Glacial diamictons are broadly found in the sediments across the Ross Sea (Domack et al., 1999), mainly containing terrigenous fine silt and clay, biogenic siliceous materials, and IRDs (Anderson et al., 1984). The sand component (grain-size > 63 μm) is commonly used as an indicator of IRD (Reimnitz et al., 1998; Wang et al., 2009), for which the seasonal sea ice carries relatively finer particles (<250 μm) and large ice chunks and icebergs transport coarser ones (>250 μm) (Darby and Paula, 2008; Phillips and Grantz, 2010). Combining the IRD data and lithological characteristics, the core ANT32-RA05C can be divided into three sections (Figure 2) and described as follows.

1) Section-1, above 16 cm in depth. This section is the coarsest layer, rich in sands and gravels. The average IRD (>63 μm and >250 μm) contents are about 44 and 32%, respectively. The maximum diameter of the gravels is ∼8 cm found at 3–8 cm in depth. 2) Section-2, 16–52 cm in depth. The sediments are brownish clay and silt, which are relatively fine, with only 10% IRDs (>63 μm). 3) Section-3, below 52 cm in depth. This section contains olive and gray clay and silt, and the grainsize changes are small. The average IRD (>63 μm and >250 μm) contents are about 32 and 19%, respectively, and the diameter of gravels ranges from 0.5 to 4 cm.

The chemical composition of 70 sediment samples is determined for the core ANT32-RA05C, and the average content of SiO₂ is 66.9%, followed by 11.2% Al₂O₃ and 4.3% Fe₂O₃. For stratigraphic changes (Figure 3), the SiO₂ content is higher in coarse-grained sediments, and Al₂O₃ and Fe₂O₃ have an opposite pattern.

There are two main sources of element Si, namely, terrigenous and biogenic sources, and the former is from the quartz debris and clay minerals, while the latter is biogenic, such as diatom frustules and sponge spicules. A correlation analysis can reveal the relationship between the different sources, and as a result (Table 1), SiO₂ is positively correlated with the IRD, biogenic silica, and CaO but negatively correlated with the rest. The positive relationship between biogenic silica, CaO, and element Si indicates a close relationship between them, likely relating to marine productivity changes. Previous studies show that over 90% of minerals in coarse sediments of the Ross Sea are feldspathic quartz (Li et al., 2021), which are strongly resistant to physical weathering (Cook et al., 2017; Hobbs et al., 2016), as observed in the range of the chemical index of alteration (CIA, 47–67) in the core ANT32-RA05C. Considering the negative relationship between SiO₂ and Al₂O₃, it is inferred that the terrigenous debris with element Si is mainly from quartz.

Subjected to the alternating field demagnetization, characteristic remanence can be separated between 15 and 60 mT (Figure 4), and all the samples obtain reliable results using 4–11 steps with a maximum angular deviation (MAD) of <15°. The remanence intensity of the samples changes little from the bottom to top (Figure 5), with three peaks at 236–242 cm, 207–217 cm, and 13–58 cm. The magnetic inclination is mainly reversed, consistent with the geographic location of the core.

The magnetic parameters can reveal the changes in magnetic minerals and/or magnetic domains. For example, ARM is usually sensitive to single-domain (SD) grains (Maher, 1988; Duan et al., 2012), and SIRM can be used to infer magnetic particles excluding the influence of superparamagnetic (SP) grains (Evans and Heller, 2003). Inferred from the magnetic parameters (Figure 5), magnetic minerals are concentrated at 236–242 cm, 207–217 cm, and 13–58 cm, consistent with the intervals of high remanence intensity. All these findings suggest a relatively stable process, which can be employed to establish an RPI curve for comparison.

The U and Th isotopes are listed in Table 2, in which the values are similar and the ²³⁰Th content gradually decreased to the bottom, generally fulfilling the base of ²³⁰Th-excess dating (Osmond, 1979; Xia, 1989; Edwards et al., 2003). Assuming that ²³⁰Th_ex flux is constant over the time of deposition, ²³⁰Th_ex decays exponentially with the depths, and the slope coefficient of the regression represents the sedimentation rate (Scholten et al., 1994). The calculation is as follows: ln 230 Th ex ( h ) = ( - λ /k ) ×h + ln 230 Th ex ( 0 ) , where ²³⁰Th_{ex (h)} and ²³⁰Th_{ex (0)} are the excess radioactivity of ²³⁰Th at depth “h” and the surface, respectively. λ is the ²³⁰Th decay constant, 9.21738 × 10⁻⁶, and the curve with −λ/V indicates the slope. Based on the estimates shown in Figure 6, the average sedimentation rate is 1.37 cm/kyr.

