Late Eocene-early Oligocene paleoenvironmental changes recorded at Lühe, Yunnan, southwestern China

1 During the late Eocene to the early Oligocene, marine records document a globally 2 congruent record of declining carbon dioxide concentrations, Antarctic icesheet growth, and 3 associated reorganisation of the global climate system. In contrast, the few existing 4 terrestrial records demonstrate high heterogeneity of environmental change and are difficult 5 to reconcile with those of the oceanic realm. Global drivers for climatic change are 6 particularly difficult to disentangle from regional ones, especially those caused by the 7 complex tectonic evolution of the Tibetan region and its influence on the Asian monsoon 8 system and vegetation. Here, we reconstruct the climatic and environmental history from the 9 late Eocene into the early Oligocene at Lühe Basin, Yunnan, China, a key sedimentary 10 repository along the SE margin of the Tibetan Plateau and an important region for assessing 11 Asian monsoon changes. We investigate a 340-m long section via a multi-proxy approach 12 and climate model simulations. The organic geochemical proxies, via n-alkanes, terpenoids, 13 and hopanes, suggest that thermally immature sediments were deposited in a terrestrial 14 flood plain basin that was primarily occupied by gymnosperms and angiosperms. Branched 15 glycerol diakyl glycerol tetraethers indicate relatively stable temperatures (ca. 10 °C) 16 throughout the section, including across the Eocene-Oligocene boundary. This temperature, 17 cooler than the modern-day average for this site (ca. 15 °C), suggests that this area has not 18 undergone significant uplift since the Oligocene. To further contextualize our data, we tested 19 a suite of climate model simulations with varying pCO2, paleogeography, and Tibetan 20 topography across the Eocene-Oligocene boundary. This data-model comparison suggests 21 that a response to regional factors might explain the absence of a pronounced cooling at 22 Lühe across the Eocene-Oligocene boundary, supporting the emerging picture that the 23 global expression of the EOT in terrestrial environments is more complex than indicated by 24 the marine record. 25 26


Highlights: 28
▪ Depositional environment primarily terrestrial flood plain basin, with gymnosperms 29 ▪ Relatively stable mean annual temperatures (ca. 10 °C) across EOT 30 ▪ Eastern Tibet at its current height since at least the EOT 31 ▪ Data-model comparison suggests regional factors may explain lack of cooling at EOT 32 1. Introduction the early Paleogene, is likely linked with regional climatic responses, especially in the Asian 61 of feldspars within volcanic ashes exposed in the lower portion of the coalmine provides an 103

Sample preparation 116
A total of 56 samples were analysed for organic geochemistry in order to determine 117 the preservation state of the sediments, the paleoclimatic conditions, and the 118 paleovegetation. Samples were extracted using a microwave-assisted extraction system 119 with dichloromethane (DCM) and methanol (MeOH) (9:1 v/v). The resulting total lipid extract 120 (TLE) was eluted with alumina column chromatography into an apolar fraction using

Indices for thermal maturity 125
The apolar fraction contains compounds predominantly derived from plant, algal, and 126 bacterial communities. Bacteria-derived hopanes and eukaryote--derived n-alkanes were 127 used to assess the degree of thermal maturity of the organic matter preserved in the 128 sediments, as high thermal maturity may bias the preservation of organic matter and thus 129 the fidelity of our reconstructions. Here we calculated the stereochemistry of the C 31 hopane 130 alkanes can be indicative of the organic matter source and is calculated as ACL = Σ(C n × n) / 142 Σ(C n ) (Eglinton and Hamilton, 1967), here based on odd n-alkane chain-lengths from C 21 143 through C 33 . The P-aqueous ratio (P aq , calculated as P aq = (C 23 + C 25 ) / (C 23 + C 25 +C 29 + 144 C 31 ), Ficken et al., 2000) and the C 23 / (C 23 + C 31 ) index (Nott et al., 2000) are generally 145 associated with wetland conditions, given that C 23 and C 25 n-alkanes are produced by 146 Sphagnum mosses and some submerged vascular macrophytes but are generally absent in 147 higher plants. CPI The apolar fractions were used to estimate thermal maturity (Figs. 2, 3). The C 31 227 hopane configuration ratio of ββ / (ββ + αβ + βα) ranges from 0.0 to 0.7 (from high to low 228 thermal maturity, respectively) with a mean of 0.4 ± 0.2 σ (Fig. 3A). Values are slightly lower 229 in the bottom ~30 m of the section. Although variable, most values are over 0.3 and there is 230 no consistent trend through the section. Instead, it appears that the organic matter is 231 relatively immature with an admixture of mature, reworked organic matter in some low-TOC 232 intervals. The CPI ranges from 1.9 to 9.4 with a mean of 4.9 ± 1.6 σ (Fig. 3B). These CPI 233 values likewise suggest that these sediments are relatively immature, although the variation 234 reflects the dynamic depositional environment. 235

