Determining the climatic boundary between the Chinese loess and palaeosol: evidence from aeolian coarse-grained magnetite

Transcription

1 Geophys. J. Int. (24) 156, doi: /j X x Determining the climatic boundary between the Chinese loess and palaeosol: evidence from aeolian coarse-grained magnetite Qingsong Liu, 1 Subir K. Banerjee, 1 Michael J. Jackson, 1 Fahu Chen, 2 Yongxin Pan 3 and Rixiang Zhu 3 1 Institute for Rock Magnetism, Department of Geology and Geophysics, University of Minnesota, 55455, USA. liux272@tc.umn.edu 2 Center for Arid Environment and Paleoclimate Research, Department of Geography, University of Lanzhou, Lanzhou, Gansu, China 73 3 Paleomagnetism Lab, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, China 111 Accepted 23 September 2. Received 23 August 4; in original form 22 August 12 1 INTRODUCTION Sediments deposited on the Chinese Loess Plateau as accumulations of windblown silt are good media for recording changes in palaeoclimate. Detailed rock magnetic records can provide excellent proxies for such palaeoclimatic fluctuations. This was clearly shown by a striking correlation between the variations in the loess/palaeosol susceptibility and the marine oxygen isotope records (Heller & Liu 1984; Kukla et al. 1988). However, increasing evidence has shown that the variations in most rock magnetic parameters are controlled by multiple factors. For example, susceptibility may be different from site to site or at different times at a given site (Sun & Liu 2). A simplified view is that the Chinese loess/palaeosol sequences were controlled by both winter and summer monsoons (An et al. 1991). Loess was deposited SUMMARY This paper proposes a new method to distinguish between interglacial deposits and pedogenically-overprinted glacial loess based on the concentration variations of coarse-grained aeolian (magnetically pseudo-single domain and multidomain) magnetite. We apply the method to a sequence from the upper part of the loess unit L2 (marine isotope stage MIS 6) to the sub-loess unit S1L2 (MIS 5d) at the YuanBo (YB) section in the Chinese Loess Plateau. The method is based on the differences in low temperature properties between the coarse-grained (multidomain and pseudo-single domain) detrital magnetite and the pedogenic magnetic particles including superparamagnetic (SP) particles and relatively larger (>SP) maghemite particles. The former is characterized by a crystallographic Verwey transition around 12 K. In contrast, the magnetization of the latter continuously decreases in intensity with increasing temperatures. The method involves two steps: (1) calculating the first-derivative of the low temperature thermal demagnetization of the saturated isothermal remanent magnetization acquired at 2 K (LT-SIRM), which enhances the behaviour related to the Verwey transition, and (2) fitting a third-order polynomial background to the data between both 5 7 and 15 3 K, and then subtracting this background from the total derivative curves. The area under the background-corrected derivative curves represents the absolute intensity drop associated with the Verwey transition ( J TV ). This is caused by the aeolian coarse-grained magnetite, which is very sensitive to the changes in the intensity of the winter monsoon, and in turn related to the changes in palaeoclimate. The results show that the sharp drop of J TV at m corresponds to the climatic boundary from L2 (MIS 6) to S1S3 (MIS 5e), and not at 39.8 m as thought previously. Thus, the palaeosol deposits just below this boundary are in fact highly altered L2 materials instead of S1S3 accumulations. Key words: aeolian, Chinese loess, low-temperature properties, pedogenesis. during glacial periods when the winter monsoon, driven by cold air masses of the Siberian High, prevailed at the Chinese Loess Plateau. In contrast, during interglacial periods, the dominant summer monsoon brought warm and moist air masses and soils developed, and aeolian deposition continued with a lower depositional rate ( 2 cm ka 1 ) than that during the glacial period ( 2 3 cm ka 1 ). In fact, at any time, these two kinds of monsoons were competing processes, and only the relative contrasts between the loess and the palaeosol reflect the fluctuations of the palaeoclimate (Verosub et al. 1993; Fine et al. 1995). Pedogenesis is an intricate process also determined by multiple factors including climate, parent material, pedoturbation, water, heat and vegetation. (Johnson et al. 199). During the formation of soil, pedogenic processes altered not only the newly deposited interglacial material, but also the underlying glacial units generally GJI Seismology C 23 RAS 267

