Strong seasonality in precipitation, with wet summers and dry winters, characterizes the monsoon climate across East Asia today. Models indicate that monsoon intensity, measured as the amount of summer rainfall, may increase with global warming (Takahashi et al., 2020), but projected changes have large uncertainties and substantial intermodel variability (Chevuturi et al., 2018), and the role of anthropogenic activity (i.e., aerosol emission) on monsoon strength is still highly debated (Bollasina et al., 2011;Dong et al., 2019;Kim et al., 2016;Mu and Wang, 2021). Paleoclimate records from intervals of warmer global climate may therefore be informative for understanding monsoon dynamics in a warmer world. Analysis of paleo-monsoon strength is commonly inferred from oxygen isotope measurements of speleothems (Cheng et al., 2016;Cosford et al., 2008;Liu et al., 2020), but these records are limited to the Quaternary, when CO 2 levels were lower than today. Studies of monsoon strength in deep time (pre-Quaternary) indicate a strong link between CO 2 and rainfall dating back to the Eocene (Licht et al., 2014), while others highlight the important role of paleogeography on monsoon intensity (Farnsworth et al., 2019). Although the mechanism is debated, recent work provided firm evidence for an East Asian Monsoonstyle system, with wet summers relative to winters, prior to the Neogene (Vornlocher et al., 2021), but quantitave estimates of the amount of wetseason rainfall in the late Oligocene are lacking. Here we provide new measurements of the oxygen isotope value of cellulose (δ 18 O cell ) extracted from modern and late Oligocene fossil wood in order to determine how monsoon rainfall during the late Oligocene compares to today. The late Oligocene is an ideal period to study the effect of global temperature on monsoon intensity because it represents the likely trajectory for Earth's climate system as CO 2 emissions continue to rise unabated (Westerhold et al., 2020). It also represents a critical period in central Tibetan uplift (Fang et al., 2020;Su et al., 2019), and therefore is an essential yet understudied interval of time for investigating Asian monsoon dynamics in deep-time.
Determination of δ 18 O cell value from living trees has long been used to quantify recent climate change (Brienen et al., 2012;Knorre et al., 2010;Loader et al., 2010;Loader et al., 2007;McCarroll and Loader, 2004;Poussart et al., 2004;Rinne et al., 2013;Saurer et al., 2008;Treydte et al., 2006;Young et al., 2015), including in monsoon regions of East Asia (e.g., Sakashita et al., 2016;Xu et al., 2016). Application of these techniques to well-preserved fossil wood collected from pre-Quaternary sediments has allowed for new quantitative climate data in deep time (Jahren and Sternberg, 2002, 2003, 2008;Richter et al., 2008a;Wolfe et al., 2012). These studies all rely on empirical relationships between the stable oxygen isotope composition of tree-ring cellulose (δ 18 O cell ) and that of meteoric water (δ 18 O MW ) (Csank et al., 2013;Olson et al., 2020;Richter et al., 2008b;Saurer et al., 2016;Sternberg et al., 2007;Waterhouse et al., 2002), which can then be related to precipitation or temperature depending on the environment in which the plant grew: δ 18 O MW values in low-latitude sites respond more to rainfall amount (Araguás-Araguás et al., 1998;Brienen et al., 2012;Dayem et al., 2010;Gonfiantini et al., 2001;Xie et al., 2011), while δ 18 O MW values in middle to high-latitude sites are affected more by changes in temperature (Bowen, 2008;Dansgaard, 1964;Schubert and Jahren, 2015). We leverage these earlier works to present the first estimates for wet-season rainfall in southern China during the late Oligocene based on the δ 18 O cell value of well-preserved wood fossils, thus providing data on seasonal rainfall intensity in this monsoon region under a warmer climate state.
Mummified fossil wood samples were collected from the upper Yongning Formation near Nanning in the Guangxi Zhuang Autonomous Region of South China (22 • 52 ′ 50" N, 108 • 25 ′ 2 ′′ E; elevation = ~80 m; Fig. 1) (Quan et al., 2016). The upper Yongning Formation is a lacustrine deposit dated to the late Oligocene based on palynology and mammalian fossils (Quan et al., 2016). For this study, we selected 64 mummified wood samples representing the genus Castanopsis (Huang et al., 2018) for cellulose extraction and oxygen isotope analysis. This genus is considered an important paleoecological indicator of subtropical evergreen forests (Gee et al., 2003), and are currently restricted to regions of east and southeast Asia (Huang et al., 2018). Living Castanopsis trees growing in the region have maximum rooting concentrations within the upper 0.3 m of the soil (Hao et al., 2006). Such shallow rooted trees may better record annual variations in δ 18 O MW (McCarroll and Loader, 2004;Waterhouse et al., 2002) than more deeply rooted species or trees growing in drier climates (e.g., Huang et al., 2019a;Huang et al., 2019b).
