Magnetoclimatological Record of Late Pleistocene Loess in the Southern Hunshandake Sandy Land, Inner Mongolia: A Threshold Response to the East Asian Summer Monsoon Variations

We report the first complete loess–paleosol record spanning the last 130 kyr from the southern extremity of the Hunshandake Sandy Land (HSL) in central‐eastern Mongolia. Our combined mineral magnetic and geochemical results demonstrate that during the last interglacial, the front of the East Asian summer monsoon (EASM) extended to central‐eastern Inner Mongolia, consistent with modern observations of climate change. However, during the last glacial, typical magnetic parameters (e.g., magnetic susceptibility and anhysteretic remanent magnetization and their ratios), which have been successfully adopted to denote the EASM variations on the Chinese Loess Plateau (CLP), show only muted temporal variations and cannot be readily correlated with equivalent records from the main body of the CLP. Remarkably, the Zr/Rb ratio, a robust geochemical indictor of the winter monsoon, is positively correlated with saturation isothermal remanent magnetization and saturation magnetization during the last glacial, demonstrating the strong control of wind vigor on high‐field magnetization signals dominated by lithogenic ferrimagnetic minerals. In contrast to the variable response of magnetic parameters to regional paleoclimatic fluctuations, the Rb/Sr and Ba/Sr ratios (two summer monsoon indicators) faithfully track weak chemical weathering processes and fine‐scale monsoon oscillations, especially during the last glacial. Additionally, we found that higher proportions of pedogenic high‐coercivity magnetic minerals were formed during the last interglacial, which may reflect relatively low rainfall but higher evapotranspiration in the southern extremity of the HSL compared with the CLP. We therefore attribute this complex magnetic record to a threshold response to the EASM variations in eastern Inner Mongolia.


Introduction
The East Asian monsoon (EAM) is an important component of the global monsoon and plays a key role in large-scale moisture transport and energy cycles (An, 2000;Kutzbach, 1981). Understanding the spatiotemporal variability of the EAM is important because of its influences on the physical environment and on the wellbeing of the large human population of the monsoon domain. Previous studies of the loess-paleosol sequences on the Chinese Loess Plateau (CLP) have significantly improved our understanding on the EAM evolution in northern China Ding et al., 1995;Guo et al., 2002;T. S. Liu & Ding, 1998;Y. B. Sun et al., 2021). However, the paleoclimatic history at the northern margin of the EAM during the Late Pleistocene is poorly documented because of the scarcity of chronologically well-constrained sedimentary records. The arid and semi-arid region of central and eastern Inner Mongolia is located along the present-day northern margin of the East Asian summer monsoon (EASM) and is particularly sensitive to the advance and retreat of the EASM (Li et al., 2023). The Hunshandake Sandy Land (HSL), covering an area of about 52,000 km 2 , is one of the four largest sandy lands in central and eastern Inner Mongolia (Figure 1). Quite a few paleoclimatic studies have been conducted on lake sediments and aeolian sand-paleosol sequences within and surrounding the HSL since the Last Glacial Maximum, focusing on the timing of the Holocene climatic optimum (S. Z. Liu et al., 2015;Mason et al., 2009;Tang et al., 2015;X. P. Yang et al., 2013), moisture variations during the middle-late Holocene (W. Y. Jiang et al., 2006;H. Y. Lu et al., 2011;Xiao et al., 2006), and their forcing mechanisms (S. Z. Liu et al., 2016;Xiao et al., 2006;Y. W. Zhou et al., 2018). However, there is lack of research on the advance and retreat of the EASM on the glacial-interglacial time scales, especially for loess deposits from the monsoonal margin, which are important for improving our understanding of the spatial variability and forcing mechanisms of the EASM on orbital-to-millennial-scale timescales.  Yang & Ding, 2008), and the locations of the Beigou section (red star) and typical loess records mentioned in the text (green stars). Arrows indicate the modern wind directions of the East Asian summer monsoon (red) and East Asian winter monsoon (blue). Deserts: A, Taklimakan; B, Gurbantunggut (Junggar); C, Kumtag; D, Qaidam; E, Badain Jaran; F, Tengger; G, Mu Us; H, Hobq; I, Hunshandake; J, Horqin; K, Hulun Buir. The inset map shows the location of the Beigou sampling site, representative lakes (blue dots), surface runoffs (purple dotted lines), and cities (black dots) in central-eastern Inner Mongolia.

