Sedimentology (2004) 51, 77 93

Transcription

1 Sedimentology (2004) 51, doi: /j x Comparison of particle size characteristics of the Tertiary red clay and Pleistocene loess in the Chinese Loess Plateau: implications for origin and sources of the red clay S. L. YANG and Z. L. DING Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing , P. R. China ( ABSTRACT The origin of the Tertiary red clay underlying the Pleistocene loess in the Chinese Loess Plateau remains controversial, although several lines of evidence have suggested a wind-blown origin. This study examines the particle-size parameters of the late Miocene and Pliocene red clay by comparing it with those of the late Pleistocene loess. The particle-size distribution of a total of loess and 6394 red clay samples taken from 12 loess sections along a north south transect and two red clay sequences at Lingtai and Jingchuan was systematically analysed. The median grain size, skewness and kurtosis of the late Pleistocene loess all show a systematic southward change and are principally influenced by distance from source region. The spatial and temporal differentiation of dust deposits is expressed in a skewness kurtosis median grain size ternary diagram, from which the distance to the source region can be inferred. The particle-size characteristics of the Tertiary red clay sediments are very similar to those of the palaeosols within the late Pleistocene loess deposits, suggesting an aeolian origin for the red clay. Based on the comparison of red clay and loess in the ternary diagrams, it is inferred that the source sink distance was greater in the Neogene than in the last and penultimate interglacials, and that the dust source region in north-western China underwent a progressive expansion during the period from at least 7Æ0 Ma to the present. Keywords Aeolian deposits, Chinese loess, grain size, kurtosis, red clay, skewness. INTRODUCTION The loess deposits of the Chinese Loess Plateau were derived from the arid and semi-arid regions of north-western China, having been transported by the East Asia winter monsoon and the westerlies (Liu, 1985). The loess palaeosol sequences provide a quasi-continuous record of regional and global climatic changes during the past 2Æ6 my (Heller & Liu, 1982; Liu, 1985; Kukla, 1987; Kukla & An, 1989; Rutter et al., 1990; Ding et al., 1994). In general, such complete sequences of loess occur in the central and southern parts of the Loess Plateau, have a thickness ranging from 130 to 200 m and consist of over 30 loess soil couplets (Liu, 1985; Kukla & An, 1989; Rutter et al., 1991; Ding et al., 1994, 1999a, 2001a). Beneath the oldest loess unit (L33), there is a set of reddish clay silt deposits with a thickness of about m, informally known as the red clay formation (Liu, 1985). According to Zhu & Ding (1994), this formation has a geographical distribution similar to that of the Pleistocene loess in the Loess Plateau. Palaeomagnetic measurements at different sites on the Loess Plateau indicate that the contact between the red clay and the overlying loess is around the Matuyama Gauss reversal boundary (Heller & Liu, 1982; Liu et al., 1988a; Kukla & An, 1989; Rutter et al., 1990; Zheng et al., 1992; Sun et al., 1998a,b; Ding et al., 1998a,b, 2001a). Ó 2004 International Association of Sedimentologists 77

2 78 S. L. Yang & Z. L. Ding In recent years, the red clay sequence has attracted much attention (Liu et al., 1988b; Zheng et al., 1992; Ding et al., 1998a,b, 1999a, 2001a,b; Sun et al., 1998a,b; Guo et al., 2001, 2002; Lu et al., 2001; Qiang et al., 2001). One important conclusion reached after investigation of the red clay is that it may be wind blown in origin; this opinion is based on anisotropy of magnetic susceptibility (Liu et al., 1988b), field characteristics (Zheng et al., 1992; Sun et al., 1998a,b), grain-size distributions (Ding et al., 1998a; Lu et al., 2001) and geochemical concentrations (Ding et al., 1998a, 2001b). Magnetostratigraphic and pedostratigraphic studies of two thick loess red clay sequences at Lingtai and Jingchuan (Fig. 1) have suggested that the relatively continuous aeolian record in the Chinese Loess Plateau may well extend back in time from 2Æ6 to 7Æ0 7Æ7 Ma (Sun et al., 1998a; Ding et al., 1998b, 2001a). However, debate continues on the origin of the red clay formation (Mo & Derbyshire, 1991; Zhang & Xue, 1996; Guo et al., 2001). Some authors suspected that hydraulic factors may have played a significant transportational role, especially in the formation of the lower part of this sequence (Guo et al., 2001). To date, the most important evidence for a windblown origin comes from comparison of grainsize data of the red clay and the overlying loess (Ding et al., 1998a; Lu et al., 2001), although several aspects of these data require further analysis. A striking difference between the loess and the red clay lies in the fact that the Pleistocene loess consists of alternating reddish soil and yellowish loess layers, whereas no typical loess horizons are seen within the red clay (Ding et al., 1999a). The alternation of loess and soils documents largescale oscillations between glacial and interglacial conditions during the Pleistocene (Liu, 1985; Kukla, 1987). In this context, the red clay may have developed under a generally warm and humid climate (Ding et al., 1999a). Geological records (Sun, J.M. et al., 1998) have shown that, during glacial periods, the deserts in northern China expanded significantly towards the southeast as a result of the growth of continental ice sheets and an intensification of the winter monsoon, with deserts retreating north-westwards as summer monsoonal rainfall increased during interglacials. Such dramatic changes in desert extent point to a changeable environment in the dust source regions. If, as has been proposed, the red clay sequence is truly of aeolian origin, there is thus a need to investigate the possible extent of the dust source region during the late Miocene and Pliocene. Using grain-size analysis, a north south transect of loess deposits for the past two glacial interglacial cycles and two thick loess red clay Fig. 1. Schematic map showing the zonation of loess in the Chinese Loess Plateau: zone I, sandy loess; zone II, loess; and zone III, clayey loess (adapted from Liu, 1985). The arrow indicates the dominant direction of the winter monsoonal winds, coinciding with the observed decrease in grain size and thickness of the loess. Also shown are the Mu Us desert (dotted) and mountains (black) around and within the Plateau. The study localities are indicated.

