Precise age for the Permian – Triassic boundary in South China from high-precision UPb geochronology and Bayesian age – depth modeling

This study is based on zircon U-Pb ages of 12 volcanic ash layers -and volcanogenic sandstones from two deep water sections with conformable and continuous formational Permian–Triassic boundaries (PTBs) in the Nanpanjiang Basin (South China). Our dates of single, thermally annealed and chemically abraded zircons bracket the PTB in Dongpan and Penglaitan and provide the basis for a first proof-of-concept study utilizing a Bayesian chronology model comparing the three sections of Dongpan, Penglaitan and the Global Stratotype Section and Point (GSSP) at Meishan. Our Bayesian modeling demonstrates that the formational boundaries in Dongpan (251.939 ± 0.030 Ma), Penglaitan (251.984 ± 0.031 Ma) and Meishan (251.956 ± 0.035 Ma) are synchronous within analytical uncertainty of ca. 40 ka. It also provides quantitative evidence that the ages of the paleontologically defined boundaries, based on conodont unitary association zones in Meishan and on macrofaunas in Dongpan, are identical and coincide with the age of the formational boundaries. The age model also confirms the extreme condensation around the PTB in Meishan, which [...] BARESEL, Bjorn, et al. Precise age for the Permian–Triassic boundary in South China from high-precision U-Pb geochronology and Bayesian age–depth modeling. Solid Earth, 2017, vol. 8, no. 2, p. 361-378 DOI : 10.5194/se-8-361-2017


