New sedimentary evidence reveals a unique history of C4 biomass in continental East Asia since the early Miocene

Pyrogenic carbon (PyC) and n-alkane data from sediments in the northern South China Sea reveal variations in material from C4 plants in East Asia over the last ~19 Ma. These data indicate the likely presence of C4 taxa during the earliest part of the record analysed, with C4 species also prominent during the mid and late Miocene and especially the mid Quaternary. Notably the two records diverge after the mid Quaternary, when PyC data indicate a reduced contribution of C4 taxa to biomass burning, whereas plant-derived n-alkanes indicate a greater abundance of C4 plants. This divergence likely reflects differences in the predominant source areas of organic materials accumulating at the coring site, with PyC representing a larger source area that includes material transported in the atmosphere from more temperate (relatively cooler and drier) parts of East Asia. Variations in the relative abundances of C3 and C4 taxa appear to be linked to a combination of environmental factors that have varied temporally and geographically and that are unique to East Asia. A major expansion of C4 biomass in warmer subtropical parts of eastern Asia from ~1 Ma and particularly from ~0.4 Ma is later than other parts of the world.

The Calvin-Benson cycle, the process through which plants convert inorganic carbon (C) and water to three C (C 3 ) sugar molecules, originated when atmospheric composition was very different from present 1, 2 . One modification to the cycle, leading to reduced photorespiration effects and improved photosynthetic efficiency under certain conditions (e.g. moisture stress and relatively low pCO 2 ), involves production of four C (C 4 ) oxaloacetate as the first-formed product of photosynthesis. Uncertainty surrounds the exact date of origin of this modification. That said, there is no widely accepted evidence of C 4 taxa pre-dating the Oligocene 3 , with molecular studies indicating that the C 4 photosynthetic pathway first appeared between 35 and 30 Ma 4 , a period that includes a major decline in pCO 2 5 . Only ~3% of angiosperm species currently support the C 4 pathway. However, despite being utilized by a relatively small proportion of the global flora, C 4 taxa -predominantly in the form of grasses -account for a substantial proportion of vegetation cover of Earth 6 , and include important food staples.
An expansion of plants utilising the C 4 photosynthetic pathway (e.g., tropical grasses) relative to those using the C 3 pathway (e.g., trees, shrubs and temperate grasses) in the late Miocene constitutes one of the most important biogeographical transformations of the Cenozoic 7-10 . The timing of this expansion appears to have varied geographically, occurring later in East Asia than other parts of the world [11][12][13][14][15] . As evidence has accumulated, however, the expansion of C 4 plants in East Asia appears to have been phased, rather than a single event, and may have occurred much earlier in some parts of the region 16,17 , than in others 18 . Determining when and where the C 4 photosynthetic pathway first appeared on the Asian continent is difficult because direct evidence, in the form of plant remains such as pollen and phytoliths, is rare. Furthermore, when records are available, they are often incomplete; C 4 taxa often occupy environments that are not conducive to the preservation of organic material in situ, and achieving a high level of certainty in distinguishing between many C 4 and C 3 taxa in pollen and phytolith records can be problematic.
Stable C isotope (δ 13 C) values for C 3 plants globally range between −37‰ and −20‰ (average −28.5‰) 19 , while the δ 13 C values for C 4 plants range from −15‰ to −9‰ (average −13‰) [20][21][22] . Pyrogenic carbon (PyC), also known as elemental carbon (EC) or black carbon (BC), is produced through the incomplete combustion of biomass and can preserve the original C isotope value of the parent plants 23 . PyC is generally resistant to decomposition, and upon extraction from sediments can be used not only to document the frequency and intensity of past fire activity, but also changes in the relative contributions of C 4 and C 3 plants in combusted biomass 15,16,[23][24][25] . Similarly, long chain n-alkanes (≥n-C 27 ), which are important building blocks of lipids forming the protective waxy cuticle covering leaf and stem tissue, are resistant to diagenetic alteration 26 . Where herbaceous vegetation predominates, n-alkanes of C 31 and C 33 are abundant components of sediments, while larger proportions of n-alkanes of C 27 and C 29 are generally indicative of wooded and forested landscapes (i.e. C 3 taxa) [27][28][29] . Thus ratios of measured levels of n-alkanes, such as C 31 /C 27 , can be used to reconstruct the relative contributions of herbaceous and woody plants to organic C preserved in sediments 27 . Furthermore, although long chain n-alkanes typically have lower δ 13 C values than bulk tissue 30 , fractionation effects can be removed, so that the PyC and long chain n-alkanes composition of organic material in sediments can be used as a basis for inferring past variations in the relative abundances of C 3 and C 4 plants in, respectively, combusted biomass and vegetation 23,30,31 .