2 268 Q. Liu et al Interglacial deposition Alteration of glacial material Glacial material Climate boundary Susceptibility boundary Figure 1. Conceptual model for the formation of loess/palaeosol cycles. The number indicates different stage of the soil development. In stage 1, glacial material continuously deposited with a relatively higher deposition rate. In stage 2, a rapid palaeoclimatic change from glacial to interglacial occurred. Under such conditions, the newly deposited interglacial and the underlying glacial material remained in a same weathering zone and developed soils. In stage 3, the interglacial material goes on depositing with a relatively lower deposition rate. In stage 4, the weather gradually became colder again, soil processes diminished, then glacial material deposited again. The climatic boundary between the glacial and the interglacial had been obscured by the pedogenesis. up to meters just below the surface, because they remained within the same weathering zone. Consequently, the highly altered glacial material and the overlying interglacial material may have identical soil properties although they were originally deposited under very different conditions (Fig. 1). The ambiguity of the exact location of the glacial interglacial boundary is less important for the investigation of large-scale loess/palaeosol cycles. However, to successfully apply magnetic-property studies to fine-scale palaeoclimatic variations, and to establish a more accurate chronology of the loess sediments, and then to calculate the sedimentary accumulation rate, it is essential to accurately identify this climatic boundary, even though it has been highly obscured by pedogenesis. Despite extensive investigations of the properties of the Chinese loess and palaeosol sequences, an accurate method for determining the glacial and interglacial boundary has not yet been developed. This requires quantitative and sensitive means of distinguishing the highly altered loess from the overlying true interglacial deposits. However, this is very difficult due to their identical soil properties. Table 1. Parameter list for distinguishing loess and palaeosol. A proxy or tracer used to determine the glacial interglacial boundary must be: (1) of aeolian origin, (2) resistant to pedogenic processes, and (3) highly sensitive to palaeoclimatic changes. Common parameters used in loess palaeoclimatic studies are summarized in Table 1. Among these, most of the chemical and geochemical parameters (e.g. Ca, Rb, Sr, U, Ce, Fe 2+ ) are highly affected by leaching during pedogenesis (Gallet et al. 1996). In contrast, Neodymium isotope composition is immobile, but rather uniform in both loess and palaeosols, strongly suggesting a uniform source of the loess/palaeosol sequences (Gallet et al. 1996). 1 Be has also been used to unravel the palaeomagnetic fluctuations (Shen et al. 1992; Heller et al. 1993; Gu et al. 1996). These results show that variations in susceptibility and 1 Be composition exhibit similar patterns even though their relative enrichments between the loess and palaeosols are different. 1 Be is not supposed to be chemically mobilized by diagenetic processes. However, during the interglacials, 1 Be could have been leached down by precipitation when accumulation of the isotope exceeded the adsorbing capacity of the sediments. Quartz is highly resistant to pedogenesis (Porter & An 1995; An & Porter 1997; Sun et al. 2). Reports from Sun et al. (2) showed that the median grain size of quartz has the potential for reflecting the variations in the intensity of the winter monsoon. Unfortunately, its sensitivity is not high enough to locate the objective climatic boundary due to the high concentration of quartz in both loess and palaeosol samples. Unlike these parameters mentioned above, coarse-grained (pseudo-single domain/multiple domain) magnetite in the Chinese loess/palaeosols originates unambiguously from the aeolian source, and can survive during pedogenesis. In addition, its concentration is highly sensitive to the depositional processes. Here arises a key question: how can we accurately determine the contribution of the coarse magnetite masked by the strong pedogenicallyenhanced background? Verosub et al. (1993) suggested that the citrate bicarbonate dithionite (CBD) method could effectively separate the information of detrital or lithogenic magnetite from the signal of pedogenic components. However, subsequent studies showed that the CBD-soluble magnetic minerals also included those relatively coarser (<14 µm) magnetic particles of detrital origin (Hunt et al. 1995; Sun et al. 1995). Recently, Van Ooschot et al. (22) reported that the acid-ammonium-oxalate/ferrous-iron (AAO Fe 2+ ) extraction technique can selectively remove very fine-grained magnetite and maghemite from coarse-grained magnetite. However, Parameter Description Reference SUS, ARM, SIRM, χ fd Highly related to the pedogenesis Heller & Liu (1984), Heller et al. (1993), Maher (1998), and references therein Quartz Resistant to pedogenesis, but low Porter & An (1995), An & Porter (1997) and sensitivity Sun & Liu (2) 1 Be Similar pattern to susceptibility, Shen et al. (1992), Beer et al. (1993), Heller et al. (1993) ambiguities exist for its origin Thompson & Maher (1995) and Gu et al. (1996) Ca, Rb, Sr, U, Ce Strongly fractionated by pedogenesis Gallet et al. (1996) Nd Unaffected by pedogenesis and can Gallet et al. (1996) indicate loess origin, but no apparent difference between loess and palaeosol CaCO 3 Strongly affected by the leaching process. Chen et al. (1999), Liu et al. (1999a) It is a function of both precipitation and time Median grain size of particles Highly altered by the pedogenesis Chen et al. (1999) Fed/Fet Highly related to pedogenesis Guo et al. (1996, 1998)

