Introduction 导言

Earth’s orbital variations drove periodic Pleistocene climate oscillations between glacial and interglacial conditions1,2,3,4. During cold glacials, continental-scale ice sheets developed in the mid- to high-latitudes of North America and Eurasia, and the Antarctic Ice Sheet expanded beyond its present-day margins5,6,7. These ice sheets “retreated” to present-day configurations, or beyond, during warm interglacials5,6. Glacial oscillations notably transitioned from a lower-amplitude 41-thousand years (kyr) cyclicity to a higher-amplitude mean ~100-kyr cyclicity between ~1.25 and ~0.6 million years ago (Ma)6,8,9,10,11,12,13. This climatic event, known as the mid-Pleistocene transition (MPT), featured substantial Northern Hemisphere ice sheet expansion during glacials6,8,9,10,11,12,13, and was accompanied by distinct ocean thermohaline circulation weakening at ~0.9 Ma14,15. The MPT was also associated with prominent inland Asian aridification and desertification, which may have enhanced global cooling through dust emission, iron fertilization of oceans16,17, and an atmospheric CO2 concentration decrease linked to increasing dust-induced oceanic productivity18,19.
地球轨道的变化推动了更新世气候在冰川期和间冰期之间的周期性振荡 1,2,3,4 。在寒冷的冰川期，北美和欧亚大陆的中高纬度地区形成了大陆规模的冰盖，南极冰盖也扩展到了今天的边缘 5,6,7 。在温暖的间冰期，这些冰盖 "退缩 "到今天的形态，或超越今天的形态 5,6 。在距今约 125 万年到约 60 万年之间，冰川振荡明显地从振幅较低的 41 千年周期过渡到振幅较高的平均约 100 千年周期 6,8,9,10,11,12,13 。这一气候事件被称为中更新世过渡（MPT），其特点是北半球冰盖在冰川时期大幅扩张 6,8,9,10,11,12,13 ，并伴随着 ~0.9 Ma 时明显的海洋温盐环流减弱 14,15 。冰期过渡还与亚洲内陆显著的干旱化和沙漠化有关，这可能通过尘埃排放、海洋铁肥化 16,17 以及与尘埃引起的海洋生产力增加有关的大气 CO 2 浓度下降 18,19 而加剧了全球变冷。

After more than two decades of research, the drivers of climate change across the MPT remain debated11,12. The MPT occurred without a concomitant shift in the orbital forcing rhythm, which suggests that the transition may have been caused by nonlinear internal feedback rather than by external (insolation) forcing11,20. Several prominent hypotheses have been proposed to explain the cause of the MPT, including gradual lowering of atmospheric CO2 concentrations, sustained regolith removal from North America and Europe, long-term nonlinear feedbacks between ice sheets and global climate, and their combined effects11,20,21,22,23. Furthermore, there is no consensus about the forcing and/or pacing mechanisms for the mean ~100-kyr glacial cycles from the MPT onward: they are interpreted (i) as short eccentricity (95-kyr and 125-kyr) cycles, (ii) as 2–3 obliquity (41-kyr) cycles, (iii) as 4–6 precession (19-kyr and 23-kyr) cycles, or (iv) as semi-random fluctuations without true periodicity11,24,25,26. In addition, the timing of the MPT remains debated; an increasing number of studies appear to support a gradual transition between ~1.25 Ma and ~0.6 Ma rather than an abrupt shift at a single point around 0.9–1 Ma11,12,20,27. Finally, our current understanding of the MPT relies primarily on marine records11,12,20,21,22. However, continuous high-resolution terrestrial records through the MPT are rare, which hinders development of a comprehensive understanding of the global nature and drivers of the MPT.
经过二十多年的研究，对跨 MPT 气候变化的驱动因素仍有争议 11,12 。MPT发生时，轨道强迫节奏没有随之改变，这表明过渡可能是由非线性内部反馈引起的，而不是由外部（日照）强迫引起的 11,20 。为解释 MPT 的成因，人们提出了几种著名的假说，包括大气中 CO 2 浓度的逐渐降低、北美和欧洲持续的碎石清除、冰盖与全球气候之间的长期非线性反馈，以及它们的综合效应 11,20,21,22,23 。此外，对于从 MPT 开始的平均 ~100-kyr 冰川周期的作用力和/或步调机制还没有达成共识：它们被解释为（i）短偏心周期（95-kyr 和 125-kyr），（ii）2-3 个斜周期（41-kyr），（iii）4-6 个前倾周期（19-kyr 和 23-kyr），或（iv）没有真正周期性的半随机波动 11,24,25,26 。此外，关于 MPT 的时间仍然存在争议；越来越多的研究似乎支持在 ~1.25 Ma 到 ~0.6 Ma 之间的逐渐过渡，而不是在 0.9-1 Ma 左右的某一点突然转变 11,12,20,27 。最后，我们目前对 MPT 的了解主要依赖于海洋记录 11,12,20,21,22 。 然而，贯穿 MPT 的连续高分辨率陆地记录非常罕见，这阻碍了对 MPT 全球性质和驱动因素的全面了解。

The ~640,000 km2 Chinese Loess Plateau (CLP) extends across the northeastern Tibetan Plateau margin from ~100 to 115°E and from ~34 to 41°N (Fig. 1). The CLP loess-palaeosol deposits yield long-term and near-continuous records of past climate change across the MPT. Dust, which makes up the CLP loess-palaeosol sequence, was transported from the inland Gobi Desert, other nearby sandy deserts, and poorly-vegetated areas by near-surface northwesterly winter monsoon winds28 (Fig. 1a). The grain size of the Quaternary CLP loess-palaeosol sequences reflects dust transport by winter monsoon winds and is generally interpreted as a winter monsoon intensity indicator29,30,31. Stronger winter monsoon winds are associated with coarser dust accumulation on the CLP. Winter monsoon winds are caused by the outflow of cold and dry air from high-pressure cells over the cool Asian continental interior, which blows toward lower-pressure cells over the warmer western Pacific and Indian oceans. Strong winter monsoons during cold global glacials resulted in the deposition of thick loess layers; they consist of a mixture of clays, silts, and fine sands, and are largely unaltered by pedogenesis, with a yellow color32,33. The Asian summer monsoon transports heat and moisture from the western Pacific and Indian oceans toward intense low-pressure (warm) cells over South and East Asia during the boreal summer, and contributes 60–75% of annual precipitation on the CLP34,35. Enhanced summer monsoon precipitation on the CLP during warm interglacials drove more intense pedogenesis, formation of abundant iron oxides (e.g., magnetite, maghemite, and hematite), and red soil development within the yellow loess sequence31,34,36. Since the concentration of pedogenically formed iron oxides can be measured by magnetic susceptibility (χ), CLP loess-palaeosol χ is a much-used proxy for summer monsoon precipitation31,36,37. Stronger pedogenesis during periods of increased precipitation accelerates fine magnetite/maghemite and hematite formation, which causes higher χ values31,36,37. The shift from yellow to red color is caused dominantly by increased pigmentary red hematite formation during intense pedogenesis38. Cyclic stratigraphic alternations between yellow loess and red palaeosol layers are comparable across North China29,31,33,34, and provide a unique, continuous continental archive of orbital- to millennial-scale Asian monsoon and environmental variability. Furthermore, they shed light on both low- and high-latitude processes (i.e., summer and winter monsoon, respectively), and can be used to understand the relationship between global and regional Asian climate changes across the MPT27,29,31,32,33,34,35,36,38,39,40,41,42.
中国黄土高原（CLP）横跨青藏高原东北边缘，从东经约100°至115°，北纬约34°至41°，长约64万公里（图1）。中国黄土高原的黄土-古沉积物提供了整个青藏高原过去气候变化的长期和近乎连续的记录。构成中电地区黄土-古沉积序列的尘埃是由近地表西北冬季季风 28 （图 1a）从内陆戈壁滩、附近其他沙质沙漠和植被稀疏地区吹来的。第四纪中电黄土-古沉积物序列的粒度反映了冬季季风的沙尘迁移，一般被解释为冬季季风强度指标 29,30,31 。较强的冬季季候风与中电地区较粗的尘土堆积有关。冬季季候风是由冷空气从冷亚洲大陆内部的高压单元流出，吹向较暖的西太平洋和印度洋的低压单元造成的。在寒冷的全球冰川时期，强烈的冬季季风导致厚厚的黄土层沉积；黄土层由粘土、淤泥和细沙混合物组成，基本未受成土作用的影响，呈黄色 32,33 。亚洲夏季季候风在北半球夏季将热量和水分从西太平洋和印度洋输送到南亚和东亚上空的强低压（暖）气室，占中華電力區全年降水量的 60-75% 34,35 。中華電力區在暖間冰期的夏季季候風降水量增加，推動了更強烈的成土作用、大量鐵氧化物（如磁鐵礦、鎂鐵礦和赤鐵礦）的形成，以及黃壤層內紅土的形成 31,34,36 。 由于可以通过磁感应强度（χ）来测量由泥沙形成的铁氧化物的浓度，中电黄土-古沉积 χ 是夏季季风降水量的常用代用指标 31,36,37 。在降水增加期间，较强的成土作用会加速细磁铁矿/磁铁矿和赤铁矿的形成，从而导致较高的χ 值 31,36,37 。从黄色到红色的转变主要是由于在强烈的造山过程中形成的红色赤铁矿色素增加所致 38 。黄色黄土层与红色古沉积层之间的周期性地层交替在整个华北地区都具有可比性 29,31,33,34 ，为亚洲季风和环境变异提供了一个独特的、连续的、从轨道尺度到千年尺度的大陆档案。此外，它们还揭示了低纬度和高纬度过程（即夏季季风和冬季季风），并可用于理解全球和亚洲区域气候变化在整个 MPT 之间的关系 27,29,31,32,33,34,35,36,38,39,40,41,42 。

