1 INTRODUCTION

The Yellow Sea Warm Current (YSWC), which generally runs along the west side of the Yellow Sea Trough, with a water depth of 50–70 m (Lin et al., 2011), brings warm and saline water into the Yellow Sea. The variations of YSWC is crucial to the formation of the modern hydrodynamics and sedimentary system in the Yellow Sea. Today, the circulation pattern in the Yellow Sea displays significant seasonal variations. YSWC is one of the most common oceanic phenomena in winter, which compensates for coastal currents of China and South Korea (Naimie et al., 2001). A series of hydrological fronts are formed when the northward warm, saline YSWC encounters the southward cold, fresh coastal currents, which are much stronger when the contrast between the YSWC and coastal currents is more apparent (Chen, 2009). The fronts work as "water barriers" to prevent coarse terrestrial materials from being transported into the central Yellow Sea muddy area (CYSM) when the YSWC is strong (Li et al., 2014; Gao et al., 2016; Hu et al., 2018). However, the fronts are weak or nonexistent when the YSWC is weak. In these cases, the coarse terrestrial materials can be transported and deposited in the CYSM. Therefore, the strength of the front could be used to indicate the strength of the YSWC to some extent. The seasonal variations of the circulation patterns determine the dispersal and deposition of riverine sediments in the Yellow Sea. The suspended sediment concentration in the coastal areas is higher than the outer shelf sea areas and in winter is higher than in summer (Bian et al., 2013). In addition, the suspended sediment deposits in summer and transports in winter (Yang et al., 2011). The weak tide currents and low energy areas are the main reason for the formation of muddy deposit in the central Yellow Sea (Wang et al., 2014; Zhou et al., 2015; Gao et al., 2016), but the YSWC in winter may be the dominate force for transporting sediments to the CYSM (Lim et al., 2015; Gao et al., 2016; Koo et al., 2018). The YSWC is distinctive due to the high surface temperature in the pathway it imparts. Although the change of sea surface temperature (SST) is also affected by other climatic factors, such as the East Asian Winter Monsoon (EAWM), SST in the pathway of the YSWC could provide valuable information on the variability of its intensity and regional significance.

Previous researches have focused on the invasion time of YSWC. It is roughly believed that the invasion of YSWC started at 7–5 ka BP (Kim and Kucera, 2000; Xiang et al., 2008; Li et al., 2009; Nan et al., 2017a), but the time of YSWC approached to the modern condition may slightly later than its invasion (Kong et al., 2006; Li et al., 2009). However, there have been few studies on the variation of the YSWC due to the difficulty in characterizing the intensity of this current system with appropriate proxies (He et al.,2014; Zhang et al., 2019). Moreover, there is a lack of data from long-term continuous ocean current observations under the impact of fishing, and therefore some important issues regarding the YSWC remain poorly understood, such as how much does the YSWC affect the regional SST, especially the temperature in the pathway of the YSWC. The heat loss caused by the EAWM and heat lateral inputs by the YSWC complicate the SST variation. Existing SST records from the YSWC pathway indicate that the variation of SST is broadly coeval with the EAWM (Jia et al., 2019), while, the warm waters carried by the YSWC could alleviate the SST decline induced by a strong EAWM (Wang et al., 2011), implying a considerable effect of the YSWC on the regional SST. Moreover, modern observations have shown that the heat loss of the Yellow Sea surface generally corresponds to increases in the surface temperature of local regions in the winter, which are tightly linked to various features of the YSWC, such as its temperature and strength (Wei et al., 2010). The records from three sites in the YSWC pathway showed a clear decreasing trend of SST during the last 200–300 years, caused by weakened YSWC (He et al., 2014). It was contrary to the variations of Northern Hemisphere temperature and solar insolation, indicating that the global temperature signal was overwhelmed by the regional process and the YSWC. Zhang et al. (2019) also found that the SST pattern of the YSWC pathway in the North Yellow Sea was opposite to the coastal Yellow Sea and the variation of total solar irradiance (ΔTSI) at multi-centennial timescales over the last 1 000 years, potentially indicate a remarkably intensified YSWC during the Little Ice Age. Although many domestic and international studies focus on the relationship between YSWC and regional SST in recent decades, the amount of data accumulated remains insufficient to fully understand its regional significance.