The South China Sea (SCS), a marginal sea between Asia and the Pacific Ocean, is dominated climatically by the East Asian monsoon 16,32,33 . Sediments accumulating on the bed of the northern SCS, largely supplied by major rivers but also including aeolian material transported by the East Asian Winter Monsoon (EAWM) 34 , provide a potential source of information on environmental variations affecting continental East Asia during the late Cenozoic. To date, there is only one sediment-based record of variations in PyC from the northern SCS 16 . A second core, collected from a deep water site in the southern part of the SCS, has yielded a record of variations in n-alkane content of sediment, presumably largely of low latitude origin, for the last 5 Ma 29 . Here we present new data in the form of variations in PyC and n-alkanes from the same sediment core from the northern SCS relating to the last ~19 Ma. These new data are compared with published information in order to describe and explain the unique history of C 3 and C 4 taxa in East Asia since the early Miocene.
Study area, materials and methods. The Ocean Drilling Program (ODP) Leg 184, Site 1146 (19°27.4′N, 116°16.4′E, 2092 m water depth) retrieved sediment from a small rift basin on the mid-continental slope of the northern SCS, located less than 50 km to the northeast of ODP Site 1147/1148 (Fig. 1). The sediment sequence obtained at ODP Site 1146 extended to a sub-seafloor depth of 643 m composite depth (mcd). For this study, a total of 149 2-cm-thick sample slices were collected at 4 m intervals between 641 mcd and the surface. Chronological control was established on the basis of palaeomagnetism and biostratigraphy 35 (S1, Supplementary material), with the ages of sample depths between control points established through linear interpolation (Fig. S1). According to the age-depth model for ODP Site 1146 35 , the sequence of sediments between 641 mcd and the surface covers the last ~19 Ma, with the sampling interval of 4 m adopted in the current research equating to a temporal resolution of ~100 ka.
PyC was extracted using the method of Gustafsson et al. 36 . Sediment samples were pretreated with 1 M NaOH, HCl and HNO 3 to remove inorganic C and some of the organic forms that might lead to BC formation during the pre-combustion step. PyC was then measured following combustion at 375 °C in air for 24 h, with the δ 13 C values for PyC (δ 13 C PyC ) values determined using a MAT-253 mass spectrometer. Isotopic compositions are expressed as deviations relative to the V-PDB standard with a precision of ±0.2‰ or better. This procedure differs from the oxidation method used to measure PyC (therein termed BC) in sediments from ODP Sites 1147/1148 and the Lingtai Section of the Chinese Loess Plateau (CLP) 16,37 .
For analysis of high molecular weight n-alkanes, samples were freeze-dried before being ground manually. After adding an internal standard (nC 24 D 50 ), a ~3 g sample was ultrasonically extracted in dichloromethanemethanol (3:1, v/v). Extracts were dried and saponified with 6% KOH-methanol. Hexane was then used to extract hydrocarbons, which were separated by column chromatography into the polar (i.e. alkanol) and non-polar (i.e. alkane) fractions. The δ 13 C values for individual alkanes (δ 13 C Alk ) were determined using a GV Isochrom II system interfaced to a Hewlett-Packard 5890 gas chromatograph. The gas chromatograph was fitted with a fused Si column HP-5 MS (30 m-0.32 mm-0.25 mm) connected to the combustion interface. Helium was used as the carrier gas with a flow rate of 1.2 mL min −1 . The temperature was programmed to remain at 80 °C for 2 min, before rising to 220 °C at a rate of 10 °C min −1 , and then to 290 °C at a rate of 3 °C min −1 . The temperature was then maintained at 290 °C for 15 min. CO 2 was used as a reference gas, which was automatically introduced into the isotope ratio mass spectrometer before and after each analysis. The δ 13 C values were calibrated against a standard mixture of n-alkanes (nC 12 -nC 32 ) of known isotopic composition (Indiana University, USA) and reported as ‰ relative to Vienna Peedee Belemnite (V-PDB). Replicate analyses showed that the standard deviation for each compound was less than 0.3‰. In order to maintain the reproducibility and accuracy of results, standards were run between samples, and each sediment sample was analyzed at least twice.