Fig. 1: Site location map.
图 1：场地位置图。
figure 1

a Map of Asian dust sources (Tibetan Plateau, Gobi Desert, sandy deserts, and wind-eroded land) and aeolian loess deposits in North China28. The distributions of Gobi Desert, sandy deserts, wind-eroded land, and loess were from a recent study28, which refers to the Environmental and Ecological Science Data Center for western China, National Natural Science Foundation of China (http://westdc.westgis.ac.cn). b Map of the Chinese Loess Plateau with the locations of our studied Chaona and Luochuan sections (red stars), and other loess-palaeosol sections discussed here (black stars). The Yellow River system in North China is indicated by the dark blue line. We created these maps with ArcGIS (version 10.7) and Illustrator 2020 software.
a 亚洲沙尘源（青藏高原、戈壁沙漠、沙质沙漠和风蚀地）和华北地区风化黄土沉积分布图 28 。戈壁滩、沙 漠、风蚀地和黄土的分布来自最近的一项研究 28 ，该研究参考了国家自然科学基金委 员会中国西部环境与生态科学数据中心（http://westdc.westgis.ac.cn）的数据。 b 中国黄土高原地图，包括我们研究的朝纳和洛川剖面（红星）以及本文讨论的其他黄土-古沉积剖面（黑星）的位置。深蓝色线条表示华北地区的黄河水系。我们使用 ArcGIS（10.7 版）和 Illustrator 2020 软件绘制了这些地图。

Full size image

Here we present new continuous millennial-resolution grain size records for the last 1.6 Myr from two parallel loess-palaeosol sections on the Central CLP to improve understanding of orbital-scale Asian climate variability and dynamics during the transition to the “~100-kyr world”. We combine them with existing χ records from the same sections42,43 and with grain size and χ records from other loess-palaeosol sections29,33,44,45,46 to investigate glacial-interglacial changes in the winter monsoon, summer monsoon, dust and moisture transport, and Asian interior climatic conditions across the MPT, with particular focus on two extreme pulses. We consider these CLP palaeoclimate records within a broader context of existing terrestrial and global palaeoclimate records. We also perform new Earth System model simulations to assess the influence of Northern Hemisphere ice sheet expansion in driving the observed continental-scale Asian glacial climate anomalies across the MPT.
在这里，我们展示了来自中南半岛中部两个平行黄土-古沉积断面的新的连续千年分辨率粒度记录，以加深对过渡到"~100-kyr 世界 "期间轨道尺度亚洲气候变率和动态的理解。我们将它们与来自同一断面的现有 χ 记录 42,43 以及来自其他黄土-古溶胶断面的粒度和 χ 记录 29,33,44,45,46 结合起来，研究了冰川-间冰期冬季季风、夏季季风、尘埃和水汽输送以及整个大陆坡季风区亚洲内部气候条件的变化，尤其关注两个极端脉冲。我们在现有陆地和全球古气候记录的大背景下考虑这些中柱古气候记录。我们还进行了新的地球系统模型模拟，以评估北半球冰盖扩张在推动所观测到的跨越 MPT 的大陆尺度亚洲冰川气候异常方面的影响。

