The South China Sea (SCS) is the largest marginal sea and a semi-enclosed basin which exchanges water with the Pacific Ocean via the Luzon Strait, and with the Indian Ocean over the shallow Sunda Shelf via Indonesian throughflow (Xiang et al., 2009). According to the current atmospheric conditions, the East Asian Monsoon (EAM) regulates the SCS water hydrology by controlling the northern hemisphere's atmospheric heat budget (Jian et al., 2001; Wang et al., 2005). The East Asian summer monsoon (EASM) transports moist wind toward northern China, from the north of Australia through the Warm Pool (Figure 2B). The East Asian winter monsoon (EAWM) also carries cold and dry Siberian air southward along eastern China (Figure 2A) (An, 2000; Hu et al., 2000; Fernando et al., 2007). A major governing factor for this phenomenon is unequal surface distribution of land and ocean (Shaw and Chao, 1994). In other words, sea level has been a critical factor that determines monsoonal changes throughout history (Hu et al., 2002).

During the summer, surface currents of the southern SCS produce a coastal current from the southwest to the northeast along the coast, causing the East Asian monsoon and the western branches of the Kuroshio Current, which are formed due to the southwest monsoon (Figure 2D). A summer monsoon also leads to the formation of an anticyclonic vortex at the center of the SCS. During the winter, a periodic cyclonic vortex at the center of the SCS and a current from the northeast to southwest direction forms, caused by the northeast monsoon. On the east side of the cyclonic vortex, there is a weak anticyclonic vortex, and there is also a flow directed away from the Natuna islands, through the middle of two circulations, expanding to the north and opposite to the wind pattern (Fang et al., 1998) (Figure 2C).

Islands in the southern part of the SCS, such as Borneo, Indochina, The Malay Peninsula, Sumatra and the Philippine Islands, are covered with tropical and seasonal rainforests. Among those, the islands around the Sunda Shelf are covered with tropical rainforests, such as Borneo, Java and Sumatra (Florin, 1963) (Figure 2). In these tropical rainforests, herbs can be seen as a lower layer of the forest (Zhang et al., 2011). At the same time, woody vines are abundant in the tropical rainforests. Coastal regions of these islands are dominated by shrubs and herbs (Poliakova and Behling, 2016) (Figure 2).

Gravity core NS07-25 (6°39.945′ N, 113°32.936′ E) was extracted from the northwest slope of the Nansha Trough, southern SCS, at a water depth of 2,006 m (Figure 1). The core was 556 cm in length and consisted of greenish gray to olive-gray silty-clay. Planktonic foraminifera from 11 horizons in core NS07-25 were chosen for AMS¹⁴C dating at the Xi-an Accelerator Mass Spectrometry (AMS) Center, China, and approximately 20 mg of shells were obtained from the N150 μm fraction of the measurements. The 556 cm of the core has been dated to 37 ka B.P. The age model of the gravity core NS07-25 based on the AMS¹⁴C dates (Xiang et al., 2009) are presented in Table 1. Global reservoir effect of ~400 years, was assumed to have no effect (Bard, 1988) in the southern SCS.

The core was sub-sampled at 1 cm length intervals, and 27 sub-samples were chosen (representing major climatic changes during the past 37,000 years) for the element (C and N) and isotope (δ¹³C_org) analyses at the Sun Yat-sen University, Guangzhou. The sub-samples were freeze-dried, powdered and homogenized. The TOC, TN and δ¹³C_org were analyzed after sieving through an 80 μm mesh to remove large debris, acidified with 10% HCl to remove inorganic carbon, rinsed with deionized water to remove salts, and then dried. The samples were fed to the combustion tube from a solid sample tray and were rapidly converted to CO₂ by dynamic combustion at 1,800°C. The water was removed and then separated by gas chromatography and magnetic fields. The isotope ratio was determined by ¹³C/¹²C isotope ratio mass spectrometer (IRMS). The precisions of duplicate analyses of samples were ±2 and ±3% of the means for TOC and TN, respectively. Stable carbon compositions were determined with a Delta V Advantage isotope ratio mass spectrometer and expressed in the delta notation (δ¹³C_org) relative to Peedee belemnite (PDB). The precision of duplicate analyses was 0.06‰ for δ¹³C_org. The precision of the duplicate analyses was 0.01% for TOC and TN.