3 Comparison of particle size characteristics of Chinese red clay and loess 79 sequences developed in the last 7Æ0 7Æ7 my is first characterized. Then, on the basis of comparison of the loess with the red clay, we consider the implications for changes in the extent of the dust source region in north-western China since the late Miocene are considered. SITE LOCATIONS AND LITHOSTRATIGRAPHY The north south loess transect runs from Hongde about 88 km south of the Mu Us desert margin to Yangling in the southernmost part of the Loess Plateau, which is about 360 km away from the southern margin of the Mu Us Desert. Twelve loess sections, located at Hongde, Huanxian, Mubo, Quzi, Qingyang, Baimapu, Xifeng, Ningxian, Binxianbei, Binxian, Yongshou and Yangling (Fig. 1), were logged and sampled. At present, the mean annual temperature at Hongde is 8Æ0 C and the annual precipitation 400 mm, the values for Yangling being 12Æ9 C and 667 mm. Both mean annual temperatures and rainfall increase southwards along the loess transect. All loess sections except Quzi consist of the loess (L) soil (S) sequence S0, L1, S1, L2 and S2. In the Quzi section, only loess units above S1 are exposed. An obvious advantage of this loess transect over the previously studied Yulin Weinan transect (Ding et al., 2000; Rokosh et al., 2002) (Fig. 1) lies in the fact that the loess sections along the Hongde Yangling transect are closer together and more evenly spaced. The Holocene soil S0 is dark in colour because of its relatively high organic matter content. The upper part of the Holocene soil has been partly eroded or disturbed by agricultural activities at some sites along the transect. The loess units L1 and L2 were deposited during the last and penultimate glacial periods respectively. Both L1 and L2 are yellowish in colour and massive in structure, ranging in thickness from over 20 m in the north to several metres in the south (Fig. 2). The L1 loess unit can generally be subdivided into five subunits, termed L1-1, L1-2, L1-3, L1-4 and L1-5 (Fig. 2). L1-2 and L1-4 are weakly developed soils, and the others are typical loess horizons. Previous studies (Kukla, 1987; Kukla et al., 1988) have shown that L1-1 is correlated with ocean oxygen isotope stage 2, L1-5 with Fig. 2. Median grain-size (Md) and magnetic susceptibility (SUS) records for the 12 sections along the north south loess transect. Subdivision of the loess soil sequences is indicated. The stratigraphic positions for the selected samples in Figs 4, 7 and 8 are marked by solid triangles in each section. The axes for the grain-size curves are the obverse of the SUS curves. Note that the depth scale varies from one section to another. The shaded zones indicate interglacials.

4 80 S. L. Yang & Z. L. Ding stage 4, and L1-2, L1-3 and L1-4 together with stage 3. The fivefold subdivision of L1 is clearly expressed in the grain-size and magnetic susceptibility curves shown in Fig. 2. The L2 loess unit is also composed of three typical loess layers (L2-1, L2-3 and L2-5) and two weakly developed soils (L2-2 and L2-4), as suggested by the grainsize and magnetic susceptibility curves (Fig. 2). The L2 loess unit is correlative with marine oxygen isotopic stage 6. The soil units S1 and S2 developed in the last and penultimate interglacial periods and correlate with marine oxygen isotope stages 5 and 7 respectively (Kukla, 1987; Kukla et al., 1988). Both are brownish or reddish in colour and have an A-Bw-C or A-Bt-C horizon sequence. The thickness of S1 at Hongde is 6Æ5 m, and two thin loess layers are present within it. At Yangling, the southernmost site in the transect, the thickness of S1 decreases to 1Æ8 m, and no loess unit is visible. Soil unit S2 is composed of two soils (S2-1 and S2-2) and a thin intervening loess horizon. Only the upper soil (S2-1) was sampled in this study. The Lingtai and Jingchuan loess red clay sections (Fig. 1) are situated in the central part of the Loess Plateau, with mean annual temperature and precipitation values of 9 10 C and mm respectively. The Lingtai loess red clay sequence has a thickness of 305 m and consists of 175 m of Pleistocene loess and 130 m of Tertiary red clay (Fig. 3). The Jingchuan loess red clay sediment, with a thickness of 325 m, is composed of 199 m of loess and 126 m of red clay (Fig. 3). In the Pleistocene loess of both the Lingtai and Jingchuan sections, each of the S0 L33 loess soil units (Rutter et al., 1991) is readily identified in the field. Within the Tertiary red clay, over 100 reddish soil horizons are distinguishable at both Lingtai and Jingchuan (Ding et al., 1999a, 2001a). Palaeomagnetic studies (Ding et al., 1998b, 2001a) suggest a basal age of 7Æ05 Ma for the Lingtai section and 7Æ7 Ma for the Jingchuan section (Fig. 3). Pedostratigraphic and magnetostratigraphic correlation between the two sections suggests that sedimentation of the red clay was semi-continuous at both sites (Ding et al., 2001a). For the 12 loess sections along the north south transect, a total of 6694 samples were collected at 5 10 cm intervals. In the Lingtai and Jingchuan sections, sample spacing was 3 5 cm, a total of samples being taken (Ding et al., 1999a, 2001a). Bulk magnetic susceptibility and grain size were measured for all samples with a Bartington MS2 susceptibility meter and a SALD laser diffraction particle analyser. The particle analytical procedures were as detailed by Ding et al. (1999b). The susceptibility and grain-size data are shown in Fig. 2 for the loess transect and in Fig. 3 for the Lingtai and Jingchuan sections. Palaeosols are characterized consistently by higher susceptibility values and finer particle sizes compared with the loess horizons above and below them. The two red clay sequences have very fine grain-size distributions, with a median grain size centred around 4 8 lm (8 7 F). SPATIAL CHANGES IN PARTICLE-SIZE PARAMETERS OF LOESS DEPOSITS ALONG THE NORTH SOUTH TRANSECT The grain-size records from the loess transect (Fig. 2) clearly display an overall southward decrease in particle size in the case of both loess and soil horizons. From Hongde southwards to Baimapu, the median grain size decreases rapidly from lm (4Æ3 4Æ1 F) to lm (5Æ1 4Æ6 F) for the typical loess units within L1 and L2, then decreases relatively gradually from lm (5Æ3 5Æ1 F) at Xifeng to lm (6Æ1 5Æ6 F) at Yangling. For the soil units S1 and S2-1, the median grain size shows a more gentle southward decrease along the north south transect, i.e. from lm (6Æ2 6Æ0 F) at Hongde to 8 9 lm (7 6Æ8 F) at Yangling. Figure 4 illustrates the grain-size distributions of some representative samples from horizons L1-1 (loess), L1-4 (weak soil) and S1 (soil) in the 12 loess sections. Almost all the samples show a bimodal character, with a majority falling between 3Æ5 and 6Æ5 F (88 11 lm), and a minority between 11 and 12 F (0Æ5 0Æ25 lm). For loess unit L1-1, the principal mode becomes broader and flatter from Hongde to Yangling, whereas it is consistently broad and flat in the L1-4 and S1 soils. Also, the principal mode gradually shifts to the finer part of the curve from north to south. The minority mode invariably lies between 11 and 12 F (0Æ5 0Æ25 lm). At present, the sedimentological significance of this population is unclear. A possible reason for its occurrence is the peculiar function of the SALD laser particle analyser itself. Grain-size analyses using this equipment were conducted at 0Æ25 F intervals in the range 0Æ25 lm (12 F) to 2000 lm()1 F). It appears that particles finer than 0Æ25 lm cannot be detected by the granulometer and are automatically added to the component between 11 and 12 F, a process that may explain the nature of the minority mode in the size