10.1029/2023GC010905
3 of 20 Magnetoclimatological studies of Chinese loess-paleosol sequences have provided abundant information on the EAM evolution, Asian aridification, and abrupt climatic changes, from tectonic to millennial timescales (Deng et al., , 2006Hao et al., 2008;. For example, the strong correlation between the magnetic susceptibility (χ) of Chinese loess and the deep-sea oxygen isotope record suggests a close relationship between the EAM fluctuations and global ice volume (Heller & Evans, 1995;. In the hinterland of the CLP, the formation of superparamagnetic (SP) and single domain (SD) ferrimagnetic minerals during intervals of strengthened pedogenesis (interglacials) results in a significant magnetic enhancement of paleosols (Q. S. Liu et al., 2005Liu et al., , 2007Maher & Thompson, 1992). In contrast, the magnetic signals of the loess-paleosol sequences in the western CLP and westerlies-influenced region are controlled by both pedogenesis and the eolian supply of magnetic minerals (reflecting wind vigor and/or the distance to the source region), and thus coarse-grained detrital ferrimagnetic minerals may even dominate the magnetic properties (X. S. Wang et al., 2015;Zan et al., 2012Zan et al., , 2020. Additionally, spatiotemporal variations of dust sources may further complicate the interpretation of the linkages of loess magnetic properties to long-term monsoonal fluctuations (X. S. Wang et al., 2015;Zeng et al., 2017). In contrast to the rich magnetoclimatological studies for the hinterland of the CLP, systematic mineral magnetic studies in central-eastern Inner Mongolia are sparse, hindering a comprehensive understanding of the paleoclimatic significance of magnetic properties of eolian deposits in the northern margin of the EASM.
The geochemical composition of Chinese loess is controlled mainly by three processes: the deflation of primary materials from different source areas, wind sorting of dust during transport, and post-depositional weathering (J. Chen et al., 1999Chen et al., , 2006Gallet et al., 1996). Geochemical proxies of loess-paleosol sequences have been widely used for paleoclimatic reconstruction (Lei et al., 2021;L. Liang et al., 2013; and for provenance discrimination (Jahn et al., 2001;J. M. Sun, 2002;Zhang et al., 2012). For example, elemental ratios such as Zr/Rb, SiO 2 /TiO 2 , and SiO 2 /Al 2 O 3 , which are relatively unaffected by pedogenesis but are sensitive to grainsize sorting, can be used to reflect the East Asian winter monsoon (EAWM) strength (J. Chen et al., 2006;L. W. Liu et al., 2004); while Rb/Sr, Ba/Sr, and Na/Al ratios are sensitive to post-depositional weathering and pedogenic processes, and therefore to the EASM variations (J. Chen et al., 1999;Gallet et al., 1996). Furthermore, the distribution patterns of rare earth elements (REE) may reflect the homogeneity and eolian origin of Chinese loess (Gallet et al., 1998;Jahn et al., 2001). Therefore, the combination of mineral magnetic and geochemical analyses can potentially help decipher mineral magnetic-paleoclimatic relationships and the chemical weathering history in the northern margin of the EASM.
Recently, we discovered a complete Late Pleistocene loess-paleosol sequence in the southern extremity of the HSL that provides an excellent opportunity to reconstruct paleoclimatic variability in the monsoonal margin. Here, we present the results of detailed measurements of environmental magnetism, geochemistry, and diffuse reflectance spectroscopy (DRS) for this Late Pleistocene loess-paleosol sequence. The objectives of the study are (a) to characterize the magnetic properties of loess deposits in the northern margin of the EASM; (b) to define reliable magnetic indicators for paleoclimatic reconstruction in central-eastern Inner Mongolia since the Late Pleistocene; and (c) to determine possible forcing mechanisms of climatic variations in the region.

Geographic Setting and Sampling
The Beigou loess section (42.40°N, 115.71°E) is located ∼30 km northwest of Zhenglanqi County, on the southern margin of the HSL (Figure 1). The mean annual rainfall and temperature are 359 mm and 2.5°C, respectively (30-year average for 1981-2010). Rainfall occurs mainly during the warmer months of May through September, which accounts for ∼85% of the annual total. The maximum temperature is 19.4°C in July and the minimum is −17.2°C in January. The exposed loess-paleosol sequence has a thickness of 4.72 m and spans the last glacialinterglacial cycle. The pedostratigraphic boundary between the upper loess and the lower paleosol is at the depth of 2.8 m, recognized by lithological and color changes ( Figure 2). 118 bulk samples were continuously taken at 4-cm intervals from the entire section for magnetic, geochemical, and diffuse reflectance spectroscopic analyses, and three samples for optically simulated luminescence (OSL) dating were collected at the depths of 0.3, 1.57, and 2.53 m ( Figure 2).