3 Determining the climatic boundary between loess and palaeosol 269 their results did not show sensitivity to such a palaeoclimatic boundary. Banerjee et al. (1993) took an alternative approach by taking advantage of the distinct low-temperature properties of multidomain (MD) magnetite and the extremely fine (<2 nm) superparamagnetic (SP) magnetite. In this paper, we propose an improved rock magnetic method of quantitatively separating the coarse-grained (MD and pseudo-single domain PSD) magnetite signal against the strong maghemite and SP magnetite background based on Banerjee et al. s early efforts. We then apply the new method to the Yuanbo (YB) section to test its validity. We aim to address the difference between the low temperature features of the coarse-grained magnetite in loess and palaeosol. Based on the spatial variations in the concentration of the coarsegrained magnetite, we try to clearly distinguish the highly altered glacial material from the overlying interglacial alterations. 2 SAMPLING AND MEASUREMENTS The Yuanbo (YB) section is located in the northwestern margin of the Chinese Loess Plateau on the fourth terrace of the Daxia River in the Linxia Basin. The mean annual precipitation is 5 mm. The palaeosol unit S1 (MIS 5) consists of three well-developed sub-palaeosol units (S1S1, S1S2, and S1S3) and two interbedded sub-loess layers (S1L1 and S1L2). Among them, the best-developed S1S3, corresponding to MIS 5e, is characterized by a reddish-brown colour. Bag samples were collected at 2 cm intervals and shipped to the University of Minnesota. This study focused on a sandwichlike sequence including the upper part of L2 (MIS 6), sub-palaeosol S1S3 (MIS 5e) and the overlying sub-loess unit S1L2 (MIS 5d). For a detailed description of the site s pedostratigraphy and chronology refer to the work of Chen et al. (1999). Thermal demagnetization of the low temperature saturation isothermal remanent magnetization in 2.5 T (LT-SIRM) acquired at 2 K was performed by a Quantum Design Magnetic Properties Measurement System susceptometer (MPMS). Remanence was measured at temperature increments of 2 K, with an error of ±1K. To separate the intensity drop at the Verwey transition (T V, 12 K) produced by magnetite from the pedogenically-driven magnetic background, two steps were involved: (1) The first derivatives of the mass normalized intensity of LT-SIRM (Fig. 2a) were calculated to enhance the features of the Verwey transition (Fig. 2b); (2) The background (dashed line in Fig. 2b) was removed by subtracting the third-order polynomial fit between 5 7 K and K to avoid effects of the Verwey transitions. Root mean square deviations (σ rms ) were calculated to evaluate the goodness of the fits. The area under the residual derivative curves, represents the absolute intensity drop at T v ( J TV ). The remanence data between 2 and 5 K were not included for fitting the background for two reasons. First, the remanences decrease sharply between 2 K and 5 K due to unblocking of the smaller SP particles, and required higher order polynomial fits. Second, there can be large changes of remanence in this temperature interval for partially oxidized PSD magnetite (Özdemir & Dunlop 1993). Hysteresis loops were measured with an automated Princeton vibrating sample magnetometer (VSM). M fr (Liu et al. 22) was also measured with the VSM to investigate variations in the concentration of antiferromagnetic minerals (e.g. hematite/goethite). M fr was obtained by hysteretic demagnetization of a SIRM with a maximum demagnetization field of.3 T. M fr is sensitive to the absolute concentration of the antiferromagnetic minerals in samples. Detailed information of the M fr method was described in Liu et al. (22). 3 RESULTS Low-temperature thermal demagnetization of SIRM (LT-SIRM) curves for four selected samples from L2 to S1L2 at 4.64, 39.64, and m are shown in Fig. 2(a). The gradual decay of remanence between 5 and 3 K reflects the unblocking of SP particles as well as the smoothly decreasing magnetic moments of maghemite and goethite. Sharp transitions around 12 K are present in the remanence data for all samples, indicating large (>SP), nearly stoichiometric magnetite not only in the loess but also in the palaeosol units. The characteristics of the transition were highly enhanced in the first-order derivative of the LT-SIRM curves (Fig. 2c). After subtracting the continuous background (dashed lined in Fig. 2c), the corrected derivatives of loess and palaeosol samples show distinct patterns. The loess samples have a broad derivative peak between 11 K and 12 K, whereas the palaeosol samples are characterized by a dominant peak at 12 K, which corresponds to nearly stoichiometric magnetite. We note that the 11 K peak gradually diminishes from loess to palaeosol. The σ rms between the third-order polynomial fits and the raw derivative data are small and less than Am 2 kg 1 K 1. Fig. 2(b) shows that the third-order polynomial is good enough to successfully fit the background. Depicted in Fig. 3 are comprehensive results including both magnetic and nonmagnetic investigations. Generally, the relatively higher susceptibility and SIRM and lower concentration of CaCO 3 correspond to the