Distributions of TOC, TN, C/N and δ¹³C_org from the NS07-25 southern SCS sediment core was clearly divided into three stages, where pollen data strongly confirmed these three stages by Hierarchical Cluster Analysis using Coniss (Figure 4). These stages are strongly correlated with climate-derived marine isotope stages. During these stages sediment provenance varies according to the geodynamic and climatic events, such as rainfall pattern and sea level changes. Similarly vegetation type and vegetation distribution also followed these geodynamic and climatic changes. The C/N and δ¹³C_org distribution patterns could be used as a source identification method for organic matter (Meyers, 1994; Lamb et al., 2006; Mackie et al., 2007) while some studies have used N/C ratios with δ¹³C_org distribution (Zheng et al., 2017) (Figure 6). δ¹³C_org of marine particulate organic carbon (mainly plankton) varies between −18 and −21‰ (Peters et al., 1978) and C3 plant δ¹³C_org varies between −21 and −32‰ (Tieszen et al., 1979), whereas in the NS07-25 core, δ¹³C_org data ranges between −21.6 and −26.4‰, revealing the availability of both marine OM and terrestrial C3 plant OM (Figure 6). δ¹³C_org value of C4 plants varies between −16 and −10‰ (Tieszen et al., 1979). The C/N ratio values from sediment cores are distributed in an intermediate value of 24.9 to 7.6 between marine particulate organic carbon and C3 plants. [C3 plants have C/N values above 12 (Tyson, 1995) and marine phytoplankton varies between 5 and 7 (Meyers, 1994)].

In the terrestrial environment, pollen and geochemical proxies (TOC, TN, C/N, and δ¹³C_org) show a higher correlation signified by their same origin [δ¹³C_org from C4 plant prominent OM, and the total herb pollen from the same core shows a higher correlation of R = 0.615, from Lake BarombiMbo, West Cameroon (Giresse et al., 1994)]. In the deep ocean sediment, above correlation only depends on two factors being assumed; (1) Different distribution methods of pollen and geochemical proxies between source and fate effect the correlation. Similar distribution pattern increased the correlation while different distribution patterns decrease the correlation [pollen: wind and hydrological pathways, geochemical proxies: mainly hydrological pathways (Blair and Aller, 2012)]. (2) Alteration of terrestrial plant geochemical proxy values due to the mixing of in situ and ex situ OM sources (phytoplankton, algae, etc.…).

According to our findings, in marine environments, the correlation between these vegetation proxies tend to decrease due to complex mixing and different distribution patterns, even though the terrestrial vegetation signal is still observable in the deep ocean. In this study, we have corrected bulk data according to the similar distribution patterns for pollen and geochemical proxies (separated the pollen and geochemical data for riverine pathway prominent time periods).

As geochemical proxies are mainly distributed through riverine pathways, the same distribution patterns for the pollen were considered. Among three zones of geochemical and pollen data, during stage I, wind and riverine distribution were contributed. In Stage II, the riverine pathway was prominent, except during the LGM, where distribution through wind has similarly contributed. During stage III, the riverine pathway was prominent. Even during the time periods where the riverine path is prominent, there is a considerable distribution of pollen through aeolian influx, altering the correlation (Figure 7).

Even though terrestrial plant and geochemical proxies influence their values due to mixing, generally, some proxies such as marine TOC and marine TN maintain a positive correlation with the terrestrial OM. This positive correlation occurs, firstly when terrestrial OM input is increased in the marine environment, the preserved portion of terrestrial TOC in the marine environment also increases. And secondly, along with the higher OM input, the ocean receives a higher inorganic nutrient content as well, leading to a higher marine productivity that results in a higher marine TOC. (Terrestrial OM inputs to the ocean increase in humid periods). According to these two factors, when terrestrial OM inputs are increasing, marine TOC also increased, showing the positive correlation between terrestrial OM and marine TOC.

According to the correlation between pollen and geochemical proxies, it was evident that the total abundance of C4 plant pollen and TOC shows a very low correlation of R² = 0.0004 for data that contains three stages (including all distribution methods) (Figure 7). The stage III and II have a calculated correlation of R² = 0.187. In these two stages, the majority of pollen has been distributed to the ocean through the riverine pathway, similar to the organic matter. When the data are corrected further by removing the LGM data, the correlation (R² = 0.3236) increases, revealing the importance of similar distribution patterns for the two proxies (Figure 7). TN also follows a similar correlation pattern with C4 plant pollen, as indicated by TOC. When the pollen data are correlated with the ratios between two elements (TOC and TN), it does not indicate a similar pattern with that of pollen data, where it is correlated with the elements separately. Importantly, δ¹³C_org does not show any correlation with the pollen data, as marine δ¹³C_org does not have a positive correlation with terrestrial OM (Figure 7). Correlations between C4 plant pollen and geochemical proxies reveal the availability of terrestrial plant OM signals in the oceanic environment in the study area. Pollen data was corrected by removing C3 plant pollen data. According to the observations, the relative abundance of C3 plant pollen increases with arid time periods (Figure 5), and during these arid windy periods, wind considerably affects pollen similar to riverine pathways.