Results and discussion 结果和讨论

Insights from CLP palaeoclimate records
中欧和东欧古气候记录的启示

To investigate winter monsoon dust transport evolution over the last 1.6 Myr, 982 and 1115 samples were collected for grain size analysis from the Luochuan (109°24′E, 35°48′N; 98.3-m thick) and Chaona (107°12′E, 35°7′N; 110-m thick) loess-palaeosol sections on the Central CLP, respectively (Fig. 1b). The 10-cm sampling interval is equivalent to a temporal spacing of ~1–2 kyr. Analogous to the global chronology for benthic foraminiferal δ18O records from marine sediment cores47, the loess-palaeosol chronology has been established based on different sections/cores across the CLP using orbital tuning, land-sea correlation, and/or grain-size age models, which result in similar ages and the same cycle-to-cycle correlations of loess and palaeosol layers to glacial and interglacial periods defined by the marine benthic δ18O record27,29,31,33,43. Our CLP loess-palaeosol chronology was established by correlating both new median grain size and existing χ records42,43 from the Luochuan and Chaona sections to the marine benthic foraminiferal δ18O record47, with additional support from palaeomagnetic constraints, including boundaries between the Brunhes and Matuyama polarity chrons and the Jaramillo subchron (see Methods and Supplementary Figs. 1, 2). Other CLP loess-palaeosol sections used for comparison were also synchronized to this chronology (Figs. 2 and 3). Our new median grain size records from both the Luochuan and Chaona sections have consistent glacial-interglacial variations that also correlate with those observed in other sections across the CLP (Fig. 2). In addition to median grain size records, we calculated the U-ratio (ratio of 16–44 μm versus 5.5–16 μm particle concentrations)48 and grain-size index (GSI, the ratio of 26–52 μm versus <16 μm particle concentrations)49 for the Luochuan and Chaona sections to assess Asian winter monsoon dust transport variability (Fig. 2). Median grain size, U-ratio, and GSI records from both Luochuan and Chaona have almost identical variability patterns (Fig. 2a–f). To reduce the effects of local changes, and to better reveal large-scale glacial-interglacial winter monsoon dust transport across the CLP, we compiled a median grain size stack based on our new records from Chaona and Luochuan and existing records from the Lingtai33, Jingchuan33, and Baicaoyuan44 sections (Fig. 2; Methods). Our median grain size stack varies consistently with the previous stack of mean grain size of quartz particles (MGSQ) based on data from the Lingtai and Zhaojiachuan sections29 (Fig. 2j, k). Here we use our new multiple grain size records to offer a new, more global, perspective on the dynamics of extreme Asian climate events across the MPT, in contrast to previous CLP grain size studies that focused on the general orbital variability and/or land-sea correlations29,30,31,33,43,44.
为了研究过去160万年冬季季风沙尘运移的演变过程，分别在洛川（东经109°24′，北纬35°48′；厚98.3米）和朝那（东经107°12′，北纬35°7′；厚110米）黄土-古沉积剖面采集了982和1115个样品进行粒度分析（图1b）。10 厘米的取样间隔相当于约 1-2 千年的时间间隔。与海洋沉积物岩芯中底栖有孔虫δ 18 O记录的全球年代学 47 类似，黄土-古沉积年代学是根据中电地区不同断面/岩芯，利用轨道调谐、陆海相关和/或粒度年龄模型建立的，其结果是黄土层和古沉积层与海洋底栖有孔虫δ 18 O记录所定义的冰期和间冰期具有相似的年龄和相同的周期-周期相关性 27,29,31,33,43 。我们的中黄土-古沉积年代学是通过将洛川剖面和朝那剖面新的中值粒度和现有的χ记录 42,43 与海洋底栖有孔虫δ 18 O记录 47 相关联而建立起来的，另外还得到了古地磁约束的支持，包括Brunhes和Matuyama极性时系和Jaramillo亚时系之间的边界（见方法和补充图1、2）。用于对比的其他中电黄土-古沉积剖面也与这一年代学同步（图 2 和图 3）。我们在洛川和朝那断面新发现的中位粒度记录与中电地区其它断面观察到的冰川-间冰期变化一致（图2）。除了中值粒度记录外，我们还计算了U比（16-44 μm与5.5-16微米颗粒浓度之比） 48 和粒度指数（GSI，26-52微米与<16微米颗粒浓度之比） 49 。在洛川和朝纳断面上的粒径指数（GSI，26-52 μm 与 <16 μm 颗粒浓度之比） 48 和粒径指数 49 ，以评估亚洲冬季季风沙尘传输的变异性（图 2）。洛川和朝纳的中值粒径、U比和GSI记录具有几乎相同的变化模式（图2a-f）。为了减少局部变化的影响，并更好地揭示中华北电力站大尺度冰川-间冰期冬季季风沙尘运移，我们根据朝纳和洛川的新记录，以及灵台 33 、泾川 33 和百草园 44 剖面的已有记录，编制了中值粒径叠加图（图2；方法）。我们的中值粒度叠加与之前基于灵台和赵家川断面数据的石英颗粒平均粒度（MGSQ）叠加 29 （图2j，k）变化一致。在此，我们利用新的多粒度记录，从一个新的、更具全球性的视角来研究亚洲极端气候事件在MPT上的动态变化，这与之前的中电粒度研究不同，之前的中电粒度研究侧重于一般的轨道变率和/或陆海相关性 29,30,31,33,43,44 。

Fig. 2: Grain size variations of loess-palaeosol sequences across the Chinese Loess Plateau.
图 2：中国黄土高原各地黄土-古沉积序列的粒度变化。
figure 2

Median grain size records (black lines) from the a Luochuan (this study), d Chaona (this study), g Lingtai33, h Jingchuan33, and i Baicaoyuan44 sections on our newly refined loess-palaeosol chronology. b, c U-ratio (ratio of 16–44 μm versus 5.5–16 μm particle concentrations) and grain-size index (GSI, a ratio of 26–52 μm versus <16 μm particle concentrations) from the Luochuan section. e, f U-ratio and GSI from the Chaona section. In both sections, U-ratio records are depicted with magenta lines and GSI records in light blue. j Chinese Loess Plateau median grain size stack. k Chinese Loess Plateau mean grain size of quartz particles (MGSQ) stack based on data from the Lingtai and Zhaojiachuan sections29. l LR04 marine benthic δ18O record47. L-numbers refer to consecutive loess horizons counting back from the present day. Numbers on the benthic δ18O record refer to Marine Isotope Stages, counting back from the present day. Pink bars indicate the correlation of the median grain size of marker loess layers (L15 and L9-1) to glacial stages (MIS 38 and MIS 22).
a 洛川（本研究）、d 朝那（本研究）、g 灵台 33 、h 泾川 33 和 i 百草园 44 断面的粒度中位数记录（黑线）在我们新近完善的黄土-古沉积年代学上。b, c 洛川断面的 U 比率（16-44 μm 与 5.5-16 μm 颗粒浓度之比）和粒度指数（GSI，26-52 μm 与 <16 μm 颗粒浓度之比）；e, f 朝纳断面的 U 比率和 GSI。j 中国黄土高原中值粒度叠加。 k 基于灵台和赵家川断面数据的中国黄土高原石英颗粒平均粒度（MGSQ）叠加 29 。 l LR04 海洋底栖生物 δ 18 O 记录 47 。L 数字是指从现在算起的连续黄土层。底栖生物 δ 18 O 记录上的数字指海洋同位素阶段，从现在开始往前数。粉红色条表示标记黄土层（L 15 和 L 9-1 ）的中位数粒度与冰川期（MIS 38 和 MIS 22）的相关性。

Full size image
Fig. 3: Magnetic susceptibility variations of loess-palaeosol sequences across the Chinese Loess Plateau.
图 3：中国黄土高原黄土-古沉积序列的磁感应强度变化。
figure 3

Magnetic susceptibility (χ) records from the a Luochuan43, b Chaona42, c Jingchuan45, d Zhaojiachuan29, e Lantian46, and f Lingtai29 loess-palaeosol sections across the CLP on our refined loess-palaeosol chronology. g Chinese Loess Plateau χ stack (this study). h LR04 marine benthic δ18O record47. S-numbers refer to consecutive palaeosol horizons counting back from the present day. Numbers on the benthic δ18O record refer to Marine Isotope Stages, counting back from the present day. Pink bars indicate the correlation of χ of marker loess layers (L15 and L9-1) to glacial stages (marine isotope stages 38 and 22).
h LR04 海洋底栖生物 δ 18 O 记录 47 。S 数字指从现在开始往前数的连续古沉积层。底栖生物 δ 18 O 记录上的数字是指海洋同位素阶段，从现在开始向前追溯。粉红色条表示标记黄土层（L 15 和 L 9-1 ）的 χ 与冰川阶段（海洋同位素阶段 38 和 22）的相关性。

Full size image

To assess summer monsoon variability on the CLP coeval to our grain-size-based winter monsoon record, we developed a new loess-palaeosol χ stack by compiling existing χ records from Luochuan43, Chaona42, Jingchuan45, Zhaojiachuan29, Lantian46, and Lingtai29 (Fig. 3). Both our grain size and χ stacks involve more CLP loess-palaeosol sections than previous stacks29,33, which further minimizes the impacts of local changes. As is the case for grain size records, temporal χ variability matches well among sections and generally correlates cycle-by-cycle with glacial-interglacial cycles in the benthic foraminiferal δ18O record (Figs. 2, 3). Nevertheless, the records contain subtle differences in detail, largely because each represents distinct climate features. In short, χ reflects summer monsoon precipitation, whereas grain size is affected by dust transport, and benthic foraminiferal δ18O is a function of both deep-sea temperature and global ice volume.
为了评估与基于粒度的冬季季风记录同时代中电地区的夏季季风变化，我们将洛川 43 、朝那 42 、泾川 45 、赵家川 29 、蓝田 46 和灵台 29 的χ记录汇总，建立了一个新的黄土-古沉积χ堆积（图 3）。我们的粒度和χ叠加比以前的叠加涉及更多的中电黄土-古沉积剖面 29,33 ，这进一步减少了局部变化的影响。与粒度记录的情况一样，时间χ变异性在不同断面之间匹配良好，并且总体上与底栖有孔虫δ 18 O记录的冰川-间冰期周期逐周期相关（图 2、图 3）。尽管如此，这些记录在细节上还是有细微的差别，这主要是因为它们各自代表了不同的气候特征。简言之，χ 反映了夏季季风降水，而粒度则受尘埃迁移的影响，底栖有孔虫δ 18 O 是深海温度和全球冰量的函数。

