force an oceanic recovery of Li in less than ~200 kyr.

Details of modern Li isotope behaviour
Lithium isotopes have been shown not to be fractionated by plant growth or primary productivity (56,110,111) and are not sensitive to carbonate weathering (112,113), given the low Li content of carbonates. Although evaporite weathering can be locally significant (114), on a global scale the evaporite contribution is relatively minor in modern rivers (20).
The δ 7 Li of primary silicate rocks defines a narrow range (continental crust ~0.6 ± 0.6 ‰, basalt ~3-5 ‰ (54, 115)) compared to the high variability of dissolved Li in modern rivers (2-44 ‰) (22,23,25,114,(116)(117)(118). Therefore, fluvial values overwhelmingly reflect weathering processes (37,(119)(120)(121)(122) , and in particular the relative importance of rock dissolution (driving riverine values to low, rock-like, δ 7 Li compositions) to secondary mineral formation (driving riverine values to high δ 7 Li)(118). This ratio is generally known as the weathering congruency (23,24,118). As a result, lithium isotopes can also inform on weathering intensity, that is the ratio of weathering rates to denudation rates (W/D) (25), where the denudation rate is the sum of the chemical weathering rate and the physical erosion rate. Weathering intensity is tied closely to congruency. Low intensity weathering (low W/D) results in low riverine d 7 Li; as weathering intensity increases, clay formation and riverine d 7 Li also increase (Fig. 5, main text). Very high intensity regimes also exist, where d 7 Li is low due to the redissolution of (or desorption from) pre-formed secondary clays.
These regimes only exist in strongly supply-limited areas, such as tropical lowlands. Overall, both the riverine Li yield and flux decrease by almost two orders of magnitude from low to high intensity regimes (Fig. 5b, c). Hence, globally-pervasive high-intensity weathering would have very little influence on the oceans' Li budget, which would then be dominated by the hydrothermal input (23,25). In the modern oceans, ~60 % of the Li inputs are via rivers with a mean δ 7 Li ~23 ‰, and ~40 % is via mid-ocean ridge hydrothermal solutions with a mean δ 7 Li ~7 ‰ (27). Lithium is predominantly removed by incorporation into lowtemperature clays, both in altered oceanic crust (AOC) and marine aluminous authigenic clays (MAAC). These clays both preferentially remove light Li and together these sinks impose an isotopic fractionation of ~15 ‰, which drives modern seawater higher to 31 ‰ (123,124). Marine carbonates represent a negligible sink for Li, but have frequently been used as an archive for past seawater Li isotope ratios in order to constrain past silicate weathering behaviour (8,24,27,34,42,45,125). Given that the global ocean residence time for Li of ~1 million years (126,127) is significantly longer than the ocean mixing time, we anticipate agreement between Li isotope reconstructions from each location if the composition of palaeo-seawater is being recorded.
The references used for Figure

Dynamic box models
While there is no direct correlation between silicate weathering rates or fluxes and riverine Li isotope ratio s(23), it is possible to reconstruct changes in weathering by modelling the observed seawater isotope ratios (34,42,45). Here we use a dynamic ocean box model to constrain possible mechanisms for changing both seawater Li concentration and isotope ratio. This model, in different degrees of complexity, is well-described in several other publications (34,42,45).
It is theoretically possible to reproduce the observed changes in seawater d 7 Li by only changing the hydrothermal input for 100 kyrs by 3.2´ relative to pre-excursion values ( Fig. S8), or for a 200 kyr excursion with a 2.2´ increase. However, such a change in hydrothermal input is significantly greater than has been reported for any time during the Cenozoic, and is therefore discounted in our interpretation. Assuming that the hydrothermal input can be constrained from mid-ocean ridge spreading rates, the PETM hydrothermal input is between 1.15´ and 1.4´ that of the present (46,47,133 because the first half of the excursion is due to a major Li addition, driving a rapid seawater d 7 Li change. The subsequent recovery back to pre-excursion seawater values is strongly dependent on residence time (Fig. S9). As described in the main text, we take the riverine Li flux as a known parameter, by mapping the silicate weathering outputs of two Earth System models onto the Li flux. One of these models is a GENIE model, while the other is a LOSCAR model (3,5 In LOSCAR, carbonate and silicate weathering fluxes are direct outputs. The silicate weathering rate is parameterised as a function of pCO2. Importantly, the strength of the weathering feedback and the initial (steady-state) weathering flux are also controlled by the model (49,50). This allowed for initial testing of the feedback strength for the PETM in the original model (5,9), which in turn effectively allows for parameterisation of regolith thickness.
Both models suggest an initial (steady-state) weathering flux similar to the present day, which also agrees with scenarios suggested by Li and West (2014) (36). However, some other models (47) have suggested that the global weathering flux was low prior to the PETM, because of enhanced soil cover relative to the present day. We test such a scenario by applying an initial riverine weathering flux that is 60% of the modern flux (36,47).
However, in this case, the riverine flux is then low compared to the hydrothermal flux (36,46,133), which results in a weaker control on seawater d 7 Li by river d 7 Li. Therefore, even decreasing d 7 Li of rivers to the lowest possible value (0‰, with a decrease from pre-PETM values of 17‰) only results in a 0.5‰ negative seawater excursion, which is not sufficient to explain the observed data. In other words, the observed data cannot be modelled by starting with a low riverine weathering flux (Fig. S10).  Equally, the observed excursion cannot be reproduced if the trends for modern rivers between d 7 Li, weathering intensity, d 7 Li, and dissolved Li fluxes are followed (main text Fig. 5), and the global regime evolves to a higher intensity regime (i.e. riverine d 7 Li decreases, but flux also decreases - Fig. S11). In modern rivers for which the Li flux is sufficient to affect seawater compositions (i.e. where the weathering intensity is not very high; see Section 6)(25), Li flux and isotope composition co-vary with weathering intensity (Fig. 5, main text). We assume that a similar relationship also existed in temporal trends of the global riverine average across the PETM, as suggested for other time periods (34,42,45), and therefore scale the riverine d 7 Li composition to the modelled riverine Li flux through time. The precise gradient of this scaling varies slightly according to the model, so that the required negative d 7 Li excursion is achieved for a given weathering flux increase. We determine the magnitude of riverine d 7 Li changes required to generate a seawater Li isotope excursion of ~1 ‰ in each model.
We assess model sensitivity by applying a ±10 % uncertainty to the riverine Li flux and a ±1 ‰ uncertainty to the riverine Li isotope ratio, and propagating these uncertainties through the calculations (Fig. S13 and S14).  The change in W/D is then derived from the modern river relationship between W/D, Li yield and d 7 Li (Fig. 5). Given the 'known' increase in W from the Earth System models, we 24 can quantitatively calculate a corresponding increase in D, which equals the sum of W and E (the weathering and erosion rates).   Figure S17. Comparison of the Svalbard section Li isotopes and d 13 C with the secondary clay mineralogy (K = kaolinite; C = chlorite) and initial osmium isotope ratios (Osi) (17,40). The results demonstrate no direct correlation with either secondary mineralogy or weathering source (see Fig. S15).