Core MABC–11 (155.53° E, 15.22° N, 5,840 m water depth) was retrieved from the eastern base of the Caiwei Guyot ( Figure 1 ) using a box corer aboard R/V Haiyang Liu Hao in July 2012, with a core length of 59 cm.

The age model of core MABC–11 was developed using magnetostratigraphy and by tuning changes in element Ca intensity (from XRF scanning) with the deep-sea benthic δ¹⁸O stack LR04 (Lisiecki and Raymo, 2005), which reflects global ice volume changes. The resulting average sedimentation rate is 0.73 mm/kyr (Yi et al., 2021a). To further refine the chronology, this age-depth model was adjusted based on ¹⁰Be/⁹Be data to integrate deep-sea paleoenvironmental records from the Mariana Trench and Magellan Seamounts (Yi, 2023). For this study, the core was sampled for the depth interval of 7.0–49.5 cm at 5 mm resolution, and 86 subsamples in total were obtained for magnetic and grain-size analyses between 221 and 904 ka ( Figure 2 ).

Hysteresis loops were conducted on all 86 samples using a Princeton Measurements Inc. MicroMag 3900 Vibrating Sample Magnetometer (VSM). A peak field of 0.3 T was set for hysteresis loops, and saturation magnetization (Ms), saturation remanence (Mrs), coercive force (Bc), and the coercivity of the remanence (Bcr) were determined from the hysteresis loops, after calibration using the data from between 0.25 and 0.30 T. All magnetic measurements were conducted at the Paleomagnetism and Geochronology Lab (PGL), Institute of Geology and Geophysics, Chinese Academy of Sciences.

The mathematical unmixing of hysteresis loops can provide detailed information about different coercivity spectra (Jackson et al., 1990). The polymodal distribution for unmixing in this study is expressed as follows:

where f_i represents the function for component i where i = 1 to n components, and p_i is the percentage contribution of the components. A series of target functions has been proposed for unmixing (Heslop, 2015); here, we used the normal function to identify the potential end members of magnetic minerals in the sediments (Heslop, 2015; Heslop and Roberts., 2012). The normal function has the following form:

Here, x is the independent variable and represents the magnetic field, and the dependent variables are the second derivatives of the hysteresis loop data, which were first standardized to the interval of [0, 4]. The coefficient p represents the relative ratio between three components (namely Cnt1–3), α determines the distribution shape (namely Shp1–3), and β controls the position of the central tendency of the curve, herein, the magnetic coercivity (namely Coe1–3).

Grain-size samples were placed in an ultrasonic vibrator with sodium hexametaphosphate [(NaPO₃)₆] for several minutes to facilitate dispersion and were measured using a Malvern Mastersizer 2000 grain size analyzer in the Key Laboratory of Engineering Oceanography, Second Institute of Oceanography, Ministry of Natural Resources of China.

Fifty grain-size classes between 0.1 and 2000 μm were exported for further analysis. The grain-size distributions were then analyzed by using mathematical methods, including the varimax-rotated principal component analysis (VPCA), environmentally sensitive components, and lognormal-based unmixing (modified from Equation 2), and the common signal of deep-sea dynamics was extracted by a PCA on the studies cores for paleoenvironmental inferences, following the procedures reported in previous studies (e.g., Chen et al., 2021; Paterson and Heslop, 2015; Yi et al., 2022).

The median grain-size value (M) of the sediment is 3.3 ± 0.2 μm, indicating a low-dynamic sedimentary environment that remained relatively stable throughout the middle Pleistocene ( Figure 3B ). The proportions of clay (< 4 μm) and silt (4~63 μm) particles display minimal variation, with average values of 52.8 ± 1.8%, and 38.2 ± 1.6%, respectively, while sand particles (> 63 μm) exhibit greater variability, averaging 9.0 ± 2.6%. Notably, coarse components, likely authigenic micro-nodules (>200 μm), are present in several samples, consistent with slow sediment accumulation in marine environments (Wang et al., 2016a), and are observed in surrounding regions (Yi et al., 2022, 2020).