Introduction
The Permian-Triassic boundary mass extinction (PTBME) is considered as the largest mass extinction within the Phanerozoic. About 90 % of all marine species suffered extinction (Raup, 1979;Stanley and Yang, 1994;Erwin et al., 2002;Alroy et al., 2008) and terrestrial plant communities underwent major ecological reorganisation (Hochuli et al., 2010). This major caesura in global biodiversity marked the end of the Palaeozoic faunas and the inception of the modern marine and terrestrial ecosystems (e.g., Benton, 2010;Van Valen, 1984). Several kill mechanisms has been proposed, such as global 1 regression (e.g., Erwin 1990; Yin et al., 2014), marine anoxia (e.g., Feng and Algeo, 2014), ocean acidification (e.g., Payne et al., 2010) or a combination thereof. Rapid global warming (e.g., Svensen et al., 2009), high nutrient fluxes from continent into oceans (Winguth and Winguth, 2012) and increased sedimentation rates (Algeo and Twitchett, 2010) also came into the play, but their respective relations with the global regression near the PTB and the main extinction peak at the PTB remain unclear. In spite of the rapidly growing amount of data, the detailed timing of available diversity estimates and environmental proxies is still lacking, and the ultimate triggers of the PTBME remain elusive. The most likely cause derives from the temporal coincidence with plume-induced massive volcanism of the Siberian Traps (e.g., Burgess and Bowring, 2015) that injected excessive amounts of volatiles (H  (CH 4 ) and contact metamorphism of organic carbon-rich sediments (Retallack and Jahren, 2008;Svensen et al., 2009) are likely to have contributed additional volatiles into the atmosphere, thus deeply altering the climate and the chemical composition of the ocean. This presumably close chronological association has led many authors to support a cause-effect relationship between flood basalt volcanism and mass extinctions. Constraining the timing and duration of the PTBME in a precisely and accurately quantified model that combines relative (i.e., biostratigraphy, environmental changes) and sequences of absolute (zircon geochronology) ages is key to reveal the cascading causes and effects connecting rapid environmental perturbations to biological responses.
The South China block provides a few exceptional marine successions with a continuous stratigraphic record across the PTB (e.g., Yin et al., 2014). Among these is the Global Stratotype Section and Point (GSSP) in Meishan D (Yin et al., 2001), where the PTB is defined by the first occurrence (FO) of the Triassic conodont Hindeodus parvus. Additionally, these South Chinese sections reflect intense regional volcanic activity during Late Permian and Early Triassic times as manifested by many intercalated zircon-bearing ash beds (Burgess et al., 2014;Galfetti et al., 2007;Lehrmann et al., 2015;Shen et al., 2011). High-precision U-Pb zircon geochronology can be applied to these ash beds by assuming that the age of zircon crystallization closely approximates the age of the volcanic eruption and ash deposition. Earliest U-Pb geochronological studies (e.g., Bowring et al., 1998;Mundil et al., 2004;Ovtcharova et al., 2006;Shen et al., 2011) are not sufficiently precise for the calibration of magmatic and biological events. Recent improvements of the U-Pb dating technique by the development of the chemical abrasion-isotope dilution-thermal ionization mass spectrometry (CA-ID-TIMS; Mattinson, 2005), by the revision of the natural U isotopic composition (Hiess et al., 2012), by the development of data reduction software McLean et al., 2011) and by the calibration of the EARTHTIME 202 Pb-205 Pb-233 U-235 U tracer solution (Condon et al., 2015) now provide more accurate weighted mean zircon population dates at the <80 ka level (external uncertainty) for a PTB age, which allow for more precise calibration between biotic and geologic events during mass extinctions and recoveries. Two of the cases benefiting from this improved technique is the highly condensed GSSP defining the PTB at Meishan (Burgess et al., 2014) and the Early-Middle Triassic boundary in Monggan (Ovtcharova et al., 2015).
The aim of this work is 1) to date the PTB in two sedimentary sections that are continuous and significantly more expanded than Meishan, using the highly precise and accurate dating technique of CA-ID-TIMS, and 2) to test the age consistency 2 Solid Earth Discuss., doi:10.5194/se-2016-145, 2016 Manuscript under review for journal Solid Earth minor limestone beds (Fig. 2). This facies association is in agreement with the vast majority of reported occurrences of this formation within the South China block. The Dalong Fm. is interpreted as a basinal depositional environment with restricted circulation and an estimated water depth of 200 m to 500 m (He et al., 2007;. In Guangxi and Guizhou, the thickness of the typical Dalong Fm. is highly variable and ranges from a couple of meters to ca. 60 m. Rocks assigned to the Dalong Fm. in Penglaitan markedly diverge from those of the typical Dalong Fm. In Penglaitan, rocks assigned to Dalong Fm. reach an unusual thickness of ca. 650 m and are lithologically much more heterogeneous, with a marked regressive episode in its middle part (Shen et al., 2007). Moreover, in Penglaitan the Dalong Fm. contains numerous volcanogenic sandstones distributed within the entire succession, a distinctive feature when compared to other sections. Only the lower part of the "Dalong Fm." in Penglaitan can be unambiguously assigned to this formation. The middle and upper part of this section are notably shallower, showing cross bedding and ripple marks in the uppermost 30 m of the Permian, which are underlain by upper shoreface to foreshore facies deposits containing coal seams and abundant plant fossils (Shen et al., 2007). This disparate depositional setting is here interpreted as that of a fault-bounded block successively thrown down and up. Hence, Penglaitan stands in marked contrast with the homogenous deeper water facies of the typical Dalong Fm. in other sections. In Penglaitan, the topmost few meters of the Permian Dalong Fm. comprise thin bedded dark grey limestone intercalated with thick volcanogenic sandstones and thin volcanic ash beds (Fig. 3).
The conformably overlying Early Triassic rocks have been previously assigned to the Luolou Fm. in both Penglaitan and Dongpan He et al., 2007;Shen et al., 2012;Zhang et al., 2006). At its type locality and elsewhere in northwestern Guangxi and southern Guizhou, the base of the Luolou Fm. is invariably represented by shallow water microbial limestone. In contrast, the onset of the Triassic at Dongpan and Penglaitan is represented by ca. 30 m of laminated black shales overlain by several hundred meters of thin bedded, mechanically laminated, medium to light grey limestone. In Dongpan, edgewise conglomerates and breccias are occasionally intercalated within the platy, thin bedded limestone unit. This succession of facies illustrates a change from basinal to slope depositional environments and is identical to that of the Ziyun Fm. at its type locality 3 km East of Ziyun County city, Guizhou Province (Guizhou Bureau of Geology and Mineral Resources, 1987). Therefore, Early Triassic rocks in Dongpan and Penglaitan are here reassigned to the Ziyun Fm., whose base is of Griesbachian age. In most sections in Guangxi and Guizhou, where latest Permian rocks are represented by the Dalong Fm., these are consistently and conformably overlain by basal black shales of the Early Triassic Ziyun Fm. or Daye Fm. (e.g., Feng et al., 2015). In these downthrown blocks, the effects of the Permian-Triassic global regression were comparatively negligible in comparison to those observed in adjacent, up-thrown blocks that recorded pronounced unconformities or condensation..