palaeosol maximum S1S3 (MIS 5e). The sharp drop of the median diameter of particles at 39.8 m had been regarded as the boundary between the glacial period L2 (MIS 6) and the interglacial period S1S3 (MIS 5e) by Chen et al. (1999). However, the new parameter J TV (absolute intensity drop at Tv) shows a constant value between 4.4 and 39.4 m, which cuts through this suggested boundary, and suddenly drops at 39.4 m from Am 2 kg 1 to Am 2 kg 1 within 1 cm. At m, susceptibility and J TV show a consistent boundary between S1S3 and S1L2. M fr is sensitive to the variations in the concentration of antiferromagnetic minerals (hematite/goethite). Fig. 3(f) shows that the M fr changes from a convex to a concave shape on going from above to below 39.4 m, where the maximum of M fr exactly corresponds to the sharp boundary suggested by J TV. Depth plots of the derivative of LT-SIRM at 12 K ( dj/dt 12K ) are shown in Fig. 3(h). The derivative at 12 K is an indicator of the concentration of stoichiometric magnetite. Just below the J TV drop at m, it slightly increases, and then shows a mirror relationship compared to susceptibility. Fig. 4 shows a comparison of the background-corrected derivatives for four selected samples. Two samples (4.64 and 38.4 m) represent the loess units L2 and S1L2, respectively. The other two samples, at and 39.3 m, correspond to the zone of sharp drop of J TV. The normalized derivatives revealed that the loess and palaeosol samples show a self-consistent pattern even though their absolute intensity drop at Tv, calculated from the area under the derivative curves between 8 and 15 K, are different. 4 DISCUSSION 4.1 Framework for interpretation of J Tv For the loess/palaeosol samples, the smooth background in the LTD-SIRM demagnetization curves could be caused by either the

4 27 Q. Liu et al. J (1 2 Am 2 kg 1 ) (a) 4.64 m m m m σrms (1 2 Am 2 kg 1 ) (b) -dj/dt (1 2 Am 2 kg 1 k 1 ) Temperature (K) (c) Temperature (K) -dj/dt (1 2 Am 2 kg 1 k 1 ) Order (d) Temperature (K) Figure 2. (a) Low temperature thermal demagnetization of SIRM, (b) Root mean square deviations (σ rms ) of the difference between the first-derivative of LT-SIRM and the corresponding polynomial fits by using data between 5 7 K and 15 3 K as a function of polynomial order; (c) First-order derivative of LT-SIRM. The background (dashed line) was fitted by the third-order polynomial. Numbers in (c) are σ rms for each fit; (d) Background-corrected derivative of LT-SIRM. Depth (m) (a) χ fd (%) CaCO3 (% ) Median diameter (µm) (c) SIRM (1 3 Am 2 kg 1 ) 7 14 χ (1-3 SI kg 1 ) (e) 5 1 M fr J Tv (1 3 Am 2 kg 1 ) Reduction Zone Oxidation Zone (g) (1 3 Am 2 kg 1 ) -dj/dt 12K (1 3 Am 2 kg 1 K 1 ) Altered L2 S1S3 S1L2 L2 (i) S1S3 S1L2 L2 (j) Chen et al (b) (d) (f) (h) Figure 3. The temporal variations in: (a) the percent frequent-dependent susceptibility (χ fd per cent). (b) the median grain diameter of particle size, (c) CaCO 3 concentration, (d) low field susceptibility, (e) SIRM, (f) hematite proxy M fr, (g) Intensity drop at Tv, (h) the first derivative of the intensity at 12 K, (i) new lithology interpreted in this study, (j) interpretation from Chen et al. (1999). Data for (a), (b), and (c) come from Chen et al. (1999). The horizontal dashed lines indicate the possible boundaries, and the vertical dashed lines represent the background values of less altered loess. The horizontal grey bars stand for the oxidation zone.

5 Determining the climatic boundary between loess and palaeosol 271 -dj/dt (1 2 Am 2 kg 1 K 1 ) (a) 4.64 m m 39.3 m 38.4 m Normalzied -dj/dt (b) Temperature (K ) Temperature (K) Figure 4. (a) Background-corrected derivatives of LT-SIRM for selected samples at 4.64, 39.44, 39.3, and 38.4 m. (b) Normalized data from (a) by the background-corrected derivative at 12 K. unblocking of SP grains, or the linear decrease of the magnetic moment of relatively coarse-grained (>SP) maghemite. The derivative of LT-SIRM serves two purposes: to enhance the local Verwey transition anomaly, and to transform the step-like Verwey transition to a spike-like function and make it feasible to fit the background by utilizing the data at temperatures both higher (145 3 K) and lower (5 7 K) than the Verwey transition. In this study, thirdorder polynomials adequately fit the background, as shown by the RMS deviation (σ rms ) between the raw values and the fitted data (Fig. 2b). SD particles in the Chinese loess/palaeosol sequences had been recognized as maghemite by numerous studies (Evans & Heller 1994; Eyre & Shaw 1994; Sun et al. 1995; Liu et al. 1999b). With decreasing grain-size, the influence of surficial oxidation becomes stronger and stronger. For SP/SD magnetic particles, the most stable phase is maghemite (Ayyub et al. 1988; Eyre & Shaw 1994), which is also the main factor that enhances the susceptibility of the palaeosol (Eyre & Shaw 1994; Sun et al. 1995). Those finer maghemite particles (SP/SD) do not contribute to the Verwey transition. Even though the more magnetite-like SD grains may survive in samples, the corresponding Verwey transition must have been strongly suppressed due to its high oxidation degree (Aragon et al. 1993; Honig 1995; Dunlop & Özdemir 1997), and have minor contributions