The Holocene climate had experienced a series of rapid changes on the centennial-scale (Mayewski et al., 2004; Wanner et al., 2011). The variation of ΔTSI had been proposed as the external driving factor (Bond et al., 2001; Fleitmann et al., 2003; Wang et al., 2005), and the North Atlantic thermohaline circulation served as a climate amplifier and conveyor (Bond et al., 1997, 2001; Renssen et al., 2006; Kang et al., 2018). In this hypothesis, the North Atlantic signal was transmitted to the western Pacific region through the thermohaline circulation or the East Asian monsoon system (Hong et al., 2009; Sun et al., 2012).However, some works had shown that the rapid climate changes in mid and low latitudes were not strictly synchronized with North Atlantic events and had different periodicities (Khider et al., 2014; Sagawa et al., 2014; Ruan et al., 2015). Therefore, other climate factors, such as El Niño-Southern Oscillation (ENSO), were believed to have mainly driven the climate variation during mid-late Holocene in East Asia (Lim and Fujiki, 2011; Park et al., 2016; Lim et al., 2017b).

As a branch of the Northwest Pacific boundary current, the YSWC transports large amounts of heat from the tropics to middle latitudes (Song et al., 2009; Xu et al., 2009). It is also a compensating current of Yellow Sea Coastal Current (YSCC), which was mainly caused by the EAWM (Huang et al., 2005; Yuan and Hsueh, 2010). EAWM could dominate the variation of SST in the Yellow Sea and the intensity of YSWC both on millennial and centennial-scale during the mid-late Holocene, (Wang et al., 2011; Zhao et al., 2013; Ge et al., 2014; Wu et al., 2016; Nan et al., 2017a, b; Jia et al., 2019; Zhang et al., 2019). However, the Kuroshio could also play a crucial role in the millennial-scale changes of Yellow Sea SST and the YSWC (Wang et al., 2011; Nan et al., 2017a, b; Jia et al., 2019). Moreover, The SST in the YSWC pathway and the YSWC could also be modulated by ENSO through the Kuroshio on centennial scale (Zhang et al., 2019). Due to deficiency records of the evolution of YSWC, research on the evolution mechanism of Yellow Sea SST is insufficient. Its teleconnection with high and low latitude forcing, especially low latitude forcing, worth further study.

In this study, we presented new U₃₇^k′-SST and grain size records from the sediment core Z1 under the YSWC pathway from central Yellow Sea. The main objective was to trace the development of the YSWC and the variation of SST in the YSWC pathway since 6.1 ka BP, and to estimate the evolution mechanisms of the SST involving high latitude climate and tropical ocean during the mid-late Holocene.

2 REGIONAL BACKGROUND

The Yellow Sea is surrounded by land on the three sides and has a broad and flat seafloor, with a water depth of 55 m on average. This shallow marginal sea is surrounded by the Chinese mainland and the Korean Peninsula, and it is in connection with the Pacific Ocean through the East China Sea. In winter, the modern circulation system of the Yellow Sea is featured by the northward, warm and salty YSWC on the west side of the Yellow Sea Trough and southward, cold and fresh coastal currents along the eastern China and western South Korea coasts. Due to the natural differences between YSWC and other coastal currents, a series of hydrological fronts form between them. These fronts have a significant impact on the hydrological conditions in the Yellow Sea.

The hydrology of the Yellow Sea displays apparent seasonal variations. During winter, the nature of shelf water is consistent, and the YSWC could intrude into the Yellow Sea and even into the Bohai Sea. However, in summer, the YSWC is weak and rarely reaches north of 35°N. The seasonal variations of the Asian monsoon and ocean currents (YSWC and coastal currents) result in considerable differences in SST patterns between different seasons in the Yellow Sea (Wei et al., 2010). The annual north to south SST gradient can be larger than 3 ℃ in the southern Yellow Sea. The spatial variation of annual SST in the southern Yellow Sea is mainly determined by the changes of winter SST. In winter, strong northerly wind rapidly cools the SST, while the northward, extending YSWC can form a warm tongue from the south to north, leading to a steeper south-north SST gradient (4–6 ℃). In summer, without the invasion of the YSWC, surface heating combined with stratified water column results in a more uniform SST in the southern Yellow Sea, which varies between 22.5 and 24 ℃ with a north-south gradient of about 1.5 ℃ (Schlitzer, 2012).