Radial cores (QXS21A and QXS24A, Fig. 1) were also collected from two living Pinus massoniana trees growing at nearby Qingxiushan Hill, Nanning (22 • 47 ′ 23.35" N, 108 • 23 ′ 4.26 ′′ E, elevation = 223 m). The site has a humid, monsoon climate, with 63% of the total annual rainfall falling in only four months (May through August) (Fig. 2). Annual growth rings (entire earlywood and latewood) were sampled by hand using a razor blade across the years 1990-2000 to produce an annually resolved δ 18 O cell record. This period was chosen because it: 1) represented a wide range of wet-season (May through August) rainfall (482 to 1267 mm), and 2) avoided potential juvenile effects on δ 18 O cell value in the early portion of the trees' growth (Büntgen et al., 2020;Duffy et al., 2019). A total of 22 annual growth rings (11 per tree) were collected for stable isotope analysis.
We extracted α-cellulose from bulk samples of the fossil wood and individual tree rings of the modern wood following a method modified from Brendel et al. (2000). Because of the low cellulose yields in the fossil wood relative to the modern wood, we increased the amount of fossil wood starting material and proportionally adjusted the volume of reagent used. Lignin was removed by nitric acid and acetic acid, and lipids were removed by ethanol and acetone, and then treated with 17% sodium hydroxide solution to obtain α-cellulose. (Ramage, 1971) and (Wang and Ding, 2008)). Note this area includes the East Asia summer monsoon, western North Pacific summer monsoon, and Indian summer monsoon (Wang and LinHo, 2002). All of these locations experience high rainfall seasonality today. b-c) Photographs of mummified fossil wood samples: b) NNW015 and c) NNW037, respectively. d-e) Photographs of the two modern cores: d) QXS21A and e) QXS24A. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) coupled with a Delta-V Advantage Mass Spectrometer (Thermo Fisher Scientific, Inc., USA). Samples were analyzed with three internal laboratory reference materials (ACELL = 32.33 ± 0.06‰, JCELL-01 = 17.64 ± 0.09‰, and SigmaCell = 28.46 ± 0.07‰) calibrated against International Atomic Energy Agency (IAEA) benzoic acid reference materials: IAEA 601 (23.24 ± 0.19‰) and IAEA 602 (71.28 ± 0.36‰). A quality assurance sample (JCELL 02, δ 18 O = 20.44 ± 0.10‰) was analyzed within each batch and analyzed as an unknown. The δ 18 O cell values had an analytical error of ±0.3‰ and ± 0.1‰ for the fossil and modern samples, respectively. The δ 18 O cell values are reported in units of per mil (‰) relative to the VSMOW standard.
We recovered α-cellulose from 67% of the fossil wood samples (43 out of 64 samples); the remaining 21 samples (33%) contained no recoverable α-cellulose, and are not discussed further. The δ 18 O cell value of the 43 fossils ranged from 21.0 to 24.1‰ with a mean (± 1σ) = 22.4 ± 0.7‰ (Table 1). The measured variance in the δ 18 O cell values across the 43 fossil specimens was small (0.5‰), consistent with a common water source and common environment during growth (Csank et al., 2013;Sternberg et al., 2007). Because the δ 18 O value of seawater during the late Oligocene was lower than today (Westerhold et al., 2020), we added 2.5‰ to the fossil values to allow for an unbiased comparison with modern δ 18 O cell values. The δ 18 O cell value of the modern samples averaged 27.4 ± 1.0‰ (n = 22) (Table 2), which was similar to the most recent 50 years of δ 18 O cell values measured independently on Pinus massoniana and Picea schrenkiana trees growing in monsoon regions of southern China (Shi et al., 2020;Xu et al., 2013) (Two-sample t-test, t = 0.6, p = 0.53). The different sampling resolution of the modern (annual) and fossil (bulk wood) samples do not allow for comparison of variance between the late Oligocene and present-day; however, the δ 18 O cell values of the modern samples is significantly higher than both the raw (22.4 ± 0.7‰) and seawater-corrected (24.9 ± 0.7‰) fossil δ 18 O cell values (two-sample t-tests, t = 20.8 and 10.3, respectively; p < 0.0001 for each). We note that all further analysis and discussion of the fossil δ 18 O cell data use the seawater corrected values.
