Reconstruction of South China Sea Deep Water Salinity During the Last Glacial Maximum (LGM)

Abstract

Reconstructing the deep water salinity during the Last Glacial Maximum (LGM, 26.5~19 ka BP), corresponding to Marine Isotope Stage 2, the most recent and coldest period, is crucial for understanding glacial deep ocean circulation variation and its effect on the climate. The South China Sea (SCS) is one of the largest marginal seas in the western Pacific Ocean, where LGM deep water salinity reconstruction remains unexplored. This study employs pore water [Cl⁻] profiles acquired from boreholes of Site U1499 of IODP Expedition 367 and Sites U1431 and U1433 of IODP Expedition 349 to reconstruct the LGM salinity in the deep SCS. Utilizing a one-dimensional diffusion-advection numerical model, the LGM salinity of the deep northern SCS is determined to be 35.68 ± 0.04 g/kg, and that of the deep central SCS is 35.61 ± 0.03 g/kg, revealing an intra-basin salinity gradient of ~0.07 g/kg. LGM salinity gradients within the SCS were reduced relative to modern ones, indicating attenuated deep circulation within the SCS during the LGM. Furthermore, a diminished salinity gradient (Δ = 0.02 g/kg) across the Luzon Strait between the SCS and Pacific and an enhanced vertical stratification between Upper Circumpolar Deep Water (UCDW) and Lower Circumpolar Deep Water (LCDW) collectively support a sluggish deep Pacific circulation during the LGM.

1. Introduction

From 26.5 to 19 ka BP, corresponding to Marine Isotope Stage 2 (MIS 2), the Earth entered the Last Glacial Maximum (LGM), widely regarded as the coldest geological episode in recent Earth history [1,2,3]. Global temperatures were 3–6 °C lower than the present temperature, and ice sheet volumes reached their maximum extent, resulting in a global sea level drop by 120–135 m compared with the current level [4,5,6,7,8,9]. Atmospheric CO₂ concentrations during the LGM were 80–90 ppm lower than pre-industrial levels, a magnitude of change comparable with the observed CO₂ increase since the Industrial Revolution [10,11,12,13,14]. Consequently, the LGM serves as a benchmark period for global paleoclimate research, proving essential for understanding Quaternary glacial-interglacial cycles and broader geological-scale climate variability and for informing projections of future climate change.

Currently, the oceans cover approximately 71% of Earth’s surface, with the deep ocean (water depth > 2 km) accounting for over 60% of the global ocean volume. This vast volume of deep seawater plays a crucial role in global matter and heat transport through meridional overturning circulation [15,16]. Excluding the carbon stored in rocks, the dissolved inorganic carbon stored in the deep sea accounts for about 90% of the total dissolved inorganic carbon in all oceans, the atmosphere, and the land organic carbon, which is approximately 60 times that of the atmosphere [17]. Studies of glacial-interglacial cycles demonstrate that atmospheric CO₂ concentrations have varied in close correspondence with climate changes over the past 730 ka [18]. These findings collectively suggest that alterations in deep ocean circulation must exert significant impacts on atmospheric carbon budgets and global climate variability. To predict future changes in deep ocean circulation and their climatic impacts, it is helpful to conduct quantitative modeling studies of LGM deep ocean circulation.

Deep ocean circulation is primarily governed by seawater density, which is determined by temperature and salinity. Thus, the critical challenge in modeling lies in reconstructing the paleotemperature and paleosalinity of LGM deep waters [19]. Proxies for reconstructing LGM deep-sea paleotemperatures are relatively abundant (e.g., benthic foraminiferal δ¹⁸O values, Mg/Ca ratios, and carbonate clumped isotopes) [20,21,22,23,24,25], whereas proxies for paleosalinity remain scarce, making quantitative reconstruction of paleosalinity significantly more challenging. A breakthrough emerged when McDuff [26] identified a pore water chloride concentration ([Cl⁻]) peak at ~25 m depth in sediment cores from Deep Sea Drilling Program (DSDP) Site 576 in the western equatorial Pacific and proposed that this feature might represent a residual signal of LGM deep seawater migration preserved in the sediments, because water was bound in ice during glaciation and the salt was becoming enriched in the remaining seawater. Pore water [Cl⁻] in marine sediments is well known as a conservative tracer, largely retaining its original seawater signature except under influences such as volcanic ash alteration, methane hydrate formation/dissociation, smectite/illite transformation, halite dissolution, phase separation, or evaporation [27]. Furthermore, seawater [Cl⁻] exhibits a linear relationship with salinity [28]. However, pore water salinity can be modified by early diagenetic processes, such as sulfate reduction or mineral dewatering reactions [29]. Therefore, pore water [Cl⁻] may serve as a viable proxy for reconstructing LGM seawater paleosalinity.