Due to unique topography and large amounts of terrestrial input from the surrounding rivers, a series of mud patches have developed including the Shandong mud wedge, the northern Yellow Sea mud area, the CYSM, and the southeastern Yellow Sea mud area. The CYSM is the largest of these mud patches. The total area is about 6.1×10⁴ km² and deep than 70 m. It does not extend from estuaries, but away from the coastal zones. The modern sediment accumulation rates of the central Yellow Sea mud area range from 0.3 to 2.7 mm/a (Alexander et al., 1991). A large number of fine particles and organic mass gather here due to the weak tidal current and the activities of the YSWC in this region (Wang et al., 2014), which provide an ideal site for paleoenvironment reconstruction.

3 MATERIAL AND METHOD

The gravity core Z1 (35°18′N, 123°42′E, water depth: 78.5 m, core length: 196 cm, Fig. 1), with undisturbed sediments, was retrieved from the central Yellow Sea in 2010. The upper 114 cm of the core was mainly composed of silt and clay, containing almost no sand (Fig. 2). The lowermost part of the core was dominated by silt, with a certain amount of sand and clay (Fig. 2). Subsamples were taken at 1-cm intervals for grain size and alkene analyses in the Key Laboratory of Marine Geology and Environment Science, Institute of Oceanology, Chinese Academy of Sciences.

For the analysis of grain size, the soybean size samples of core Z1 were first pretreated with 10% excess hydrogen peroxide (H₂O₂) to remove the organic matter. Then an appropriate amount of hydrochloric acid (HCl) (1 mmol/L) was added to remove the carbonate fraction and the samples were washed four times with deionized water to a pH value of 7.0. The samples were dispersed and homogenized with sodium metaphosphate (NaPO₃)₆ solution (6 nmol/L) by ultra-sonication before measurement. A CILAS 1 190-L laser particle size analyzer was used to measure the grain size of the samples, with measurement ranged from 0.04 to 2 500 μm. The repeated error of the instrument measurement was less than 0.5%.

For alkenones analysis, about 4–34 g of dried, pulverized, and homogenized samples with two internal standards (C₂₄ deuterium-substituted n-alkane and n-C₁₉ alcohol) were first extracted using mixed dichloromethane (DCM) and methanol (3꞉1, v/v) by ultra-sonication. The supernatant was extracted at least four times until it was colorless and then dried with N₂ gas. To remove terrestrial wax esters, a hydrolysis step was performed for the residue, with about 4-mL 5% potassium hydroxide (KOH) solution in methanol, and three times of sonication for 10 min each, then left overnight. The hydrolysate was extracted four times with hexane and dried with N₂ gas again. The hydrolytic lipid fraction was separated by silica gel chromatography, and the lipids were eluted with hexane and 5% DCM in methanol successively to get 10-mL alkane and 8-mL alkenone fractions, respectively. The alkenone fractions were dried with N₂ gas a third time, then 20-μL N, O-bis (trimethylsilyl) trifluoroacetamide (BSTFA) and 20-μL isooctane were added for derivatization for 1 h at 70 ℃. Excessed BSTFA and isooctane were dried with N₂ gas a fourth time and the residue was dissolved in 20-μL isooctane. An Agilent-6890N gas chromatograph equipped with a 50-m fused silica column (HP1, 0.32 mm×0.25 μm), splitless injection device, and flame ionization detector was used to analyze the extracts. The carrier gas is helium and the flow rate is 1.2 mL/min. The temperature of the inlet and detector are 280 ℃ and 320 ℃, respectively. An optimized temperature program was applied to separate C₃₇ alkenones ideally (Nan et al., 2012), from 50 to 150 ℃, 30 ℃ per minute; then from 150 to 230 ℃, 8 ℃ per minute; from 230 to 320 ℃, 1.2 ℃ per minute, and finally keep 320 ℃ for 15 min. Alkenones were identified by comparing characteristic peak and elution time. A blank experiment was carried out simultaneously with each batch of samples and there were no detectable alkenone peaks on the gas chromatographic baseline, indicating reliable experimental data. The formula for calculating the unsaturation index is U₃₇^k′=C_37:2/(C_37:2+C_37:3) (Prahl and Wakeham, 1987; Sikes et al., 1991), where C_37:xrepresents the content of alkenones of 37 carbon atoms with x double bonds. The U₃₇^k'-SST was calculated both by the global (SST=(U₃₇^k′–0.044)/ 0.033) (Müller et al., 1998) and local (SST=(U₃₇^k′+0.350)/0.059) (Tao et al., 2012) calibration equations.