The grain-size distributions are multi-modal, with modal sizes of approximately 3 μm (dominant), ~40 μm, and 400-500 μm ( Figure 3 ). Only minor differences were found in the grain-size distributions across different samples, suggesting a stable sedimentary environment during the examined interval. Following the method of Boulay et al. (2003), which has proven effective for identifying sedimentary processes and dynamics (e.g., Hu et al., 2021; Sun et al., 2003), we identified three environmentally sensitive grain-size components (CC1-CC3), with modal sizes of 2.6-4.8 μm, 42.2-51.5 μm, and 676-1000 μm, respectively ( Figure 3C ).

Polymodal grain-size spectra can be mathematically partitioned (Ashley, 1978), enabling the separation of orthogonal modes (independent grain-size components/factors) to identify potential changes in input functions and/or sedimentary dynamics (e.g., Chen et al., 2020, 2021; Yi et al., 2012b). Using a three-component lognormal function following the method of Paterson and Heslop (2015), which is similar to Equation 2, we obtained three components, EM1, EM2, and EM3 ( Figure 3C ), with modal sizes of 3.2 μm, 34.6 μm, and 455 μm, respectively. VPCA can also be used to identify the processes controlling sediment grain-size changes and to extract paleoenvironmental signals (e.g., Hu et al., 2021; Yi et al., 2012a). Similarly, the results of VPCA also identify three characteristic components, VF1-VF3 ( Figure 3D ), accounting for 78.9% in total ( Table 1 ).

Combining all the grain size results, including the environmentally sensitive components (CC1-CC3), lognormal-based unmixing (EM1-EM3), and VPCA results (VF1-VF3), there are three grain-size components similarly identified with a major group (modal sizes at ~3-4 μm), suggesting a single dominant factor controlling sedimentary dynamics in the study area during the depositional interval.

Previous studies on magnetic minerals in the Caiwei Guyot sediments, including hysteresis loop, IRM acquisition, and first-order reversal curve analyses, demonstrate that low-coercivity magnetite is the dominant magnetic mineral, with fine grains (Lin et al., 2019; Yi et al., 2021b). Similar mineral properties have been documented at IODP Site U1337 in the eastern Pacific (Yamazaki, 2012), and core XTGC1311 from the middle Pacific (Li et al., 2020).

Hysteresis loop analysis of all 86 samples reveals that the magnetic loops are closed below 200 mT ( Figure 4A ), indicating a dominance of low-coercivity magnetic minerals. The coercivity values (Bc and Bcr) average 11.4 ± 0.3 mT and 31.5 ± (< 0.05) mT, respectively. On the Day plot (Day et al., 1977), the samples plot within the PSD range, close to the SD field ( Figure 4B ). Despite minimal variation on the Day plot, a higher Bcr/Bc and Mrs/Ms ratio prior to ~500 ka suggests slightly coarser magnetic grains (Roberts et al., 2018). A three-component lognormal function was applied to mathematically unmix the hysteresis loops (Heslop, 2015; Heslop and Roberts., 2012), yielding coercivity components (Coe1-Coe3) of 6.1 ± 0.5 mT, 25.7 ± 1.0 mT, and 65.2 ± 2.1 mT, respectively ( Figure 5 ). A similar analysis was also applied to unmix the IRM acquisition curves (Maxbauer et al., 2016), and a comparison between hysteresis loop-based and IRM acquisition-based results shows no significant differences (Yi et al., 2021b), and thus is not plotted here.

In a log-normal based unmixing (Equation 2), we employed three parameters to describe a subpopulation of magnetic coercivity, α, β, and p, in which α determines the shape of the distribution (Shp1–Shp3), representing each magnetic type/source/mineral, β controls the position of the central tendency of the curve (Coe1–Coe3), herein representing magnetic coercivity or degree of early diagenesis for each magnetic group, and p indicates the relative percentage of each component in a population (Cnt1–Cnt3), representing the ratio of each magnetic group against the total.