Consistent with previous studies27,29,31,33,39, glacial loess layers have overall larger median grain sizes, U-ratios, and GSI values (stronger winter monsoons), and lower χ values (weaker summer monsoons and lower precipitation) than interglacial palaeosol layers (Figs. 2–4). Loess layers L15 (correlating to MIS 38 at ~1.25 Ma) and L9-1 (correlating to MIS 22 at ~0.9 Ma) have notably lower χ values and exceptionally larger median grain sizes, U-ratios, and GSI values than other loess layers (Figs. 2–4). The distinct grain size increases in L15 and L9-1 are observed in all loess-palaeosol sequences from both the eastern and western CLP (Fig. 2), which indicates dominant and widespread dust transport changes during MIS 38 and MIS 22 across the entire CLP. We infer that winter monsoon conditions over Asia during these periods were amplified (i.e., cooler, drier, and windier) compared to preceding and succeeding glacials.
与之前的研究 27,29,31,33,39 一致，冰期黄土层与间冰期古沉积层相比，总体上具有较大的中值粒度、U比和GSI值（较强的冬季季风）以及较低的χ值（较弱的夏季季风和较少的降水）（图2-4）。与其他黄土层相比，L 15 黄土层（与 MIS 38 相关，约 1.25 Ma）和 L 9-1 黄土层（与 MIS 22 相关，约 0.9 Ma）的 χ 值明显较低，中值粒度、U 比和 GSI 值也特别大（图 2-4）。在中华北电力大学东部和西部的所有黄土-古沉积序列中都观察到了L 15 和L 9-1 粒度的明显增大（图2），这表明在MIS 38和MIS 22期间，整个中华北电力大学都发生了主要和广泛的粉尘迁移变化。我们推断，与之前和之后的冰川期相比，这两个时期亚洲冬季季风条件增强（即更冷、更干、风更大）。

Fig. 4: Terrestrial and global climate variability on glacial-interglacial timescales across the mid-Pleistocene transition.
图 4：整个更新世中期过渡时期冰川-间冰期时间尺度上的陆地和全球气候变异性。
figure 4

Our newly established Chinese Loess Plateau a χ and b median grain size stacks. S-numbers and L-numbers refer to consecutive palaeosol and loess horizons counting back from the present day, respectively. c Mediterranean sea level record8, d Global sea level reconstruction based on Pacific benthic foraminiferal δ18O and Mg/Ca records10. e Seawater δ18O from ODP site 1123, South Pacific Ocean9. f Global sea level reconstruction based on the LR04 benthic foraminiferal δ18O stack6. g LR04 benthic foraminiferal δ18O record47. Numbers refer to Marine Isotope Stages, counting back from the present day. h global average sea surface temperature (SST) stack70. i Enhanced Asian aridification, new desert formation, and desert expansion from ~1.25 Ma onward inferred from geological records19,28,32,56,58,59. Pink bars indicate the correlation of coarse marker loess layers (L15 and L9-1) to glacial stages (marine isotope stages 38 and 22).
我们新建立的中国黄土高原 a χ 和 b 中值粒度堆积。c 地中海海平面记录 8 , d 基于太平洋底栖有孔虫δ 18 O 和 Mg/Ca 记录的全球海平面重建 10 .g LR04 底栖有孔虫δ 18 O 记录 47 .h 全球平均海面温度（SST）叠加 70 . i 根据地质记录推断，从 ~1.25 Ma 开始，亚洲干旱化加剧，新沙漠形成，沙漠扩展 19,28,32,56,58,59 .粉红色条表示粗标记黄土层（L 15 和 L 9-1 ）与冰川阶段（海洋同位素阶段 38 和 22）的相关性。

Full size image

To investigate the main (orbital) periodicities of the Asian summer and winter monsoon, and to compare these with global (i.e., high-latitude) climatic variations, we present time-evolutive spectral analyses of the χ, grain size, and marine benthic δ18O stacks. Spectral analyses of the median grain size and χ stacks suggest a major transition from a predominant 41-kyr to a mean ~100-kyr periodicity across the MPT, albeit with subtle differences in the exact expression of the transition (Fig. 5a, b). These differences likely reflect more nuanced regional responses to MPT climate change. CLP precipitation (indicated by χ) is dominated by moisture transport from the western Pacific and Indian oceans to inland Asia by the summer monsoon, which is a low-latitude process, whereas transport of cold and dry air from Eurasia toward the tropical oceans by the winter monsoon (indicated by median grain size) represents a high- to mid-latitude process. Spectral analyses of the benthic δ18O record47 reveal a prominent switch from predominant 41-kyr to mean ~100-kyr cycles across the MPT, with co-occurring 41-kyr and ~100-kyr cycles between ~1.2 Ma and ~0.6 Ma (Fig. 5c). A weak obliquity (41-kyr) cyclicity until ~0.6 Ma suggests that the shift in MPT periodicity was more gradual and delayed in the global mean glacial cycle pattern reflected in the benthic foraminiferal δ18O record than in the CLP precipitation (χ) and winter monsoon (median grain size) records (Fig. 5). The low-frequency spectral power in all three records is not centered exactly on the 100-kyr band before ~0.65 Ma and after ~0.2 Ma (Fig. 5). Spectral power splits broadly into two branches close to 80-kyr and 120-kyr, respectively, during the MPT in all three records, although the timing and subtle features of the split vary. These two branches converge toward a band close to 100-kyr by the MPT termination in all three records, which gradually evolves into a ~120-kyr band. The dominance of mean ~100-kyr power after the MPT coincides with a substantial decline in the 41-kyr band. These observations are broadly consistent with previous findings for the MPT11,25. In addition, we find a 50- to 60-kyr band in the χ stack around 1.3 Ma and the median grain size stack between ~1.3 Ma and ~1.4 Ma, which is not distinct in the benthic δ18O stack (Fig. 5). Overall, an apparent shift from a predominantly 41-kyr to mean ~100-kyr orbital periodicity across the MPT in the CLP χ, median grain size, and marine benthic δ18O stacks suggests that orbital-scale Asian summer and winter monsoon variations are closely linked to glacial-interglacial pacing of Northern Hemisphere ice sheets.
为了研究亚洲夏季和冬季季风的主要（轨道）周期性，并将其与全球（即高纬度）气候变异进行比较，我们对χ、粒径和海洋底栖生物δ 18 O堆集进行了时间演化光谱分析。中位粒度和 χ 叠加谱分析表明，整个 MPT 的主要周期从 41 天过渡到平均约 100 天，尽管过渡的确切表现形式存在细微差别（图 5a、b）。这些差异可能反映了区域对 MPT 气候变化更细微的反应。CLP降水（用χ表示）主要是由夏季季风从西太平洋和印度洋向亚洲内陆输送的水汽，这是一个低纬度过程，而冬季季风从欧亚大陆向热带海洋输送的干冷空气（用粒径中值表示）则代表了一个高纬度到中纬度过程。底栖 δ 18 O 记录 47 的频谱分析表明，在整个 MPT 中，41-kyr 周期为主向平均 ~100-kyr 周期转变的现象十分明显，在 ~1.2 Ma 到 ~0.6 Ma 之间，41-kyr 周期和 ~100-kyr 周期同时出现（图 5c）。直到 ~0.6 Ma 才出现的弱斜度（41-kyr）周期表明，与中电降水（χ）和冬季季风（中位粒径）记录相比，底栖有孔虫δ 18 O 记录所反映的全球平均冰川周期模式中，MPT 周期的变化更为渐进和延迟（图 5）。在~0.65Ma之前和~0.2Ma之后，这三个记录的低频谱功率并不完全集中在100kyr频带上（图5）。 在所有三个记录中，MPT期间的频谱功率大致分裂成两个分支，分别接近80-kyr和120-kyr，但分裂的时间和细微特征有所不同。在所有三个记录中，这两个分支在大跃变结束时汇聚成一个接近 100-kyr的波段，并逐渐演变成一个 ~120-kyr 波段。在 MPT 结束后，〜100-kyr 的平均功率占主导地位，同时 41-kyr 波段的功率大幅下降。此外，我们还发现在 1.3 Ma 左右的 χ 叠加带和 ~1.3 Ma~1.4 Ma 之间的中值粒度叠加带中有一个 50~60kyr 带，这在底栖的δ 18 O 叠加带中并不明显（图 5）。总之，中电χ、中位数粒度和海洋底栖生物δ 18 O叠加的轨道周期在整个MPT中明显从主要的41-kyr转变为平均的~100-kyr，这表明亚洲夏季和冬季季风的轨道尺度变化与北半球冰盖的冰期-间冰期步调密切相关。