Magnetic, Geochemical, and Diffuse Reflectance Spectroscopic Measurements
All samples were packed in 2 × 2 × 2 cm plastic boxes for measurements of magnetic susceptibility (χ), anhysteretic remanent magnetization (ARM) and isothermal remanent magnetization (IRM). χ (normalized by mass) was measured using an AGICO MFK1-FB Kappabridge at frequencies of 976 and 15,616 Hz in a peak magnetic field of 200 A/m in the Institute of Geophysics, China Earthquake Administration. Frequency-dependent magnetic susceptibility (χ fd ) is defined as χ 976Hz − χ 15,616Hz . Anhysteretic remanent magnetization was imparted using a 2G Enterprises alternating field demagnetizer in a DC field of 0.05 mT and a maximum AC field of 100 mT. Saturation IRM (SIRM) was imparted in a field of 2 T, and a 0.3 T backfield of IRM −0.3T was then applied to obtain the "hard" isothermal remanent magnetization (HIRM), calculated as (0.5 × (SIRM + IRM −0.3T )), and S −0.3 (− IRM −0.3T /SIRM). Both ARM and IRMs were measured using a 2G 755-4 K superconducting rock magnetometer in the Institute of Geomechanics, Chinese Academy of Geological Sciences (CAGS). Hysteresis measurements were performed on a MicroMag 3900 VSM with a maximum field of 1 T in the Institute of Geophysics, China Earthquake Administration. Hysteresis parameters, including coercivity (B c ), saturation magnetization (M s ), and saturation remanence (M rs ), were calculated after subtracting the high-field paramagnetic contribution using a line fit through data points between 70% and 100% of the maximum applied field. Remanent coercivity (B cr ) was obtained from the back-field demagnetization curves. Unmixing of IRM acquisition curves was performed using the MAX UnMix web application (Maxbauer et al., 2016). Low-temperature magnetic experiments were conducted using a Quantum Designs Magnetic Properties Measurement System (MPMS3) in Institute of Geomechanics, CAGS. Zero-field cooled (ZFC) and field cooled (FC) curves were obtained by cooling the sample in zero field and in a 2.5 T field from 300 to 1.8 K, respectively, imparting a saturation remanence in a 2.5 T field at 1.8 K and then measuring the remanence during warming from 1.8 to 300 K in zero field. A cooling-warming cycle of a saturation remanence imparted in a 2.5 T field at 300 K was conducted in zero field. All low-temperature remanence measurements were performed at 2-5 K intervals. Diffuse reflectance spectra (DRS) were measured in the Institute of Geology and Geophysics, Chinese Academy of Sciences using a Cary 5000 UV-vis-IR spectrophotometer (Varian Inc., Palo Alto, CA), equipped with an integrating sphere accommodating a PMT/PbS detector Figure 2. Photograph of the Beigou loess section (left), and age-depth model for the section (right). Red solid circles represent three optically simulated luminescence (OSL) sampling sites at the depths of 0.3, 1.57, and 2.53 m, respectively. The black circle data points (right) represent tie points determined by correlating χ ARM /χ and Rb/Sr ratios with the LR04 benthic δ 18 O stack over the past 150 kyr (Lisiecki & Raymo, 2005) (see dashed lines in Figure 8). Note that while both the BG-1 and BG-2 OSL dates are very close to the correlation-based ages, the OSL age of ∼50 ka for BG-3 is far away from the age-depth line (∼70 ka), suggesting an obvious age underestimation due to OSL signal saturation. at a scan rate of 30 nm/min from 200 to 2600 nm in 0.5-nm steps. The reflectance values were then transformed into the Kubelka-Munk (K-M) remission function and the band intensities of the second derivative of the K-M function spectrum are proportional to the concentrations of goethite and hematite. The intensities of the bands at 425 nm (I 425 ) and 535 nm (I 535 ) are used as proxies for relative changes in the mass concentration of goethite and hematite, respectively. The calibration curve Y = −0.133 + 2.871X − 1.709 × X 2 (Y is the Hm/(Hm + Gt) mass ratio and X is the I 535 /(I 425 + I 535 ) ratio) and the regression equation Fe d = GI 425 + HI 535 (the regression coefficients G = 8.56 × 10 3 and H = 22.2 × 10 3 ) were applied to quantify the relative concentrations of hematite and goethite, as described in detail by Torrent et al. (2007). The concentrations of major elements for bulk samples were determined using a PW4400 X-ray Fluorescence spectrometer, and trace and rare-earth element (REE) analyses were performed using an ICP-MS-PE300D in National Research Center for Geoanalysis, CAGS. Additionally, 32 loess and paleosol samples, previously collected from Fanshan (northeastern CLP), Jiuzhoutai and Baicaoyuan (western CLP) (X. S. Wang et al., 2014), were also utilized for the determinations of major, trace, and REE compositions and comparison of geochemical characteristics with the Beigou record.

Luminescence Dating
Well-sealed sample tubes were processed under red low intensity light conditions (655 ± 30 nm), with loess from the central part of the tube used for isolation of quartz for OSL dating. The fresh samples (approximately 100 g) were first treated with 30% hydrochloric acid (HCL) and 30% hydrogen peroxide (H 2 O 2 ) to remove carbonates and organic matters, respectively. Then the fine silt (4-11 μm) was obtained using sedimentation procedures based on Stokes' Law. These polymineral fine grains were immersed in hydrofluorosilicic acid, H 2 SiF 6 (30%), for 3 days in an ultrasonic bath to extract the fine-grained (4-11 μm) quartz component. Finally, the purified quartz was deposited on 9.7-mm-diameter stainless steel discs for experiments.