The δ 13 C value of atmospheric CO 2 has not remained constant over time 9,23,38,39 . Moreover, fractionation effects can occur during the production, transport and sedimentation of organic matter 23,40,41 , and between atmospheric CO 2 and plant organic C, as well as varying between different taxa 42,43 . These sources of variation introduce additional uncertainty into calculations of the relative contribution to combusted biomass and relative abundance in vegetation of C 4 taxa based on, respectively, δ 13 C PyC and δ 13 C Alk , and hence need to be accounted for when interpreting records of ancient δ 13 C. Carbon isotopic enrichment factors for PyC and n-alkanes relative to atmospheric CO 2 (ε PyC-CO2 and ε Alk-CO2 , respectively) were thus applied to calculate C 4 abundance in the current study (S2 and Fig. S2, Supplementary material).

Results
The PyC content of sediments from ODP Site 1146 (Fig. 2a, Table S1) shows an overall slight increase from the early Miocene through to the Quaternary, ranging from 0.014% to 0.16% (average 0.06%). This trend is superimposed upon strong variability in the data, with the amplitude of variation particularly large from the mid Quaternary. The latter period includes two intervals of high PyC content, centred upon ~1 Ma and 0.2 Ma. δ 13 C PyC values range from −25‰ to −17.2% (average −21.2‰), and also show an overall increase since the early Miocene, before declining after ~2 Ma (Fig. 2b, Table S1). Relatively high (less negative) δ 13 C PyC values date to the early Miocene (~18- 19 Ma), mid Miocene (~14 Ma), late Miocene/early Pliocene (~7-4.5 Ma), and to the early to mid Quaternary (~2- 1 Ma).
Trends in C 31 /C 27 ratios (Fig. 2c, Table S1) are broadly similar to those exhibited by PyC; C 31 /C 27 ratios vary over the last ~19 Ma, showing an overall slight increase to ~1 Ma, with an increased frequency of above average values after ~7. 5 Ma, exhibiting greatest variability and highest values from ~1 Ma. In addition, C 31 /C 27 ratios fluctuate through the early and mid Miocene, with four peaks around ~18 Ma, ~15.5 Ma, ~14 Ma and ~11 Ma.
Variations in δ 13 C value of atmospheric CO 2 39 (Fig. 2e), interpolated to correspond directly with dated δ 13 C PyC and δ 13 C Alk records, were used to derive estimates of ε PyC-CO2 and ε Alk-CO2, respectively. Conservative estimates of the relative contribution of C 4 taxa to combusted biomass and vegetation ranged from, respectively, 0-34.7% (average 12.9%) and 0-41.6% (average 9.9%) ( Fig. 2f and g). Values of both ε PyC-CO2 and ε Alk-CO2 indicate the likely presence of C 4 taxa in East Asia during the early Miocene, later part of mid Miocene and the late Miocene through to the Late Quaternary. Some differences are evident in the two sources of information, however. The ε PyC-CO2 data tend to indicate a relatively significant and continuous, though strongly varying, C 4 contribution to combusted biomass, especially when compared with the corresponding estimated abundance of C 4 taxa in terrestrial vegetation based on ε Alk-CO2 data. The largest differences between the two datasets date to the Pliocene and Quaternary; ε PyC-CO2 data indicate an overall substantial contribution of C 4 grasses to combusted biomass from the late Miocene through to the Quaternary, though with an overall decline evident after ~2 Ma. Lower ε Alk-CO2 values indicate a predominance of C 3 grasses from around the beginning of the Pliocene through to ~1 Ma. This is supported to an extent by the enriched C 31 /C 27 ratios, which suggest a greater herbaceous contribution to organic C from the late Miocene. An expansion of herbaceous vegetation generally (including a greater proportion of C 4 grasses at times) from ~1 Ma, particularly from ~0.4 Ma, is suggested by enriched high δ 13 C Alk data values and generally higher average C 31 /C 27 ratios.