5 Comparison of particle size characteristics of Chinese red clay and loess 81 Fig. 3. Magnetic susceptibility (SUS) and median grain-size (Md) records of the red clay loess sequences at Lingtai and Jingchuan, together with the magnetic reversal polarities (modified from Ding et al., 1998b, 2001a). Most of the soil units (Si) and major loess beds (Li) are indicated. The contacts between the loess and the red clay at Lingtai and Jingchuan are at a depth of about 175 m and 199 m, respectively, as marked by the dashed lines. The axes for the grain-size curves are the obverse of the SUS curves. distribution diagrams. Accordingly, specific significance to the finest portions of the grading curves can be attributed. All the loess and palaeosol samples from the transect show a positively skewed distribution (Fig. 4), a typical characteristic of dust deposits in the Loess Plateau (Liu, 1966). Kurtosis displays more complicated features. The frequency distribution curves for samples from the L1-1 loess unit all show a leptokurtic pattern, especially those

6 82 S. L. Yang & Z. L. Ding Fig. 4. Grain-size distributions of some representative samples from loess unit L1-1, from the weakly developed soil unit at L1-4 and from palaeosol unit S1 of the loess transect. The stratigraphic positions of the selected samples are indicated in Fig. 2. from the northern part of the Loess Plateau. On the other hand, the weakly developed soil L1-4 and the S1 soil unit are platykurtic. Skewness and kurtosis were calculated for all samples from the loess transect (Fig. 5). In each of the sections, the loess units consistently exhibit higher skewness and kurtosis values than the soil units. Even within loess units L1 and L2, the typical loess layers such as L1-1, L1-5, L2-1, L2-3 and L2-5 are clearly indicated by the skewness and kurtosis peaks. Along the loess transect, the values of skewness and kurtosis show an overall southward decrease for both loess and soil units. Both values decrease more rapidly southwards in the loess units compared with the palaeosols. For example, the skewness and kurtosis values of L1-1 decrease rapidly from about 3Æ0 and 15 at Hongde to only 1Æ0 and 3Æ1 at Yangling respectively. For the S1 soil unit, however, the two values decrease gradually from about 1Æ0 and 3Æ0 at Hongde to 0Æ6 and 2Æ3 at Yangling. Both the spatial and temporal changes in median grain size, skewness and kurtosis of the loess deposits closely parallel each other (Fig. 5). To test this relationship further, correlation coefficients for the three parameters down the loess transect were calculated (Fig. 6). Both the skewness and kurtosis show a good inverse correlation

7 Comparison of particle size characteristics of Chinese red clay and loess 83 Fig. 5. Skewness (phi scale), kurtosis (phi scale) and median grain-size (lm) records of the 12 sections along the loess transect. The skewness and kurtosis are calculated using the method of moments (Friedman & Sanders, 1978). The subdivisions of the loess soil sequences are clearly expressed. Note that the depth scales along the transect are variable. with the median grain size (phi scale) (R 2 ¼ 0Æ932 and 0Æ889 respectively) (Fig. 6A and B), indicating that the coarser the grain size, the higher the skewness and kurtosis values. There is a strong positive correlation between skewness and kurtosis (R 2 ¼ 0Æ973; Fig. 6C). The incorporation of skewness, kurtosis and median grain size in a ternary diagram provides a means of assessing spatial and temporal changes in the particle-size characteristics of these aeolian materials. The two end-members of each parameter must be fixed at 0 and 100 as a precondition in constructing such a diagram; skewness, kurtosis and median grain size were normalized accordingly. It should be noted that this type of ternary diagram is quite different from those Fig. 6. Median grain size (phi scale), skewness (phi scale) and kurtosis (phi scale) coplots for all samples from the north south loess transect.