The Dongpan section
Numerous litho-, bio-and chemo-stratigraphic studies (e.g., Feng et al., 2007;He et al., 2007;Luo et al., 2008;Zhang et al., 2006) (Meng et al., 2002) can easily be recognized in the field. Based on the conodont alteration index (CAI), Luo et al. (2011) established that the section shows only a low to moderate thermal overprint equivalent to a maximal burial temperature of 120°C. Our own estimation of the CAI of conodont elements obtained from the same beds points toward values around 3, thus confirming the estimation of Luo et al. (2011).
Beds 2 to 5 consist of thin (dm to cm) siliceous mudstones, mudstones, minor lenticular limestone horizons and numerous intercalated volcanic ash beds. These beds yield radiolarians, foraminifera (Shang et al., 2003), bivalves (Yin, 1985), ammonoids (Zhao et al., 1978), brachiopods (He et al., 2005), ostracods (Yuan et al., 2007), and conodonts (Luo et al., 2008) of Changhsingian age. Chinese authors have provided very detailed studies of radiolarian occurrences from the top of the Dongpan section, documenting about 160 species belonging to 50 genera Wu et al., 2010;Zhang et al., 2006). Most of these radiolarians belong to the Neoalbaillella optima assemblage zone of Late Changhsingian age (Feng and Algeo, 2014), although it is unclear if some of the Permian taxa reported from the top of the section by previous authors (i.e. above bed 6, Feng et al., 2007) still belong to this assemblage or to a provisional ultimate Permian biozone (Xia et al., 2004).
We collected five samples with visible radiolarians (DGP-1 to DGP-5, see Fig. 2) for this study. Our goal was not to duplicate the detailed faunal studies performed at Dongpan by previous authors, but essentially was to correlate these previous results with our U-Pb ages using own radiolarian data. A selection of well-preserved taxa is illustrated in Appendix C. We also report the occurrence of morphotypes belonging to the genus Hegleria which was previously reported from the section but not illustrated. Our data confirm that radiolarians of the Dongpan section belong to the Neoalbaillella optima assemblage zone.
The conodont fauna obtained from beds 3 and 5 was assigned to the Neogondolella yini interval zone by Luo et al. (2008).
Neogondolella yini is also a characteristic species of the UAZ1 zone, which is the oldest zone of a new high accuracy zonation around the PTB constructed by means of unitary associations (Brosse et al., 2016). Bed 6 is composed of a yellow fine-grained volcanic ash bed and thin-bedded siliceous mudstone. Beds 7 to 12 contain more frequent mudstone and yield a diverse Permian fauna He et al., 2007;. Additionally, He et al. (2007) showed that end-Permian brachiopods underwent a size reduction in the upper-most beds of the Dalong Fm., which they linked with a regressive trend.
The sharp and conformable base (bed 13) of the Early Triassic Ziyun Fm. consists of brown-weathering black shales containing a few very thin (mm-to-cm) volcanic ash beds and volcanogenic sandstones. Previous studies did not recognize how recent weathering superficially altered these black shales. Bed 13 contains abundant bivalves and ammonoids of Griesbachian age He et al., 2007), which are also known from other sections where the equivalent black shales are not weathered. Therefore, the formational boundary placed between beds 12 and 13 is reasonably well constrained in terms of paleontological ages. Even in the absence of any close conodont age control, this boundary has been unanimously acknowledged as the PTB in all previous contributions, thus emphasizing the significance of this formational change.
5 Solid Earth Discuss., doi:10.5194/se-2016-145, 2016 Manuscript under review for journal Solid Earth The Penglaitan section is well known for its Guadalupian-Lopingian boundary (Capitanian-Wuchiapingian GSSP;Jin et al., 2006;Shen et al., 2007). However, the part of the section that straddles the PTB has not been the focus of any detailed published work. Shen et al. (2007) report Changhsingian Peltichia zigzag-Paryphella brachiopod assemblage from a volcanogenic sandstone bed at ~28 m below the PTB. In addition, Palaeofusulina sinensis is abundant in the uppermost limestone units of the Dalong Fm. and conodonts in the topmost part were assigned to the Clarkina yini Zone. A poorly preserved Permian nautiloid was recovered from the volcanogenic sandstone 1.3 m below the PTB (Fig. 3). About 0.3 m above the PTB, concretionary, thin-bedded micritic layers intercalated within the basal black shales of the Ziyun Fm. yielded one P1 element of Hindeodus parvus (Fig. 3). Pending the age confirmation of new paleontological data, and in full agreement with Shen et al. (2007), we place the PTB at this sharp but conformable formational boundary.