We have compared δ¹³C_org along the AB axis (Figure 1) from the shallow ocean to the deep ocean for the last 20 ka. SO 18323 and SO 18300 represent the continental shelf while SO 18323 is located closer to the paleo-North Sunda River (Wang et al., 2009). These sites may not have experienced the sea level incensement beginning at 17 ka. The terrestrial environment of these sites might have changed to a marine environment at ≈15 ka (Figure 1) due to their submergence by ocean according to the bathymetry of the shelf. Both SO 18323 and SO 18300 sites have almost similar water depths (92 and 91 m).

Terrestrial OM mainly comes through hydrological pathways, unlike pollen that are dispersed through riverine paths and air. Therefore, the δ¹³C_org values of SO 18323 and SO 18300 was mainly affected by the hydrology of the area (before submergence, fresh water hydrology and after submergence marine hydrology) and the pre-submergence in situ vegetation. Core SO 18300 is situated near the coastal low land area in a river bank, where we may find C4 vegetation rather than C3 vegetation (Wang et al., 2009). When the shore line retreated, at the beginning of deglaciation after 17 ka, increasing δ¹³C_org values were observed in SO 18300. However, after the submergence at 15 ka, SO 18300 showed a stable C3 plant signal due to the OM input from above river catchments and the distribution in marine environment. When observing SO 18323, which is located deeper in the shelf, more C3 plant-based OM distribution is shown locally. In addition, after submergence of core location, the signal of δ¹³C_org tends to decrease toward C3-δ¹³C_org values. If there were C4 plants in the North Sunda River catchment, approximately 15 ka, the δ¹³C_org signal of SO 18323 should shift toward C4 plants (higher δ¹³C_org values) (Figure 8).

Continental slope is a more dynamic environment than the continental shelf and the deep ocean. Sediment core 18284-3 (226 m) and GC 18287-3 (598 m) was in a costal environment when sea level was lower at 17 ka. Both North Sunda River and Saim River fell into the ocean through this area. Submergence of the majority of the Sunda shelf was recorded in pollen data within 12 to 17 ka (Wang et al., 2009). δ¹³C_org distribution patterns on the Continental slope was discussed using previously published data of core GC 18284-3 (226 m), GC 18287-3 (598 m) and GC 18294-4 (849 m) (Steinke et al., 2003). A lower δ¹³C_org value (≈-27‰) from the shallower core GC 18284-3 (226 m) can be observed at the beginning of Sundaland submergence. Advancing deeper into the ocean, δ¹³C_org values become heavier due to the mixing with marine organic matter, GC 18287-3 (≈-24‰) and GC 18294-4 (≈-21‰).

Continental rise sediment core GC 17964-3 is located near the Paleo Baram river mouth area, northwest of Borneo (Figure 1). This core shows a more deviated pattern for pollen distribution. The total pollen count of this core is lower, and pollen usually do not show a decreasing trend after 12 ka. The presence of pollen from Borneo and Sumatra area could be the reason for this observation (Sun et al., 2000). δ¹³C_org from this sediment core varies similarly to that of GC 18294-4 but with lighter δ¹³C_org values, showing another riverine input with higher C3 plant materials, other than north Sunda River the Saim River (Steinke et al., 2003), and Paleo Baram River which should have the accountability of this signal.

δ¹³C_org distribution in the deep ocean shows an opposite pattern to that of continental slope's distribution. Two deep ocean cores were selected. Core 17961 (Kienast et al., 2001), located in the middle of the southern SCS, has a lesser impact from land areas and shows heavier δ¹³C_org values, during the glacial period (≈-19.5‰) and the Holocene period (≈-20‰). Core NS07-25, located landward in comparison, is closer to northern Borneo (Figure 1). Core NS07-25 tends to show more negative values compared with core 17961 possibly due to the C3 plant signal from Borneo and the Sumatra region as discussed.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.