8 84 S. L. Yang & Z. L. Ding conventionally used in the study of sedimentary rocks, a particular advantage of the former type of diagram being that the changes in all three parameters can be displayed. In order to examine variations in these parameters along the dust transport pathway, four adjacent samples were selected from typical loess layers (L1-1, L1-5, L2-1 and L2-3), weakly developed soil units (L1-4 and L2-4) and palaeosol units (S1 and S2-1). The basis of this sample selection was the median grain-size record shown in Fig. 2. For typical loess units, samples were consistently selected from units containing the coarsest grain sizes, whereas samples from soil units were taken from the finest grain-size units (marked in Fig. 2). The three parameters of the four adjacent samples in each unit were then averaged before plotting on the ternary diagrams. It is assumed that this set of samples was deposited at approximately the same time in each of the loess and soil horizons. Three features are clearly visible in the diagram for each of the selected horizons (Fig. 7). First, spatial changes in the three parameters all form a narrow arc-like band for each horizon. Secondly, for a specific loess or soil horizon, the samples from the northern part of the transect are generally located in the lower portion of the arc-like band relative to the samples from the southern part. This is consistent with downwind sorting of dust particles. Thirdly, soil horizons clearly lie higher in the diagrams than the loess horizons. In fact, there is an upward progression from the coarsest loess horizons (L2-3, L2-1) to the finer loess horizons (L1-1, L1-5), then the weakly developed soils (L1-4, L2-4) and, finally, the welldeveloped soils (S1, S2-1) (Fig. 7). Thus, it appears that this type of ternary diagram effectively represents the spatial and temporal differentiation of loess deposits. POTENTIAL FACTORS INFLUENCING THE PARTICLE SIZE OF CHINESE LOESS The spatial and temporal changes in particle-size characteristics, as illustrated above, may be Fig. 7. Skewness (Sk) kurtosis (Kr) median grain size (F 50 ) ternary diagrams showing spatial changes in particlesize characteristics in loess units (L2-3, L2-1, L1-1 and L1-5), weakly developed soil units (L1-4 and L2-4), and palaeosol units (S1 and S2-1) along the loess transect. The stratigraphic positions of the selected samples are indicated in Fig. 2.

9 Comparison of particle size characteristics of Chinese red clay and loess 85 caused by several factors such as source sink distance, transporting wind intensity, post-depositional weathering and particle aggregation. Clearly, the principal factor needs to be identified before the implications of loess and red clay grain-size distributions can be assessed. Source sink distance and wind intensity In earlier work, the grain size of Chinese loess was first used as a proxy indicator for the intensity of the winter monsoon (An et al., 1991; Xiao et al., 1995), but was later regarded as reflecting both the winter monsoon intensity and the source sink distance (Ding et al., 1999c). In fact, it is hard to evaluate wind intensity changes in the geological past, because the dust deposits are not the product of a single wind event but reflect a combined and averaged history of many and varied wind situations. In this study, the median grain size, skewness and kurtosis all show a consistent southward change along the loess transect, suggesting that the three parameters may be closely related to the dust transport distance. In order to determine the relationship between distance and grain-size parameters during different periods, four adjacent samples were selected from each of L1-1, L1-4, L1-5 and S1 (marked in Fig. 2). The sample selection is the same as that shown in Fig. 7. The sand content (> 63 lm%) and median grain size of the four adjacent samples were then averaged. The grain-size changes vs. the distance from the present desert margin along the Hongde Yangling transect are shown in Fig. 8. The sand content and median grain size of loess units L1-1 and L1-5 decrease rapidly from 20 32% to 0Æ5 2% (Fig. 8A), and from to lm (Fig. 8B), respectively, with the increase in distance from 90 to 360 km. As for the weakly developed soil L1-4 and soil unit S1, the two grain-size parameters show a gradual southward decrease, although the gradient of L1-4 is a little steeper than that of S1. For example, the sand content decreases from 5 7% to nearly zero (Fig. 8A) and from lm to6 9lm (Fig. 8B) for median grain size in soil unit S1 between 90 and 360 km from the source. The spatial gradation in particle-size parameters of both loess and soils indicates that southward sorting of aeolian dust during subaerial transport was a regional process during both glacial and interglacial periods and that local dust input from within the Loess Plateau may have been a minor component. It is presumed that the location of the southeast margin of the Mu Us desert during the last glacial maximum (L1-1 in the loess stratigraphy) was similar to that of the present (Sun, J.M. et al. (1998). Thus, a strong inverse relationship between grain-size parameters and the minimum distance from the desert margin during the last glacial maximum can be obtained (R 2 >0Æ95; Fig. 8). For L1-4, L1-5 and S1, the distance to the source region is unknown. However, it is significant that grain-size parameter distance curves closely match the lines of best fit of L1-1 if shifted to the right along the distance axis (Fig. 9). This implies that the difference in grainsize gradient among the four units is a result of changes in source sink distance rather than wind intensity. Weathering effects The process of weathering tends to produce progressively finer grained particles, especially Fig. 8. Sand content (> 63 lm%) and median grain size vs. distance from desert margin in L1-1, L1-4, L1-5 and S1 of the loess transect. The stratigraphic positions of the selected samples from the loess transect are indicated in Fig. 2. The dashed lines are the best fit curves for grain-size parameters vs. distance in L1-1.

