Median particle size is commonly used to represent the average particle size of sediments (Hu et al., 2017; Li et al., 2018). The sediments in core 973-4 were fine-grained in most layers, but the grain size in the 4.3–6.0 m layer was significantly coarser (Figure 2). Core 973-5 contained fine-grained sediment in most layers, without any significant differences in the particle size being observed throughout the core; only a minor increase in coarser content was found in several layers (e.g., 1.3–2.7 m, 3.9–4.0 m, 5.9–6.0 m, 8.2–8.4 m, Figure 3).

The foraminifera-rich layer (4.3–6.0 m) in core 973-4 is consistent with the high content of the sediment coarse fraction (Figures 2B,C). The numbers of the two species were almost always less than five per gram below 6 m bsf. This number increases abruptly to more than 10 per gram at about 6 m bsf and then gradually decreases upward until 4.3 mbsf. Low values are observed in the uppermost 4.3 m bsf of the core (Figures 2B,C). In core 973-5, the numbers of benthic foraminifera were far less than core in 973-4 and few foraminifera were seen in the whole core, except in a few layers (e.g., at depths of about 2.5 m, 4.0–5.5, and 6 m; Figures 3B,C).

The δ¹⁸O composition of Uvigerina spp. in the core 973-4 varied from 2.57 to 5.26‰, with lighter δ¹⁸O values in the upper part of the core (Figure 2D). The δ18O depletion fits well with the Deglaciation transition layer at about 4.3 m bsf and shows stable values in the uppermost core. In core 973-5, δ¹⁸O values show a broad range of variation below 5.5 m bsf (Figure 3D), and values between 2.93 and 4.49‰ at the 3.3–3.6 m bsf layer, corresponding to the deglaciation transition. The δ¹³C values for Uvigerina spp. in core 973-4 ranged from −1.97 to 0.25‰ with an average value of −1.15‰ (Figure 2E). The δ¹³C values below 4.3 m bsf are obviously negative, with only small fluctuations; above this depth they become heavier until the surface. In core 973-5, the δ13C values were consistently negative below the depth of 3.3 m bsf and showed heavier trend from -2.9 to 0.18‰ in the upper part of the core (Figure 3E).

In core 973-4, the CRS contents vary between 0.01 wt‰ and 0.95 wt‰. A high CRS content was observed in the interval 6.8–9.0 m (Figure 2G). The δ³⁴SCRS values show a wide variation, ranging from −43.8 to 32.6‰. The sulfate isotope compositions show extremely positive values at the CRS enrichment depth (6.8–9.0 m), and then are lower elsewhere, with a minimum value of −43.8‰ in the interval 2.8–5.6 m (Figure 2F). In core 973-5, high CRS contents and positive δ³⁴S_CRS values are observed below 4 m bsf (Figure 3G), and the δ³⁴S_CRS values above this depth become lighter, ranging from -46‰ to -15.7‰ (Figure 3F).

In the typically anoxic subseafloor marine sediments, the consumption of porewater sulfate is controlled by two microbially mediated processes: 1) organo clastic sulfate reduction (OSR) (Berner, 1982); and 2) anaerobic oxidation of methane (AOM) (Boetius et al., 2000). The two net reactions are expressed stoichiometrically as follows: 2CH₂O + SO₄ ^2-→2HCO₃ ⁻ + H₂S 1) CH₄+SO₄ ^2-→HCO₃ ⁻ + HS⁻ + H₂O 2) However, these two different processes always result in different average S/C ratio in sediments. In oxic and suboxic marine sediments (OSR dominate), the reduced sulfur and total organic carbon (TOC) contents typically show a positive correlation with an average S/C ratio of 0.36 (Berner, 1982). In our studied cores, the TOC content was typically low (Zhang et al., 2018), and the S/C ratios at 6.0–9.0 m bsf (973-4) and below 4.0 m bsf (973-5) were higher than 0.36 (Figure 4). These situations were referred to an organically-limited and methane-rich environment, in which AOM was the dominant process, contributed to a significant fraction of sulfides (Kaneko et al., 2010; Lim et al., 2011; Sato et al., 2012) and may have significantly increased the S/C ratios of sediments. This AOM origin for sulfides explains why there was no correlations between the CRS and TOC contents in the sediments of both cores (Figure 5).