Vegetation and environmental reconstruction 236
Throughout the section, the n-alkane distribution shows a strong odd-over-even 237 preference (Fig. 3B), with a CPI ranging from 1.9 to 9.4 with a mean of 4.9 ± 1.6 σ, 238 suggesting this is primarily terrestrial in origin. In most of these sediments, the apolar 239 fractions are dominated by the C 29 n-alkane, followed by a high abundance of the C 27 and 240 then C 31 n-alkanes (Figs. 2A; 4), suggesting dominance of higher plants. The ACL ranges 241 from 26.1 to 29.6 with a mean of 28.4 ± 0.6 σ (Fig. 3C). This relatively high CPI (Fig. 3B), 242 high ACL (Fig. 3C), and dominance of the C 29 n-alkane ( Fig. 2A, 4) suggests that the 243 vegetation at this site was likely dominated by woody angiosperms and gymnosperms. More 244 specifically, the ACL of 28.4 is more likely associated with deciduous rather than evergreen 245 angiosperms. 246 Several samples also contained diterpenoids and triterpenoids ( Fig. 2A) and 25.3 respectively, relative to the average of 27.6 ± 0.8 σ) and the P aq being particularly 287 high (0.9 and 0.8 respectively, relative to the average of 0.3 ± 0.2 σ). Although these two 288 sediment depths still contain n-alkanes with a strong odd-over-even preference and long 289 chain-lengths associated with higher plants (i.e., C 27 , C 29 , and C 31 ), they show a clear C 23 290 and C 25 dominance, which is considered a robust signature for either Sphagnum peat 291 mosses (Nott et al., 2000) or aquatic plants (Ficken et al., 2000). Therefore, these two outlier 292 horizons may represent swampy environments or even open water conditions. Interbedded 293 lignites found throughout the section further confirm that this was (at times) a peat-forming 294 environment. This environment is consistent with a riverine floodplain, as also supported by 295 the presence of sedimentary structures of river channels and sedimentological evidence of 296 intervals of water-logging conditions. 297 Taken together, our biomarker results are compatible with a terrestrial environment 298 (likely a flood plain) with organic-rich soils derived from swamps, colluvium, occasional peat-299 forming, and wet areas. We do see evidence for abundant terrestrial biomarkers (e.g., leaf 300 waxes, terpenoids indicative of woody gymnosperms and angiosperms, and soil bacterial presence of algal biomarkers). The specific higher plant biomarkers are also consistent with 303 this area being covered in deciduous and evergreen broad-leaved mixed forests, as 304 observed in the nearby Lühe town section (Tang et al., 2020). 305

Climate reconstruction using GDGT indices 306
Lithologies in the Lühe coalmine section vary; lithologies include sands, mudstones, 307 and coal/lignite layers, and are interbedded with fossil remains of wood (logs and branches) 308 and leaves (Fig. 3). Such lithological variability is indicative of a dynamic paleoenvironment, 309 likely a flood plain, where deposition of swamp-derived organic-rich soils were interspersed 310 with colluvium, occasional peat mire, and shallow stagnant environments (Xu et al., 2008). 311 This dynamic environment poses a challenge for the application of a univocal brGDGT 312 paleothermometer calibration. Therefore, we rely on a series of characteristics stemming 313 from field observations, TOC (wt%) data, and organic geochemical analyses to constrain the 314 type of lithology and paleoenvironment, and then apply three different brGDGT-temperature 315 calibrations (Fig. 5). 316 Most sediments (n = 46) have a TOC (wt%) ranging between 0.1-23 % with the 317 majority < 3% (Fig. 5), consisting of the lithologies categorised from mudstone to silty 318 sandstones. The remaining sediments (n=10) have TOC (wt%) ranging between 40-63 %; 319 these high TOC ranges are indicative of organic-rich environments, consistent with the 320 presence of coal layers identified in the stratigraphy. Samples from sand lithologies were 321 tested and excluded from the sample set as they did not yield sufficient organic matter for 322 analysis (Fig. 5). 323 Of all samples analysed (n = 56), 38 yielded sufficient brGDGTs for 324 paleotemperature estimates (Fig. 5). For samples with TOC (wt%) <23 % (n = 33), we 325 cannot further constrain the type of paleoenvironment and/or the source of bacterial 326 production (e.g., lacustrine vs soil). Thus, we apply both the MAT and pH soil calibration by samples with TOC (wt%) >23 % (n = 5), identified as lignites/coals, we further apply the 329 peat-specific temperature and pH calibration by Naafs et al (2017b) (Fig. 5).

Climate model results 395
We employed a fully coupled atmosphere-ocean GCM with a range of perturbed 396 Priabonian and Chattian boundary conditions to investigate potential mechanisms for our 397 temperature proxy results. We tested the effects of different parameters on temperature and 398 precipitation in this region (Table 1) (0°N-60°N, 60°E-120°E), as seen in Table 1, considering changes in pCO 2 , global 405 paleogeography, and the site elevation as variables across the E-O boundary. For the 406 regional impact, a decrease from 4x to 2x pre-industrial pCO 2 across the E-O boundary results in regional temperatures cooling by ~6 °C, regardless of topographic changes of the 408 Tibetan Plateau (valley, plateau, or valley to plateau; Table 1, simulations 1-3). When 409 assuming no change in pCO 2 , global changes in paleogeography from a Priabonian to 410 Chattian configuration produce a reduced impact on the climate of Asia, with a slight 411 increase in MAT of 1.5 °C (Table 1, simulations 4-6). This is the result of regional changes to Assuming that the stratigraphy at this site does encompass the E-O boundary, the 450 temperature response at Lühe might reflect paleogeographic changes rather than the effect 451 of a rapid and large drop in pCO 2 , as observed in the regional scale simulations (Table 1, (Table 3). However, the overall precipitation is likely 472 overestimated in CLAMP, and particularly for the dry months in warm climates, because 473 water is not a limiting growth factor for plant growing near to aquatic depositional sites 474 (Spicer et al., 2011).