CA-ID-TIMS analysis
Sample preparation, chemical processing and U-Pb CA-ID-TIMS zircon analyses were carried out at the University of Geneva. Single zircon grain dates were produced relative to the EARTHTIME 202 Pb-205 Pb-233 U-235 U tracer solution (Condon et al., 2015). All uncertainties associated with weighted mean 206 Pb/ 238 U ages are reported at the 95 % confidence level and given as ±x/y/z, with x as analytical uncertainty, y including tracer calibration uncertainty, and z including 238 U decay constant uncertainty. The tracer calibration uncertainty (y) of 0.03 % (2σ) has to be added if the calculated dates are to be compared with other U-Pb laboratories not using the EARTHTIME tracer solution. The 238 U decay constant uncertainty (z) of 0.11 % (2σ) should be used if compared with other chronometers such as Ar-Ar. All 206 Pb/ 238 U single grain ages have been corrected for initial 230 Th-238 U disequilibrium assuming Th/U magma of 3.00 ± 0.50 (1σ). Th-corrected 206 Pb/ 238 U dates are on average 80 ka older than the equivalent uncorrected dates when applying this correction and are presented as mean ages of selected zircon populations and their associated ±2-sigma uncertainties in Figs. 2 and 3, and as single grain 206 Pb/ 238 U age ranked distribution plots in Fig. 4. The full data table and analytical details are given in Appendix B.

Bayesian chronology
In this study we use Bayesian interpolation statistics to establish a probabilistic age model based on our high-precision U-Pb zircon dates of each individual ash bed and its stratigraphic position, as it is incorporated in the free Bchron R software package (Haslett and Parnell, 2008;Parnell et al., 2008) to constrain the chronological sequence and sedimentation history of the investigated sections. By assuming normal distribution of our U-Pb dates within one sample, and based on the principle of stratigraphic superposition, which requires that any stratigraphic point must be younger than any point situated below in the stratigraphic sequence, it models the age and its associated 95 % confidence interval for any depth point within the studied sedimentary sequence. The model is based on the assumption of constant sedimentation rates, which is allowed to change several times between each dated stratigraphic point. The number of such changes is modeled by a Poisson distribution, and the size of the sedimentation rates by a gamma distribution. The strength of this approach is its flexibility that allows changes in sedimentation rate from zero (hiatus in sedimentation) to very large values (sedimentation event at high rate). In contrast to standard linear regression models, this approach leads to more realistic uncertainty estimates, with increasing uncertainty at growing stratigraphic distance from the dated layers. The model also detects and excludes outliers, which conflict with other evidence from the same sequence in order to produce a coherent and self-consistent chronology; no predefined outlier determination is required from the user. One of the drawbacks of this Bayesian approach is that a change in the sedimentation rate is assumed to occur at each dated stratigraphic position, though it is unlikely that the change in sedimentation occurs exactly at the depth of a dated bed. Another drawback is that the sedimentation parameters are shared across the whole sequence. In consequence, Bchron does not allow much opportunity for users to individually influence the chronology behaviour.
In this study we use the Bayesian Bchron model as it is part of the Bchron package (http://cran.rproject.org/web/packages/Bchron/index.html). This model outperforms other Bayesian age depth-models, as shown by a extensive comparison conducted on radiocarbon dates from Holocene lake sediments (Parnell et al., 2011). It provides a non- iterations.

Samples
In total, 12 volcanic ash beds and volcanogenic sandstones were sampled from the Dalong Fm. of Late Permian age and from the overlying Ziyun Fm. of Early Triassic age at Dongpan and Penglaitan (see Appendix A). Most of the dated samples exhibit 206 Pb/ 238 U age dispersions that exceed the acceptable scatter from analytical uncertainty and are interpreted as reflecting magmatic residence or a combination of the latter with sedimentary recycling. Only in two cases (DGP-16, PEN-22) we find single grain analyses younger than our suggested mean age and interpret them as unresolved Pb loss since they strongly violate the stratigraphic order established by the chronology of the volcanic ash beds.
At Dongpan, six fine-to medium-grained volcanic ash beds (DGP-10, DGP-11, DGP-12, DGP-13, DGP-16 and DGP-17) in the upper-most 10 m of the Dalong Fm., one fine-grained ash bed (DGP-21) just 10 cm above the base of the Ziyun Fm., and one thin-bedded volcanogenic sandstone (DGP-18) 40 cm stratigraphically higher were collected for geochronology. At Fm., were dated. A single fine-grained and extremely thin (2-3 mm) volcanic ash bed (PEN-22) was sampled 50 cm above the base of the Ziyun Fm. and thus closely brackets the formational boundary. U-Pb CA-ID-TIMS geochronology following procedures described above and in the appendix was applied to a number of single crystals of zircon extracted from these volcanic ash beds. Trace element and Hf isotopic compositions of these dated zircons will be presented elsewhere.

Results
The U-Pb isotopic results are presented in Fig. 4 as 206 Pb/ 238 U age ranked plots for each individual sample, and in Table B1 (Appendix).

Sample DGP-10
This volcanic ash bed was sampled 9.7 m below the formational boundary. All ten dated zircons are concordant within analytical error, where the seven youngest grains define a cluster with a weighted mean 206 Pb/ 238 U age of 252.170 ± 0.055/0.085/0.28 Ma (mean square of weighted deviates [MSWD] = 1.18) for the deposition of DGP-10.