Positive δ³⁴S (up to 32.6‰) values of sulfide minerals could represent a present- or paleo-SMTZ (sulfate-methane transition zone) where AOM process occurred strongly (Aharon and Fu, 2003; Jorgensen et al., 2004; Borowski et al., 2013; Zhu et al., 2013). In the studied cores, the δ³⁴S_CRS values below 6.0 m bsf in core 973-4 and below 4.0 m bsf in the core 973-5 are positive. Especially in the 6.0–9.0 m bsf layer of core 973-4, a wide SMTZ with positive sulfur isotope compositions up to 20‰ represents a sustained and stable methane flux. As shown in the Figure 2G, at the end of the LGM period (4.0–6.0 m) with a low sea-level stage, the δ³⁴S_CRS became highly negative (−43.8‰ to −39.4‰). The low δ³⁴S values in seep-impacted sediments may be attributed to iron limitation caused by low sedimentation rates (Formolo and Lyons, 2013) or disproportionation of microbial sulfur occurring close to the sediment-water interface (Canfield and Thamdrup, 1994; Borowski et al., 2013). Because the sedimentation rate of our study area is typically high at low sea level stands (Zhang et al., 2018), disproportionation of microbial sulfur instead of iron limitation might be the reason for the low δ³⁴S_CRS values. Generally, disproportionation of microbial sulfur happen in an open system, which can be caused by intense methane flux (Li et al., 2018). Herein, we suppose the low δ³⁴S values might be also caused by this phenomenon in this period.

According to previous research, the δ¹³C values of Uvigerina spp. in normal seawater range from −0.1 to 1.0‰ (Rathburn et al., 2003; Schmiedl et al., 2004; Fontanier et al., 2006). However, in our studied cores, the δ¹³C values of benthic foraminifera below 3.9 m bsf in core 973-4 and 2.5 m bsf in core 973-5 (corresponding to the end of the Last Glacial period) were almost all lower than -1.0‰, showing an obvious negative carbon bias. Although some researchers considered that the δ¹³C values in the Glacial period were more negative than that in Deglacial period (Wei et al., 2006), the δ¹³C values of benthic foraminifera in the northern SCS always vary from about -0.6 to 0.2‰ (range variation of approximately 0.8‰ in the past 90 ka) under the influence of climate change (Wei et al., 2006). These indicate that such wide range of the carbon negative bias (e.g.,1.9‰ in core 973-4) in our study area was not caused by climate change and seem to be caused by AOM reaction. In addition, when recording methane seepage activity, a wide range of δ¹³C values of benthic foraminifera are more appropriate than absolute negative values (Rathburn et al., 2003). Thus, the carbon negative bias can be attributed to methane seepage in the studied cores. Because ¹⁸O is higher in cold seep fluids, and the heavier δ¹⁸O values were thought to be another evident of gas hydrate dissociation or cold seep activity (Uchida et al., 2004). In core 973-4 and 973-5, the δ¹⁸O values were obviously heavier during the Glacial period in both cores, which is consistent with the δ¹³C results. Thus we can tentatively deduce the methane seepage activities in the core 973-4 by analyzing the δ¹³C and δ¹⁸O values change. Figure 2E shows negative carbon isotope compositions in MIS2 (4.0–9.0 m), indicating a sustained methane seepage during this period. Furthermore, other reports have suggested that cold seeps were particularly active at low sea-level stands due to a hydrostatic pressure reduction in the SCS (Tong et al., 2013; Han et al., 2014), hence it supports our hypothesis that an intense methane seepage happened during LGM (4.0–6.0 m). Since the end of the LGM (above 4.0 m), the δ¹³C values in core 973-4 were quite stable and comparable to normal sea-water values typical of non-seep environments. This implies that the gas hydrate dissociation events gradually weakened in this core since the Holocene. As the temperature rose gradually in the Deglaciation periods, the changes in global ice volume and bottom water temperature were not the dominant factors for hydrate decomposition. Instead, as the sea level and the hydrostatic pressure on the sediment have gradually increased since the Last Glacial, the gas hydrate stability zone has thickened, thus causing a weakening of gas hydrate dissociation and methane seepage. Uvigerina spp. and Bulimina spp. are considered as specific endophytic benthic foraminifera that can adapt to the modern cold seep environments with high TOC and low oxygen content (Abu-Zied et al., 2008), and they were the two dominant species of benthic foraminifera in cores 973-4 and 973-5. Moreover, as shown in Figures 2B,C, their numbers increased to different extents during the Last Glacial period (4.0–9.0 m of core 973-4), which may be under the influence of the methane flux related to the gas hydrate dissociation during that time (Chen et al., 2007). However, previous research indicated that normal cold seep activity cannot significantly change the total abundance of benthic foraminifera (Panieri et al., 2009). Therefore, the large increases in the number of benthic foraminifera at the end of the LGM period (4.0–6.0 m of core 973-4), especially the abrupt increase in the layer at about 6 m bsf in core 973-4, is more likely associated with geological events rather than to large quantities of methane fluid. Sustained methane seepage can provide a rich food source for benthic foraminifera (Schonfeld, 2001), and Bulimina spp. can adapt well to a low oxygen and high sulfide cold seep environment, even in shallow waters (Panieri, 2006). The high density of this species related to methane seepage lead to another increase above the layer in which gas hydrate dissociated.