Fig. 5: Orbital-scale climate variability across the mid-Pleistocene transition.
图 5：上新世中期过渡时期的轨道尺度气候变异。
figure 5

Evolutive spectra of the a Chinese Loess Plateau χ stack, b Chinese Loess Plateau median grain size stack, and c LR04 benthic foraminiferal δ18O record47 with a 320-kyr sliding window and 2.6-kyr step. Red dashed lines indicate 41-kyr and 100-kyr periodicities.
a 中国黄土高原 χ 叠加谱、b 中国黄土高原中位粒度叠加谱和 c LR04 底栖有孔虫δ 18 O 记录 47 的演化谱，滑动窗口为 320kyr，步长为 2.6kyr。红色虚线表示 41kyr 和 100kyr 周期。

Full size image

Mechanisms for CLP loess coarsening across the MPT
中电黄土粗化跨越 MPT 的机制

Our multiple new grain size records reveal both orbital-scale variability and extreme pulses across the MPT, in agreement with previous records29,33,39,40. Two prominent loess grain size anomalies at ~1.25 Ma (L15) and ~0.9 Ma (L9-1) have been explained previously in terms of phased Tibetan Plateau uplift39. However, evidence for major plateau uplift during the MPT is tenuous. The Tibetan Plateau was already close to its present-day elevation and configuration by at least the late Miocene, with only limited and more regional Quaternary adjustments50,51,52,53,54,55. Thus, plateau uplift cannot explain the distinct coarsening of L15 and L9-1 across the CLP, nor their astronomical pacing. This leaves the cause(s) of L15 and L9-1, and their palaeoclimatic significance, open to further investigation33.
我们的多个新的粒度记录揭示了整个MPT的轨道尺度变化和极端脉冲，这与以前的记录是一致的 29,33,39,40 。在 ~1.25 Ma (L 15 ) 和 ~0.9 Ma (L 9-1 ) 出现的两个突出的黄土粒度异常，以前曾被解释为青藏高原的阶段性隆起 39 。 然而，青藏高原在大跃进期间发生重大隆起的证据并不充分。至少在中新世晚期，青藏高原就已经接近于现在的海拔高度和构造，只是在第四纪进行了有限的、更具区域性的调整 50,51,52,53,54,55 。因此，高原隆起无法解释整个中国大陆坡L 15 和L 9-1 的明显粗化，也无法解释它们的天文步调。因此，L 15 和 L 9-1 的成因及其古气候意义还有待进一步研究 33 。

Fundamentally, these exceptionally coarse loess layers must reflect a combination of (i) widespread wind strength increase, (ii) transport pathway shortening due to enhanced and expanded Central Asian aridity, (iii) enhanced coarse dust production through increased aridity and sediment availability, and/or (iv) reduced vegetation cover with lower soil stability and greater soil erosion by wind during glacials MIS 38 and MIS 22. These terrestrial changes probably became prominent at the onset of (~1.25 Ma) and halfway through (~0.9 Ma) the MPT. For example, new sandy deserts (e.g., Badain Jaran Desert, Tengger Desert) are known to have formed at ~1.2–0.9 Ma to the north of the CLP28,56, while existing sandy deserts (e.g., Mu Us Desert) expanded southward at ~1.25 Ma32 (Fig. 4i). Sandy desert environments first appeared in the Hobq Desert at ~1.3–1.2 Ma, replacing preceding fluvio-lacustrine environments57. The Tarim and Qaidam basins (northwestern China) also experienced increased aridity from ~1.25 Ma onward19,58,59 (Fig. 4i). Increasing carbonate δ13C values of the Kazakhstan loess sequence suggest that Central Asia aridified across the MPT60. Glacial loess deposits in Tarim Basin, which were deposited under the control of mid-latitude Westerlies, became generally coarser across the MPT18. The loess coarsening and parallel dust flux increases in the Tarim Basin are consistent with the expansion of Central Asian arid regions across the MPT18. Moreover, Tibetan Plateau cooling across the MPT increased physical weathering intensity, which in turn produced more sand-sized material available for transport to western China through riverine and rain erosion19. These Asian arid regions, especially the neighbouring Badain Jaran, Tengger, Mu Us, and Hobq deserts (see Fig. 1a), provided coarse dust sources for the CLP28. We propose that desert expansion to the west and north of the CLP together with winter monsoon wind strengthening and potential dust source changes led to loess coarsening on the CLP. This agrees with previous results of Sr and Nd isotopic compositions and detrital zircon U-Pb dating of CLP loess-palaeosol sediments, which track dust provenance changes during the MPT, with increased contributions from the NE Tibetan Plateau and Gobi Altay Mountains61,62,63,64,65. The inferred Asian aridification and winter monsoon strengthening is also supported by the extension of aeolian loess deposits to the south of the Qinling Mountains and Yangtze River catchment in southeast China across the MPT66.
从根本上说，这些异常粗大的黄土层必须反映以下因素的综合作用：(i) 广泛的风力增强；(ii) 中亚干旱加剧和扩大导致迁移路径缩短；(iii) 干旱和沉积物供应增加导致粗大尘埃产生增多；和/或 (iv) 在 MIS 38 和 MIS 22 冰期，植被覆盖减少，土壤稳定性降低，风对土壤的侵蚀加剧。这些陆地变化很可能在MPT开始（约1.25Ma）和中期（约0.9Ma）变得突出。例如，已知新的沙质沙漠（如巴丹吉林沙漠、腾格里沙漠）形成于大约1.2-0.9Ma的中柱以北 28,56 ，而现有的沙质沙漠（如穆乌斯沙漠）则在大约1.25Ma向南扩展 32 （图4i）。沙质沙漠环境最早出现在约 1.3-1.2 Ma 的霍布克沙漠，取代了之前的河流-湖泊环境 57 。塔里木盆地和柴达木盆地（中国西北部）从 ~1.25 Ma 开始也经历了干旱的加剧 19,58,59 （图 4i）。哈萨克斯坦黄土序列碳酸盐 δ 13 C 值的增加表明，中亚在整个大沙漠时期都是干旱的 60 。塔里木盆地的冰川黄土沉积是在中纬度西风控制下沉积的，在整个 MPT 18 期间，黄土沉积普遍变粗。塔里木盆地的黄土粗化和沙尘通量的平行增加与中亚干旱地区跨越 MPT 的扩张是一致的 18 。此外，青藏高原跨 MPT 的冷却增加了物理风化强度，这反过来又产生了更多的沙粒大小的物质，可通过河流和雨水侵蚀运往中国西部 19 。 这些亚洲干旱地区，特别是邻近的巴丹吉林沙漠、腾格里沙漠、木乌斯沙漠和霍布克沙漠（见图1a），为中原盆地提供了粗大的尘源 28 。我们认为，沙漠向中国铝土坝西部和北部扩展，加上冬季季风的加强和潜在尘源的变化，导致了中国铝土坝的黄土粗化。这与之前对中原黄土-古沉积物的Sr和Nd同位素组成以及非晶锆石U-Pb年代测定的结果相吻合，这些结果追踪了MPT期间尘源的变化，其中来自青藏高原东北部和戈壁阿勒泰山脉的尘源有所增加 61,62,63,64,65 。亚洲干旱化和冬季季风加强的推断也得到了秦岭以南和中国东南部长江集水区的风化黄土沉积延伸的支持 66 。