Dating sediments in polar seas is much more difficult than that in other areas (Wang et al., 2009), and an integrated result of various dating methods is commonly used. According to an average sedimentation rate of 1.37 cm/kyr from the ²³⁰Th method, the bottom age of the core ANT32-RA05C is about 200 ka, offering a preliminary constraint for RPIs.

To refine the age model, RPI is employed, as suggested in previous studies (Sagnotti et al., 2001; Branchfeld et al., 2003; Macrì et al., 2005; Macrì et al., 2006; Collins et al., 2012; Xiao et al., 2016). However, NRMs are sensitive to changes in magnetic minerals, and to remove this potential influence, various normalizations have been proposed (King et al., 1983; Meynadier et al., 1992; Tauxe, 1993), including MS, ARM, and IRMs. Accordingly, the RPIs of the core ANT32-RA05C can be obtained, and all the curves show a similar pattern (Figure 7). Subsequently, the derived RPIs are correlated with the SINT800 RPI curve (Guyodo and Valet, 1999), to obtain additional age controls. Specifically, the low RPI at 274 cm could be correlated with the magnetic excursion of Jamaica/Pringle Falls at ∼191 ka (Guyodo and Valet, 1999), and other constraints are listed in Table 3. Based on this correlation, there is a high consistency between ANT32-RA05C RPI and SINT800, yielding an average sedimentation rate within 66–274 cm in depth as 1.34 cm/kyr and 1.42 cm/kyr within 66–246 cm in depth. This estimate is close to the ²³⁰Th result (1.37 cm/kyr), confirming the reliability of the age model of the studied core in the Ross Sea.

Based on the age model from ²³⁰Th dating and RPI changes, the paleoenvironmental processes in the study area since MIS 7 can be recovered (Figure 8). The changes in IRD, biogenic silica, and element Si are selected as the paleoenvironmental proxies, indicating ice-sheet melting and marine productivity, respectively, as aforementioned. For comparison, the temperature reconstruction of the well-studied EDC ice core is displayed for reference (Jouzel et al., 2007). However, because of the little consistency between the magnetic parameters and the Antarctic temperature and other proxies of the core ANT32-RA05C (Figure 8), magnetic properties are not included in this discussion.

As shown in Figure 8, element Si can be well correlated with the changes in the Antarctic temperature. Specifically, the Si content fluctuated during MIS 6, increased sharply at the MIS 6/5 boundary, decreased slowly during MIS 3, and increased again since ∼20 ka. It is then inferred that the marine productivity in the Ross Sea covaried with the Antarctic temperature changes on the glacial-interglacial timescales, agreeing well with the previous studies in Antarctica (Pudsey and Howe, 1998; Diekmann, 2007; Anderson et al., 2009; Xiao et al., 2016). During the LGM, the Antarctic temperature declined to a much low level, and sea ice largely expanded, causing a significant decrease of the IRD transport and thus the fine-grained sediments. However, the regional environment within this key interval may be unstable since all the proxies changed greatly and frequently (Figure 8). After the LGM, the Antarctic temperature increased rapidly, icebergs collapsed, and IRD increased. Meanwhile, the warming led to an increase in marine productivity, inferred from biogenic silica content of the core ANT32-RA05C. Moreover, the changes in IRD and biogenic silica are not synchronous, while IRD changes are consistent with element Si, indicating that biogenic processes may precede changes in the ice sheet in the Ross Ice Shelf during the last deglaciation, thus highlighting the role of oceanic circulation in ice-sheet retreat (Hillenbrand et al., 2009; Williams et al., 2019). In the Holocene, the Antarctic temperature rises as high as present, and IRD increased significantly to the highest level (up to 54%) over the past 200 kyr, suggesting that ice sheet, regional temperature, and marine productivity are coupled again as similar in other intervals.

As an important proxy of iceberg drift and ice sheet evolution, IRD has been broadly used in the palaeoceanographic research in the Arctic and Antarctic regions (Darby and Paula, 2008; Wang et al., 2009), representing a climatic transition from cold to warm. In the core ANT32-RA05C, three intervals with evident IRDs are recognized: 29–32 ka, 15–17 ka, and 5 ka to present. However, IRDs did not exactly covary with the temperature changes, such as inconsistent cases in 29–32 ka and MIS6/5 boundary. The difference between the IRDs of the core ANT32-RRA05C and Antarctic temperature might be caused by proxy sensitivity or a different mechanism during climatic transitions. For example, IRD events could be influenced by factors such as ice shelf stability, sea-level changes, regional sedimentary dynamics, sea surface temperature, and shelf topography (Teitler, 2010; Patterson et al., 2014; Li et al., 2021). Considering the similarity of change patterns of the Antarctic temperature between different intervals, it is inferred that the ice shelf stability could be a major reason in the diverse IRD records, which are worthy of further investigation in the future.