Sample DGP-11
This volcanic ash bed was sampled 7.9 m below the formational boundary. Eleven zircon crystals were analyzed, resulting in scattered 206 Pb/ 238 U dates of 251.662 ± 0.263 Ma to 252.915 ± 0.352 Ma. The six youngest zircons yield a weighted mean 206 Pb/ 238 U age of 251.924 ± 0.095/0.12/0.29 Ma (MSWD = 1.80) that is too young with respect to the stratigraphic sequence defined by over-and underlying ash beds. Therefore, we have to assume that abundant unresolved lead loss affected these zircons, despite application of the same chemical abrasion procedure as for all other samples. It is worth noting that all zircons from DGP-11 were almost completely dissolved after chemical abrasion and show elevated 206 Pb/ 238 U age uncertainties of ~0.30 Ma compared to other volcanic ash beds from Dongpan.

Sample DGP-12
This volcanic ash bed was sampled 7.3 m below the formational boundary. The weighted mean age of 252.121 ± 0.035/0.074/0.28 Ma (MSWD = 1.04) is derived from eight concordant grains representing the youngest zircon population of this ash bed. (MSWD = 0.13). Because zircon dates from this bed spread over almost 2 Ma, recycling of older volcanic material via sedimentary processes appears more likely than via magmatic recycling.

Sample PEN-70
This volcanic ash bed was sampled 0.6 m below the formational boundary. Eighteen zircon grains were analyzed. As in the case of PEN-6, they yield a scatter of 206 Pb/ 238 U dates spanning 1.5 Ma, ranging from 251.994 ± 0.169 Ma to 253.371 ± 0.165 Ma. The weighted mean age of 252.125 ± 0.069/0.095/0.29 Ma (MSWD = 0.59) for the deposition of this ash bed is calculated by using the seven youngest concordant grains.

Sample PEN-28
This sample was taken 0.3 m below the formational boundary. It is derived from the base of a 30 cm thick volcanogenic sandstone which represents the youngest Permian bed in Penglaitan. Analyses of seven zircon grains yield a cluster with a weighted mean 206 Pb/ 238 U age of 252.062 ± 0.043/0.078/0.28 Ma (MSWD = 0.49), reflecting the last crystallization phase of this zircon population. Six older grains ranging from 252.364 ± 0.156 Ma to 253.090 ± 0.375 Ma indicate either magmatic or sedimentary recycling. The U-Pb data of PEN-28 was already published in Baresel et al. (in press).

Sample PEN-22
This 2 mm thick volcanic ash bed was sampled 0.5 m above the formational boundary. Eight zircons yield a weighted mean 206 Pb/ 238 U age of 251.907 ± 0.033/0.073/0.28 Ma (MSWD = 0.10). One zircon grain shows a significantly younger age suggesting lead loss. Two slightly older grains reflect noticeable pre-eruptive crystallization. Incorporation of these grains into the weighted mean calculation would lead to an excessive MSWD of 3.6 and 1.9, respectively. However, we noticed that some volcanic ash beds and volcanogenic sandstones in these sections show a large age dispersion of up to 2 Ma, incompatible with recycling of zircon that previously crystallized within the same magmatic system and became recycled into later melt batches, leading to few 100 ka dispersion of dates (e.g., Broderick et al., 2015;Samperton et al., 2015), but pointing to sedimentary reworking. The U-Pb data of PEN-22 was already published in Baresel et al. (in press). Figure 5 shows a comparison of three different age-depth models based on linear interpolation, cubic spline interpolation and Bayesian statistics, each applied to exactly the same U-Pb dataset of Dongpan (Fig. 5a) and Penglaitan (Fig. 5b). It is visible that, the Bayesian Bchron model produces a slightly increased uncertainty of the model age with increasing distance from the stratigraphic position of a U-Pb dated sample (as discussed in the Methods' section). Due to the well constrained U-Pb dates of Dongpan and Penglaitan, all three age-depth models predict (within uncertainty) similar ages for the PTB in Dongpan (Fig. 5a) and Penglaitan (Fig. 5b). Given that the Bayesian Bchron model evaluates the age probability distribution In contrast to the other two models, the Bayesian Bchron model can identify U-Pb dates that violate the principle of stratigraphic superposition, as shown for the Dongpan ash beds DGP-11 (outlier probability of 67 %) and DGP-18 (outlier probability of 100 %). Including them into the age-depth chronology of Dongpan results in unrealistic negative sedimentation rates, as reflected by the linear and cubic interpolation models for the interval between DGP-11 and DGP-12, and for the interval between DGP-21 and DGP-18 (Fig. 5a).

