3.1. Mineralogical and Oxygen Isotopic Composition of the Authigenic Carbonate From BY Core
 Carbonate content is abundant in the BY core (Figure 2c), increasing from 8–40% before 10,800 calendar years (cal years) B.P. to 40–55% between 10,800 and 7200 cal years B.P., and subsequently decreasing gradually from 55% to 34% from 7200 cal years B.P. to the present. XRD results show that calcite predominates in the carbonate fraction between 12,400 and 6000 cal years B.P. (depth 180–48 cm), and dolomite occurs after 6000 cal years B.P. (depth 47 cm) (Figure 3). Using SEM, the calcite fraction consists of small (1–5 μm long), flaky and lenticular idiomorphic crystals. Dolomite occurs as a knobbly coating (<1 μm) on the surface of feldspar (Figure 4). The morphological features of calcite and dolomite indicate rapid carbonate precipitation [Fontes et al., 1996].
Figure 3. X-ray diffractogram (28° to 32° 2θ) for the <40 μm fraction from the core BY. Below depth of 48 cm, the major peak occurs at 3.01Å (29.62° 2θ), reflecting dominance of calcite. The flat peak at depth of 137 cm represents a relatively high content of magnesium in calcite. At depth of 47 cm, the small increase in the intensity at 2.89Å (30.96° 2θ) indicates the presence of dolomite. Above a depth of 43 cm, the major peak occurs at 2.89Å (30.96° 2θ), indicating an increase of dolomite [Vasconcelos et al., 1995].
Download figure to PowerPoint
Figure 4. SEM photographs of <40 μm fraction from BY core showing (a) an overview of calcite aggregates on the surface of feldspar, (b) an enlarged picture of the rectangle in Figure 4a, (c) two calcite aggregates, (d) calcite crystals precipitated on the surface of quartz as marked by circles, (e) knobbly dolomite coating on the surface of feldspar, and (f) an enlarged picture of the rectangle in Figure 4e. The dolomite marked by circles has morphological characteristics similar to those of dolomite precipitated in experiments [Vasconcelos et al., 1995].
Download figure to PowerPoint
 From 12,400 to 10,800 cal years B.P., all δ18O values of authigenic carbonate are between −3 and −1‰ VPDB (Figure 2d). From 10,800 to 7800 cal years B.P., the oxygen isotope value decreases with lowest δ18O values (averaging −7.0‰) observed between 7800 and 7200 cal years B.P. From 7200 to 6000 cal years B.P., the oxygen isotope value gradually increases to between −6.2 and −5.3‰. This trend is followed by a rapid increase of 4.1‰ in δ18O values from 6000 to 4800 cal years B.P. After 4800 cal years B.P., the δ18O values show small variations between −1.5 and −0.2‰.
3.2. Mid-Holocene Drying Inferred From Dolomite Precipitation and δ18O Record in BY Core
 In lake environments, carbonate phases are controlled by salinities and Mg/Ca ratios of lake water [Müller et al., 1972]. As salinity increases, low-Mg calcite precipitates first, followed by high-Mg calcite, aragonite, and finally dolomite. In core BY, the presence of dolomite around 6000 cal years B.P. provides depositional evidence for salinization, indicative of increasing aridity since the mid-Holocene in southern Inner Mongolia.
 Increased aridity beginning in the mid-Holocene is supported by the δ18O record of the authigenic carbonate. δ18O values of authigenic carbonate are controlled by temperature and the oxygen isotope value of lake water [e.g., Fontes et al., 1996], and carbonate phases [Land, 1980; McKenzie, 1981]. Previous studies have shown that dolomite is enriched in 18O relative to cogenetic calcite by ∼3.0‰ [Land, 1980; McKenzie, 1981]. Assuming a maximum percentage of 100% for dolomite after 4800 cal years B.P., then a minimum variation of ∼3.0‰ in δ18O values is expected for cogenetic calcite during the time interval 7200–4800 cal years B.P. (Figure 2d). If temperature were the major factor controlling carbonate δ18O values, then the minimum variation of ∼3.0‰ would have resulted from at least a 14°C temperature change [Kim and O'Neil, 1997]. Because such a large change in temperature is unlikely, we suggest that the δ18O values of authigenic carbonate are mainly controlled by the oxygen isotope value of lake water, i.e., the balance between precipitation and evaporation [Leng and Marshall, 2004]. Increased monsoonal precipitation produces lower water δ18O values while decreased precipitation generates higher δ18O values, as demonstrated in previous studies [e.g., Wei and Gasse, 1999]. Evaporation of lake water results in higher δ18O values in authigenic carbonate [Leng and Marshall, 2004]. Our δ18O values suggest that precipitation gradually decreased from 7200 to 6000 cal years B.P., and then decreased more rapidly from 6000 to 4800 cal years B.P., indicating progressive drying between 7200 and 4800 cal years B.P. Here we consider the presence of dolomite at 6000 cal years B.P. as marking the beginning of the mid-Holocene drying event at core BY. The mineralogical and oxygen isotopic changes are consistent with pollen data and mean annual precipitation estimated from the same core (Figures 2e–2g) [Jiang et al., 2006].