OSL measurements were performed using a Risø TL/OSL-DA-20 reader equipped with a 90 Sr/ 90 Y beta irradiation source in Institute of Geology, China Earthquake Administration. The quartz OSL signal was stimulated at 125°C for 60 s with blue LEDs and detected using 9235QA photomultiplier tube coupled in front of which is 7.5 mm Hoya U-340 glass filters. A sensitivity-corrected multiple-aliquot regenerative-dose (SMAR) dating protocol was used for determining the equivalent dose (D e ) (Y. C. Lu et al., 2007;X. L. Wang, Lu, & Wintle, 2006). The net OSL signal was calculated using the initial 5 s integration of the OSL decay curve subtracted with that of the last 5 s. Two sets of natural aliquots are prepared for obtaining D e . Nine aliquots are used to determine the natural OSL intensity, and 10 aliquots are used to construct the dose-response curve that brackets the natural OSL intensity. The regenerative-dose aliquots are individually irradiated with different regeneration doses after the natural OSL signal is optically removed using a SOL2 30 min bleaching. The natural aliquots and the regenerative-dose aliquots are preheated at 260°C for 10 s to remove thermally unstable signals. Then, the natural and regenerative-dose OSL intensities (LN and Li, respectively) are obtained. The test dose OSL signal (TN and Ti) is applied to correct for sensitivity changes caused by the thermal and optical treatments and to normalize the differences between aliquots. D e is finally obtained by projecting the corrected natural OSL intensity (LN/TN) onto the dose-response curve constructed using the corrected regenerative-dose OSL intensity (Li/Ti).
The concentrations of uranium (U) and thorium (Th) were measured using neutron activation analysis, and potassium (K) content was determined by flame spectrum analysis (Table 1). The in situ water content was derived by weighing the samples before and after drying (105°C for 8 hr in an oven) and an error of ±3% was assumed for all samples. The fine-grained quartz efficiency was measured for each sample by comparing the OSL signals regenerated by α and β irradiation after an initial exposure to the SOL2 solar simulator for 3 min, as introduced by L. P. Zhou et al. (1995)  consistent with values given by Rees-Jones (1995) for fine grained quartz OSL. Dose rates were calculated using the conversion factors (Adamiec & Aitken, 1998) for U, Th and K concentrations, the measured efficiency and water content, and these dose rates are listed in Table 1.

Age Model Construction
OSL dating is widely used to provide independent absolute chronology of the Late Pleistocene loess on the CLP (e.g., Y. C. Lu et al., 2007;Stevens et al., 2008). The three OSL samples collected at the depths of 0.3, 1.57, and 2.53 m yielded the respective ages of 16.15 ± 0.95, 41.37 ± 2.69, and 49.99 ± 2.40 ka (see details in Table 1). Combined with the OSL ages and detailed field observations, comparison of the χ record with those of typical loess sections in the CLP (Figure 3) indicates that the upper loess unit and the lower paleosol unit, respectively, represent the last glacial loess (L 1 ) and the last interglacial paleosol (S 1 ). Owing to its very marginal location relative to the wind systems supplying dust to the CLP, the corresponding thickness of ∼4.72 m since the last interglacial at Beigou is significantly thinner than at sections in the CLP, such as Jingyuan (∼45 m) (Y. B. Sun et al., 2019), Baicaoyuan (∼28 m) (Huang & Sun, 2005), Luochuan (∼11 m) (H. Y. Lu et al., 1999), and Weinan (∼12 m) (Pan et al., 2002) (Figure 3).
Owing to a striking resemblance between deep-sea benthic δ 18 O records and paleoclimatic proxies (e.g., χ, grain size, and Rb/Sr ratio) of Chinese loess, chronological frameworks of loess-paleosol sequences during the last glacial-interglacial cycle are commonly generated by matching rapid changes of χ (Kukla et al., 1988) and/or grain size (Vandenberghe et al., 1997) with deep-sea δ 18 O records, which describe global ice volume  (Huang & Sun, 2005), Luochuan (Lu et al., 1999), and Weinan (∼12 m) (Pan et al., 2002), since the last interglacial. The gray horizontal bars denote the three sub-paleosol layers of S1 (S 1−1 , S 1−2 and S 1−3 ). OSL ages for the depths of 0.3, 1.57, and 2.53 m are also shown. and deep-water temperature change (Lisiecki & Raymo, 2005). Relying on three independent OSL ages that provide a preliminary chronological constraint for the Beigou record, a more precise polynomial (third-order) age model (Figure 2) was developed by tie points determined by correlating high/low χ ARM /χ and Rb/Sr ratios with low/high values of the LR04 benthic δ 18 O stack over the past 150 kyr (Lisiecki & Raymo, 2005) (see dashed lines in Figure 8). χ ARM /χ and Rb/Sr ratios are used because they are sensitive to post-depositional weathering and pedogenesis and has proved to be reliable indicators of the EASM strength (J. Chen et al., 2006;Maher & Thompson, 1992). Based on our age model, the estimated average accumulation rate of L 1 (4.6 cm/ka) is approximately twice higher than that of S 1 (2.4 cm/ka) (Figure 2), which is consistent with that on the CLP, further reinforcing the validity of our age control.