Since the source of magnetic minerals in the western Pacific was relatively stable (Chen et al., 2023), the coercivity spectrum can be expressed by these three parameters, and plotting these three parameters together can provide useful information to assess how magnetic minerals changed after deposited ( Figure 5 ). As shown, the major component is Coe2, accounting for about 47%–50% of total magnetic grains in the sediment. Moreover, there is no distinct difference between the pattern of α-β for components Coe2 and Coe3, indicating that these two magnetic groups were linked to a similar source and/or experience similar post-deposition changes (early diagenesis). However, for component Coe1, a more complex relationship between the three magnetic parameters is observed ( Figure 5A ). Considering all of these observations, together with the close relationship between the derived magnetic components ( Figure 5D ), component Coe2 was employed for later analysis.

There are several proxies for studying deep-sea ventilation, including the enrichment of oxygen-sensitive metals (such as Mn, Zn, Ni, and V), sediment grain size, and magnetic coercivity. For instance, element Mn migrates from reducing to oxidizing environments, making it highly sensitive to sedimentary redox changes (Costa et al., 2018; Löwemark et al., 2014; Slemons et al., 2012; Yi et al., 2023). Conversely, sediment grain size reflects the intensity of bottom-water flows, where coarser particles suggest higher sedimentary dynamics and intensified bottom waters, as reported in previous studies (e.g., Hall et al., 2001; Yi et al., 2022). Magnetic properties of deep-sea sediments also offer potential insights into deep-water redox conditions (e.g., Chang et al., 2016; Kissel et al., 2020; Kruiver and Passier, 2001) due to conversions between Fe²⁺ and Fe³⁺ in crystal lattices in the context of oxygen-rich bottom water. However, the exact influence of early diagenesis expressed by changes in magnetic mineral properties may vary between sites because investigations with different oxidation states show that PSD magnetite could be significantly negatively correlated to partial oxidation in a shell-only model, or positively correlated in cases of changing magnetite sizes (Ge et al., 2014; Chen et al., 2023). Hence, considering that the Mn record of core MABC–11 has been cross-validated in previous studies (Yi et al., 2021a; Yi, 2023), we take the Mn record as a reference and compare all of the three proxies to reveal the potential linkages between them ( Figures 6 – 8 ).

The Mn-enriched sedimentary record from core MABC–11 shows a significant increase in Mn concentration, concurrent with a decrease in sediment median grain size ( Figure 6 ). This observation is contrary to the typical linkage between sediment grain size and bottom-water intensity reported in previous studies (e.g., Hall et al., 2001; Lamy et al., 2024), suggesting that other processes beyond bottom-water intensity are involved.

Aeolian inputs, which are increasingly recognized as important contributors to deep-sea sedimentation in the western Pacific, may have played a major role, particularly since the middle Pleistocene (Yi et al., 2020, 2022; Yao et al., 2021). The grain-size variation observed in core MABC–11 is generally consistent with records of drying processes in the Asian interior ( Figure 7A ), inferred from the δ¹³C record of the Taklimakan Desert (Liu et al., 2020), and eolian transport to the Japan Sea ( Figure 7B ), implied from the K content of IODP Site U1422 (Zhang et al., 2018). Fine grains in deep-sea sediments in the North Pacific were mainly carried by the westerlies and/or winter monsoons from the Asian interior (e.g., Rea, 1994; Jiang et al., 2019; Xu et al., 2015), inferring that aridification in inner Asia would result in more eolian particles in deep-sea sediment and a decrease in sediment grain size.

Moreover, seamounts, such as Caiwei Guyot, exert unique influences on regional circulation, vertical mixing, and sediment transport (Bograd et al., 1997; Chen et al., 2015; Yang et al., 2017; Zhang and Boyer, 1993, 1991). For example, a series of complex responses, such as the anticyclonic cap (Lavelle and Mohn, 2010), are generated to modulate regional circulation when currents flow across a seamount (Perfect et al., 2018; Robertson et al., 2017). By studying sediment grain-size properties in the central Philippine Sea, the agreement between deep-sea sedimentary dynamics and ENSO-like changes was highlighted in the Quaternary, suggesting the long-term influence of upwelling and unique submarine topography (Yi et al., 2022). Similarly, in a 3-year monitoring study, a deep anticyclonic cap over the studied guyot was proposed (Guo et al., 2020), and this topography-induced downwelling could have imparted evident precessional signals into the sedimentary Mn of core MABC–11 (Yi et al., 2021a). Based on this vertical connection in the study area, the downwelling processes could result in more eolian particles being deposited into deep-sea sediments, agreeing with the relationships between the Mn and grain-size records observed in core MABC–11 ( Figure 6 ) and between the zonal SST difference (ENSO-like changes) and MABC–11 grain-size record ( Figure 7C ). Therefore, it is inferred that eolian inputs may be the dominant factor controlling the sedimentary dynamics in the Caiwei Guyot during the middle Pleistocene.