Six accelerator mass spectrometry radiocarbon (AMS¹⁴C) ages were obtained from four samples (2cm slices) of well-preserved calcareous material of mixed benthic foraminifera and/or shell fragments in the size fractions larger than 150 μm and two bulk samples (1-cm slices) of organic carbon due to the very low biogenic carbonate content, measured at Beta Analytic Inc. (Miami, FL, USA). The six AMS¹⁴C dates were calibrated against calendar ages with Marine13 Radiocarbon Age Calibration Curve (Reimer et al., 2013) using the online Calib 7.0.4 program (Stuiver, 2017). A regional marine reservoir age of 139 years (ΔR=-139±59), was used in the calibration, following Nan et al. (2017b). The ages were reported in calibrated year before present (a BP; 0=AD 1950), with 2-sigma uncertainty ranges (Table 1). The core Z1 covered the period from ca. 6092 a BP to the present day, based on the six calibrated AMS¹⁴C data. An age-depth model was constructed based on a linear interpolation of the calendar ages (Fig. 2).

4 RESULT 4.1 Grain size and end-member modelling

The core Z1 consisted mostly of silt (48.9%–86.6%), with a high sand content (average 24.8%) from 6.1 to ~3.9 ka BP and a high clay content (average 21.4%) between ~3.9 and 0.0 ka BP (Fig. 2). The mean grain size ranged from 8.9 to 62.4 μm throughout the core. Generally, the grain size fluctuated heavily before ~3.9 ka BP, which was a period with large sand content. It was in sharp contrast to the stable grain size as the content of sand significantly declined after ~3.9 ka BP.

The transportation of sediments to the sea is affected by factors such as provenance, topography, and hydrodynamic conditions. Therefore, it is necessary to separate the sub-population of grain size when using the grain size proxy for marine sedimentation paleo-environment research. The end-member analysis model is a powerful tool to unmix the grain size data of multi-component sediments. The parametric lognormal end-member analysis was performed on the grain size data of core Z1 to identify the unimodal sub-population of terrestrial sediments. To determine the minimal end-members, which are necessary for a better interpretation of the data, the coefficient of determination (R²) and mean angular deviation (θ) are calculated. Relatively high R² and low θ values suggest a good statistical fit (Paterson and Heslop, 2015). The R² for the two end-members model for the grain size in the core Z1 is 97.6%, indicating the variance of each grain size could be reproduced (Fig. 3a). The θ value is 7.1° for the model with two end-members, indicating the difference between the estimated and true values (Fig. 3b). The two end-members model meets the requirement of minimal end-members and maximum reproducibility. The grain size distributions of the two end-members are shown in Fig. 3c. The mode of end member 1 is 50.03 μm, which belongs to coarse silt, and end member 1 represents coarse-grained terrestrial materials (Zhang et al., 2016). The mode of end member 2 is 8.70μm, which belongs to fine silt, and end member 2 represents fine-grained terrestrial materials (Zhang et al., 2016).

The content variation of end-member 1 and end-member 2 was opposite along the core since 6.1 ka BP (Fig. 4). The content of end-member 1 shows a significantly decreased trend of range from 11.1% to 89.0% during 6.1–~3.9 ka BP, then present rather stable values with an average value of 4.9%.

4.2 The U₃₇^k′-SST sequence

The unsaturation index of alkenones for the core top sample was 0.53, equivalent to the SST of 14.6 or 14.9 ℃ based on the global and local equations, respectively. And the average U₃₇^k′ of the whole core was 0.57, equivalent to the SST of 16.0 and 15.6 ℃ based on the global and local equations, respectively. Although, the two different calibration equations produced quite similar values, the amplitude of SST variation of core Z1 from the global equation (up to 9.3 ℃) was much higher than that from the local equation (about 5.1 ℃) during 6.1 ka BP (Fig. 5). Because of the quite smaller temperature fluctuation during the mid-late Holocene than the last deglaciation period (Dang et al., 2020), and the response of haptophytes which are the major producers of C₃₇ alkenones to temperature might be different between the southern Yellow Sea and the open ocean (Tao et al., 2012). The local equation could be more suitable for the SST calibration of the core Z1.