Magnetic Characteristics of the Beigou Loess and Their Paleoclimatic Implications
The χ values across the whole investigated interval range between 33.6 × 10 −8 and 95.6 × 10 −8 m 3 /kg, with a mean of 43.3 × 10 −8 m 3 /kg. These values are significantly lower than those of the southern (Weinan; Pan et al., 2002) and central CLP (Luochuan; H. Y. Lu et al., 1999) (Figure 3), suggesting much weaker pedogenesis on the northern margin of the EASM compared with the coeval L 1 and S 1 on the CLP. The three sub-paleosol layers of S 1−1 , S 1−2 and S 1−3 , corresponding to MIS 5a, 5c, and 5e, respectively, have higher χ values but lower coercivity values (Figure 4), indicating pedogenesis-induced magnetic enhancement during the last interglacial. χ fd and ARM also exhibit notable increases in S 1 and decreases in L 1 (Figure 4). Since χ fd and ARM are responsive to viscous SP and smaller SD ferrimagnetic grains, respectively (Dearing et al., 1996;, the consistent variations of χ, χ fd , and ARM, as well as of the magnetic grain-size parameter χ ARM /χ, confirm that the magnetic enhancement of S 1 is primarily due to in situ formation of SP/SD ferrimagnetic grains during pedogenesis, although the pedogenic intensity in central-eastern Inner Mongolia is much weaker than that in the CLP. In contrast to the lower S −0.3 values in loess layers, but higher values in the more weathered paleosols (  ranging between 0.77 and 0.85 (Figure 4), are systematically lower than those of L 1 , and are lower than those of S 1 in typical loess sections in the CLP, such as Luochuan and Caijiapo (>0.85) (Bloemendal & Liu, 2005;Deng et al., 2004;Hu et al., 2013). This indicates that relatively high proportions of high-coercivity minerals (e.g., hematite and goethite) are present in S 1 . Higher HIRM, together with the high concentration of hematite in S 1 indicated by the DRS data, further demonstrate that the concentration of pedogenically formed hematite increased during the last interglacial (Figure 4). The IRM decomposition results of 16 samples reveal four components based on their coercivity spectra, among which the major IRM carrier of ferrimagnetic assemblage (Comp. 1), with a low coercivity, is well defined and accounts for >80% of the total IRM (Figures 5a and 5b). Compared with the higher coercivity of Comp. 1 (average coercivity of 70 mT) in L 1 , the samples from S 1 have a much lower coercivity of Comp. 1 (average coercivity of 40 mT), indicating that the pedogenic ferrimagnetic minerals have a relatively low coercivity Spassov et al., 2003). The higher intensity of Comp. 2 (average coercivity of ∼150 mT) in the S 1 samples, compared with those from L 1 , may reflect either pedogenic maghemite of higher oxidation state (Spassov et al., 2003) or the increased contribution of nano-sized pedogenic hematite . The consistently lower S −0.3 but higher HIRM values during the last interglacial ( Figure 4) suggest that Comp. 2 may primarily represent maghemite contributions, and that the pedogenic hematite population would not be significantly included in the soft fraction of S −0.3 . Strong positive correlations between χ and χ fd , ARM, and SIRM (Figuresure 6a, 6c and 6e), and between Rb/Sr and ARM (Figure 6d), but a negative correlation between SIRM and Zr/Rb (Figure 6f), are also observed during the last interglacial. Since the Rb/Sr and Zr/Rb ratios have been widely adopted as reliable indicators of the EASM and EAWM, respectively (J. Chen et al., 1999;Gallet et al., 1998;, these correlations further confirm that the enhanced values of χ, χ fd , χ ARM , and SIRM during the last interglacial result from the pedogenic formation of SP/ SD ferrimagnetic minerals. The combined rock magnetic and geochemical results provide strong evidence that the exceptional increases of SIRM in L 1 are caused by the significant input of coarse-grained ferrimagnetic minerals. The hysteresis results show that the L 1 samples have lower M rs /M s but much higher B cr /B c values compared with the samples from S 1 (Figures 5c, 5d, and 7), implying the systematic coarsening of ferrimagnetic grains in L 1 (Roberts et al., 2018). The low-temperature magnetic experiments reveal a sharp decrease in remanence across the Verwey transition of ∼120 K in L 1 (Figure 5e) but a muted remanence transition in S 1 (Figure 5f), indicating severe maghemitization of magnetite in S 1 (Özdemir et al., 1993). The FORC diagram of the sample from S 1 has a large vertical spread and triangular contours, indicating the presence of both finer SP/SD particles and coarser PSD particles (Figure 5h), while