We note that intensification of glacial Asian climate and environmental conditions coincided with Northern Hemisphere ice sheet expansion at the onset and middle of the MPT, when expression of mean ~100-kyr glacial cyclicity initiated and enhanced, respectively6,8,9,10,13,67. Both marine and terrestrial data suggest that glacial Northern Hemisphere ice sheets expanded substantially at the beginning of, and halfway through, the MPT. For example, various sea level reconstructions suggest notable low stands during MIS 38 and MIS 22 relative to preceding glacials, albeit with subtle amplitude differences among reconstructions that relate to variable uncertainties in different methods6,8,9,10 (Fig. 4c–f). In addition, 26Al-10Be burial dating of tills suggests that the Laurentide Ice Sheet advanced to its extreme southern limit (~40°N) by ~1.3 Ma67. The ODP Site 887 magnetic susceptibility and Deep Sea Drilling Project (DSDP) Site 607 carbonate concentration records suggest that Northern Hemisphere ice sheets expanded and shed more ice-rafted debris into the Gulf of Alaska at ~1.3 Ma and Central North Atlantic Ocean at ~0.9 Ma, respectively68,69. Associated with ice sheet expansion, glacial global average sea surface temperatures decreased markedly after ~1.25 Ma70 (Fig. 4h). We infer that Asian glacial climate intensification at ~1.25 and ~0.9 Ma as indicated by CLP loess coarsening may be linked to concomitant shifts to greater glacial Northern Hemisphere ice sheet expansion in the context of enhancing global cooling across the MPT.
我们注意到，亚洲冰川期气候和环境条件的加强与北半球冰盖在大跃进开始和中期的扩张是一致的，当时平均 ~100 千年冰川周期性的表达分别开始和加强 6,8,9,10,13,67 。海洋和陆地数据都表明，北半球冰川冰盖在大跃进初期和中期大幅扩张。例如，各种海平面重建结果表明，相对于之前的冰川期，MIS 38 和 MIS 22 期间的海平面明显偏低，尽管重建结果之间存在微妙的振幅差异，这与不同方法的不确定性有关 6,8,9,10 （图 4c-f）。此外， 26 Al- 10 Be埋藏测年表明，劳伦泰德冰原在 ~1.3 Ma 67 前已推进到其极南极限（~40°N）。ODP 887 号站点的磁感应强度和深海钻探项目（DSDP）607 号站点的碳酸盐浓度记录表明，北半球冰盖在 ~1.3 Ma 和北大西洋中部在 ~0.9 Ma 分别扩张并向阿拉斯加湾和北大西洋中部抛撒了更多的冰蚀碎屑 68,69 。随着冰盖的扩张，冰期全球平均海面温度在 ~1.25 Ma 70 之后明显下降（图 4h）。我们推断，中电黄土粗化所显示的 ~1.25 Ma 和 ~0.9 Ma 亚洲冰川气候的加剧，可能与北半球冰川冰盖在整个 MPT 全球变冷加剧的背景下同时向更大的冰川冰盖扩张有关。

To assess the effect of Northern Hemisphere ice sheet expansion on Asia’s interior glacial conditions, aridification and desertification, we performed climate model sensitivity experiments. We used the Community Earth System Model (CESM 1.2) to perform two experiments to examine the Asian (hydro-) climate response to Northern Hemisphere ice sheet expansion (see Methods). One experiment (Large Ice Sheet (LIS) experiment) was conducted with Northern Hemisphere ice sheet distributions, orography, vegetation, lakes, aerosol conditions, and orbital parameters, set at Last Glacial Maximum (LGM, ~20 ka) values5,71, and with 220 ppm CO2 and 450 ppb CH4 concentrations that are similar to values during glacial MIS 50 (~1.5 Ma) as reconstructed from Antarctic ice cores72. The second experiment (Small Ice Sheet (SIS) experiment) was conducted with the same boundary conditions as the LIS experiment, but with a smaller Northern Hemisphere ice sheet distribution, for which we employ the early Holocene/post-LGM configuration at 13 ka5 (Fig. 6a). Northern Hemisphere glacial ice sheet expansion across the MPT was similar in amplitude to the change between 13 ka and the LGM (~20 ka) according to sea level and benthic foraminiferal δ18O records6,8,9,10,47, although exact ice sheet configurations were different between the ~20–13 ka and MPT periods67. Hence, our experiments are not representative of the full range of changing MPT boundary conditions, which remain poorly constrained, but are instead sensitivity experiments that provide insights into the likely climate variations during the intense glacial events, similar to those that coincided with L15 and L9-1.
为了评估北半球冰盖扩张对亚洲内部冰川条件、干旱化和荒漠化的影响，我们进行了气候模式敏感性实验。我们使用共同体地球系统模式（CESM 1.2）进行了两次实验，以研究亚洲（水文）气候对北半球冰盖扩张的响应（见方法）。其中一个实验（大冰盖实验）是在北半球冰盖分布、地形、植被、湖泊、气溶胶条件和轨道参数设置为末次冰川极值（LGM，约 20 ka）值 5,71 ，以及 220 ppm CO 2 和 450 ppb CH 4 浓度的情况下进行的，这些浓度与南极冰芯重建的冰川 MIS 50 期间（约 1.5 Ma）的值相似 72 。第二个实验（小冰原（SIS）实验）的边界条件与 LIS 实验相同，但北半球冰原分布较小，我们采用了全新世早期/LGM 后 13 ka 5 （图 6a）。根据海平面和底栖有孔虫δ 18 O 记录 6,8,9,10,47 ，北半球冰川冰盖在整个 MPT 期间的扩张幅度与 13 ka 和 LGM（约 20 ka）之间的变化幅度相似 67 ，尽管在约 20-13 ka 和 MPT 期间冰盖的具体配置有所不同 67 。因此，我们的实验并不代表变化的 MPT 边界条件的全部范围，因为对这些条件的约束仍然很差，但我们的实验是敏感性实验，它提供了对强烈冰川事件期间可能的气候变异的见解，类似于那些与 L 15 和 L 9-1 相吻合的事件。