During the past 6.1 ka BP, the estimated U₃₇^k′-SST fluctuated between 12.6–17.7 ℃, with an average of 15.6 ℃ (Fig. 5), which was consistent with the results obtained from nearby cores in the southern Yellow Sea (Wang et al., 2009, 2011; Wu et al., 2016; Jia et al., 2019). The SST time series revealed a gradual increase since 6.1 ka BP. During ~5.3–4.2 ka BP, the SSTs were much higher, which was then followed by a period of lower values during ~4.2–2.8 ka BP. There were 10 SST peaks at centennial-scales, centering at 5.1, 4.3, 3.6, 3.3, 3.0, 2.4, 2.0, 1.4, 0.8, and 0.4 ka BP, respectively (Fig. 5). Three minimum SSTs centered at 5.7, 4.7, and 4.0 ka BP were also apparent within the whole core (Fig. 5).

5 DISCUSSION 5.1 Regional pattern of SST evolution in the YSWC pathway

The almost same abundance of dominant haptophyte species Emiliania huxleyi and Gephyrocapsa oceanica in the surface water of the southern Yellow Sea in winter and summer (Sun et al., 2014), indicate that their production of C₃₇ in our core represented a roughly consistent annual level. Also, U₃₇^k′ values have a strong linear correlation with mean annual SSTs rather than seasonal SSTs (Tao et al., 2012). In our results, the U₃₇^k′-SST of core top sample (14.9 ℃) was similar to modern annual mean SST of 15.43 ℃ but quite different from summer average SST of 22.83 ℃ and winter average SST of 9.41 ℃ (Schlitzer, 2012). Thus, the U₃₇^k′-SST along the core could represent the annual mean SST of Yellow Sea since 6.1 ka BP (Wang et al., 2011; Xing et al., 2012; Wu et al., 2016; Jia et al., 2019; Zhang et al., 2019). However, the present spatial pattern of annual SSTs in the southern Yellow Sea was mainly determined by the winter SSTs since the summer SSTs had uniform values (Schlitzer, 2012). Therefore, the variations of U₃₇^k′-SST in core Z1 could be mostly involved with the variations of winter SST in the region since 6.1 ka BP.

The pattern of the SST variation from the YSWC pathway since mid-Holocene indicates a regional pattern more likely. The SST variation along the core Z1 during the past 6.1 ka BP (Fig. 6a) was consistent with other SST records (Fig. 6b–d) from the YSWC pathway (Wu et al., 2016; Jia et al., 2019; Zhang et al., 2019), with a gradually warming trend. From a global perspective, there were quite large differences between SST variation from the YSWC pathway with the evolution of temperatures in the Northern Hemisphere (Fig. 6g), the Chinese mainland (Fig. 6h), and the East China Sea (Fig. 6i) during the mid-late Holocene (Marcott et al., 2013; Zheng et al., 2017; Yuan et al., 2018). The stacked Northern Hemisphere temperature had displayed decreasing trends closely related to obliquity and total insolation since the mid-Holocene, suggesting the action of orbital forcing on climate change (Marcott et al., 2013). The reconstructed air temperature from the northeast and southwest of the Chinese mainland also displayed a similar evolutionary trend and was influenced by a reduction in insolation (Zhao et al., 2011; Zheng et al., 2017). The SST of the East China Sea decreased broadly with the changes in global temperature under the influence of reducing summer insolation during the mid-late Holocene; however, the abrupt decrease in SST was caused by the regional forces of strengthened circulation system and the stronger cold eddy during the last two thousand years (Yuan et al., 2018). The differences between the patterns of temperature variation in different regions indicated the divergences of climate or oceanic conditions caused by different drivers. The distinctive SST pattern in the YSWC pathway, which was almost the exact opposite of the temperature variations in the above regions, indicated its separate background during the mid-late Holocene.

5.2 Influence of EAWM on SST variation in the YSWC pathway

The reduced temperature gradient between high and low latitudes resulted in a weakened EAWM since 6.1 ka BP (Huang et al., 2011; Hu et al., 2012; Li and Morrill, 2015). An increasing SST between 6.1–~1.8 ka BP may be related to the gradually weakened EAWM which directly cools the Yellow Sea by promoting heat loss from the sea surface. However, the stalely high SST values in our core after ~1.8 ka BP are in contrast to the strengthened EAWM indicated by the sensitive grain size of sediments of three cores from CYSM (Hu et al., 2012), implying that EAWM was not the only factor affecting SST variation in the YSWC pathway.