10 86 S. L. Yang & Z. L. Ding Fig. 9. Sand content (> 63 lm%) and median grain size vs. distance from desert margin curves for L1-4, L1-5 and S1. The three curves have been differentially moved to the right along the distance axis to facilitate comparison with the lines of best fit of L1-1 (dashed lines, see Fig. 8). clay minerals, at the expense of the coarser grained material. However, there is good evidence indicating that the Chinese loess palaeosol deposits, even the best-developed soil unit (S5), have experienced only the incipient and early stages of chemical weathering characterized by acid leaching and carbonate dissolution, with the alteration of silicate minerals, in particular, being limited (Chen et al., 1998, 2001; Han et al., 1998). Furthermore, many of the clay minerals in loess and palaeosol units have been shown to be mainly of detrital, rather than pedogenic origin (Liu, 1985; Ji et al., 1999). Thus, it appears that weathering processes have had a very restricted effect on the grain-size characteristics of the Chinese loess. Particle aggregation Previous studies (e.g. Derbyshire et al., 1998) have shown that modern airfall dust and Chinese loess contain some silt-sized aggregates. The diameters of these aggregates fall in the range lm, and most of them have diameters between 10 and 20 lm. It follows from this that relatively coarse particles (i.e. the sand fraction)

11 Comparison of particle size characteristics of Chinese red clay and loess 87 in dust deposits may contain few aggregates. As mentioned earlier, the sand content of all the sections along the Hongde Yangling transect shows a consistent decrease with increase in source sink distance for both glacials and interglacials (Fig. 8). The median grain-size data for the transect present a southward-fining pattern similar to the sand content. This leads to the conclusion that loess aggregation does not have a measurable effect upon the spatial differentiation pattern in the Loess Plateau. Based on the discussion above, it is concluded that the source sink distance is a dominant factor influencing the grain-size characteristics of deposited loess. Such particle-size characteristics of aeolian deposits, as mentioned earlier, can be clearly shown in the skewness kurtosis median grain size ternary diagram. Moreover, source sink distance changes can also be inferred from the ternary diagram. COMPARISON OF PARTICLE-SIZE PARAMETERS OF LOESS AND RED CLAY The particle-size distributions of some representative loess palaeosol and red clay samples from the Lingtai and Jingchuan sections are shown in Fig. 10. Again, all show a bimodal pattern, with a principal mode in the coarse fraction and a secondary mode in the finer fraction. The Fig. 10. Grain-size distributions of some representative loess palaeosol and red clay samples from the Lingtai and Jingchuan sections.

12 88 S. L. Yang & Z. L. Ding secondary mode in the soil and red clay samples is more prominent than in most of the loess samples from the two loess palaeosol sequences (Fig. 10) and the north south transect (Fig. 4), reflecting the fact that a higher percentage of the finest particles is found in the soil and red clay samples. Overall, the red clay from the Lingtai and Jingchuan sections shows grain-size distributions very similar to the loess-derived soils such as S5, S26 and S32, all being broad curves, in contrast to the relatively sharp and peaked grain-size distribution pattern in the loess samples from the two sections. Therefore, the Tertiary red clay sediments may have experienced depositional and post-depositional processes similar to those evident in the well-developed Pleistocene palaeosols. Figure 11 shows changes in skewness, kurtosis and median grain size for the Lingtai and Jingchuan loess red clay sequences. In both, the skewness and kurtosis values are much higher for the loess beds than for the palaeosols, and quite closely match the pattern shown by the median grain size. The highest values of skewness and kurtosis in the two sequences occur in sandy loess units, such as L9, L15 and L33. The skewness and kurtosis of the two red clay sequences centre, respectively, on 0Æ2 0Æ6 and 2Æ2 2Æ6, being similar to or slightly lower than those found in the loessic palaeosol units. It is also evident that skewness and kurtosis are very consistent in the red clay parts of the two sections. Averaged median grain size, skewness and kurtosis all show a stepwise increase upsection at both sites. As indicated by the vertical dashed lines in Fig. 11, changes in these parameters can be roughly subdivided into three portions, i.e. the red clay, the lower Pleistocene loess (S15 L33) and the upper Pleistocene loess (L15 S0). The ternary plots of the three parameters for the Pleistocene loess, the loessic palaeosol and the Tertiary red clay from the Lingtai and Jingchuan sections are shown in Fig. 12. They form narrow, long arc-like bands, the location of these on the plots showing a progression from loess at the base (Fig. 12A and E), through a palaeosol zone (Fig. 12B and F) to the red clay at the top (Fig. 12C and G). Plotting of all samples on a single ternary diagram shows that the three distributions overlap, the central zone consisting of palaeosols, relatively fine loess and relatively coarse red clay samples (Fig. 12D and H). It is inferred from this that the location of the red clay in the upper part of the arc indicates that the red clay underwent sorting over greater distances than was the case for the loess palaeosol deposits, and that the three populations may have shared a common origin. To investigate further changes in source sink distances since the late Miocene, ternary diagrams for the entire Lingtai and Jingchuan loess red clay sequences have been constructed (Fig. 13). Both aeolian sequences were subdivided into six intervals mainly on the basis of the median grainsize curves (Fig. 11), and the averaged values of the three parameters were computed for each interval. The six selected intervals are S0 L5 ( 0Æ5 0 Ma), S5 L15 ( 1Æ2 0Æ5 Ma), S15 L24 ( 1Æ6 1Æ2 Ma), S24 L33 ( 2Æ6 1Æ6 Ma), the upper red clay ( 4Æ0 2Æ6 Ma) and the lower red clay (> 4Æ0 Ma). As shown in Fig. 13, the two lowest intervals of the Pleistocene loess (i.e. S15 L24 and S24 L33) in both sections occupy almost the same position in both diagrams. From the uppermost part of the loess down to the lowermost red clay, the averaged values of the three particle-size parameters shift consistently upwards, suggesting that, in the last 7Æ0 7Æ7 my, the dust source region in north-western China expanded progressively towards the south and east. DISCUSSION The red clay sequence at Lingtai and Jingchuan has been subjected to relatively stronger pedogenic processes than the overlying loess and can be regarded as an extremely thick soil complex (Ding et al., 1999a). However, geochemical studies indicate that the chemical weathering of the red clay materials mainly occurred in the source area (Gu et al., 1997, 1999; Ding et al., 2001b). According to Ding et al. (1999a), even the bestdeveloped portion of the red clay at Lingtai has pedogenic characteristics similar to or slightly more advanced than the palaeosol units above S8. A recent study (Han et al., 2002) has shown that the late Pliocene red clay experienced only mechanical translocation of clays, chemical alteration being rather weak. In this context, the weathering process in the red clay, like that in the loess, may have played only a minor role in alteration of the particle-size distribution. The Tertiary red clay sediments at Lingtai and Jingchuan have particle-size characteristics similar to those found in the loess deposits of the last two glacial interglacial periods along the north south transect across the Loess Plateau. This provides strong circumstantial evidence for an aeolian origin for both red clay sequences.