Age-depth models
The aim for applying Bayesian age modelling to the dated volcanogenic beds from these two sections was to obtain an age model for the PTB. The age-depth models yield ages of 251.938 ± 0.029 Ma (Dongpan; Fig These two ages overlap within uncertainties and thus demonstrate the synchronicity of the PTB in the two sections. Making the reasonable assumption of absence of significant gaps in these two sections, the new U-Pb dates can be used to infer sedimentation rates. The age-depth model of Dongpan suggests increased sedimentation rates in the upper-most part of the Dalong Fm. from bed 6 (DGP-17) upwards. Below bed 6, calculated sedimentation rates appear to be relatively constant with 3.6 ± 1.2 cm ka -1 , but above bed 6 they jump to 6.0 ± 2.4 cm ka -1 . In Penglaitan, the sedimentation rate of the uppermost Dalong Fm. and basal-most Ziyun Fm. is significantly lower than in Dongpan with 0.7 ± 0.3 cm ka -1 . Previously published U-Pb zircon geochronology from Penglaitan (Shen et al., 2011), including a weighted mean date of 252.16 ± 0.09 Ma from a volcanogenic sandstone at 26.7 m below the PTB, was not considered in our age model, since substantial improvements in the analytical protocol hamper comparing these dates with our U-Pb results.

The change of the PTB age through analytical improvement of U-Pb dating
The first geochronological studies in the GSSP Meishan D have been carried out on bed 25, immediately underlying the PTB, by U-Pb sensitive high-resolution ion microprobe (SHRIMP) analysis of zircons yielding a 206 Pb/ 238 U age of 251.2 ± 3.4 Ma (Claoué-Long et al., 1991) and by 40 Ar/ 39 Ar dating of sanidine at 249.91 ± 0.15 Ma (Renne et al., 1995). However, these dates are either not sufficiently precise to allow calibrating magmatic and biological timescales at resolution adequate for both groups of processes, or are biased by a systematic age offset between the U-Pb and Ar-Ar systems of ~1.0 % (Schoene et al., 2006). In order to properly compare the two systems, all older 40 Ar/ 39 Ar data have to be corrected for the revised age of the standard Fish Canyon sanidine of 28.201 ± 0.046 Ma (Kuiper et al., 2008) and the decay constant uncertainty has to be added to U-Pb and Ar-Ar ages which would drastically expand the 40 Ar/ 39 Ar age error and recalculate the 40 Ar/ 39 Ar age of Renne et al. (1995) to 252.1 ± 1.6 Ma. In a first detailed ID-TIMS study, U-Pb ages of mechanically abraded zircons were published by Bowring et al. (1998)  0.3 Ma. Though much more precise than former studies, these ages are mainly based on multi-grain zircon analyses. That this approach might disguise complexity of zircon population ages, as pervasive lead loss and inheritance, was shown by Mundil et al. (2001) by confining data selection to single-crystal analyses of the same horizons. In a second attempt, driven by further improvements of the U-Pb ID-TIMS technique (e.g., chemical abrasion of zircon grains by hydrofluoric acid exposure to remove zircon domains with lead loss; reduced procedural common Pb blanks), the PTB extinction horizon in

Comparison of lithostratigraphy
All three interpolated ages of the formational boundary in Dongpan (251.938 ± 0.029 Ma), Penglaitan (251.982 ± 0.031 Ma) and Meishan (251.956 ± 0.033 Ma) are in agreement within errors (Fig. 6). They support the synchronicity of the conformable boundary between the Dalong Fm. and the Ziyun Fm. in Dongpan and Penglaitan, and also demonstrate their temporal coincidence with the conformable boundary in Meishan between the Changhsing Fm. and Yinkeng Fm. The age model also confirms the extreme condensation around the PTB in Meishan, with a maximal sedimentation rate of 0.4 cm ka -1 as reported by Burgess et al. (2014) for the 26 cm thick interval bracketed by beds 25 and 28. In this respect, Dongpan and Penglaitan offer a greater potential for higher resolution studies of environmental proxies around the PTB with maximal sedimentation rates for the same interval of 8.4 cm ka -1 and 1.0 cm ka -1 , respectively. The increased sedimentation rate above bed 6 in Dongpan is in agreement with the previously inferred sedimentary fluxes deduced from elemental chemical analyses . From bed 7 upward, He et al. (2007) showed a clear increase of Al 2 O 3 and TiO 2 indicating increased fluxes of terrestrial input into this trough. The accompanying size reduction of brachiopods (He et al., 2007)  parvus (bed 27c) does. Available conodont data from Meishan allow the assignment of bed 24a-e to UAZ1 (UAZ1 might reach further down as indicated by a dashed line in Fig. 6), bed 25 to UAZ2, bed 27a-d to UAZ3 and bed 28 to UAZ4 (Brosse et al., 2016). The stratigraphic thickness comprised between the base of UAZ1 and the top of UAZ4, amounts to 1.22 m. By using the three section age-depth models, we attempted to project the respective thickness corresponding to the UAZ1-UAZ4 interval in Meishan onto the two other sections. This projection resulted into a pronounced, artificial lengthening of UAZs in Dongpan and Penglaitan. UAZ1 is the penultimate Permian conodont UAZ in Meishan (Brosse et al., 2016). When projected onto the age-depth models of Dongpan and Penglaitan, this UAZ1 is artificially expanded and even crosses the PTB in Penglaitan (Fig. 6). In Penglaitan, the last Permian UAZ2 projects correctly above UAZ1 without overlap but is completely within the Triassic. The cause of these contradictions stems from the irreconcilable conjunction of i) extreme condensation in Meishan, ii) high evolutionary rates of conodonts, and iii) the ca. 30 ka precision of the last generation of U-Pb dates.
In Dongpan, the onset of a protracted radiolarian diversity decline in bed 5 reported by Feng and Algeo (2014) is here interpolated at 251.990 ± 0.027 Ma, occurring 52 ± 40 k.y. before the formational boundary (Fig. 2). Excess SiO 2 values of this bed  suggest a genuine diversity pattern at the local scale, which seems to be unrelated to any