the samples from L 1 have typical FORC diagrams of PSD and MD mixtures with a low-coercivity, vertically extended component and a weak, high-coercivity, horizontally extended tail ( Figure 5g) (Egli, 2013;Harrison & Feinberg, 2008). The IRM component analyses also indicate the presence of PSD/MD magnetite in the samples from L 1 , with the midpoint coercivity of ∼70 mT (Figure 5a), which accords with the characteristic coercivity values obtained for detrital magnetite in typical loess (Spassov et al., 2003). More importantly, several intervals of high SIRM and M s values in L 1 are positively correlated with the Zr/ Rb ratio, which is a reliable indicator of the EAWM, since the last interglacial-glacial transition (Figure 6f; Figure 8), further confirming that the obvious enhancements both of SIRM and M s in L 1 is caused by significant increases in the content of coarse-grained eolian magnetite associated with the intermittent strengthening of the EAWM during the last glacial. Positive correlations between χ, SIRM, and the Zr/Rb ratio are also observed in L 1 (Figure 8), implying that the magnetic properties in this interval are dominated by the coarse-grained lithogenic magnetite component, compared with the pedogenesis-related SP/SD ferrimagnetic grains. We therefore conclude that the variations in both χ and SIRM were dominated by pedogenesis during the last interglacial and by wind vigor as well as associated variations in the distance to the dust sources during the last glacial. This is because both the ultrafine pedogenic (SP-SD) and coarse-grained eolian (PSD-MD) fractions may result in the enhancement of χ and SIRM (Hartstra, 1982;Maher, 1988). The complex response of χ to lithological and climatic variations indicates that during the last glacial it could not be used as a reliable proxy for pedogenic intensity at the northern boundary of the EASM; instead, Rb/Sr and χ ARM /χ are more robust indicators of post-depositional weathering and pedogenesis, and can be used to construct a chronological framework (Figure 8).
Precipitation has been recognized as a dominant factor for pedogenic processes, and a strong linkage of soil magnetic properties to regional precipitation in soils on the CLP has been identified (Heller et al., 1993;Maher & Thompson, 1992, 1995Nie et al., 2010Nie et al., , 2014. Soil water balance, which depends on the frequency and intensity of rainfall events and on water loss by evapotranspiration, can also control the formation of magnetic minerals by dominating the annual soil "wetting" and "drying" cycle (Orgeira et al., 2011). The Beigou section is located to the northwest of the 400-mm isohyet (isopleth of mean annual precipitation-MAP), that is, the precipitation boundary between the semi-humid to semi-arid regions in North China. Although the summer temperature in the southern HSL is slightly lower than that of Luochuan, the modern MAP of 365 mm/yr in the region is significantly lower than that of Luochuan (592 mm/yr) (Figure 9). The much lower χ values in the southern HSL compared with those in the CLP substantiate the view that the MAP is a key factor for the enhancement of ferrimagnetic signals (e.g., Gao et al., 2018;Nie et al., 2007;Song et al., 2014). Besides the MAP, the mean annual temperature is another factor for the formation of magnetic minerals, and the magnetic characteristics and content variations of hematite have been used to characterize environmental and climatic evolution Z. X. Jiang et al., 2012Z. X. Jiang et al., , 2014Z. X. Jiang et al., , 2022. Abundant evidence from synthetic experiments suggests that the production of pedogenic hematite is more dependent on temperature than precipitation (Barron & Torrent, 2002;Gao et al., 2018;Torrent & Guzman, 1982). The higher proportion of pedogenic hematite in S 1 of the Beigou section suggests that dry climatic conditions and limited effective precipitation favor the formation and preservation of weakly magnetic minerals, such as hematite, at the northern limit of the EASM. This speculation is consistent with the conclusions of Schwertmann (1985) that a high soil temperature and low moisture content favor the transformation of ferrihydrite to hematite, because a dehydration step is involved during which ferrihydrite loses its water and the Fe atoms attain a consistent hematite structure. We therefore prefer the view that the pedogenic production of ferrimagnetic minerals (maghemite and magnetite) is primarily controlled by precipitation, while the pedogenic production of hematite is likely more dependent on temperature than on precipitation.