When a [Cl⁻] peak (20–35 m) was identified in the pore water depth profile of deep-sea sediments at Ocean Drilling Program (ODP) Site 1063 near the Bermuda Rise in the Atlantic, Adkins and Schrag [30], inspired by McDuff’s hypothesis, first applied a one-dimensional diffusion-advection model to reconstruct the salinity of LGM deep seawater in this region as 35.76 ± 0.04 g/kg, approximately 0.89 g/kg higher than modern deep water salinity. Subsequent studies have extended this approach to key regions across global oceans by analyzing pore water [Cl⁻] depth profiles and applying the same model: In the North Atlantic, simulated LGM deep water salinities at ODP Sites 981 and 1063 were 36.10 ± 0.10 g/kg and 35.83 ± 0.03 g/kg, respectively. In the Southern Ocean, ODP Site 1093 yielded a simulated salinity of 37.08 ± 0.17 g/kg, indicating significantly saltier LGM deep waters in the Southern Ocean compared with the North Atlantic [29]. This contrasts with the modern salinity gradient, where North Atlantic Deep Water (NADW) is saltier than Southern Ocean water [31], implying a reversed deep water salinity gradient between these basins during the LGM. Homola et al. [32] further constrained LGM deep water salinity in the subtropical North Atlantic (35.51–36.21 g/kg) using relationships between pore water density, [Cl⁻], and salinity. Their results, combined with the water depths of these sites, revealed that LGM NADW salinity increased notably with depth, supporting the hypothesis of reduced vertical mixing and intensified stratification in the glacial Atlantic. The enhanced vertical salinity gradient also suggests that the saltier LGM Atlantic bottom water was sourced from high-salinity Antarctic Bottom Water (AABW), whereas the fresher, shallower intermediate/deep water originated from NADW. In contrast, LGM Pacific deep water salinity exhibited relative homogeneity (36.10 ± 0.10 g/kg) [33], with similar values along deep ocean circulation pathways.

Current reconstructions of LGM deep seawater salinity are largely confined to some regions in the North Atlantic, Southwest Pacific, and Northeast Pacific [29,32,33], with a notable lack of studies in the Western Pacific and Southern Ocean and a complete absence of research on marginal seas. The South China Sea (SCS), one of the largest marginal seas in the Western Pacific, serves as a critical conduit for material and energy exchange between the Pacific and Indian Oceans [34]. Existing LGM salinity studies on the SCS have focused primarily on the surface water of northeastern and southwestern continental-oceanic transition zones [35,36,37], whereas salinity data pertaining to the deep SCS remain scarce. This study addresses these research gaps by analyzing pore water [Cl⁻] profiles from the deep ocean (>3700 m) at International Ocean Discovery Program (IODP) Sites U1499, U1431, and U1433 in the SCS. We reconstructed LGM deep water salinity in the SCS using a one-dimensional diffusion-advection model. Specifically, we elaborated in detail on various parameters of the model, including the porosity, effective diffusion coefficient, advection velocity, and initial and boundary conditions of the model. Based on the simulated salinity results, we further analyzed LGM deep water circulation pathways and flow velocities within the SCS. We also analyzed the LGM salinity gradients cross the Luzon Strait by supplementing Upper Circumpolar Deep Water (UCDW) salinity simulation result, providing insights into the strength of western Pacific deep water intrusion into the SCS during the LGM. The reconstruction of SCS deep water salinity not only fills a critical gap in marginal sea research but also supplements the existing database of UCDW salinity during the LGM. This work contributes to a more comprehensive framework for studying glacial-interglacial changes in thermohaline circulation and their implications for global climate dynamics.

2. General Situation of the SCS and Sites Information

The SCS, located at the junction of the Pacific Plate, the Eurasian Plate, and the Indo-Australian Plate, has emerged as a hotspot of global climate change research in recent years due to its unique geographical position and tectonic evolution. The SCS is a semi-enclosed basin, connected to the deep Pacific Ocean solely through the Luzon Strait between Taiwan Island and Luzon Island, and the deepest sill in the strait is approximately 2600 m [38,39,40,41]. Below this depth, there is no direct seawater exchange between the Pacific and the SCS. On the eastern side of the Luzon Strait, within the depth interval spanning approximately 2000 m to the sill’s maximum depth, the lower potential temperature and higher salinity of Pacific deep water resulting in higher density exceeds that of the SCS [41]. This meridional density difference generates a persistent baroclinic pressure gradient oriented westward across the Luzon Strait, providing the primary forcing mechanism for the overflow from the Pacific into the SCS basin. Consequently, this water sinks after it crosses the Luzon Strait, forming a deep water overflow [42,43]. Therefore, SCS deep water is relatively homogenous and exhibits characteristics similar to Pacific water of 2000~2600 m depth [44]. Once inside the basin, this deep seawater circulation within the SCS constitutes a basin-scale cyclonic gyre, characterized by a counterclockwise flow pattern [39].