Comparison of chemostratigraphy
Organic carbon isotope chemostratigraphy of Dongpan (Fig. 7) extending from the Permian bed 5 to the Triassic bed 13 was provided by Zhang et al. (2006) and for Meishan (Fig. 7) extending from the Permian bed 24 to the Triassic bed 29 by Cao et al. (2002). The correlation of these δ 13 C org records by Zhang et al. (2006), based on the occurrence of ash beds in both sections, is largely over-interpreted. With the exception of a short negative excursion followed by a more prominent positive excursion between beds 9 and 11, the Permian part of the δ 13 C org record in Dongpan is relatively stable and oscillates between -28 ‰ and -27 ‰. With the exception of a negative excursion culminating in beds 25 and 26, the Permian part of the δ 13 C org record in Meishan shows a sustained positive trend from -30 ‰ to -26 ‰. The basal Triassic part of these two records is also incompatible in that they display opposed trends. With the possible exception of the Xinmin section (Shen et al., 2013a), the δ 13 C org record of Dongpan does not correlate with that of any other South Chinese section, but even Xinmin shows a ~3 ‰ offset of the base trend in comparison to Dongpan. However, we note that in Meishan an abrupt decline in δ 13 C carb occurs in bed 24e at 251.950 ± 0.042 Ma (Burgess et al., 2014) and slightly above in bed 26 in δ 13 C org at 251.939 ± 0.032 Ma, which is temporally coincident with the main negative δ 13 C org excursion in bed 9 in Dongpan at 251.954 ± 0.027 Ma. The second smaller negative δ 13 C org excursion at the PTB in Dongpan at 251.942 ± 0.028 Ma and in Meishan at 251.892 ± 0.045 Ma cannot be distinguished within uncertainty from the main excursion, which hampers the correlation of the δ 13 C org records based on U-Pb ages. However, interpreting organic carbon records requires the simultaneous analysis of palynofacies, which are not documented in Dongpan. Shen et al. (2012) also showed that the total organic carbon (TOC) never exceeds 0.2 %, thus indicating a generally poor preservation of the organic matter in this section. As shown by Shen et al. (2012), this preservation bias is further supported by coincident peaks in both terrestrial (spore and pollens) and marine (algae and acritarchs) organic material (see Fig. 7). This uneven preservation of the organic matter further hampers the understanding of the δ 13 C org signal in Dongpan. More generally, the consistency and lateral reproducibility of the Late Permian carbonate and organic carbon isotope records in South China remain equivocal (e.g., Shen et al., 2013b). These records are probably influenced by the local graben and horst paleotopography that hampered efficient circulation of water masses with the open ocean, thus reflecting more local than global changes. • Applying Bayesian age modelling to sections with such high-precision time information allows to compare disparate information from different sections, quantitatively. Along with the work of Ovtcharova et al. (2015) at the Early-Middle Triassic boundary, these are the two first proof-of-concept studies adopting age-depth modelling to compare coeval sections with different fossil contents, different facies and disparate sedimentation rates at highest temporal resolution. We anticipate that this approach will need to become the future standard in the assessment of the Geologic Time Scale. parvus.
• The higher sedimentation rates of Dongpan and Penglaitan provide a much better prospect for the high-resolution study of environmental proxies around the PTB than the condensed GSSP section in Meishan. Our age-depth models also reveal that the combination of condensed deposition with high evolutionary rates of conodonts and the ~30 ka resolution of the last generation of U-Pb ages makes it impossible to project stratigraphic data points or intervals of Meishan onto expanded PTB sections without distortions. This intrinsic problem of the Meishan GSSP section should stimulate the search of alternative sections with more expanded records.
• The seemingly erratic Late Permian carbon isotope record in South China does not allow laterally reproducible intercalibration with the newly obtained U-Pb dates. This stands in sharp contrast with the Early Triassic carbon isotope record which is of global significance (e.g., Galfetti et al., 2007).