East Asian Monsoon Variations in Central-Eastern Mongolia During the Last Glacial-Interglacial Cycle
During the last interglacial-glacial transition, the Zr/Rb ratio (a proxy of the EAWM intensity), M s , and SIRM show a concordant pattern of fluctuations, indicating a gradual strengthening of the EAWM after MIS 5a. In contrast, the Rb/Sr ratio and ARM-derived magnetic parameters (proxies of the EASM intensity) are characterized by an initially slow and then an abrupt weakening, suggesting a general weakening of the EASM. The "anomalous" troughs recorded by both SIRM and M s at a depth of ∼3.2 m, together with an obvious inverse relationship between these two high-field magnetizations and the ARM and χ records during the last interglacial-glacial transition (Figures 4 and 8), suggest that the EAWM-derived SIRM signal is evidently superimposed on a weak EASM-dominated magnetic signal. We therefore propose that the climatic conditions during the last interglacial-glacial transition were characterized by rapid retreat of the EASM but the significant strengthening of the EAWM, indicating an anti-phased relationship between the EASM and EAWM in central-eastern Inner Mongolia. This is consistent with several susceptibility and bulk grain size-based Figure 5. Results of isothermal remanent magnetization component analysis (a, b), hysteresis loops (c, d), normalized low-temperature remanence versus temperature (e, f), and FORC diagrams (g, h) for typical loess (left) and paleosol samples (right). The gray dots, blue, purple, green, red, and yellow curves represent the raw data, Comp. 1 (ferrimagnetic assemblage), Comp. 2 (pedogenic maghemite with higher oxidation state?), Comp. 3 (hematite), Comp. 4 (goethite), and modeled coercivity distribution, respectively (a, b). Blue (black) loop is before (after) paramagnetic correction (c, d). The blue and pink curves, respectively, represent cooling and warming curves after imparting a 2.5 T SIRM at room temperature; and the black and red solid curves, respectively, represent the warming curves of a 2.5 T remanence after cooling in zero field and in a field of 2.5 T (e, f). B c , coercive force. B u , distribution of interaction fields. Smoothing was performed using FORCinel (Harrison & Feinberg, 2008), with the following parameters: S c0 = 4, S c1 = 7, S b0 = 3, S b1 = 7, λ c = 0.1, and λ c = 0.1 (g, h). paleoclimatic records from the CLP which indicate that the EAWM is inversely correlated with the EASM on the glacial-interglacial timescale (An, 2000;T. S. Liu & Ding, 1998).
Based on our age model, we compared the Rb/Sr and Ba/Sr records from Beigou, both of which faithfully track weak chemical weathering processes, with various other paleoclimatic signals: the benthic δ 18 O record (Lisiecki & Raymo, 2005), the Greenland ice core (NGRIP) δ 18 O record (NGRIP members, 2004), the stalagmite δ 18 O record from Hulu cave (Y. J. Wang et al., 2001), records of grain size and reconstructed summer precipitation from the Yuanbao section in the western CLP (Rao et al., 2013), and North Hemisphere summer insolation (NHSI) at 65°N, during the last glacial (Berger & Loutre, 1991). Reference to Figure 10 shows that intervals of strengthened/weakened summer monsoon, evidenced by high/low values of Rb/Sr and Ba/Sr, correspond to strong/weak NHSI, demonstrating that insolation is the primary control of summer monsoon variations in central-eastern Inner Mongolian on the orbital scale. Additionally, on the millennial timescale, grain size records from the northern and western CLP and Chinese stalagmite δ 18 O records document weak summer monsoon events that correspond well to the Heinrich events of the North Atlantic and cold phases of Dansgaard-Oeschger cycles evident in Greenland ice cores (e.g., F. H. Chen et al., 1997;Porter & An, 1995;Y. B. Sun et al., 2021). As shown in Figure 10, the intervals of 57-51 and 15-10 ka, with higher values of Rb/Sr and Ba/Sr in the Beigou section, correspond well to MIS 3 and the last deglaciation, respectively, and two short interstadial peaks during 46-44 and 38-36 ka are also evident. Additionally, five Heinrich events (H2-H6), reflected by lower values of Rb/Sr and Ba/Sr, are also evident in the Beigou record ( Figure 10). In summary, the broad comparability among the records of Rb/Sr and Ba/Sr at Beigou, summer precipitation variations in the western CLP, the Chinese speleothem δ 18 O record, and NHSI, demonstrates that eolian deposits in the southern HSL still have the potential to record sub-orbital-scale EASM variations, as well as some features of millennial-scale climatic events, although the average sediment accumulation rate of the Beigou section (∼4.6 cm/ka) was much lower than that of typical sections in the western CLP (e.g., Beiguoyuan, with the rate of ∼40.6 ± 7.5 cm/ka) during the last glacial (Stevens et al., 2006).

Geochemical Characteristics of the Beigou Loess: Climatic and Provenance Implications
Based on the differential mobility of elements, weathering processes can be characterized sequentially as an early stage of Na and Ca loss, an intermediate stage of K loss, and a final stage of Si loss (Nesbitt et al., 1980). The relative enrichment of Na compared with K can be clearly evaluated using the K 2 O/Al 2 O 3 -Na 2 O/Al 2 O 3 diagram (Garrels & Mackenzie, 1971). The Na 2 O/Al 2 O 3 and K 2 O/Al 2 O 3 values at Beigou are 0. 075-0.1982 and 0.1942-0.2422, respectively, and these two ratios for the CLP (e.g., at Baicaoyuan, Fanshan, Jiuzhoutai, and Xifeng) have the ranges of 0.1158-0.1452 and 0.1891-0.2015, respectively (Figure 11a). Compared with the global average abundance for the upper continental crust (Taylor & McLennan, 1985), the readily weatherable Na and Ca are depleted by leaching at both Beigou and in typical loess sections on the CLP, and the loss of K is also evident (Figure 12a). The comparatively higher Na, Ca, and K content of the Beigou loess samples indicates that the chemical weathering processes in the southern margin of the HSL are characterized by the early weathering stage of Na and Ca removal. Compared with typical loess on the CLP, the Beigou loess-paleosol samples have lower Rb/Sr but higher Sr/Al ratios (Figure 11b), also suggesting weaker chemical weathering in the southern margin of the HSL.