We used geophysical and pore water [Cl⁻] data collected from Site U1499 of IODP Expeditions 367 and Sites U1431 and U1433 of IODP Expedition 349 to reconstruct the salinity of LGM deep water in the SCS. IODP 367-Site U1499 (18°24.5698′ N, 115°51.5881′ E) is located on Basement Ridge A within the continent–ocean transition zone of the SCS (Figure 1), at a water depth of 3760 m, with a bottom water temperature of 2.5 °C (potential temperature θ ≈ 2.18 °C). The upper 162.4 m of sediment at Hole U1499A, drilled using an advanced piston corer (APC), comprises dark greenish gray bioclast-rich clay (including foraminifers, nannofossils, radiolarians and diatoms) interbedded with thin clayey silt layers, greenish gray clay-rich calcareous ooze, and dark greenish gray nannofossil-rich clay deposited since the early Pleistocene. Sedimentation rates are stratigraphically variable: average of 9.7 cm/ka in the mid-upper Pleistocene interval (0–48.06 m), ~8.4 cm/ka in the lower Pleistocene interval (100.25–162.4 m), and undeterminable within a slumped section (48–100 m) [45]. The temperature profile demonstrates a linear increase with depth, yielding an anomalously high geothermal gradient of 93 °C/km (

Table 1. Site descriptions and model parameters in the numerical simulations of IODP Leg 349—Site U1431 [47], Site U1433 [48], IODP Leg 367—Site U1499 [46], and ODP Leg 189—Site 1172 [50].

Site		IODP U1499	IODP U1431	IODP U1433	ODP 1172 1
Site description	Latitude	18°24.5698′ N	15°22.5379′ N	12°55.1380′ N	43°57.5854′ S
	Longitude	115°51.5881′ E	117°00.0022′ E	115°02.8345′ E	149°55.6961′ E
	Water depth (m)	3760	4240	4380	2620
	Bottom water temperature (°C)	2.5 (θ ≈ 2.18 °C)	2.5 (θ ≈ 2.13 °C)	2.5 (θ ≈ 2.11 °C)	2.46 (θ ≈ 2.26 °C)
	Geothermal gradient (°C/km)	93	14.8	78	46
	Sedimentation rate (cm/ka)	8.4	5	20	3.2
Simulation parameters	Simulation depth (m)	162.4	168.9	188.3	100
	Simulation time (ka)	410	410	410	410
	Porosity (φ0, φ∞, β/cm−1)	0.851, 0.498, 2.15 × 10−4	0.794, 0.523, 1.31 × 10−4	0.770, 0.516, 1.31 × 10−4	0.699, 0.574, 2.41 × 10−4
	Upper boundary condition (x = 0)	sea level (410 ka~0)	sea level (410 ka~0)	sea level (410 ka~0)	sea level (410 ka~0)
	Lower boundary condition (g/kg)	34.89	34.39	34.25	34.96
	Initial condition (g/kg)	34.63	34.38	34.37	34.71
Simulation results	LGM salinity (g/kg)	35.68 ± 0.04	35.61 ± 0.03	35.63 ± 0.31	35.70 ± 0.14
	LGM salinity difference with modern salinity (g/kg)	1.05 ± 0.04	1.23 ± 0.03	1.26 ± 0.31	0.99 ± 0.14
	Relative difference (%) 2	3.03 ± 0.12	3.58 ± 0.09	3.67 ± 0.90	2.85 ± 0.40

¹ Site 1172 was used for comparison with other three sites in the SCS and is located next to Tasmania ² Relative difference is the ratio of the difference between the LGM and modern salinity to the modern salinity.

) [46]. IODP 349-Site U1431 (15°22.5379′ N, 117°00.0022′ E) is located near the relict spreading ridge, where the youngest crustal magnetic anomalies are observed in the eastern subbasin of the SCS (Figure 1), at a water depth of 4240 m, with a bottom water temperature of 2.5 °C (θ ≈ 2.13 °C). The upper 168.9 m of Pliocene–Pleistocene sediment, drilled by the APC, consist of dark greenish gray clay, silty clay, and greenish gray nannofossil ooze. Borehole temperature increases linearly with depth, corresponding to a geothermal gradient of 14.8 °C/km, and the sedimentation rate is ~5 cm/ka (Table 1) [47]. IODP 349-Site U1433 (12°55.1380′ N, 115°02.8345′ E) is situated in the southwest subbasin near the relict spreading center (Figure 1), with a water depth of 4380 m. The upper 188.3 m of Pleistocene sediment recovered by the APC primarily comprised dark greenish gray clay with interbedded very thin clayey silt layers. The seafloor temperature is about 2.5 °C (θ ≈ 2.11 °C), and the linear geothermal gradient reaches 78 °C/km. The sedimentation rate at this site is relatively high, reaching 20 cm/ka (Table 1) [48]. Despite the lithological variations among these three sites, the clay-rich sediments are characterized by low permeability and high porosity, which are typical for fine-grained marine deposits. Low permeability results from the small particle size and cohesive nature of clays, which is consistent with the analysis of Reece et al. [49] that high clay content corresponds to low permeability. Conversely, high porosity of ~0.8 near the seafloor decreases rapidly to ~0.5 within the upper 100–150 m due to compaction [46,47,48].