13 Comparison of particle size characteristics of Chinese red clay and loess 89 Fig. 11. Changes in skewness (phi scale), kurtosis (phi scale) and median grain size (lm) in the red clay loess sequences at Lingtai and Jingchuan. The skewness and kurtosis were calculated using the method of moments (Friedman & Sanders, 1978). The major soil (Si) and loess (Li) units are indicated. The horizontal dashed lines indicate the contact between the loess and red clay at Lingtai and Jingchuan. The vertical dashed lines indicate the averaged values of these records in the red clay, the lower Pleistocene loess (L33 S15) and the upper Pleistocene loess (L15 S0) respectively.

14 90 S. L. Yang & Z. L. Ding Fig. 12. Skewness (Sk) kurtosis (Kr) median grain size (F 50 ) ternary diagrams for samples of loess (A and E), palaeosol (B and F) and red clay (C and G) and for all samples (D and H) at Lingtai and Jingchuan. On the basis of studies at Xifeng and Jingle (Fig. 1), some authors concluded that the red clay at both sites contains waterlain sediments (Mo & Derbyshire, 1991; Guo et al., 2001). It appears that the accumulation and preservation of Tertiary aeolian deposits at a specific site depends largely on the local sedimentary environment. In general, aeolian sediments can accumulate continuously only on flat, broad-relief units with a relatively dense vegetation cover capable of acting as a dust trap. It has been shown (Liu, 1985; Zhu & Ding, 1994) that, during the early Cenozoic, the Loess Plateau was an erosional environment that gave rise to several flat, broad basin-like features. Subsequently, these were filled with aeolian Tertiary red clay and Quaternary loess forming huge loess yuan (plateaus). However, in other areas within the Loess Plateau, aeolian accumulation was affected by the palaeodrainage system, leading to red clay sediments of mixed origin. By way of comparison, it is known that local reworking was also active during the accumulation of the Pleistocene loess. For example, the so-called secondary loess deposits, many of which contain coarse sands and bedding structures, are seen in many places on the Loess Plateau. It follows from this that detailed sedimentological studies must be undertaken on all new red clay sections on the Plateau in order to verify their precise sedimentary origins. Systematic changes in the particle-size characteristics of the aeolian deposits along the loess

15 Comparison of particle size characteristics of Chinese red clay and loess 91 Fig. 13. Skewness (Sk) kurtosis (Kr) median grain size (F 50 ) ternary diagrams showing temporal changes in particle-size characteristics of the red clay loess sequences at Lingtai and Jingchuan. transect indicate that differential downwind sorting mainly reflects differences in the distance between source areas and depositional sites. However, two underlying assumptions must be addressed in order to estimate changes in source sink distance using grain-size parameters from the two aeolian sequences and the loess transect: (1) the regional wind intensity, on average, is relatively constant between glacial and interglacial periods; and (2) the production of grains over all size ranges in the source region is relatively constant over the last 7Æ7 my. During the last glacial maximum, the minimum distance from the desert margin to the Yangling section was about 360 km, whereas it was much more remote during the last interglacial. In general, the skewness, kurtosis and median grainsize values of the two red clay sequences are lower than those for the soil units S1 and S2-1 from the loess transect (Figs 5 and 11). In the ternary diagrams (Figs 7 and 13), the two red clay data points are lower down the skewness axis than those for soil units S1 and S2-1 along the Hongde Yangling transect. This suggests that the source sink distance in the case of the Tertiary red clay was greater than for the late Pleistocene palaeosols. CONCLUSIONS The particle-size characteristics of the red clay at Lingtai and Jingchuan are similar to the palaeosols within the overlying loess deposits, supporting the aeolian origin of the two red clay sequences. Chronological and stratigraphical studies have shown that there is close correlation between the Pleistocene loess and the Tertiary red clay successions at the two sites studied (Ding et al., 2001a), suggesting that the two red clay loess sequences can be regarded as the most complete and continuous aeolian sedimentary series for the last 7Æ0 7Æ7 my in the Chinese Loess Plateau. Given the quality of the data on regional and global climate changes provided by the Pleistocene loess in this region, it is reasonable to expect that a longer and more detailed climatic history for eastern Asia will emerge from further study of the red clay. The spatial changes in the particle size of the loess transect suggest that distance to the source region is the principal factor influencing the grain-size characteristics of loess deposits. Thus, the particle-size parameters of aeolian dust provide a valuable indicator of changes in distance to source through time. On the basis of comparison of red clay and loess, it is inferred that the distance to source was greater during the Neogene than in the last and penultimate interglacials. The dust transport distance was greatest for the red clay and least for the loess, implying that there has been progressive expansion of the dust source region in north-western China since the late Miocene.