Fig. 6: Simulated Asian climate and atmospheric circulation responses to ice volume increase.
图 6：冰体积增加对亚洲气候和大气环流的模拟响应。
figure 6

a Northern Hemisphere ice sheet distribution used in the Small Ice Sheet (SIS, upper) and Large Ice Sheet (LIS, lower) experiments. The yellow dot represents the Chinese Loess Plateau (CLP). Simulated changes in b annual temperature, c annual precipitation, d annual net moisture (precipitation minus evaporation), e Asian high-pressure cell during winter, and f winter monsoon (700 hPa winds) due to ice sheet expansion (SIS – LIS differences). LIS:SIS ratios of atmospheric g annual dust emission flux and h annual dust loading. Solid green contours in b–h denote the 3000 m topographic contour, which includes the Tibetan Plateau. Red stars represent the Chaona and Luochuan loess-palaeosol sections. Small red dots in b–h denote regions with statistical significance above the 95% confidence level (Student’s t-test). Winter in the model is represented by December to February.
小冰盖（SIS，上图）和大冰盖（LIS，下图）实验中使用的北半球冰盖分布图。黄点代表中国黄土高原（CLP）。冰盖扩张导致的 b 年气温、c 年降水量、d 年净湿度（降水量减去蒸发量）、e 冬季亚洲高压单元和 f 冬季季风（700 hPa 风）的模拟变化（SIS - LIS 差异）。大气 g 年尘埃排放通量和 h 年尘埃负荷的 LIS:SIS 比率。b-h 中的绿色实线等值线表示 3000 米地形等值线，其中包括青藏高原。红星代表朝纳和洛川黄土-古沉积剖面。b-h 中的小红点表示统计显著性高于 95% 置信度的区域（学生 t 检验）。模型中的冬季为 12 月至次年 2 月。

Full size image

Differences between our SIS and LIS experiments suggest that Northern Hemisphere glacial ice sheet expansion led to a lowering of Asian mean annual temperature, precipitation, and net surface moisture (precipitation minus evaporation) (Fig. 6b–d), which facilitated intensification and expansion of Central Asian aridity and increased dust production. Mean annual precipitation in the LIS experiment decreased by ~14% in arid inland regions (60–100°E, 30–60°N) and by ~9% in East Asian monsoon regions (100–120°E, 20–40°N) relative to the SIS experiment. Furthermore, ice sheet expansion strengthened Asian high-pressure cells and winter monsoon circulation (Fig. 6e, f), which enhanced winter monsoon dust transport capacity toward the CLP. Overall, annual dust fluxes sourced from arid regions to the north and east of the Tibetan Plateau increased by up to an order of magnitude from the SIS to LIS experiments (Fig. 6g). This is associated with an approximate doubling of the annual atmospheric dust loading over East Asian down-wind regions (Fig. 6h), which is comparable to the largely doubled median grain sizes of L15 and L9-1 relative to adjacent loess layers observed across the CLP (Figs. 2, 4b). Dust changes are substantially larger than precipitation changes between the SIS and LIS experiments (Fig. 6), consistent with the markedly larger grain size changes compared to χ changes at ~1.25 Ma and ~0.9 Ma (Figs. 2, 3).
我们的 SIS 实验和 LIS 实验之间的差异表明，北半球冰川冰盖的扩张导致亚洲年平均气温、降水量和地表净湿度（降水量减去蒸发量）下降（图 6b-d），从而促进了中亚干旱的加剧和扩大，并增加了沙尘的产生。与 SIS 试验相比，LIS 试验中干旱内陆地区（60-100°E，30-60°N）的年平均降水量减少了约 14%，东亚季风区（100-120°E，20-40°N）的年平均降水量减少了约 9%。此外，冰盖扩张加强了亚洲高压单元和冬季季风环流（图 6e、f），从而提高了冬季季风向中南半岛的沙尘输送能力。总体而言，从 SIS 到 LIS 试验期间，来自青藏高原北部和东部干旱地区的年沙尘通量增加了一个数量级（图 6g）。这与东亚下风向地区大气尘埃年负荷增加了约一倍有关（图6h），这与在整个中国黄土高原观测到的L 15 和L 9-1 相对于相邻黄土层的粒径中值增加了一倍相当（图2，4b）。在 SIS 和 LIS 实验之间，尘埃的变化要比降水的变化大得多（图 6），这与在 ~1.25 Ma 和 ~0.9 Ma 的 χ 变化相比，粒度的变化明显更大（图 2、3）是一致的。

Our adopted CESM 1.2 experiments do not include dynamic vegetation, nor glaciogenic dust responses. Hence, our dust inferences represent minimum responses because (i) Northern Hemisphere ice sheet expansion was associated with increased glaciogenic dust73 and (ii) accompanying temperature and precipitation lowering (with arid zone expansion) may have further decreased vegetation cover46,74, which in turn would have reduced soil stability, thereby facilitating erosion and dust production and availability. Other potential dust-producing processes not included in the model, such as enhanced physical weathering and rock fracturing through intensified frost wedging and/or glacial grinding, would also produce more dust material that becomes available for transport under colder MPT conditions19. Regardless, our model estimates, together with our land-sea palaeoclimate correlation, strongly support the hypothesis that intensification and southeastward expansion of Asian aridity increased coarse dust availability, and that strengthening of winter monsoon winds increased coarse dust transport, leading to an overall loess coarsening across the CLP, in response to phases of Northern Hemisphere ice sheet expansion during the MPT.
我们采用的 CESM 1.2 试验不包括动态植被，也不包括冰川源尘埃响应。因此，我们的沙尘推断代表了最小响应，因为（i）北半球冰盖扩张与冰川源沙尘增加有关 73 ，（ii）伴随的温度和降水降低（干旱区扩张）可能会进一步减少植被覆盖 46,74 ，这反过来又会降低土壤稳定性，从而促进侵蚀和沙尘的产生与供应。模型中未包括的其他潜在粉尘产生过程，如通过强化霜楔和/或冰川碾压而加强的物理风化和岩石断裂，也会产生更多的粉尘物质，在更冷的 MPT 条件下可用于迁移 19 。无论如何，我们的模型估算结果以及我们的陆海古气候相关性都有力地支持了这一假设：亚洲干旱的加剧和向东南的扩展增加了粗粒尘埃的可利用性，冬季季风的加强增加了粗粒尘埃的迁移，从而导致整个中太平洋地区的黄土整体变粗，以应对 MPT 期间北半球冰盖的阶段性扩张。