3. Methods

3.1. Sampling and Measuring Methods

Sediment pore water samples were extracted using a titanium squeezer system operated via a hydraulic press, which processed 5–10 cm long whole-round sediment core sections [51,52]. At Site U1431, sampling intervals were set at 0.5–1.5 m for the upper 30 m of the sediment column from Hole U1431A and approximately 10 m below 30 m depth from Hole U1431D, giving a total of 52 samples. At Site U1433, intervals were ~5 m above 70 m and ~10 m below 70 m, giving a total of 25 samples from only Hole U1433A. Building on pore water data acquired from IODP 349, IODP 367 implemented enhanced sampling resolution protocols. For Hole U1499A, sampling intervals were refined as follows: ~3 m above 45 m, ~5 m between 45 and 90 m, and ~10 m below 90 m, resulting in a total of 33 samples. Pore water [Cl⁻] was determined through two analytical approaches: (1) argentometric titration using a Metrohm 785 DMP automatic titrator with silver nitrate (AgNO₃) solution for Sites U1431 and U1499, and (2) ion chromatography (IC) performed on a Metrohm 850 Professional IC system for Site U1433. Analytical precision, derived from replicate analyses, yielded relative standard deviations of 0.10% (2σ, Site U1431) and 0.12% (2σ, Site U1499) for titrimetric measurements, and 0.9% for chromatographic determinations [51,52]. All analyses were calibrated against the IAPSO Standard Seawater reference material. Pore water salinity values were calculated from measured [Cl⁻] concentrations using the established stoichiometric relationship: S = [Cl⁻] × 1.80655 × 35.45/1000/ρ, where S (g/kg) represents practical salinity, and ρ (g/cm³) is seawater density [32].

3.2. Diffusion-Advection Numerical Model

To accurately reconstruct the change in deep seawater salinity within the SCS since the LGM and to generate depth-dependent salinity curves in pore water of marine sediments of Sites U1431, U1433, and U1499, we employed a one-dimensional diffusion-advection numerical model following Schrag and DePaolo [53]. The governing partial differential equation is expressed as:

$${\displaystyle \frac{\partial \phi C}{\partial t}}={\displaystyle \frac{\partial }{\partial z}}\left(\phi D_{eff}{\displaystyle \frac{\partial C}{\partial z}}+v\phi C\right),$$

(1)

where C represents the salinity of pore fluids, derived from [Cl⁻], φ denotes the porosity of seafloor sediments as dimensionless volume faction, D_eff (cm²/s) is the effective diffusion coefficient of [Cl⁻] in pore water, v (cm/ka) represents the advective velocity of downward burial, t (years) signifies time, and z (cm) represents depth (with the seawater-sediment interface located at z = 0, and positive z indicating depth in the downward direction). Cl⁻ transport in pore water is governed by diffusion (first term on the right side) and advection (second term), with the model assuming steady-state sedimentary conditions and excluding secondary diagenetic processes such as volcanic ash diagenesis or methane hydrate dissociation. Subsequent sections detail the methods for determining the porosity, the effective diffusion coefficient, and advection velocity, as well as the initial condition and boundary conditions required to solve Equation (1).

3.2.1. Porosity in Sediments

Porosity (φ) in sediments refers to the total void space per volume of the bulk sediments [54]. In upper fine-grained marine sediments, porosity diminishes exponentially with depth because of sediment compaction. Under steady-state compaction conditions (constant sedimentation rate), the empirical exponential decay of porosity with depth is expressed as [55,56]:

$$\phi \left(z\right)={\phi }_{\infty }+\left({\phi }_0-{\phi }_{\infty }\right)e^{-\beta z},$$

(2)

in which ${\phi }_{\mathrm{\infty }}$ denotes the seemingly constant porosity achieved at a certain depth, ${\phi }_0$ represents the porosity at the seawater-sediment interface, and β is an empirical attenuation constant, with units of cm⁻¹.

3.2.2. Effective Diffusion Coefficient

The effective diffusion coefficient (D_eff) represents the diffusion coefficient of Cl⁻ in pore water of marine sediments. Its magnitude critically influences the morphology of simulated profiles, particularly affecting both the amplitude and depth of concentration peaks. Consequently, precise determination of D_eff is essential for ensuring simulation accuracy [57]. The effective diffusion coefficient is mathematically expressed as D_eff = D/θ², where θ denotes the tortuosity parameter of porosity. For fine-grained sediment lithologies, the tortuosity satisfies the empirical relationship ${\theta }^2=1-2\times \mathrm{l}\mathrm{n}\left(\phi \right)$ [58]. The parameter D denotes the molecular diffusion coefficient of seawater Cl⁻, calculated using the following formula [59]:

$${\displaystyle \frac{D}{D_0}}={\displaystyle \frac{{\mu }_0}{\mu }},$$

(3)

where D₀ is the molecular diffusion coefficient of Cl⁻ in pure water, expressed as a function of temperature (T, °C): D₀ = (9.60 + 0.438 × T) × 10⁻⁶ [60]. The variables μ and μ₀ represent the viscosity of seawater and pure water, respectively. The seawater viscosity μ is determined by an empirical equation developed by Mattaus [60], which is a function of temperature, salinity (S, g/kg), and pressure (P, bar). When the salinity is set to 0 in this empirical equation, the resulting viscosity corresponds to that of pure water. Pore water temperature can be calculated from the seafloor temperature and the linear geothermal gradient. The current average salinity of deep seawater in the SCS is ~34.62 g/kg [61,62]. Pressure can be determined based on the seafloor depth and simulated depth.