For magnetic proxies, it is observed that as the Mn contents increase in core MABC–11, the coercivity values of the magnetic components decrease ( Figure 8 ). This inverse relationship between magnetic coercivity and Mn record suggests that redox conditions influenced the preservation and alternations of magnetic minerals in deep-sea sediments.

The relationship between magnetic coercivity and deep-sea redox conditions has been demonstrated using surficial sediments in the Philippine Sea and its surrounding area, and the results show that for PSD magnetite, in higher deep-water oxidation conditions, early diagenesis could result in a lower coercivity of the sediments, and vice versa (Chen et al., 2023). In such a case, prolonged exposure of magnetic minerals to oxygen-rich bottom waters leads to the maghemitization of magnetite grains, which reduces their coercivity. This observation is consistent with experimental studies, in which the partial oxidation of PSD magnetite grains in a core-shell model can decrease the effective diameter of magnetite grains, resulting in a lower coercivity value (Ge et al., 2014; Özdemir and Dunlop, 2010). However, offsets between the Coe2 and Mn_f1 records are also evident ( Figure 8 ), which can be attributed to that the relationship between magnetic coercivity and redox conditions may be influenced by other factors, such as the magnetic grain size, mineralogy, and concentration (Chen et al., 2023). Hence, the observed relationship between magnetic coercivity and sedimentary Mn likely reflects changes in bottom-water oxygenation in the study area and can serve as a proxy for deep-sea ventilation ( Figure 9 ).

Integrated evidence suggests that significant changes in regional ventilation occurred in the Magellan Seamounts during the middle Pleistocene ( Figure 9 ). To exclude potential dominant influences from marine productivity on abyssal redox conditions, a comparison was conducted between the sedimentary Mn of core MABC–11 and the planktonic δ¹³C record from ODP Site 806 (Schmidt et al., 1993). No significant correlation was observed between these records (Yi, 2023), suggesting that ventilation changes were the primary driver of redox conditions rather than changes in productivity in the study area. These proxies reveal a weak but observable in-phase relationship between abyssal ventilation and the LR04 record (Lisiecki and Raymo, 2005), likely indicating intensified deep-sea ventilation during interglacial intervals. This finding aligns with other records from the North Pacific (Jacobel et al., 2017), and further highlights the complex relationship between global glacial-interglacial cycles and deep-sea circulation.

Moreover, glacial intensification of abyssal ventilation is clearly evident during MIS 12, consistent with similar findings in the eastern Pacific (Yi et al., 2023) and the southwestern Pacific (Hall et al., 2001). The diversity of abyssal ventilation in the study area between glacial and interglacial alternations may be attributed to the redistribution of bottom/deep water masses within the deep Pacific (Yi, 2023), which is worthy of further investigation in the future.

Furthermore, a long-term trend of decreasing abyssal ventilation since ~430 ka ( Figure 8 ) coincides with the Mid-Brunhes Event (MBE), a period characterized by enlarged amplitudes in the glacial-interglacial cycles ( Figure 9B ). Whether the MBE represents multiple equilibria in the climate system (Jansen et al., 1986; Paillard, 1998), or a transition between two distinct climate states singly responding to astronomical forcing (Tzedakis et al., 2017; Yin, 2013) remains debated. A latitudinal shift of the Southern Hemisphere westerlies inducing CO₂ to respire from the Southern Ocean (Kemp et al., 2010) and/or a slowdown of AABW formation (Yin, 2013) have been proposed to be the potential mechanisms for the MBE. In our study, the reduced ventilation observed in this study is consistent with the predicted slowdown in AABW formation after the MBE, corroborating a recent reconstruction of AABW variability in the eastern Pacific (Yi et al., 2023).