16 92 S. L. Yang & Z. L. Ding ACKNOWLEDGEMENTS This study is funded by the Chinese Academy of Sciences (KZCX2-SW-133), the National Natural Science Foundation of China (grant ) and the NKBRSF project (G ). The authors wish to thank Professor E. Derbyshire and Dr C. D. Rokosh for their critical comments and language improvements on an earlier version of this paper. Thanks are also extended to S. S. Hou, X. Wang and Dr S. F. Xiong for field assistance, and to Drs J. T. Han, Z. Y. Gu and Z. T. Guo for valuable discussions. REFERENCES An, Z.S., Kukla, G., Porter, S.C. and Xiao, J.L. (1991) Late Quaternary dust flow on the Chinese Loess Plateau. Catena, 18, Cande, S.C. and Kent, D.V. (1995) Revised calibration of the geomagnetic polarity timescale for the Late Cretaceous and Cenozoic. J. Geophys. Res., 100, Chen, J., Ji, J.F., Qiu, G. and Lu, H.Y. (1998) Geochemical studies on the intensity of chemical weathering in Luochuan loess-paleosol sequence, China. Sci. China (Series D), 41, Chen, J., An, Z.S., Liu, L.W., Ji, J.F., Yang, J.D. and Chen, Y. (2001) Variations in chemical compositions of the eolian dust in Chinese Loess Plateau over the past 2.5 Ma and chemical weathering in the Asian inland. Sci. China (Series D), 44, Derbyshire, E., Meng, X. and Kemp, R.A. (1998) Provenance, transport and characteristics of modern aeolian dust in western Gansu Province, China, and interpretation of the Quaternary loess record. J. Arid Environ., 39, Ding, Z.L., Yu, Z.W., Rutter, N.W. and Liu, T.S. (1994) Towards an orbital time scale for Chinese loess deposits. Quatern. Sci. Rev., 13, Ding, Z.L., Sun, J.M., Liu, T.S., Zhu, R.X., Yang, S.L. and Guo, B. (1998a) Wind-blown origin of the Pliocene red clay formation in the central Loess Plateau, China. Earth Planet. Sci. Lett., 161, Ding, Z.L., Sun, J.M., Yang, S.L. and Liu, T.S. (1998b) Preliminary magnetostratigraphy of a thick eolian red clayloess sequence at Lingtai, the Chinese Loess Plateau. Geophys. Res. Lett., 25, Ding, Z.L., Xiong, S.F., Sun, J.M., Yang, S.L., Gu, Z.Y. and Liu, T.S. (1999a) Pedostratigraphy and paleomagnetism of a 7.0 Ma eolian loess-red clay sequence at Lingtai, Loess Plateau, north-central China and the implications for paleomonsoon evolution. Palaeogeogr. Palaeoclimatol. Palaeoecol., 152, Ding, Z.L., Ren, J.Z., Yang, S.L. and Liu, T.S. (1999b) Climate instability during the penultimate glaciation: Evidence from two high-resolution loess records, China. J. Geophys. Res., 104, Ding, Z., Sun, J., Rutter, N.W., Rokosh, D. and Liu, T. (1999c) Changes in sand content of loess deposits along a north south transect of the Chinese Loess Plateau and the implications for desert variations. Quatern. Res., 52, Ding, Z.L., Rutter, N.W., Sun, J.M., Yang, S.L. and Liu, T.S. (2000) Re-arrangement of atmospheric circulation at about 2.6 Ma over northern China: evidence from grain size records of loess-palaeosol and red clay sequences. Quatern. Sci. Rev., 19, Ding, Z.L., Yang, S.L., Hou, S.S., Wang, X., Chen, Z. and Liu, T.S. (2001a) Magnetostratigraphy and sedimentology of the Jingchuan red clay section and correlation of the Tertiary eolian red clay sediments of the Chinese Loess Plateau. J. Geophys. Res., 106, Ding, Z.L., Sun, J.M., Yang, S.L. and Liu, T.S. (2001b) Geochemistry of the Pliocene red clay formation in the Chinese Loess Plateau and implications for its origin, source provenance and paleoclimate change. Geochim. Cosmochim. Acta, 65, Friedman, G.M. and Sanders, J.E. (1978) Principles of Sedimentology. John Wiley & Sons, New York, 792 pp. Gu, Z.Y., Lal, D., Liu, T.S., Guo, Z.T., Southon, J. and Caffee, M.W. (1997) Weathering histories of Chinese loess deposits based on uranium and thorium series nuclides and cosmogenic 10 Be. Geochim. Cosmochim. Acta, 61, Gu, Z.Y., Ding, Z.L., Xiong, S.F. and Liu, T.S. (1999) A seven million geochemical record from Chinese red-clay and loess-paleosol sequence: Weathering and erosion in northwestern China. Quatern. Sci., 19, (in Chinese with English abstract). Guo, Z.T., Peng, S.Z., Hao, Q.Z., Biscaye, P.E. and Liu, T.S. (2001) Origin of the Miocene-Pliocene Red-Earth Formation at Xifeng in Northern China and implications for paleoenvironments. Palaeogeogr. Palaeoclimatol. Palaeoecol., 170, Guo, Z.T., Ruddiman, W.F., Hao, Q.Z., Wu, H.B., Qiao, Y.S., Zhu, R.X., Peng, S.Z., Wei, J.J., Yuan, B.Y. and Liu, T.S. (2002) Onset of Asian desertification by 22 Myr ago inferred from loess deposits in China. Nature, 416, Han, J., Fyfe, W.S. and Longstaffe, F.J. (1998) Climatic implications of the S5 paleosol complex on the southernmost Chinese loess plateau. Quatern. Res., 50, Han, J., Fyfe, W.S. and Gu, Z. (2002) Assessment of the palaeoclimate during Ma registered by transition of Red Clay to loess-palaeosol sequence in central North China. Palaeogeogr. Palaeoclimatol. Palaeoecol., 185, Heller, F. and Liu, T.S. (1982) Magnetostratigraphic dating of loess deposits in China. Nature, 300, Ji, J.F., Chen, J. and Lu, H.Y. (1999) Origin of illite in the loess from the Luochuan area, Loess Plateau, Central China. Clay Miner., 34, Kukla, G. (1987) Loess stratigraphy in central China. Quatern. Sci. Rev., 6, Kukla, G. and An, Z.S. (1989) Loess stratigraphy in central China. Palaeogeogr. Palaeoclimatol. Palaeoecol., 72, Kukla, G., Heller, F., Liu, X.M., Xu, T.C., Liu, T.S. and An, Z.S. (1988) Pleistocene climates in China dated by magnetic susceptibility. Geology, 16, Liu, T.S. (1966) Composition and Texture of Loess. Science Press, Beijing, 132 pp. (in Chinese). Liu, T.S. (1985) Loess and the Environment. China Ocean Press, Beijing, 251 pp. Liu, X.M., Liu, T.S., Xu, T.C., Liu, C. and Chen, M.Y. (1988a) The Chinese loess in Xifeng. I. The primary study on magnetostratigraphy of a loess profile in Xifeng area, Gansu Province. Geophys. J., 92,