3.2.3. Advection Velocity

Due to compaction from overlying sediments, some physical properties of marine sediments undergo rapid changes within specific depth intervals. For instance, sediment porosity decreases sharply within the upper 100–200 m before plateauing to a stable state [46,47,48]. Under steady-state compaction and the absence of external overpressure, the vertical advection velocity of pore water is controlled by the downward burial of sediment [60]:

$$v\left(z\right)={\displaystyle \frac{{\omega }_{\infty }{\phi }_{\infty }}{\phi \left(z\right)}},$$

(4)

where ${\omega }_{\mathrm{\infty }}$ (cm/ka) represents the sediment burial rate (equivalent to the sedimentation rate) at a depth where the porosity remains constant, and hence $v_{\mathrm{\infty }}={\omega }_{\mathrm{\infty }}$.

3.2.4. Initial Condition and Boundary Conditions

If φ, D_eff, and v are known, Equation (1) is a parabolic partial differential equation in z and t. To obtain the numerical solution of this equation, we must specify (a) the initial condition, (b) the upper boundary condition at the seawater-sediment interface, and (c) the lower boundary condition at the deepest simulated depth. The initial condition establishes the simulation’s starting time, selected under the principle that it should not affect the magnitude of the peak value of the pore water salinity profile in the simulation. In this simulation, 410 ka BP is chosen, corresponding to Marine Isotope Stage 11 (MIS 11) [63]. Because the sea level at 410 ka BP is similar to the current sea level, the salinity of the current bottom seawater in the SCS is taken as the initial condition (Table 1). The upper boundary condition is defined as a time-dependent salinity curve at the seawater-sediment interface, derived by scaling reconstructed sea level variations since 410 ka BP using the empirically established ratio between sea level change and salinity variation since the LGM [4]. Given a sea level rise of ~130 m in the SCS since the LGM, the model iteratively adjusts trial LGM deep water salinity values to generate a suite of salinity-depth profiles. Optimal matching of simulated curves and measured data is achieved through minimizing the sum of squared difference between measured and simulated data (Chi-square, χ²). The lower boundary condition is determined based on the trend in pore water salinity at the lower simulated sediment intervals, typically assigned as the mean of the last few measured data points or the last pore water salinity value. Equation (1) is solved numerically via the finite difference method by MATLAB R2020a’s “pdede” solver, employing a depth resolution of 50 cm. The solver automatically adapts the internal time step to meet convergence criteria, and the temporal output is provided at intervals of 0.5 ka. This discretization ensures computational stability while resolving millennial-scale hydrodynamic processes.

4. Results

4.1. Simulation Results of Porosity

The compaction of sediments is reflected in the depth-dependent variations in physical properties within the uppermost 100 m at Site U1499, and 150 m at Sites U1431 and U1433. Porosity shows an obvious decrease with depth in this uppermost depth range, from 80–90% to 50%, and remains relatively constant in the deeper sediments [46,47,48]. To maintain consistency, the uppermost 200 m of sediments is selected as the simulation depth of porosity analysis for these three sites. The porosity equation is modeled via MATLAB by lsqnonlin function, yielding ${\phi }_0$ = 0.851 ± 0.031, ${\phi }_{\mathrm{\infty }}$ = 0.498 ± 0.016, $\beta$ = (2.15 ± 0.43) × 10⁻⁴ cm⁻¹, and the Chi-square (χ²) is 0.0525 at Site U1499; ${\phi }_0$ = 0.794 ± 0.033, ${\phi }_{\mathrm{\infty }}$ = 0.523 ± 0.046, $\beta$ = (1.31 ± 0.59) × 10⁻⁴ cm⁻¹, and χ² = 0.304 at Site U1431; and ${\phi }_0$ = 0.770 ± 0.031, ${\phi }_{\mathrm{\infty }}$ = 0.516 ± 0.034, $\beta$ = (1.31 ± 0.53) × 10⁻⁴ cm⁻¹, and χ² = 0.191 at Site U1433 (Figure 2).

4.2. Simulation Results of Salinity

The simulated depth intervals for salinity profiles are selected based on sediment cores recovered by the APC, with simulated depths of 162.4 m (U1499), 168.9 m (U1431), and 188.3 m (U1433). Because they are unaffected by secondary processes such as volcanic ash alteration or methane hydrate formation/dissociation, these three sites exhibit “standard” pore water salinity-depth profiles within the modeled intervals. “Standard” salinity profiles display a characteristic pattern: salinity increases gradually from lower values at the sediment-water interface, reaching a distinct peak within the upper 20–50 m of the sediment column. This peak corresponds to the LGM residual signal, with variations in peak magnitude and depth-specific positioning (ranging between sites). The LGM salinity output from the diffusion-advection model is largely controlled by this peak value. Below the peak salinity values, a decreasing trend characterized by different slopes in salinity is observed.