3.3. Timing and Spatial Distribution of Mid-Holocene Drying Over Northern China and Its Implications for the East Asian Monsoon Changes
 Paleoclimatic records that range from Tengger Desert to Bohai Bay (Figure 1) over northern China (Table 1, selection criteria are the same as An et al. ) have been collected and compiled from lacustrine deposits, peat, loess and stalagmites. Most sites are located in the transitional area between desert and steppe, a climatically sensitive zone, approximately paralleling the 400 mm isohyet of mean annual precipitation.
Table 1. Paleoclimatic Records Collected in This Studya
|1||Hongshui River||fluvial-lacustrine||pollen, TOC, δ18O, δ13C, CaCO3, CE||Zhang et al. |
|2||Yema Lake||lake||grain size, carbonate||Shi et al. |
|3||Alashan Plateau||lakes||pollen||F.-H.Chen et al. |
|4||Yanhaizi Lake||lake||grain size, MS, TOC, C/N||C.-T. A.Chen et al. |
|5||Midiwan||loess-soil-sand||pollen, δ13C||Zhou et al. , Li et al. |
|6||Loess Plateau||loess-soil||pollen, δ13C, TOC||Zhou et al. |
|7||Dadiwan||loess-wetland||pollen, mollusk, TOC, grain size||An et al. |
|8||Sujiawan||loess-wetland||pollen, mollusk, TOC, grain size||An et al. |
|9||Baxie||loess-soil||MS, grain size, TOC, δ13C||An et al. |
|10||Zoige||peat||TOC, gray scale||Zhou et al. |
|11||Shanbao Cave||stalagmite||δ18O||Shao et al. |
|12||Mianyang||lake||pollen||Yang et al. |
|13||Qidong||Yangtze delta||pollen, TOC, CaCO3||Liu et al. |
|14||Huanghe delta||Huanghe delta||pollen||Yi et al. |
|15||Taishizhuang||peat||pollen, δ18O||Jin and Liu , Tarasov et al. |
|16||Bayanchagan Lake||lake||pollen, Pediastrum, carbonate mineral, δ18O||Jiang et al. , this study|
|17||Haoluku||ancient lake||pollen, grain size, LOI, CE||Liu et al. |
|18||Xiaoniuchang||ancient lake||pollen, grain size, LOI, CE||Liu et al. |
|19||Gushantun||peat||pollen||Liu , Sun et al. |
|20||Jinchuan||peat||pollen||Sun et al. , W. Y. Jiang and T. S. Liu (unpublished data, 2006)|
|21||Sandaolaoyefu||peat||pollen||Yuan and Sun , Sun et al. |
|22||Hulun Lake||lake||pollen, diatom, Ostracode||Yang et al. |
 We identified mid-Holocene dry events through vegetation changes, the termination of soil development and the decline in lake levels [e.g., An et al., 1993; Zhou et al., 2001; Liu et al., 2002; Shi et al., 2002; F.-H.Chen et al., 2003], as these changes are largely related to precipitation [An et al., 2000]. For vegetation records from arid and semiarid regions, the date of increase in desert vegetation (e.g., Ephedra and Chenopodiaceae) was taken as the time of drying [e.g., Zhang et al., 2000], while the date of decrease in broadleaved deciduous/evergreen trees and/or increase in steppe vegetation was selected for those from semihumid and humid regions [e.g., Liu et al., 1992; Yang et al., 1998; Liu et al., 2002; An et al., 2003; Tarasov et al., 2006].
 Humidity reconstruction in deserts was obtained mainly from pollen and geochemical records. A fluvial-lacustrine record from the terrace of the Hongshui River in the Tengger Desert (site 1, Figure 5) [Zhang et al., 2000] suggests a significant increase in desert vegetation (Ephedra and Chenopodiaceae) at 7500 cal years B.P., reflecting enhanced aridity. Lake records indicate a severe drought starting between 7200 and 7000 cal years B.P. in the Tengger Desert (sites 2 and 3, Figure 5) [Shi et al., 2002; F.-H.Chen et al., 2003], and 8800 cal years B.P. in the Hobq Desert (site 4, Figure 5) [C.-T. A.Chen et al., 2003]. A drought that started at 8300 cal years B.P. is also indicated by a pollen record near the margin of Mu Us Desert (site 5, Figure 5) [Li et al., 2003].
Figure 5. Map showing the timing (solid lines, cal years B.P.) and spatial distribution of mid-Holocene drying over northern China. Sites are numbered according to Table 1. Dashed lines are isohyets of mean annual precipitation (mm). Deserts are marked as follows: A, Taklimakan; B, Gurbantunggut (Junggar); C, Kumtag; D, Qaidam; E, Badain Jaran; F, Tengger; G, Mu Us; H, Hobq; I, Ulan Buh; J, Otindag; K, Horqin; and L, Hulun Buir. The arrows indicate the retreat of the summer monsoonal rainfall belt.