to the Verwey transition. Therefore, we believe that J TV is determined solely by PSD/MD magnetite. The value of J Tv is affected mainly by three factors: concentration, grain-size, and non-stoichiometric degree of magnetite in samples. Fig. 3 indicates that the normalized Verwey transition for two samples at and 39.3 m (which defined the sharp drop of J Tv ) shows a consistent pattern even though the latter has a much lower J TV ( Am 2 kg 1 ) compared to the former ( Am 2 kg 1 ). Both of them have a unique and dominant derivative peak at 12 K, indicating that pedogenesis altered the 11 K material without strongly affecting the nearly stoichiometric MD magnetites responsible for the 12 K peak. We also notice that pedogenesis enhanced the 12 K transition just below 39.44m. Thus, the significant drop of J Tv can not be caused by pedogenesis. This can also be supported by a similar pedogenic environment suggested by the similar susceptibility and χ fd around m. Given the same pedogenic environment and the consistent pattern of the 12 K Verwey transitions, we believe that the relatively sharp drop of J Tv at m is entirely due to a decrease in the concentration of the nearly stoichiometric MD magnetite. The same reasoning can also be applied to the comparison of the two characteristic loess samples at 4.64 m (L2) and 38.4 m (S1L2) (Fig. 4). They also show identical bimodal patterns suggesting that similar magnetic minerals cause their Verwey transitions. Thus we also interpret the difference between the J TV of these two loess samples in terms of the changes in the concentrations of the coarse magnetite particles. Compared to the bottom boundary of S1S3, its upper boundary seems more straightforward to interpret because the soil developed downwards during pedogenesis. When the temperature gradually decreases, the pedogenesis has less and less effects on the newly deposited material. Fig. 3 shows that there is a consistent boundary at m suggested by nearly all parameters. We, therefore, confidently regard this as the boundary between S1S3 (MIS 5e) and S1L2 (MIS 5d). Unlike SP magnetite/maghemite particles, the MD/PSD magnetite particles are unambiguously of aeolian origin, and their concentration is strongly related to the strength of the winter monsoon. As the climate changed, changes in the location of the source area and the winter monsoon patterns could have caused the variations in the magnetic minerals incorporated in the loess material. Generally, the higher the strength of the winter monsoon, the more MD magnetite particles can be transported to the loess plateau. In turn, the variations in the winter monsoon reflect the changes in the palaeoclimate. For example, a general consensus is that the interglacial and glacial periods are characterized by relatively weaker and stronger winter monsoon, respectively (An et al. 1991; Chen et al. 1999). Thus, the systematic variations in the concentration of MD/PSD magnetite revealed by J Tv, the strength of the winter monsoon, and the palaeoclimate are highly linked. We conclude that the sharp drop of J TV at m corresponds to the palaeoclimate boundary from the glacial period (MIS 6) to the interglacial period (MIS 5e). Therefore, the material between and 4.3 m is in fact a pedogenically-overprinted glacial deposit (L2 with high deposition rate) instead of an interglacial accumulation (S1S3 with low deposition rate). 4.2 Effects of non-stoichiometry The presence of two peaks in the first derivative of LT-SIRM curves of the loess indicates a mixture of two kinds of magnetite with different non-stoichiometric degrees. The 11 K Verwey transition peak was caused by non-stoichiometric magnetite (partially

6 272 Q. Liu et al. oxidized and/or cation-substituted), and the 12 K transition is due to more stoichiometric magnetite. Under the same oxidation environment, the larger size (MD) magnetite has a higher potential ability to keep its overall stoichiometric properties than the relatively smaller size magnetite. Therefore we postulate that the magnetite particles that produced the 11 K transition are smaller than those causing the 12 K transition. After alteration by pedogenesis, the morphology of the transition is systematically changed. The effects of titanium substitution on T v have been well documented (Kakol et al. 1992, 1994; Moskowitz et al. 1998). These earlier results revealed that the Verwey transition can be significantly suppressed to a lower temperature below 8 K with only minor Ti-substitution (x <.4). Our results show that the Verwey transition is limited within K indicating that titanium variation is less important in our samples, while, low temperature oxidation is the dominant factor that affects the transition of loess samples. Fig. 2(c) shows that the 11 K T v gradually diminishes with increasing pedogenic degree. This can be equivalently caused by: 1) reduction of the 11 K non-stoichiometric magnetite to more stoichiometric magnetite, which would enhance the amplitude of the 12 K Verwey transition; or 2) further oxidation of the 11 K non-stoichiometric magnetite to maghemite. Samples below 39.8 m (Fig. 3h) show a significant decrease of J TV from Am 2 kg 1 for the raw loess to Am 2 kg 1 for the weakly altered loess, but display stable J LTD and dj/dt 12K. This indicates that 11 K Verwey transition had been significantly depressed but has had no effects on the 12 K transition. We, therefore, believe that this is caused by the second mechanism (further oxidation). In contrast, between 39.5 and 39.8 m, the 12 K transition was gradually enhanced indicating more stoichiometric MD magnetite particles had been introduced by the pedogenesis. This can be successfully explained by the first case (reduction). Thus we find that the oxidation and reduction processes prevailed at different zones. 