This study utilizes the average pore water salinity from the upper 3 m of sediment column, representing modern bottom water salinity, i.e., initial condition (Table 1). The upper boundary condition is defined by salinity evolution curves reconstructed from sea level variations since 410 ka. For the lower boundary condition, the terminal measured salinity value is generally selected, except at Site U1433 where the last salinity datum deviates from the trend observed at the base of the sediment column. Consequently, the penultimate measured salinity value is adopted as the lower boundary condition for this site (Table 1). Integrating porosity equations, effective diffusion coefficients and advection velocities into the one-dimensional diffusion-advection numerical model, we assume a suite of LGM salinity with 0.05 g/kg increment firstly, generating a suite of simulated salinity-depth profiles and calculating the sum of the squared difference between modeled values and measured data (χ²). Then, refining the simulation increment of 0.01 g/kg within the minimal χ² neighborhood, we obtain the salinity corresponding to the χ² minimum as the LGM salinity value. Modeled uncertainty is based on the difference in the average of differences between the measured [Cl⁻] ± 1 standard deviation and the modeled [Cl⁻]. Finally, we obtained reconstructed LGM deep water salinity of 35.68 ± 0.04 g/kg, 35.61 ± 0.03 g/kg, and 35.63 ± 0.31 g/kg at U1499, U1431, and U1433, respectively (Figure 3). The substantial uncertainty of Site U1433 is primarily attributable to significant measurement errors in [Cl⁻] analysis by ion chromatography. These LGM salinity values exceed modern deep water salinities by 1.05 ± 0.04 g/kg, 1.23 ± 0.03 g/kg, and 1.26 ± 0.31 g/kg, corresponding to relative increases of 3.03 ± 0.12%, 3.58 ± 0.09%, and 3.67 ± 0.90%, respectively.

5. Discussion

5.1. LGM Deep Water Salinity of the SCS

The deep SCS circulation constitutes a basin-scale cyclonic gyre, characterized by a counterclockwise flow pattern (Figure 1). Modern low-temperature and high-salinity Pacific deep water intrudes into the SCS through the Luzon Strait and moves southwestward along bathymetric contours near the northern continental slope. Influenced by complex submarine topography, this deep current turns southward within the northwest subbasin, flowing along the western boundary of central basin in the SCS. Near the Zhongsha Island, the southward-flowing current bifurcates into two branches. The southward branch continues flowing along the western boundary of the basin, forming a weak subbasin-scale cyclonic circulation within the southwest subbasin, and the eastward branch, namely the primary current, develops a subbasin-scale cyclonic circulation around seamounts in the central SCS modulated by internal waves, mesoscale eddies, and complex topography. Ultimately, these two branches converge at the basin’s southern periphery, forming a very weak northward-flowing current along the eastern boundary of the SCS central basin [64,65,66,67]. Consequently, the salinity of deep SCS decreases progressively along the deep circulation pathway. Deep water salinity at Site U1499 in the northern SCS is ~34.63 g/kg. In contrast, the central SCS, characterized by numerous seamounts and the resultant complex topography, undergoes enhanced vertical mixing of the upper and lower water layers because of topographic upwelling of bottom currents along seamount flanks, resulting in the lowest regional salinity in the deep SCS [61,64]. For example, Sites U1431 and U1433 near seamounts in the central SCS, the deep water salinity are ~34.38 g/kg and 34.37 g/kg, respectively. Consequently, the maximum salinity gradient reaches ~0.26 g/kg from the northern SCS to the central SCS.

During the LGM, the spatial distribution of deep SCS salinity remained analogous to modern patterns, but with an attenuated salinity gradient. Simulated results indicate a maximum salinity gradient of ~0.07 g/kg from the northern SCS to the central SCS. These confirm that the LGM salinity decreased from northern SCS to the central SCS along the modern circulation pathway, indicating the deep water circulation during the LGM is consistent with the present. However, it is noteworthy that the salinity gradient during the LGM was substantially lower than that observed today. Assuming LGM deep water temperature of the SCS more homogenous than modern as the deep open ocean [29,68], then the deep water circulation of SCS was much controlled by salinity. Consequently, the reduced LGM salinity gradient within deep SCS suggests that a much sluggish circulation during the LGM.

5.2. Deep Water Exchange Between the Pacific and SCS During the LGM

Current physical oceanographic studies have established that the deep seawater in the SCS originates from the Southern Ocean. The Southern Ocean deep water (consisting of the UCDW and Lower Circumpolar Deep Water (LCDW)) flows northward along the western boundary of the Pacific Ocean and enters the SCS via the Luzon Strait [69,70,71,72]. The boundary between the LCDW and UCDW in the Pacific is defined at a potential temperature of θ = 1.5 °C, with this boundary situated at an approximate depth of 3000 m [72]. As the maximum sill depth of the Luzon Strait is around 2600 m, the LCDW is precluded from entering the SCS, leaving UCDW as the predominantly constituent of the SCS deep water [38,65]. The Luzon Strait serves as the sole deep-sea conduit connecting the SCS and the western Pacific. Modern hydrographic data show deep water from the western Pacific flows through the Luzon Strait into the SCS at depths exceeding 2000 m [41], with SCS waters at depths greater than 2000 m exhibiting hydrographic properties analogous to Pacific waters at depths ranging from 2000 to 2600 m [42,64,65]. On the eastern flank of the Luzon Strait within the Philippine Sea, deep water salinity averages ~34.65 g/kg at depths 2000–2600 m [73], whereas the interior SCS basin below 2000 m maintains a mean salinity of 34.62 g/kg [64], yielding the modern salinity gradient across the Luzon Strait of ~0.03 g/kg.