UCC-normalized abundances of major and trace elements reveal a similar pattern between the Beigou loess and typical loess records from the CLP, except for the slightly lower concentrations of U, Fe, and Mg, and the higher concentrations of Ca and Sr (Figure 12a), reflecting very weak chemical weathering of the Beigou record. Bivariate plots of U/Pb versus Th/Pb and Ba/Rb versus Zr/Hf also show similar distribution between the Beigou and typical loess from the CLP (Figures 11c and 11d). The similarity of the UCC-normalized major and trace element abundances between Beigou and typical loess records from the CLP indicates that the eolian deposits at Beigou may have experienced multiple recycling and thorough mixing. The REE content of clastic rocks is controlled mainly by the lithologic composition of the source area (Mclennan, 1989;X. P. Yang et al., 2007), and thus it can be used to trace the provenance of eolian sediments (Gallet et al., 1998). The spatially uniform REE patterns between the loess deposits studied herein and those for the main body of the CLP (Figure 12b) further confirm that the loess source materials at Beigou have been thoroughly mixed prior to deposition. Geological evidence Figure 8. Representative magnetic and geochemical parameters for the Beigou section and comparison with a benthic δ 18 O stack (Lisiecki & Raymo, 2005). The orange and gray bars denote the intervals of relatively strong East Asian summer and winter monsoon, respectively (left). The blue and red arrows represent the trends of winter monsoon strengthening and summer monsoon weakening, respectively. has demonstrated that most of the modern dunes in the HSL are of very Late Pleistocene or even Holocene age (X. P. Yang et al., 2013), although the age of formation of the modern dunes in the HSL remains to be determined. The preservation of a complete loess-paleosol record spanning the last interglacial-glacial cycle on the modern southern limit of the HSL implies that loess accumulation was not significantly affected by proximal dune-fields, and the Beigou loess is primarily derived from relatively distant source areas. However, given the multiple and complex sources of Chinese loess, more detailed work is needed to constrain the exact source and transport dynamics of loess deposition in the southern HSL.

Conclusions
We have documented for the first time the preservation of complete loess-paleosol sequences spanning the last interglacial-glacial cycle on the present-day southern limit of the HSL, central-eastern Inner Mongolia. Our combined mineral magnetic and geochemical results demonstrate that: (a) during the last interglacial, magnetic enhancement is primarily controlled by precipitation-driven pedogenic processes, indicating that the front of the EASM extended to central-eastern Inner Mongolia; (b) during the last glacial, however, both SIRM and χ show positive correlations with the Zr/Rb ratio (a robust indicator of the winter monsoon), reflecting the strong control of wind vigor and/or significant dust inputs from proximal sources on magnetic properties dominated by lithogenic ferrimagnetic component; (c) the climatic conditions during the last interglacial-glacial  Wang et al., 2001) (e); reconstructed summer precipitation (f) and the grain-size record from the Yuanbao section in the western Chinese Loess Plateau (Rao et al., 2013) (g); and Northern Hemisphere July insolation (Berger & Loutre, 1991) (h). Red horizontal bars indicate sub-orbital-scale warm events, and blue bars denote Heinrich-like events identified in the records. transition were characterized by the rapid retreat of the EASM but the significant strengthening of the EAWM, indicating an anti-phased relationship between the EASM and EAWM; (d) higher proportions of high-coercivity magnetic minerals were formed during pedogenesis at the northern limit of the EASM, as a result of climatic conditions of relatively lower rainfall but higher evapotranspiration; (e) geochemical proxies of summer monsoon (e.g., Rb/Sr and Ba/Sr ratios) may faithfully track weak chemical weathering processes and fine-scale monsoon oscillations, and both ratios show broad comparability with summer precipitation variations in the western CLP, the Chinese speleothem δ 18 O record, and NHSI; and (f) the loess source materials developed in the southern margin of the HSL may have been thoroughly mixed prior to deposition. Our results highlight this "atypical" mineral magnetic record to be a threshold response to the EASM variations in central-eastern Inner Mongolia.  (Taylor & McLennan, 1985), Beigou loess (blue circles), Beigou paleosol (black circles), Fanshan (green circles), Jiuzhoutai (yellow circles), and Baicaoyuan (purple circles).

Data Availability Statement
Magnetic and geochemical data of the Beigou loess section in the present study can be archived online in a public repository (Sheng et al., 2023).  Sun and McDonough (1989).