Gravity flows such as submarine landslides and turbidity currents are ubiquitous in Dongsha area (Li et al., 2020). And these gravity flows are proved to be able to drive erosion and deposition in Taiwan submarine canyons. As shown on the bathymetric and seismic profiles close to the research areas. Recurrent MTDs are identified in the slope, which indicate Dongsha slope might be eroded by slope failures. And Li et al. (2020) concluded that a combined action of recurrent slope instability and turbidity currents drive the erosion and deposition of submarine asymmetry in our studying area. But from the C-M plot (Figure 6) (Passega, 1964), a common method to identify the force of deposition using the sediment size. The grain size distribution in the 4.4–6.0 m bsf layer in core 973-4 is not parallel to the RQ line in the, which indicated that sedimentation was not impacted by turbidity currents. Thus we contribute the gravity flows recorded in our sediment cores to submarine landslides.

The sediment particle size results in the three cores, 973-3 (Chen et al., 2014), 973-4 and 973-5, which were located at the top, middle and bottom of the Dongsha area continental slope, were comprehensively investigated. The submarine landslides in this study area mainly developed at the top of the continental slope, with lower intensity in the middle slope, and only landslide deposits were present at the bottom of the slope. Furthermore, the age reversal interval (Figure 2) coincided with the high foraminifera and sand content in layer 4.4–6.0 m bsf in core 973-4, which is consistent with frequent submarine landslides influencing this area (Zhong et al., 2015; Lin et al., 2016; Wu et al., 2018).

Based on a comprehensive analysis of geochemical and sediment records, gas hydrate dissociation events and submarine landslides were recognized in two cores. We take 973-4 for analyzing (Figure 7). There was sustained and stable methane seepage during the beginning of MIS2 (6.0–9.0 m, Figure 7A) and the methane flux was relatively low. Therefore, iron sulfides with high δ³⁴S_CRS and benthic foraminifera with negative δ¹³C were found in the sediment layer (Figure 7A). At the end of MIS2, responding to the LGM period (4.0–6.0 m), the methane flux was very high and the SMTZ was near the seafloor (Figure 7B). In this case, the enrichment of TOC might be due to the lower rates of AOM and higher methane flux into the water column (Consolaro et al., 2015). More importantly, the massive decomposition of gas hydrate during the low-stand sea level period reduced slope stability (Sultan et al., 2010; Kwon et al., 2011), and herein led to the formation of the submarine landslides and sediment transportation from the top to the flat area of the slope (Figure 7B). The abrupt increase in benthic foraminifera quantity also verify this geological activity. Because methane diffused into the atmosphere before it had completely reacted, the δ³⁴S_CRS values of iron sulfides were low but the δ¹³C values of benthic foraminifera showed negative anomalies (Figure 7B). Since the Last Glacial period (above 4.0 m), rising sea level has prevented the decomposition of gas hydrates and gradually weakened cold seep activity. Thus, the geochemical and benthic foraminifera records showed no further anomalies (Figure 7C).