Discussion
Deviation of PyC and n-alkane records in ODP 1146 from northern SCS. Differences in n-alkane and PyC data from the same core likely reflect differences in both provenance of the C on which the measurements are based, and in the mode of transport of the C from the terrestrial source to the site of deposition. PyC is formed at high temperatures during the combustion of biomass and emitted to the atmosphere before a proportion is deposited in sedimentary environments 44 , with the amount deposited dependent on atmospheric transport (influenced by wind velocity and direction and topography), precipitation and depositional processes. Several studies have indicated that PyC preserved in deep-sea sediments in the SCS originates via long-range transport from burning in central and southeastern Asia 45-51 (S3, Supplementary material). Atmospheric transport from these sources is particularly active during EAWM, when dry continental conditions can also enhance the likelihood of vegetation fires 52,53 . In general, highly seasonal and semi-arid climate conditions will result in biomass that is prone to burning, at least towards the end of the dry season, and, on occasion, in intense fires 24 . Thus levels of PyC accumulating in sediments in the SCS are affected not only by the extent and composition of vegetation ; (f) The contribution of C 4 taxa to combusted biomass based on carbon isotopic enrichment factor for PyC (ε PyC-CO2 ); (g) C 4 abundance based on carbon isotopic enrichment factor for n-alkanes (ε Alk-CO2) . Arrows indicate the overall trend in proxies. Two yellow bars show the area of uncertainty in reconstructing the relative prominence of C 3 and C 4 taxa based on ε PyC-CO2 and ε Alk-CO2 , and thus the basis for the very conservative estimates of prominence of C 4 taxa mentioned in the text of this paper. available for combustion, but also by burning regime and by variations in the EAWM in particular 24 . Variations in PyC therefore likely represent an integration of conditions over a broad geographic area, with the source area of PyC accumulating in the northern SCS potentially including cold-adapted vegetation in more temperate, higher altitude and latitude parts of continental East Asia 16 . This is evident in similarities between variations in δ 13 C PyC measured at ODP Site 1146 discussed here and δ 13 C BC from Site 1148 16 , and from the CLP over the last ~7 Ma 15 (Fig. 3).
By comparison, n-alkanes are mainly derived from plant leaves 30 . Surface sediment n-C 29 concentrations are in good agreement with data on pollen from approximately the same suite of modern sediments in the northern SCS 54,55 . The main delivery mechanism for n-alkanes is likely to be dominated by fluvial inputs and then marine currents via the Bashi and Taiwan straits 55 . Some n-alkanes may also have been deposited from the atmosphere, along with aeolian dust and aerosols associated with biomass burning 56 . However, the fact that levels of n-alkanes and PyC measured in the same sediment samples are not well-correlated suggests the two proxies have different provenances (Fig. S3). Possibly therefore, n-alkanes accumulating in the past in sediments in the northern SCS originated from relatively well-vegetated parts of subtropical East Asia, rather than as a result of long distance aeolian input from more temperate land to the north, where the vegetation cover is comparatively sparse 24 . Catchments draining into the northern SCS are likely to be the dominant means by which organic material containing plant lipids is transported mainly from sub-tropical latitudes 55 , with variations in δ 13 C Alk therefore affected by a different set of drivers to PyC abundance, including changes in monsoonal precipitation and fluvial discharge, catchment conditions and the extent and proximity of land relative to ocean 57 .