During the LGM, the deep water circulation pathway in the Pacific Ocean remained consistent with modern configuration [70,74]. UCDW flowed northward along the western boundary of the Pacific, entering the SCS via the Luzon Strait [71]. Consequently, comparing the salinity differences between the UCDW in the western Pacific and the deep SCS during the LGM, we can elucidate the characteristics of deep water exchange between the western Pacific and SCS. However, up to now, no absolute paleo-salinities of UCDW during LGM are available. This is because low-temperature alteration of young basaltic crust and volcanic detritus within sediments, altering the original pore water [Cl⁻] [75], resulted in no proper data sets for reconstructing LGM salinities of UCDW in western Pacific by one-dimensional diffusion-advection model yet.

Given these constraints, Site 1172 (43°57.5854′ S, 149°55.6961′ E) in the Southern Ocean was selected as an optimal alternative to reconstruct the salinity of the UCDW. Site 1172 is located at a water depth of ~2620 m within UCDW range and represents the upstream source region of the UCDW mass which flows through western Pacific and enters the SCS. The upper 100 m at Site 1172, which primarily consists of white foraminiferal nannofossil ooze and white and light greenish gray nannofossil ooze, is devoid of volcanic ash, ensuring a pristine pore water [Cl⁻] profile unaffected by early diagenetic alteration. The bottom water temperature is 2.46 °C (θ ≈ 2.26 °C), with a linear geothermal gradient of ~46 °C/km and a sedimentation rate of 2–3.2 cm/ka (Table 1) [50]. The simulation initial time is set at 410 ka BP. Due to the absence of pore water [Cl⁻] data in the upper 3 m of the sediment column, the first measured value of 34.71 g/kg is selected as the initial condition, which also falls within the modern average UCDW salinity range of 34.71–34.73 g/kg [76]. The upper boundary condition is defined by a salinity evolution curve reconstructed from sea level change records since 410 ka BP, accounting for a sea level rise of ~130 m since the LGM [4]. The lower boundary condition uses the deepest measured pore water [Cl⁻] value of 34.96 g/kg within the simulated interval. And the porosity equation for the upper 220 m of the sediment column at Site 1172 is modeled as: $\phi \left(z\right)=0.574+(0.699-0.574)e^{-0.000241z}$. Utilizing this equation and aforementioned conditions, the simulated LGM deep water salinity at Site 1172 is determined to be 35.70 ± 0.14 g/kg (Figure 4).

The simulated salinity result of Site 1172 is close to the LGM deep water salinity at Site U1499 in the SCS, with a smaller salinity gradient of 0.02 g/kg from UCDW to the deep SCS. Combined with more homogeneous deep water temperature during the LGM, salinity dominated potential density variation [29,68]. Consequently, the reduced salinity gradient across the Luzon Strait during the LGM might lead to a weakened potential density gradient, possibly decreasing the overflow flux of western Pacific deep water into the SCS. It is consistent with the interpretation inferred from benthic foraminifera B/Ca ratios and δ¹³C data [77], radiocarbon activity [78], and δ¹⁸O data [79]. The reduced gradients of deep water [CO₃²⁻] (reconstructed from benthic foraminifera B/Ca ratios) and δ¹³C between the Pacific and SCS indicate weakened deep water ventilation and more sluggish Pacific deep water circulation during the glacial relative to the interglacial [77]. Overall, the representative evidence of multiple proxies substantiates that the LGM deep water circulation between the western Pacific and SCS was characterized by sluggish circulation and weakened vertical mixing in deep water, as the evidence of salinity, consistent with attenuated thermohaline forcing under glacial boundary conditions.

5.3. LGM Salinity of the Deep Pacific

At present, salinity of the deep Pacific exhibits a relatively homogeneous distribution. Because of the absence of deep water formation within the Pacific, the majority of deep water originates from the Southern Ocean [71,80,81]. Consequently, the deep Pacific salinity is comparable to that of the Southern Ocean, with an average salinity of ~34.70 g/kg [31,33]. Horizontally, Pacific deep water salinity exhibits minor diminution, with a latitudinal gradient of ~0.02 g/kg observed from the South Pacific to the North Pacific at equivalent depths. Vertically, the salinity of Pacific deep water measures ~34.62 g/kg at a depth of 2000 m and increases with depth gradually, culminating in a salinity of approximately 34.70 g/kg at around 4000 m. The maximum vertical salinity gradient between UCDW and LCDW is estimated to be about 0.08 g/kg [31].

Studies of benthic foraminiferal δ¹³C, δ¹⁸O, and Nd isotopic records [70,74,77,79] demonstrate that the deep water circulation pathways of the Pacific during the LGM remained broadly consistent with the contemporary conditions and exhibited more homogeneous physical and chemical properties of LGM Pacific deep seawater. LGM deep water salinity in the Southwest Pacific, derived from Sites ODP 1123, IODP U1365, and U1370, averaged 36.12 g/kg (

Table 2. Basic information and reconstructed results of sites in the Southern Ocean, Pacific Ocean and SCS, and the relative difference in LGM deep water salinity compared with modern salinity.

