Eccentricity and obliquity paced carbon cycling in the Early Triassic and implications for post-extinction ecosystem recovery

The timing of marine ecosystem recovery following the End Permian Mass Extinction (EPME) remains poorly constrained given the lack of radiometric ages. Here we develop a high-resolution carbonate carbon isotope (δ13Ccarb) record for 3.20 million years of the Olenekian in South China that defines the astronomical time-scale for the critical interval of major evolutionary and oceanic events in the Spathian. δ13Ccarb documents eccentricity modulation of carbon cycling through the period and a strong obliquity signal. A shift in phasing between short and long eccentricity modulation, and amplification of obliquity, is nearly coincident with a 2% decrease in seawater δ13CDIC, the last of a longer-term stepped decrease through the Spathian. The mid-Spathian shift in seawater δ13CDIC to typical thermocline values is interpreted to record a major oceanic reorganization with global climate amelioration. Coincidence of the phasing shift with the first occurrence of marine reptiles (248.81 Ma), suggests that their invasion into the sea and the onset of a complex ecosystem were facilitated by restoration of deep ocean ventilation linked mechanistically to a change in the response of the oceanic carbon reservoir to astronomical forcing. Together these records place the first constraints on the duration of the post-extinction recovery to 3.35 myr.

The global Smithian-Spathian boundary has not been yet resolved anywhere in the world. In the Chaohu area, South China, the boundary is identified by the first occurrence of conodont Neospathodus pingdingshanensis in the Pingdingshan succession, which has not been correlated globally 6 . Either has ammonoid zonation been widely correlated 7 . A major positive excursion in δ 13 C carb corresponds to a lithofacies boundary at the SSB and provides a significant global correlation tool [8][9] .
Stable isotope analysis. Limestone, mudstone and calcareous shale were sampled at a 10-cm stratigraphic resolution from the uppermost Smithian through upper Spathian succession at Majiashan. Sampling resolution at the Smithian-Spathian transition was 1cm spacing. To rule out diagenetic alteration, samples were trimmed to remove veins, fractures, and the weathered surface and ground to 200-mesh size in ball mill. For limestone beds greater than 1 m-thick, rock powder of five consecutive samples was homogenized to provide 50 cm spacing. Stable isotope analysis of whole-rock samples were analyzed at 24°C ± 1°C using a MAT 253 TM stable isotope ratio mass spectrometer housed in the Isotope Lab of Nanjing Institute of Geology and Paleontology, Chinese Academy of Sciences with an analytical precision of <0.03‰. In order to evaluate for potential diagenetic alteration of the depositional δ 13 C values of bulk-rock samples, we analyzed, for a select subset of the samples used for whole-rock analysis, the δ 13 C carb and wt.% carbonate of microdrilled samples, which were petrographically identified as 'best preserved' micrite, (Fig. S4). Micro-drilled samples (30 to 60 µg) were roasted at 375 °F in vacuo to remove organic volatiles. δ 18 O carb and δ 13 C carb values were determined using a Fisons Optima IRMS with a 90°C Isocarb common acid bath autocarbonate system in the Stan Margolis Stable Isotope Lab, UC Davis. Analytical precision for both δ 18 O and δ 13 C is <±0.1‰ (1σ). All stable isotope values are reported relative to Pee Dee Belemnite (PDB) using standard delta notation (Table S1). The overlap of bulk rock and microdrilled powder values (Fig. S4A), the lack of correlation between δ 18 O carb , δ 13 C carb (Fig.   S4B), wt.% carbonate, and δ 13 C carb (Fig. S4 C) and the coupling of δ 13 C carb and δ 13 C org ( Fig. S5) indicate minimal diagenetic alteration of the observed values and provide confidence that the whole-rock analyses provide a robust record of the seawater δ 13 C carb (cf. Ref. 10,11).
Three major δ 13 C carb shifts, which are synchronous with major lithologic boundaries, define four phases of the δ 13 C carb record with average δ 13 C carb values of -5 ‰, +2‰, 0 ‰, and -2‰ (Fig. 1). These shifts are potentially linked to changes in water column stratification and oceanic circulation conditions. Major long-term δ 13 C carb minima delineate 7 large-scale oscillations in the δ 13 C carb record (Fig. S6). Global comparison of time-equivalent δ 13 C carb records reveals similar long-term features 8,9,12 .
Astrochronologic analysis. To further constrain the observed δ 13 C carb variations and the chronology of the SSB and the marine reptile beds, we performed astrochronologic testing on the high-resolution δ 13 C carb data. The δ 13 C carb data were prepared and analyzed using the R software package "Astrochron" 13 .
(1) Radioisotopic constraints. Available geochronology was used to calculate the approximate duration of the interval and to assess the potential time scale of the observed oscillations in the δ 13 C carb data. A U/Pb age of 250.55 ± 0.51 Ma (2σ analytical+tracer uncertainty) for the earliest Spathian within the ammonoid Tirolites/Columbites Zone 14 from northwestern Guangxi, South China is correlated to the Lower Nanlinghu Formation in the Majiashan section by ammonoid assemblage and a positive δ 13 C excursion (Fig. 1).
The U/Pb ages of 247.38 ± 0.10 Ma (1σ analytical uncertainty) and 247.32 ± 0.08 Ma (1σ analytical uncertainty) for the latest Spathian within the conodont assemblage of N. homeri and N. abruptus from the Great Bank of Guizhou, China 15 cannot be unambiguously correlated to the uppermost Nanlinghu Formation due to the absence of a positive δ 13 C excursion and the low resolution of conodont assemblages 6 . However, an estimated duration of < 3.23 ± 0.60 My for the Nanlinghu Formation can be calculated (see below), thus indicating that the time scale for the observed very strong long-term oscillations in the δ 13 C carb data ( Fig. 1) is 10 5 years.
(2) Data preparation. The long-term trend in the δ 13 C carb data is removed with a LOWESS smoother of 0.08, and subsequently 10 extremely negative "outliers" were removed using a threshold of -2‰. The mean sampling interval of the subsequent data set is 33 cm and the median sampling interval is 19 cm. In order to obtain evenly spaced δ 13 C carb data prior to time series analysis, the detrended data were linearly interpolated to a constant 15 cm resolution. This fine sampling grid is appropriate for the MTM Harmonic F-test that is employed to evaluate peak significance (in contrast to red noise tests, which are sensitive to over-interpolation) ( Fig. S7A-B). simulations. Results with Null Hypothesis Significance Levels (Ho-SL) less than or equal to 0.1% were identified; note that this threshold is more conservative than the critical significance level of 0.5% 16 . Detailed results are shown in Table S3 and  Table S3). The sedimentation rate curve derived from the spatial frequency modulation of the identified short-eccentricity term was integrated to obtain a depth-time map (Fig. S10, Table S4), and the tuned series was interpolated using a sampling grid of 7 kyr (the median sampling interval is 3.5 kyr and mean sampling interval is 5.8 kyr).   2B).
(6) Bandpass filtering and amplitude modulation assessment. The long-term eccentricity and short-term eccentricity cycles (Table S5)         Detailed results are shown in Table S3.