Download figure to PowerPoint
 In the desert-steppe transitional zone, paleoclimatic records mainly come from loess and lake sediments. Loess deposits have long been regarded as one of the most important archives of climatic evolution in China during the past 2.6 Ma [Liu and Ding, 1998]. Loess beds were deposited when climate was arid and cold, whereas soils developed under relatively warm and humid conditions [Liu, 1985; Kukla, 1987; Liu and Ding, 1998]. Previous studies have shown that development of Holocene soil started at 11,500–10,000 cal years B.P., and terminated at 6800–5700 cal years B.P. (sites 6 and 9, Figure 5) [An et al., 1993; Zhou et al., 2001]. Pollen records from lake sediment demonstrate that vegetation type and cover deteriorated at the same time as soil termination, characterized by decreases in pollen concentrations (sites 16 and 22, Figure 5) [Yang et al., 1995; Jiang et al., 2006] and arboreal pollen percentages (sites 16–18, Figure 5) [Liu et al., 2002; Jiang et al., 2006]. The good agreement in timing between soil termination and vegetation changes suggests a mid-Holocene drying at 7000–5500 cal years B.P. in the desert-steppe transitional zone.
 There are few paleoclimatic records, with the exception of pollen and stalagmites, available for northeastern and south-central China. Broadleaved deciduous forest was replaced by a mixed forest of deciduous and coniferous trees at 4400–4200 cal years B.P. in northeastern China (sites 19–21, Figure 5) [Liu, 1989; Yuan and Sun, 1990; Sun et al., 1991; W. Y. Jiang et al., unpublished data, 2006]. An increase in steppe vegetation occurred at 4800 cal years B.P. at the Taishizhuang peat bog (site 15, Figure 5) [Jin and Liu, 2002; Tarasov et al., 2006] and 4500 cal years B.P. at the delta of the Yellow River (site 14, Figure 5) [Yi et al., 2003]. Pollen records from south-central China have shown that a decrease in subtropical broadleaved evergreen trees, and an increase in trees with greater moisture tolerances, such as the deciduous genera Quercus, Corylus and the coniferous genus Pinus, occurred at 4300–4200 cal years B.P. (sites 12 and 13, Figure 5) [Liu et al., 1992; Yang et al., 1998]. A high-resolution oxygen-isotope record of a precisely dated stalagmite from Shanbao Cave (site 11, Figure 5) reveals an abrupt decrease in monsoonal precipitation at 4400 cal years B.P. [Shao et al., 2006]. These records suggest a dry, cold climate after 4800–4200 cal years B.P.
 On the basis of the above geological records, together with the BY record, we compiled a contour map showing the spatial changes in timing of mid-Holocene drying over northern China using Surfer software. It shows a clear zonal distribution from northwest to southeast over northern China (Figure 5). Drying began at 9000–7000 cal years B.P. in the deserts of north-central China, extending into the transitional zone between desert and steppe at 7000–5500 cal years B.P. and at ∼4500 cal years B.P. into northeastern and south-central China. This zonal distribution of drying time in the mid-Holocene roughly parallels the isohyets of mean annual precipitation in northern China (Figure 5), indicating a close relationship with the east Asian summer monsoon.
 The summer monsoonal precipitation is produced by the interaction between warm moist southerly air masses and cold northerly airflows from middle and high latitudes. In general, the more northerly the penetration of the rainfall belt into interior continental northern China, the greater is the intensity of the summer monsoon [Tao and Chen, 1987]. The Pleistocene witnessed numerous retreat-advance cycles of the east Asian summer monsoon in response to glacial-interglacial cycles [Liu and Ding, 1998; Ding et al., 2001, 2002]. For any specific site in northern China, it would usually become drier when the summer monsoon retreated southward. The progressive drying from 7200 to 4800 cal years B.P. recorded in Lake BY is consistent with the progressive weakening of the east Asian summer monsoon. In this context, the zonal distribution of the desiccation suggests that the East Asian Monsoon has significantly retreated southeastward since the mid-Holocene. Furthermore, a ∼400–550 km retreat is inferred for the summer monsoon from 6500 to 4500 cal years B.P.
 Changes in the intensity of the east Asian summer monsoon during the Holocene may be related to orbitally induced Northern Hemisphere insolation variability [An et al., 2000; Kutzbach et al., 2001; Wang et al., 2005]. According to Berger and Loutre , Northern Hemisphere summer solar insolation has been decreasing since the early Holocene. Numerical modeling experiments have shown that a decrease in insolation would have resulted in a weakened monsoon circulation [Ganopolski et al., 1998; Kutzbach et al., 2001]. The forcing mechanism lies in that the cooling of the continental interior gives rise to a decrease in thermal contrast between the east Asian continent and the adjacent ocean, thus leading to a weakened monsoon circulation, i.e., a southeastward retreat of the east Asian summer monsoon.