4.3 Other geological indications The successful identification of the post-depositionally altered L2 (MIS 6), which was originally classified as the sub-palaeosol S1S3 (MIS 5e), provides a good opportunity for investigating: (1) the chemical physical processes during pedogenesis in the past; (2) the validity of the mean grain size as a proxy of the winter monsoon intensity; (3) the ability in loess deposits of faithfully recording rapid palaeoclimatic changes; and (4) a more accurate calculation of the deposition rate of S1S3 based on the new stratigraphy proposed in this study. Pedogenesis can be understood better if the iron cycle in the loess and palaeosol can be sufficiently well understood. Processes of iron mineral authigenesis, diagenesis, transport and iron dissolution, as well as interaction with weathering inputs of iron, produce vertical variations in soil magnetic properties. Iron is released by weathering from primary Fe-bearing minerals by oxidation at mineral surfaces. The Fe 2+ ions produced may be incorporated as Fe 3+ or Fe 2+ into an iron oxide phase in situ (Maher 1998). During pedogenesis in the loess/palaeosols, plant debris and iron-reducing bacteria may also play a key role in the initial formation of ultrafine-grained magnetite. Because of the small size of many of the precipitated crystals, surface oxidation of magnetite to hematite or maghemite is likely to occur. Conversely, in a reducing environment, hematite can also be changed into magnetite (Maher 1998). The reducing process can also be caused by the burning of 1 2 per cent organic matter in the loess and palaeosol (Kletetschka & Banerjee 1995). Fig. 3(f) shows that the hematite-proxy M fr changes from Am 2 kg 1 for the less altered loess unit L2 to more than Am 2 kg 1 for the sample at the climatic boundary. Assuming that the hematite concentration in the original glacial deposition was uniform, the newly formed hematite must have originated from the pedogenesis. Therefore, about 2 3 times the amount of hematite was produced by the pedogenesis. It is also interesting to notice that the variations of the hematite concentration show an asymmetric pattern, concave and convex below and above the climatic boundary at 39.4 m, respectively. Our interpretation is that the soil development gradually progressed down into the glacial loess during the interglacial period, while the newly deposited aeolian material was quickly altered to soil, and hematite could have been simultaneously formed. The saw-tooth pattern of the hematite distribution could be a new character to locate the climatic boundary. The increase of M fr and SIRM occurring in the reduction zone supports the idea that neoformation of both antiferromagnetic and ferrimagnetic minerals occurred there. However, it is hard to determine whether there was a transformation relationship between hematite and ferrimagnetic minerals. The reason is that the formation of hematite and the transformation from hematite to magnetite, in turn oxidized to maghemite, are competing processes at the reduction zone. In fact, only a portion of the hematite can be transformed, and has minor effects on its concentration. The second important implication of this study is that the median size of the particles is not a good proxy to indicate the variations in the intensity of the winter monsoon, which is essential for us to investigate the whole monsoon system. By assuming that the particle-size distribution may have not been altered by pedogenesis, the median diameters of the particles have been used as a proxy of the winter monsoon intensity (Chen et al. 1999). However, our current studies indicate that this is not a reliable indicator. The median diameter of the altered L2 material can be decreased to the same level as the overlying interglacial material solely by pedogenesis. During pedogenesis, the grain size can be decreased by either physical disintegration or chemical alteration. For example, silt-size feldspars and other unstable silicates can be changed to clays, carbonates are dissolved, and even quartz may also suffer some size reduction (Pye 1987). Undoubtedly, pedogenesis was strong during MIS 5e, and could have greatly altered the underlying L2 material. Fig. 3(b) clearly shows that the median particle size decreased from 15 µm to 6 µm within 4 cm right below the tic boundary as a result of pedogenesis. It is our contention that the median size is not directly related to the strength of the winter monsoon because the decrease in the grain size can be alternatively caused by pedogenesis. Conversely, the changes in J Tv can be a good proxy to represent the intensity of the winter monsoon even though no quantitative relationship between J Tv and the winter monsoon intensity has been established yet. Nevertheless, Fig. 3(g) indicates that the winter monsoon had a lower intensity during S1S3 (MIS 5e) than during L2, and the intensity during S1L2 (MIS 5d) was between those during S1S3 and L2. The third implication of the study is that the Chinese loess/palaeosol sequences have the potential for recording rapid palaeoclimatic changes. The climatic variations model for the glacial/interglacial transition is an important aspect of understanding future climatic changes. A growing body of evidence shows