A comparison with the records from the Lake Huguang Maar in the coast of southeastern China revealed that the variation of SST along the core Z1 was opposite to EAWM index values, as represented by titanium content of sediments on a centennial-scale before ~3.9 ka BP (Fig. 7d) (Yancheva et al., 2007). Within the uncertain chronology, there was a good correspondence between three cold events (~4.0, ~4.7, and ~5.7 ka BP) and two warm events (~4.3 and ~5.1 ka BP) at our site with the strengthening and weakness of the EAWM. It indicated that the variation of SST could be attributed to the variability of the EAWM intensity before ~3.9 ka BP. The cold events, which resulted from the intensification of the EAWM, were also well recorded in the East China Sea and northwest Pacific (Xiao et al., 2006; Nakanishi et al., 2012; Sagawa et al., 2014). Therefore, we suggested that the EAWM played the dominant role in changing the SST of the Yellow Sea before ~3.9 ka BP because that YSWC was much weaker during this time-window (Li et al., 2009).

Today, because of the existence of the fronts between the YSWC and coastal currents, coarse-grained terrestrial sediments from both China and South Korea cannot cross the fronts, while the fine-grained sediments can successfully migrate across the fronts and deposit in the CYSM. The strength of the fronts could be represented by the content of end-member 1(i.e., coarse particles). Extremely high end-member 1 content indicated a weak YSWC during 6.1–~3.9 ka BP (Fig. 7e) when the frontal system had not yet completely formed. Based on analysis of foraminifera fauna in the southeastern Yellow Sea, Li et al. (2009) proposed that the YSWC was weak in the early stage of its formation from 6.4 to 4.2 ka BP. Multiple mineralogical and geochemical indexes of three cores from the CYSM revealed that the dominant sediment source gradually changed from the Huanghe (Yellow) River and sea bed erosion to the Changjiang (Yangtze) River, probably cause of the continuously enhancing intensities of the YSWC and YSCC, as well as the oceanic front between them (Lim et al., 2015). The content of end-member 1 was low and relative constant between ~3.9 and 0.0 ka BP, indicating that the modern YSWC had finally formed around ~3.9 ka BP. Our unpublished data of benthic foraminiferal fauna, which changed from the dominance of brackish species (Ammonia aomoriensis, Buccella frigida, and Elphidium advenum) to modern shelf-dwelling species (A. ketienziensis, Hanzawaia nipponica, and A. tasmanensis) after ~3.9 ka BP, also indicated that the development of YSWC was insufficient in the Yellow Sea until ~3.9 ka BP, even though the invasion of it had begun around 5–7 ka BP (Kim and Kucera, 2000; Xiang et al., 2008). A strong YSWC could enhance the "water barrier" effect of the frontal system and resulting in a low end-member 1 content.

Modern oceanographic observations have suggested that the intensity of the YSWC is largely controlled by the EAWM (Huang et al., 2005; Yuan and Hsueh, 2010). A strong EAWM could strengthen the YSCC, which increased the southward transport of water along the Yellow Sea coast. As a compensatory flow of the YSCC, the YSWC would enhance and then transport more warm water to the center Yellow Sea, resulting in higher SST at site Z1. It was evidenced by the distinct positive correlation between the strengthened EAWM indicated by the Ti content and the higher SSTs in the core Z1 after ~3.9 ka BP. It suggested that the variation of SST in the YSWC pathway could be attributed to the variability of the YSWC intensity after ~3.9 ka BP.

Larger fluctuations of SST after ~1.8 ka BP could result from the strengthened EAWM during this period. Due to the relatively shallow depth of the Yellow Sea, water mixing would be enhanced under a strengthened EAWM, which could change the stratification of surface water and thus enlarge SST fluctuation during the last ~1.8 ka BP.