When examining our results within the context of published δ 18 O cell values, we find that all our measured δ 18 O cell values (both fossil and modern) were within the range of δ 18 O cell values reported for trees growing across the planet today (δ 18 O cell = 14 to 34‰; Fig. 3). The δ 18 O cell values of the modern samples are similar to other records in southern China (Cai et al., 2018;Shi et al., 2020;Xu et al., 2013). However, our measured fossil δ 18 O cell values are lower than δ 18 O cell values from the modern trees in the Nanning Basin, and are more similar to values reported for trees growing at high elevation sites (3500 m) in Nepal (Xu et al., 2018); high latitude, cold sites near Lake Baikal in central Russia (Tartakovsky et al., 2012) and arctic Siberia (Holzkämper et al., 2008); and areas that receive higher summer rainfall amounts than southern China (e.g., Bangladesh, Islam et al., 2021). Nevertheless, the paleolatitude (Wu et al., 2017) and paleogeography (Quan et al., 2016) Oligocene age wood fossils from interior Siberia (Fig. 3), which further preclude temperature (Richter et al., 2008b), vapor transport distance (Jahren and Sternberg, 2002), or continentality (i.e., distance from coast) (Rozanski et al., 1993) 2021) reported δ 18 O cell values of 24.2 to 27.2‰ for Chukrasia tabularis (25.6 ± 0.8‰) and Lagerstroemia speciosa (25.9 ± 0.5‰) growing in Bangladesh; both of these studies reported data from sites with greater wet-season precipitation than our study site in Nanning.
We therefore interpret our low δ 18 O cell values in terms of the amount effect, consistent with interpretations made at other low-elevation tropical and sub-tropical sites (Schollaen et al., 2014;Shi et al., 2020;Zhu et al., 2012), in which intense monsoon precipitation becomes depleted in 18 O in summer months (Araguás-Araguás et al., 1998;Dansgaard, 1964). We achieve this by first demonstrating the utility of these relationships for quantifying growing season precipitation in monsoon climates using our modern δ 18 O cell values, and then applying these relationships towards inferring wet-season rainfall during the late Oligocene using our fossil δ 18 O cell dataset.
In order to quantify monsoon rainfall from δ 18 O cell , we first demonstrate the ability for δ 18 O cell to accurately record the δ 18 O MW value of monsoon rainfall in present-day Nanning. We did this by calculating δ 18 O MW from δ 18 O cell using the following relationship developed by Sternberg et al. (2007) from both needle leaf and broadleaf tree species growing across a wide climate gradient:
(1)
This relationship excluded samples from Arizona, USA (after Sternberg et al., 2007) because they grew under exceptionally low relative humidity (26%) that is known to affect δ 18 O cell , and contrasts with the consistently high relative humidity of Nanning Basin today (80%) (Fig. 2). We note, however, that given the strong relationship between δ 18 O cell and δ 18 O MW and large sample size of this dataset (n = 34), the slope (m) and intercept (b) do not change substantially if the data from Arizona are included (m = 0.61, b = 32.4‰, Sternberg et al. (2007)).
Using Eq. ( 1) and the δ 18 O cell values measured here for the modern trees growing in Nanning, we calculate median δ 18 O MW = -7.5 ± 1.7‰ (n = 22). These values are consistent with predicted δ 18 O MW values for summer monsoon precipitation at this site, which today range from -6.5 to -8.0‰ (Fig. 2), and demonstrate the utility of Eq. ( 1) for estimating monsoon δ 18 O MW values using δ 18 O cell .