7 Determining the climatic boundary between loess and palaeosol 273 that there were a number of climatic fluctuations occurring in a fairly short period of time during the last glacial period (Wang et al. 21). The shortest time scale could be less than 1 yr. The climate events during MIS 5a 5e have also been reported by Hammer et al. (1998). Unfortunately, the rapid palaeoclimatic changes are absent from the Chinese loess/palaeosol records (e.g. susceptibility) due to the smoothing effects of pedogenesis. However, the new parameters proposed in this study ( J TV, and dj/dt 12K ) are highly sensitive to the strength of the winter monsoon. The sharp change of J TV indicates that Chinese loess/palaeosol sequences have the potential of recording those rapid climatic events. Finally, the identification of the altered glacial material from the truly interglacial material can also help us to accurately calculate the average depositional rate during the interglacial period. As an example, the thickness of the S1S3 at YB section is about 56 cm. Assuming S1S3 lasted 1 ky, the average depositional rate during MIS 5e at YB section is 5.6 cm ky 1. 5 CONCLUSIONS Due to pedogenesis, the original rapid climatic transition from the glacial period (MIS 6) and the interglacial period (MIS 5e) has been obscured and replaced by a transition zone of about one-meterthick. However, the new parameter ( J TV ) proposed in this study can avoid those pedogenically-disturbing factors. Based on the discussion above, the main conclusions are: (1) the 12 K Verwey transition of the loess/palaeosol samples is caused by coarse-grained (MD/PSD) stoichiometric magnetite of aeolian origin; (2) Coarse aeolian magnetite suffered alteration during pedogenesis, shown by the gradually diminishing of the 11 K Verwey transition, which is produced by non-stoichiometric magnetite particles. Nevertheless, this will not cause any ambiguities when assigning the variations in the concentration of the coarse aeolian magnetite to the changes in the intensity of the winter monsoon. The sharp drop of J Tv at 34.4 m of YB profile most reasonably correspond to the rapid decrease of the intensity of the winter monsoon, which in turn indicates the climatic boundary between L2 (MIS 6) and S1S3 (MIS 5e). The identification of the highly altered glacial material from the truly interglacial material allows us to comprehensively interpret the systematic changes in both magnetic and non-magnetic parameters during pedogenic alteration. However, additional work is necessary to work on the modern soil profile (S) at the Chinese Loess Plateau to further check the validity of the new parameters. ACKNOWLEDGMENTS We are grateful to Professor E. A. Nater and Professor L. Edwards and X. F. Wang for many stimulating discussions, and to F. Lagroix and A. Eller for revising the manuscript. We also thank B. Carter-Stiglitz and Dr C. L. Deng for collecting samples. We thank F. Heller for helpful comments. This work is supported by NSF Grant No. EAR 3421 and EAR/IF All the measurements were carried on in the Institute for Rock Magnetism, which is funded by the Keck Foundation, the National Science Foundation, and the University of Minnesota. This is IRM contribution #22. REFERENCES An, Z.S. & Porter, S.C., Millennial-scale climatic oscillations during the last interglacial in central China, Geology, 25, 63. An, Z.S., Kukla G.J., Porter, S.C. & Xiao, J., Magnetic susceptibility evidence of monsoon variation on the Loess Plateau of central China during the last 13 years, Quat. Res., 36, Aragon, R., Gehring, P.M. & Shapiro, S.M., Stoichiometric, percolation, and Verwey ordering in magnetite, Phys. Rev. Lett., 7, Ayyub, P., Multani, M., Barma, M., Palkar, V.R. & Vijayaraghavan R., Size-induced structural phase transitions and hyperfine properties and microcrystalline Fe 2 O 3, J. Phys., 21, Banerjee, S.K., Hunt, C.P. & Liu, X-M., Separation of local signals from the regional paleomonsoon record of the Chinese loess plateau: A rock-magnetic approach, Geophys. Res. Lett., 2, Beer, J., Shen, C.D., Heller, F., Liu, T.S. & Kubik, P., Be and magnetic susceptibility in Chinese loess, Geophys. Res. Lett., 2, Chen, F.H., Bloemendal, J., Feng, Z.D., Wang, J.M., Parker, E. & Guo, Z.T., East Asian monsoon variations during Oxygen Isotope Stage 5: evidence from the northwestern margin of the Chinese loess plateau, Quat. Sci. Rev., 18, Dunlop, D.J. & Özdemir, Ö., Rock Magnetism: Fundamentals and Frontiers, Cambridge University Press, Cambridge. Evans, M.E. & Heller, F., Magnetic enhancement and palaeoclimate: study of a loess/palaeosol couplet across the loess plateau of China, Geophys. J. Int., 17, Eyre, J.K. & Shaw, J., Magnetic enhancement of Chinese loess the role of γ Fe 2 O 3?, Geophys. J. Int., 117, Fine, P., Verosub, K.L. & Singer, M.J., Pedogenic and lithogenic contributions to the magnetic susceptibility record of the Chinese loess/palaeosol sequence, Geophys. J. Int., 122, Gallet, S., Jahn, B.M. & Torii, M., Geochemical characterization of the LuoChuan loess-palaeosol sequences, China, and paleoclimatic implication, Chem. Geol., 133, Gu, Z.Y., Lal, D., Liu, T.S., Southon, J., Caffee, M.W., Guo, Z.T. & Chen, M.Y., Five million year 1 Be record in Chinese loess and red-clay: climate and weathering relationships, Earth planet. Sci. Lett., 144, Guo, Z.T., Liu, T., Guiot, J., Wu, N., Lü, H., Han, J. & Gu, Z., High frequency pulse of East Asian monsoon climate in the last two glaciations: link with the North Atlantic, Clim. Dynam., 12, Guo, Z.T. et al., Climate extremes in loess of China coupled with the strength of deep-water formation in the North Atlantic, Global Planetary Change, 18, Hammer, C., Mayewski, P.A, Peel, D. & Stuiver, M., Preface, Greenland Summit Ice Cores, J. geophys. Res., 12, Heller, F. & Liu, T.S., Magnetism of Chinese loess deposits, Geophys. J. R. astr. Soc., 77, Heller, F., Shen, C.D., Beer, J., Liu, X.M., Liu, T.S., Bronger, A., Suter, M. & Bonani, G., Quantitative estimates of pedogenic ferromagnetic mineral formation in Chinese loess and paleoclimatic implications, Earth planet. Sci. Lett., 114, Honig, J.M., Analysis of the Verwey transition in magnetite, J. Alloys Compounds, 229, Hunt, C.P., Banerjee, S.K., Han, J.M., Solheid, P.A., Oches, E., Sun, W.W. & Liu, T.S., Rock-Magnetic proxies of climate change in the loess paleosol sequences of the western loess plateau of China, Geophys. J. Int., 123, Johnson, D.L., Keller, E.A. & Rockwell, T.K., 199. Dynamic pedogenesis: New views on some key soil concepts, and a model for interpreting quaternary soils, Quat. Res., 33, Kakol, Z., Sabol, J., Sticker, J. & Honig, J.M., Effects of low-level titanium(iv) doping on the resistivity of magnetite near th Verwey transition, Phys. Rev., 46, Kakol, Z., Sabol, J., Sticker, J., Kozkowski, A. & Honig, J.M., Influence of titanium doping on the magnetocrystalline anistropy of magnetite, Phys. Rev., 49, Kletetschka, G. & Banerjee, S.K., Magnetic stratigraphy of Chinese loess as a record of natural fires, Geophys. Res. Lett., 22, Kukla, G., Heller, F., Liu, X.M., Xu, T.C., Liu, T.S. & An Z.S, Pleistocene climates in China dated by magnetic susceptibility, Geology, 16,

8 274 Q. Liu et al. Liu, C.Q., Zhang, J. & Li, C.L., 1999a. Variations in CaCO 3 content and Sr isotopic composition of loess and records of paleoclimatic fluctuations, Chin. Sci. Bullet., 44, Liu, X.M., Hesse, P. & Rolph, T., 1999b. Origin of maghemite in Chinese loess deposits: aeolian or pedogenic?, Phys. Earth planet. Inter., 112, Liu, Q.S., Banerjee, S.K., Jackson, M.J., Zhu, R.X. & Pan, Y.X., 22. A new method in mineral magnetism for the separation of weak antiferromagnetic signal from a strong ferrimagnetic background, Geophys. Res. Lett., 29(12) 1565, doi: 1.129/22GL Maher, B.A., Magnetic properties of modern soils and Quaternary loessic paleosols: paleoclimatic implications, Paleogeog. Paleoclimat. Paleoecol., 137, Moskowitz, B.M., Jackson, M.J. & Kissel, C., Low-temperature magnetic behavior of titanomagnetites, Earth planet. Sci. Lett., 157, Özdemir, Ö. & Dunlop, D.J., The effect of oxidation on the Verwey transition in magnetite, Geophys. Res. Lett., 2, Pye, K., Aeolian Dust and Dust Deposits, Academic Press, London. Porter, S.C. & An, Z.C., Correlation between climate events in the North Atlantic and China during the last glaciation, Nature, 375, 35. Shen, C.D., Beer, J., Liu, T.S., Oeschger, H., Bonani, G., Suter, M. & Wölfli, W., Be in Chinese loess, Earth planet. Sci. Lett., 19, Sun, J.M. & Liu, T.S., 2. Multiple origins and interpretations of the magnetic susceptibility signal in Chinese wind-blown sediments, Earth planet. Sci. Lett., 18, Sun, W.W., Banerjee, S.K. & Hunt, C.P., The role of maghemite in the enhancement of magnetic signal in the Chinese loess-palaeosol sequence: An extensive rock magnetic study combined with citrate-bicarbonatedithionite treatment, Earth planet. Sci. Lett., 133, Sun, Y.B., Lu, H.Y. & An, Z.S., 2. Grain size distribution of quartz isolated from Chinese loess/palaeosol, Chin. Sci. Bullet., 45, Thompson, R. & Maher, B.A., Age models, sediment fluxes and palaeoclimatic reconstructions for the Chinese loess and palaeosol sequences, Geophy. J. Int., 123, Van Ooschot, I.H.M., Dekkers M.J. & Havlicek P., 22. Selective dissolution of magnetic iron oxides with the acid-ammonium-oxalate/ferrousiron extraction technique-ii. Natural loess and palaeosol samples, Geophys. J. Int., 149, Verosub, K.L., Fine, P., Singer, M.J. & TenPas, J., Pedogenesis and paleoclimate: Interpretation of the magnetic susceptibility record of Chinese loess-palaeosol sequences, Geology, 21, Wang, Y.J., Chen, H., Edwards, R.L., An, Z.S., Wu, J.Y., Shen, C.C. & Dorale, J.A., 21. A high-resolution absolute-dated late Pleistocene monsoon record from Hulu cave, China, Science, 294,