Site	Latitude	Longitude	Water Depth (m)	Present Salinity (g/kg)	LGM Salinity (g/kg)	Relative Difference (%)
IODP U1499	18°24.57′ N	115°51.59′ E	3760	34.63	35.68 ± 0.04	3.03 ± 0.12
IODP U1431	15°22.54′ N	117°00.00′ E	4240	34.38	35.61 ± 0.03	3.58 ± 0.09
IODP U1433	12°55.14′ N	115°02.83′ E	4380	34.37	35.63 ± 0.31	3.67 ± 0.90
ODP 1172	43°57.59′ S	149°55.70′ E	2620	34.71	35.70 ± 0.14	2.85 ± 0.40
IODP U1370 [33]	41°51.12′ S	153°6.36′ W	5074	34.71	35.91 ± 0.09	3.54 ± 0.3
ODP 1123 [29]	41°47.16′ S	171°29.94′ W	3290	34.73	36.19 ± 0.07	4.2 ± 0.2
IODP U1365 [33]	23°51.06′ S	165°38.64′ W	5695	34.70	36.25 ± 0.09	4.53 ± 0.3
ODP 1225 [33]	2°46.26′ N	110°34.26′ W	3760	34.69	36.07 ± 0.09	4.01 ± 0.3
EQP 10 [33]	20°40.98′ N	143°21.42′ W	5412	34.70	36.21 ± 0.18	4.41 ± 0.5
EQP 11 [33]	30°21.30′ N	157°52.26′ W	5813	34.69	36.06 ± 0.18	3.96 ± 0.5

) [29,33], whereas the Northeast Pacific deep water, sampled at Sites ODP 1225, EQP10, and EQP11, yielded a comparable mean salinity of 36.11 g/kg (Table 2) [33]. The LGM salinity gradient of deep Pacific from the Southwest Pacific to the Northeast Pacific is determined to be 0.01 g/kg, smaller than modern latitudinal gradient. These results indicate uniform spatial variation in LGM Pacific deep water salinity along the deep water circulation pathway, further corroborating that the glacial Pacific deep water circulation followed modern-like trajectories and maintained more homogeneous thermohaline properties. The salinity uniformity provides an angle for understanding deep circulation in the Pacific basin across glacial-interglacial cycles.

Notably, these simulation sites referenced above for reconstructing Pacific deep water salinity are situated within the LCDW domain at depths exceeding 3000 m, reconstructing the LGM salinity 35.91–36.25 g/kg for LCDW [33]. In contrast, our simulation results from Sites U1499 and 1172, representing the UCDW salinity characteristics (35.68–35.70 g/kg), reveal a pronounced LGM salinity gradient of 0.21–0.57 g/kg between LCDW and UCDW. This LGM salinity gradient substantially exceeds the modern observed gradient of ~0.08 g/kg [31], indicating significantly enhanced vertical stratification of Pacific deep waters during the LGM. The amplified salinity gradient effectively suppresses vertical mixing of deep water masses and contributes to a more sluggish Pacific deep water circulation. These findings mutually corroborate the “reduced glacial deep water ventilation” conclusion proposed by Wan et al. [77], which was based on the reduced gradients of deep water [CO₃²⁻] and δ¹³C between Pacific and SCS during glacial periods. Collectively, these evidences demonstrate the attenuation of Pacific deep water circulation during the LGM.

6. Conclusions

To fill a knowledge gap regarding LGM salinity data of deep water in the SCS, this study thoroughly elucidates a one-dimensional diffusion-advection model for [Cl⁻] transport in pore water, detailing the calculation methods and applicable conditions for the relevant parameters. By incorporating measured [Cl⁻] data from IODP Sites U1499, U1431, and U1433 in the SCS, we simulated the salinity variations in deep seawater in the SCS during the LGM and compared these findings with the simulation results from the Southern Ocean and Pacific. The main conclusions are as follows:

(1)

The LGM deep water salinity in the northern SCS was 35.68 ± 0.04 g/kg, while in the central SCS, it was 35.61 ± 0.03 g/kg, leading to an intra-basin salinity gradient of ~0.07 g/kg. This gradient value is substantially smaller than the modern gradient of ~0.26 g/kg, demonstrating significantly weakened circulation in the deep SCS during the glacial periods.

(2)

During the LGM, the salinity gradient between UCDW—sourced from the Southern Ocean—and the deep northern SCS was marginally reduced compared to the modern gradient, indicating diminished overflow flux of UCDW into the SCS through the Luzon Strait.

(3)

The LGM salinity gradient between LCDW and UCDW reached 0.21–0.57 g/kg, significantly exceeding the modern gradient of ~0.08 g/kg. Combined with evidence of reduced glacial deep water ventilation in the western Pacific, this enhanced LGM vertical salinity gradient also demonstrates intensified deep ocean stratification and sluggish deep water circulation in the western Pacific during the glacial periods.