5.3 Influence of Kuroshio Current on SST variation in the YSWC pathway

A relatively low SST period in core Z1 occurred during ~4.2–2.8 ka BP, which could be related to the weakness of the Kuroshio Current (Jian et al., 2000; Lim et al., 2017a) (Fig. 6e), which originates from the North Equatorial Current in the tropical Pacific (Wu et al., 2012). In the Northwest Pacific boundary current system, the YSWC is one of the branches of Kuroshio Current. About two-thirds of the annual mean transport of the YSWC is determined by the Kuroshio Current (Xu et al., 2009). The abundance of Pulleniatina obliquiloculata and the enrichment of sediment mercury (Hg) has been applied to reconstruct the intensity of the Kuroshio Current (Jian et al., 2000; Lim et al., 2017a). A well-identified P. obliquiloculata minimum event and an abrupt drop in sediment Hg levels occurred during ~4.9–2.7 ka BP, implying that the intensity of the Kuroshio Current reduced at that time, corresponding to the lower SST of the core Z1 during ~4.2–2.8 ka BP. A similar SST variation has also been observed in other cores from the YSWC pathway (Wang et al., 2011; Wu et al., 2016; Jia et al., 2019) (Fig. 6), indicating that the SST record from the YSWC pathway has been affected by the variations of the Kuroshio Current on a millennial-scale through YSWC. A weakened Kuroshio Current could reduce the intensity of the YSWC and then decrease SST at site Z1, which was also consistent with the conclusions reached by other researchers (Wang et al., 2011; Nan et al., 2017a, b; Jia et al., 2019). Due to the rough resolution of these data about the evolution of Kuroshio Current (Fig. 7f), it is impossible to further discuss the impact of Kuroshio Current on a centennial-scale variation of SST in the YSWC pathway. An indirect method is that tropical Pacific conditions play an important role in regulating the intensity of Kuroshio Current. Perhaps the relationship between tropical Pacific conditions and SST in the YSWC pathway can be used to explore the impact of Kuroshio Current on a centennial-scale variation of SST in the YSWC pathway.

As the strongest inter-annual climate anomaly in the tropical Pacific, the El Niño has a significant impact on the variations of the Kuroshio Current. At the centennial-scale, SST variation at site Z1 was similar to the lithic proxy for El Niño flood events off coastal Peru (Rein et al., 2005) (Fig. 7g). Model simulations indicated a greater transport and a more northward shift of the bifurcation latitude of the North Equatorial Current during El Nino-like periods, which would increase the transport of the Kuroshio Current (Kim et al., 2004). The enhanced Kuroshio Current would increase the strength of the YSWC, carried more warm water to site Z1, and increased the SST in the area. Therefore, the good correspondence between our SST record and the ENSO index provides evidence for ENSO to regulate the variation of SST in the YSWC pathway through Kuroshio Current on a centennial-scale.

5.4 Connection of SST to Northern Hemisphere high latitude climate and the tropical ocean

The oceanic conditions of the modern Yellow Sea are sensitive to both high latitude and low latitude forcing. Several studies have explored the relationship between the SST in the Yellow Sea and climate forcing factors, such as the ΔTSI, the EAWM, and the Kuroshio Current (Wang et al., 2011; Ge et al., 2014; He et al., 2014; Nan et al., 2017a, b; Jia et al., 2019; Zhang et al., 2019). The variability of the ΔTSI had been proposed as an external driver (Nan et al., 2017a, b; Jia et al., 2019). Spectral analysis of the U₃₇^k′-SST sequence was performed using REDFIT38, which is a program for the spectral analysis of time series of samples with different resolutions (Schulz and Mudelsee, 2002). Spectral analysis of the U₃₇^k′-SST sequence revealed two periodicities of 1 010 and 538 years, with a high confidence level (95%) (Fig. 8), which corresponded to the 1 000 and 550-year periodicities previously recognized in GISP2 δ18O record during the Holocene (Stuiver et al., 1995), nitrate concentrations of the Talos Dome ice core from eastern Antarctica (Soon et al., 2014), and changes of the North Atlantic Deep Water circulation (Chapman and Shackleton, 2000). Besides, the 538year periodicity in our core could also coincide with the 500-year cycle identified in tree-ring Δ¹⁴C (Stuiver and Braziunas, 1993). The three periods were assigned to solar modulation (Stuiver and Braziunas, 1993; Stuiver et al., 1995; Chapman and Shackleton, 2000; Soon et al., 2014). Therefore, in general, SST variations in the YSWC pathway could be attributed to solar activity during the mid-late Holocene.