Variations in C 4 contribution to vegetation and combusted biomass in East Asia recorded by
PyC and n-alkane data. Variations in the prominence of C 4 taxa estimated from ε PyC-CO2 (relative contribution to combusted biomass) and ε Alk-CO2 (relative abundance in vegetation) are generally in phase with the δ 13 C PyC and δ 13 C Alk records, respectively. The δ 13 C PyC values are also in general agreement with variations in δ 13 C BC from ODP sites 1147/1148 16 (Fig. 3), notwithstanding the aforementioned systematic offset in values due to differences in the analytical techniques employed 23 . Variations in δ 13 C Alk are also within a similar range to that observed in a previously reported record from the northern SCS 57 covering a shorter period of time than that discussed here.
The presence of C 4 taxa during the early (~18. 5 Ma) and mid Miocene (~14 Ma) and from the late Miocene (~7.5 Ma) is supported by existing data 9,[15][16][17]58 . Variations in abundances of C 4 taxa have been attributed to the changes in pCO 2 10, 59 and to evolution of East Asian monsoon 12,16 , seasonality 18 , aridity and/or fires 15,58 . Intensification of the East Asian Summer Monsoon (EASM) is thought to date to about 24 Ma, with a peak in intensity during the mid Miocene, a weakening after ~8 Ma and strengthening again during the late Pliocene/ early Quaternary 29,30 , while strengthening of the EAWM is thought to have occurred 16-14 Ma, ~8 Ma, and ~3 Ma and to have further intensified from the mid Quaternary 60 . Increasing PyC content indicates rising levels of biomass burning, with periods of relatively high contributions to combusted biomass from C 4 taxa evident ~14 Ma, ~7-4.5 Ma and ~2.2-1.5 Ma. Increased contributions of C 4 taxa to biomass burning may represent a climatically promoted competitive advantage for C 4 photosynthesis, due not only to a strengthened EAWM and greater aridity 57, 61 , but also to increased seasonality as a result of renewed intensification of the EASM, especially from ~3 Ma. From the mid Quaternary, both above average PyC levels and C 31 /C 27 ratios could be a response to increased aridity, possibly due to a strengthening of the EAWM following further expansion of the northern hemisphere ice sheets 12,29,32 . The contribution of C 4 taxa to combusted biomass declines from ~2 Ma, presumably because of a greater presence of C 3 plants in the source region for PyC. Evidence from the CLP for increased prominence of C 3 taxa during much of the Quaternary may represent a combination of a more intense EAWM and reduced temperatures during the main growing season in more temperate parts of the region, with both factors favouring C 3 photosynthesis 15,62 .
At lower latitudes, C 3 taxa were prominent from the mid Miocene to the Pliocene/Quaternary boundary, under warm and humid climate conditions presumably owing to the influence of the EASM 33 . Reduced moisture and increased seasonality from the late Pliocene through to ~1 Ma, with the majority of precipitation occurring during the growing season 29 , could have facilitated sustained increases in the relative abundance of C 3 grasses, as evidenced by relatively lower ε Alk-CO2 values and higher C 31 /C 27 ratios. As seasonal conditions developed and vegetation canopies became more open, expansion of C 4 biomass could have become tightly coupled in a feedback relationship to changes in burning regime and aridity that effectively selected against shrubs and trees (i.e. C 3 species) and promoted further expansion of grasslands in which C 4 taxa were prominent. An expansion of C 4 taxa in subtropical East Asia from ~1 Ma, and especially ~0. 4 Ma, may have been triggered by increased seasonality, aridity and changes to the fire regime.
A synthesis of evidence for C 4 variations across Asia. δ 13 C values from fossil enamel and soils from the northwestern CLP (the northeastern margin of the Qinghai-Tibet Plateau) suggest increased abundance of C 4 taxa from ~3 Ma, with C 4 taxa becoming a significant component of the local vegetation around ~1.0 Ma 13 . An expansion of C 4 biomass ~1.0 Ma and, at lower altitude, ~0.4 Ma, on the southeastern parts of the CLP is evident in δ 13 C data from Yanyu (620 m amsl) and Lantian (769 m amsl) 12,18 (Fig. 3). The evidence from Lantian and Yanyu is thus consistent with our δ 13 C Alk data from the northern SCS.