17 Comparison of particle size characteristics of Chinese red clay and loess 93 Liu, X.M., Xu, T.C. and Liu, T.S. (1988b) The Chinese loess in Xifeng. II. A study of anisotropy of magnetic susceptibility of loess from Xifeng. Geophys. J., 92, Lu, H.Y., Vandenberghe, J. and An, Z.S. (2001) Aeolian origin and palaeoclimatic implications of the Red Clay (north China) as evidenced by grain-size distribution. J. Quatern. Sci., 16, Mo, D.W. and Derbyshire, E. (1991) The depositional environment of the late Pliocene red clay, Jing-Le Basin, Shanxi Province, China. Sed. Geol., 70, Qiang, X.K., Li, Z.X., Powell, C.McA. and Zheng, H.B. (2001) Magnetostratigraphic record of the Late Miocene onset of the East Asian monsoon, and Pliocene uplift of northern Tibet. Earth Planet. Sci. Lett., 187, Rokosh, C.D., Rutter, N.W., Ding, Z. and Sun, J. (2002) Regional lithofacies and pedofacies variations along a north to south climatic gradient during the Last Glacial period in the central Loess Plateau, China. Quatern. Sci. Rev., 21, Rutter, N.W., Ding, Z.L., Evans, M.E. and Wang, Y.C. (1990) Magnetostratigraphy of the Baoji loess-paleosol section in the north-central China Loess Plateau. Quatern. Int., 7 8, Rutter, N.W., Ding, Z.L., Evans, M.E. and Liu, T.S. (1991) Baoji-type pedostratigraphic section, Loess Plateau, northcentral China. Quatern. Sci. Rev., 10, Sun, D.H., Shaw, J., An, Z.S., Chen, M.Y. and Yue, L.P. (1998a) Magnetostratigraphy and paleoclimatic interpretation of a continuous 7.2Ma Late Cenozoic eolian sediments from the Chinese Loess Plateau. Geophys. Res. Lett., 25, Sun, D.H., An, Z.S., Shaw, J., Bloemendal, J. and Sun, Y.B. (1998b) Magnetostratigraphy and palaeoclimatic significance of Late Tertiary aeolian sequences in the Chinese Loess Plateau. Geophys. J. Int., 134, Sun, J.M., Ding, Z.L. and Liu, T.S. (1998) Desert distributions during the glacial maximum and climatic optimum: example of China. Episodes, 21, Xiao, J.L., Porter, S.C., An, Z.S., Kumai, H. and Yoshikawa, S. (1995) Grain size of quartz as an indicator of winter monsoon strength on the loess plateau of central China during the last 130,000 yr. Quatern. Res., 43, Zhang, Y.X. and Xue, X.X. (1996) Taphonomic characteristics of Longjiagou Hipparion fauna and genesis of Hipparion red clay in Wudu County, Gansu Province. Chin. Sci. Bull., 41, Zheng, H.B., An, Z.S. and Shaw, J. (1992) New contributions to Chinese Plio-Pleistocene magnetostratigraphy. Phys. Earth Planet. Inter., 70, Zhu, Z.Y. and Ding, Z.L. (1994) The Climatic and Tectonic Evolution in the Loess Plateau of China During the Quaternary. Geological Publishing House, Beijing, 226 pp. (in Chinese). Manuscript received 28 May 2002; revision accepted 7 October 2003.