Multitude centennial-scale variabilities were superimposed over the long-term SST trend in the core Z1 (Fig. 7c). The variations of the ΔTSI were too small to drive the significant change in ocean surface temperature (Foukal et al., 2006; Gray et al., 2010; Khider et al., 2014). Many studies have shown that abrupt changes in the EAWM were driven by the Atlantic meridional overturning circulation (Sun et al., 2012). The slowing/shutting of the thermohaline circulation could have made the entire Northern Hemisphere much colder in winter, increasing the meridional temperature gradient in the Northern Hemisphere and strengthening the EAWM (Wu et al., 2008). Within age uncertainty, the lower SST in core Z1 at ~5.7 ka BP well corresponded to the lower ΔTSI (Fig. 7a) and the increased North Atlantic ice raft debris (an indicator of Bond 4 event) (Fig. 7b). And the lower SST at ~4.7 ka BP also synchronized with the lower ΔTSI. A reduction in solar insolation would amplify the ocean-land temperature difference in winter by affecting ocean or atmosphere circulation, leading to the strengthening of EAWM, then decreased SST in the YSWC pathway when the YSWC was weak before ~3.9 ka BP. The highest SST in the core Z1 at ~1.4 and ~0.4 ka BP corresponded well to the lower ΔTSI and North Atlantic Bond 1 and 0 events, implying that the evolution of SST in the YSWC pathway after ~3.9 ka BP was dominated by the YSWC.

The good correlation between SST and ENSO index during ~4.1–1.8 ka BP, indicates that centennial-scale variation of SST could be regulated by ENSO. The 538-year cycle of SST in core Z1 could also be related to the activities of ENSO as shown by the ENSO cycle recorded in Laguna Pallcacocha Lake during ~4.0–2.0 cal. ka BP (Moy et al., 2002). Abundant records have shown more El Niño activities during this period (Rein et al., 2005; Selvaraj et al., 2012; Toth et al., 2012; Zheng et al., 2014; Park, 2017). Under El Niño-like conditions, the enhanced Kuroshio Current would increase the intensity of YSWC which would transport more warm water to site Z1 and resulting in higher SSTs. Therefore, ENSO maybe plays a significant role in the marine environment variation in the Yellow Sea during the mid-late Holocene.

The high correlation provided strong evidence that there were good relationships between the SST of Yellow Sea with the output of solar energy and the tropical Pacific. Therefore, the millennial and centennial-scale variation of SST in the Yellow Sea could be an indirect response to the reduction of solar insolation, which was amplified/transmitted by North Atlantic thermohaline circulation and EAWM.Meanwhile, the Yellow Sea was also connected to the tropical ocean environment through the Kuroshio Current. A potential mechanism explaining the evolution of SST in the YSWC pathway is provided in Fig. 9.

6 CONCLUSION

This study provided alkenone-based SST and grain size records related to changes in oceanic conditions in the YSWC pathway during the mid-late Holocene. SST variations over the last 6.1 ka BP indicate that EAWM and YSWC maybe two dominating factors that determine the evolution of SST in the YSWC pathway during the mid-late Holocene. The variation of SST in site Z1 was mainly influenced by the variation of EAWM before ~3.9 ka BP when YSWC was much weaker. However, after YSWC was fully developed around ~3.9 ka BP, which has played a key role in SST evolution in the YSWC pathway. There was a teleconnection between SST of the Yellow Sea and global climate change through EAWM and Kuroshio Current. Against the background of Holocene solar activity, the signals of the Northern Hemisphere high-latitude were amplified by the thermohaline circulation and then transmitted to the Yellow Sea by EAWM. The signal of El Niño from the tropical Pacific was also transmitted to the SST of the Yellow Sea via the warm Kuroshio Current. The double compensation effect of EAWM and Kuroshio Current on the warm water transportation of the YSWC tightly linked the SST variations in the YSWC pathway to the Northern Hemisphere high latitude climate and tropical ocean.

7 DATA AVAILABILITY STATEMENT

The datasets generated during and/or analyzed during this study are available from the corresponding author on reasonable request.

8 ACKNOWLEDGMENT

We appreciate the reviewers for their constructive comments on this manuscript and editors for the editing. We thank sampling by the experimenters of R/V Kexue 3. We are grateful to Dr. Qingyun NAN and experimenter Hongli WANG from Institute of Oceanology, Chinese Academy of Sciences for the experimental guidance. We are also grateful to Rein BERT from University of Trier for kindly providing the data of the ENSO index.