The δ 13 C values of n-alkanes obtained from palaeosols associated with the Siwalik Group and marine sediments from the Indus (ODP Site 722) and Bengal (ODP Site 717C) fans 63, 64 , dating to 12-11 Ma, indicate a predominantly C 3 flora during the mid Miocene (Fig. 3). A subsequent expansion of C 4 taxa during the late Miocene in western and southern Asia has been linked to aridification 64 , although other factors, such as increased burning 15,25 and lower pCO 2 may have played a role. Evidence of an expansion of C 4 biomass from ~8-6 Ma that peaked in the early to mid Quaternary in Asia, the South Atlantic (ODP Site 1081) 65 , and Gulf of Mexico (DSDP Site 94) 9 implies the possible effect of global decreases in moisture availability along with ice sheet formation and expansion that commenced during the late Miocene, as recorded in δ 18 O variations preserved in the tests of foraminifera 66 .
The detailed reconstruction of variations in the prominence of C 4 taxa estimated from ε PyC-CO2 and ε Alk-CO2 is limited by a number of uncertainties, including the effects of changes in atmospheric pCO 2 and fractionation due to moisture stress 42 , and potentially also O 2 /CO 2 ratios 67 . Notwithstanding these uncertainties, the available evidence is broadly consistent with the conclusion that the rise to prominence of C 4 taxa over a substantial proportion of subtropical East Asia differed in timing, likely reflecting local differences in the relative strengths of the suite of potential causal factors. Beginning in the late Pliocene, increased seasonality -with the majority of precipitation occurring during the growing season -and aridity overall could have facilitated sustained increases in the relative abundance of open vegetation in which C 4 taxa were important components. As seasonal climate conditions developed and vegetation canopies became more open, expansion of C 4 biomass may have become tightly coupled in a feedback relationship with changes in burning regime that effectively selected against shrubs and trees (i.e., C 3 species) and promoted further expansions of C 4 grasslands 15,64,68 . Lower temperatures associated with continued uplift of central Asia coupled with a strengthened EAWM and the effects of fluctuating Quaternary ice sheets after ~2 Ma could have shifted habitats previously characterized by a relatively high proportion of C 4 taxa back within an envelope of environmental conditions that were more suited to C 3 species. From around the same time, but at lower altitudinal and subtropical parts of the region, aridity, enhanced by an expanded coverage of ice in the NH and large-scale hydrological dynamics, may have forced an abrupt expansion of C 4 biomass 29 , thereby contributing to a history of C 3 /C 4 variations in East Asia that is unique for the continent and more widely.

Conclusions
The abundance and C isotope composition of n-alkanes and PyC in samples from a sediment core from the northern SCS are used to reconstruct a history of variations in the relative prominence of C 4 and C 3 taxa in East Asia over the last ~19 Ma. Although present throughout the record, C 4 taxa appear to have been prominent at times during the Miocene and especially from the late Pliocene to mid Quaternary. However, n-alkane (a proxy of relative contribution to vegetation) and PyC (a proxy of relative contribution to combusted biomass) isotope data diverge from the mid Quaternary, most likely reflecting differences in provenance, with PyC representing a larger source area that includes higher altitude and more temperate parts of the region, and atmospheric processes as the primary means of transporting organic material from the interior of the continent to the coring site. Temperatures in these already cooler parts of the continent appear to have fallen to levels too low to support C 4 taxa from the mid Quaternary. New evidence presented in this paper in the context of existing data indicates that variations in the level and seasonality of rainfall and associated changes in biomass burning appear to have been important drivers of changes in the relative contributions of C 3 and C 4 taxa to East Asian vegetation throughout much of the past ~19 Ma. An expansion of C 4 taxa at low altitude/latitude from ~1 Ma, and especially ~0. 4 Ma, may have been triggered by increased aridity and changes to the fire regime. This expansion commenced more recently than in other parts of the tropics and sub-tropics, and is one characteristic of a novel history of variations in C 3 and C 4 taxa in continental East Asia.