Fig. 2. Stratigraphy and geochronology for the Dongpan section from late Changhsingian to Griesbachian showing weighted mean 206 Pb/ 238 U dates of the volcanic ash beds and volcanogenic sandstones. U-Pb data of DGP-21 is taken from Baresel et al. (in press). Investigated radiolarian samples (DGP-1 to DGP-5) are shown in their stratigraphic positions. The Bayesian Bchron age-depth model is presented with its median (middle grey line) and its associated 95 % confidence interval (grey area). Radioisotopic dates together with their uncertainty (red horizontal bars) are presented as 206 Pb/ 238 U weighted mean dates of the dated volcanic ash beds in their stratigraphic positions. Predicted dates for the onset of the radiolarian decline (RD) and the Permian-Triassic
Boundary (PTB) are calculated with their associated uncertainty using the Bayesian Bchron age-depth model assuming stratigraphic superposition.

728
The Dongpan section is situated at 22°16'11.80''N and 107°41'31.30''E, north-east  The samples were crushed and milled, and the powder was wet-sieved to remove the clay fraction. Heavy minerals were isolated using methylene iodide. Single zircons were microscopically inspected and euhedral crystals were picked for annealing at 900°C for ~48 h, followed by chemical abrasion with 40 % HF and trace HNO 3 in pressurized 200 µl Savillex mini-capsules at 180°C for 18 h to minimize Pb loss effects (Mattinson, 2005). Since Ovtcharova et al. (2015) still revealed apparent Pb loss in some zircon grains after 15 h of chemical abrasion, this study was optimized to the longer duration of 18 h to effectively overcome this obstacle.
After several washing steps with water, 6 N HCl, and 3 N HNO 3 , single crystals were loaded in 200 µl Savillex capsules, spiked with ~4 mg of the EARTHTIME 202 Pb-205 Pb-233 U-235 U tracer solution (hereafter referred to as ET2535; Condon et al., 2015) and dissolved in ~70 μl 40 % HF and trace HNO 3 at 210°C for 48 h. After dissolution, samples were dried and redissolved in 6 N HCl at 180°C for 12 h, dried down again and re-dissolved in 3 N HCl. U and Pb were collected in 3 ml Savillex beakers after separation in a modified single 50 μl column anion exchange chemistry (Krogh, 1973) and dried down with a drop of 0.05 M H 3 PO 4 . They were loaded on a single outgassed Re filament with a Si-gel emitter modified from Gerstenberger and Haase (1997). Measurements of U and Pb isotopes were performed on a Thermo TRITON thermal ionization mass spectrometer utilizing the ET2535 tracer calibration version 3.0 defined by Condon et al. (2015). Pb isotopes were measured in dynamic mode on a MasCom secondary electron multiplier with a deadtime of 23 ns. Instrumental mass fractionation was corrected using the fractionation factor derived from the measured 202 Pb/ 205 Pb ratio relative to a true value of 0.99924. BaPO 2 interferences on mass 202 to 205 were corrected by determining 138 Hiess et al., 2012). Raw data were statistical filtered by using the Tripoli program, followed by data reduction including correct uncertainty propagation and online data visualization using U-Pb_Redux software McLean et al., 2011). U-Pb ratios and dates were calculated relative to a tracer 235 U/ 205 Pb ratio of 100.23 ± 0.046 % (2σ; Condon et al., 2015). All common Pb in the analyses was assumed to be procedural  b % discordance = 100 -(100 * (206Pb/238U date) / (207Pb/206Pb date)). c Th contents calculated from radiogenic 208Pb and the 207Pb/206Pb date of the sample, assuming concordance between U-Th and Pb systems. d Total mass of radiogenic Pb. e Total mass of common Pb. f Measured ratio corrected for fractionation and spike contribution only. g Measured ratio corrected for fractionation, tracer and blank.°° Corrected for initial Th/U disequilibrium using radiogenic 208Pb and Th/U magma = 3.00.
* Samples marked in red are from a previous study (Baresel et al., in press).