Next, we calculated average monthly monsoon precipitation using the following relationship developed by Xie et al. (2011) for nearby Guangzhou, China:
where P month is mean monthly precipitation, and δ 18 O MW is calculated using Eq. ( 1). Using our annual δ 18 O cell values from each tree core and Eqs. ( 1) and ( 2), we calculate mean (±1σ) P month in Nanning today of 230 ± 51 mm (range = 141-314 mm). These values are consistent with average monthly rainfall amount during the wettest four months of each calendar year represented by our δ 18 O cell dataset, i.e., "P wet " (1990-2000: P wet = 216 ± 46 mm, range = 158-317 mm) (Fig. 4). This result highlights the importance of rainfall in the wettest months on tree growth in this subtropical monsoon climate, and suggests δ 18 O cell values as an accurate proxy for monsoon precipitation in the region.
Applying Eqs. ( 1) and (2) to our fossil wood data allows for Global data include compilations by Sternberg et al. (2007) and Richter et al. (2008b), with updated sites including a high elevation site in the Himalayas (Xu et al., 2018), high latitude site in acrtic Siberia (Holzkämper et al., 2008), inland/high latitude site in Lake Baikal (cetral Russia; Tartakovsky et al., 2012), and Bangladesh (Islam et al., 2021). Fossil data from Siberia are from Richter et al. (2008a). Box plots show the first through third quartiles with whiskers extending a distance of the interquartile range on either side of the box. All new modern data sets were trimmed to the most recent 50 years for each δ 18 O cell series to prevent overweighting individual sites. Data were available as supplementary materials from the primary publications. Data used in this figure are available in the supplementary materials (Table S1).
estimation of monsoon precipitation during the late Oligocene. We note that this approach can be affected by changes in the δ 18 O value of source water and potential changes in δ 18 O MW gradients through time (Fricke and O'Neil, 1999) Recent efforts have aimed to improve our understanding of spatial patterns in precipitation during the late Oligocene (Utescher et al., 2021). Plant macrofossil data indicate less wet-season precipitation in northern China during this time (Li et al., 2018), in contrast with the greater wet-season rainfall we calculate here for southern China. These regional differences are consistent with observations and models indicating diminished rainfall in northern China (Li et al., 2017) and increased rainfall to the south (Bryan et al., 2019;Sooraj et al., 2015) in response to modern global warming, and lend support to our interpretation of high rainfall inferred by the low δ 18 O cell values. The increased average wet season rainfall calculated here for the late Oligocene further supports hypotheses that warmer temperatures in the future will increase the amount of precipitable water thereby intensifying rainfall in southern China (Takahashi et al., 2020). This interpretation is supported by multiple models that indicate a 50% to 100% increase in extreme precipitation frequency in southern China and southeast Asia in response to warming of only 1.5 to 2 • C (Chevuturi et al., 2018).
The Oligocene paleoflora at Nanning has previously been interpretated to reflect tropical, monsoonal climate conditions similar to areas that receive higher summer rainfall than modern Nanning (e.g., Thailand and eastern India; Huang et al., 2018;Ying et al., 2018). The results of the current study and these independent lines of evidence suggest that the East Asian monsoon was at least as strong as modern conditions-with likely higher summer rainfall levels-during the late Oligocene, and indicates that the East Asian monsoon may strengthen as CO 2 levels and global temperatures approach those of the Oligocene (Tierney et al., 2020).
We show that δ 18 O cell values can be used to provide information on summer rainfall within monsoon regions. We find that δ 18 O cell values were 5‰ lower during the late Oligocene than today, half of which is attributed to lower δ 18 O values of seawater and half to changes in climate. Comparison to Oligocene δ 18 O cell data from Siberia confirm the low δ 18 O cell values cannot be reconciled by lower temperatures because it is unlikely both sites experienced the same temperatures. Agreement between δ 18 O cell and rainfall in modern monsoon regions supports our interpretation of low fossil δ 18 O cell values inferring elevated rainfall during the late Oligocene. Using modern relationships, we calculate a 59% increase in monsoon rainfall during the late Oligocene compared to present (349 mm/month versus 220 mm/month). Such heavy rainfall in a single month is not unprecedented in Nanning at present: from 1951 to 2012, nearly 10% of the monsoon months (May through August) have experienced at least 349 mm of rainfall; however, no single year has averaged at least 349 mm across four months (i.e., P wet ≥ 349 mm, the late Oligocene average value). This result is consistent with fundamentally higher monsoon rainfall in the late Oligocene than today and indicates potential for greater monsoon rainfall in a warmer climate. We therefore conclude that continued global warming beyond 2