The CLP loess grain size records reveal different regional response intensities to boundary condition changes. Notably, loess L15 at ~1.25 Ma is coarser than L9-1 at ~0.9 Ma, which contrasts with the larger ice volume and cooler climate during the latter period (Fig. 4). This suggests that the CLP grain size response was more pronounced at the MPT onset when the global climate cooled sufficiently for Northern Hemisphere ice sheets to reach a critical size for the first time, which allowed abundant coarser particles to be transported by stronger winds. For example, the initial glaciation-induced transition from fluvio-lacustrine to desert environments in the arid Asian interior would offer greater dust material availability for transport to the CLP than the later, more sustained sandy desert conditions without fluvial-lacustrine processes. Fluvial-lacustrine conditions generally produce abundant dust for aeolian transport once water bodies desiccate and sediment becomes exposed75,76. We argue that the distinct coarsening of L15 and L9-1 across the CLP was related to a combination of greater winter monsoon intensity, shorter transport pathways, decreased vegetation cover, and increased availability of freshly eroded and transportable material in source regions. These changes responded to Northern Hemisphere ice sheet expansion to critical sizes across the MPT. These terrestrial responses were unique to the MPT, when global and Asian climates shifted from a lower-amplitude 41-kyr cyclicity to a higher-amplitude mean ~100-kyr cyclicity (Figs. 4, 5), which constituted a major global climatic reorganization.
中电黄土粒度记录显示了不同地区对边界条件变化的反应强度。值得注意的是，约 1.25 Ma 时的黄土 L 15 比约 0.9 Ma 时的 L 9-1 要粗，这与后一时期较大的冰量和较冷的气候形成鲜明对比（图 4）。这表明，CLP粒度响应在MPT开始时更为明显，当时全球气候足够冷，北半球冰盖首次达到临界粒度，这使得大量较粗的颗粒被更强的风带走。例如，在干旱的亚洲内陆地区，最初冰川作用引起的从河流-湖沼向沙漠环境的过渡，比后来没有河流-湖沼过程的、更持久的沙质沙漠条件，能提供更多的尘埃物质运往CLP。一旦水体干燥，沉积物暴露出来，冲积-湖积条件通常会产生大量的尘埃，供风化搬运 75,76 。我们认为，整个中南半岛 L 15 和 L 9-1 的明显粗化与冬季季风强度增大、迁移路径缩短、植被覆盖减少以及源区新近侵蚀和可迁移物质的增加有关。这些变化与北半球冰盖扩张到整个马普托热带雨林的临界规模有关。这些陆地响应是 MPT 所特有的，当时全球和亚洲气候从振幅较低的 41 世纪周期性转变为振幅较高的平均 ~100 世纪周期性（图 4、图 5），这构成了一次重大的全球气候重组。

Integrating observations, land-sea correlations, and model simulations, we propose that Northern Hemisphere ice sheet expansion drove large-scale amplification of Asian glacial conditions at the onset of (~1.25 Ma) and halfway through (~0.9 Ma) the MPT, when expression of mean ~100-kyr glacial cyclicity initiated and enhanced, respectively. These two glacial intensifications were marked by a combination of intensified and expanded Asian aridity, winter monsoon strengthening, and summer monsoon weakening, with distinct coarsening of loess layers L15 and L9-1 across the CLP. The shift from a predominantly 41-kyr to mean ~100-kyr orbital periodicity across the MPT is also apparent in our winter and summer monsoon records, which, more generally, reflect Northern Hemisphere ice sheet control on orbital-scale Asian climate variability, not just on extreme glacial Asian climate events at ~1.25 Ma and ~0.9 Ma. Our study supports a close relationship between the Pleistocene Asian-interior and global climate changes.
综合观测结果、陆海相关性和模型模拟，我们提出北半球冰盖扩张推动了亚洲冰川条件的大规模放大，分别发生在大冰期开始时（约 1.25 Ma）和大冰期中段（约 0.9 Ma），当时平均约 100 千年的冰川周期性分别开始和加强。这两次冰川期的加强都表现为亚洲干旱程度的加强和扩大、冬季季风的加强和夏季季风的减弱，整个中国大陆坡的黄土层 L 15 和 L 9-1 明显变厚。在我们的冬季和夏季季风记录中，整个MPT的轨道周期从主要的41-kyr到平均的~100-kyr的转变也很明显，这更普遍地反映了北半球冰盖对轨道尺度亚洲气候变异的控制，而不仅仅是对~1.25 Ma和~0.9 Ma的极端冰川亚洲气候事件的控制。我们的研究支持了更新世亚洲内部与全球气候变化之间的密切关系。

Methods 方法

Following surface outcrop removal, 1115 and 982 fresh samples, respectively, were collected from the Chaona and Luochuan sections, Central CLP, from Holocene palaeosol layer S1 (corresponding to MIS 1) to palaeosol layer S22 (corresponding to MIS 55) at 10 cm intervals. These intervals are equivalent to an averaged temporal spacing of ~1–2 kyr that is of higher sampling resolution than in many previous studies. All samples were used for grain size analyses. In preparation for grain size analyses, about 0.2–0.3 g of bulk sample was first pretreated with 30% hydrogen peroxide (H2O2) to remove organic matter and subsequently with 10% hydrochloric acid (HCl) to remove carbonates and iron oxides. After dispersing with 10 mL 10% sodium hexametaphosphate ((NaPO3)6) solution in an ultrasonic bath for ~10 min, the grain size distribution was measured using a Malvern 2000 Laser Instrument at the Institute of Earth Environment, Chinese Academy of Sciences, Xi’an (China). The relative standard deviation of grain size measurement was <3%.
在清除地表露头后，从中环中路朝纳段和洛川段全新世古溶胶层S 1 （对应MIS 1）到古溶胶层S 22 （对应MIS 55），分别采集了1115个和982个新鲜样品，间隔为10厘米。这些间隔相当于约 1-2 千年的平均时间间隔，采样分辨率高于以往的许多研究。所有样本都用于粒度分析。在准备粒度分析时，首先用 30% 过氧化氢（H 2 O 2 ）预处理约 0.2-0.3 克的块状样品以去除有机物，然后用 10% 盐酸（HCl）去除碳酸盐和氧化铁。用 10 mL 10%六偏磷酸钠（(NaPO 3 ) 6 ）溶液在超声波浴中分散约 10 分钟后，在中国科学院西安地球环境研究所使用马尔文 2000 激光仪测量粒度分布。粒度测量的相对标准偏差小于 3%。

CLP loess-palaeosol age models were established using different approaches, including orbital tuning, land-sea correlation, and grain-size age models, that are all similar and match well with the marine benthic δ18O records27,29,31,33. In particular, the extremely coarse loess layers L15 and L9-1 are consistently correlated to MIS 38 and MIS 22 in these age models27,29,33. In Luochuan, Chaona, and other CLP sections, palaeosol layers, which developed during interglacials, have higher χ values and smaller grain sizes than loess layers that accumulated during glacials27,29,31,33. Based on the correlation of loess (palaeosol) layers to glacial (interglacial) periods27,29,31,33,43, we constructed an age model for the Luochuan and Chaona loess-palaeosol sections by correlating median grain size and χ to the marine benthic δ18O stack47. We correlated iteratively until visually satisfactory cycle-to-cycle correlations were obtained. Final age models for the Luochuan and Chaona sections were established using 37 and 38 age correlation points, respectively. These age correlation points are mostly adjacent to loess–palaeosol boundaries (Supplementary Fig. 1), which facilitated consistent correlation point selection throughout the entire section. The age model is supported by an established magnetostratigraphy for each section77,78, which provides age tie points for the Brunhes and Matuyama polarity chrons, including the Jaramillo subchron, that are consistent with the geomagnetic polarity time scale (GPTS)47 (Supplementary Fig. 1), taking into account uncertainties in the post-depositional natural remanent magnetization (NRM) lock-in depth (generally <50 cm) in CLP loess-palaeosol sediments44. The age model results in a smooth and linear relationship between age and depth for both sections, without abrupt shifts, which provides additional support for our established CLP chronology (Supplementary Fig. 2).
采用轨道调谐、海陆关联、粒度年龄模型等不同方法建立的中黄土-古沉积年龄模型，与海洋底栖生物δ 18 O记录相似且吻合度较高 27,29,31,33 ，尤其是极粗黄土层 L 15 和 L 9-1 在这些年龄模型中始终与 MIS 38 和 MIS 22 相关 27,29,33 。在洛川、朝那等中电剖面，间冰期发育的古沉积层比冰川期堆积的黄土层具有更高的 χ 值和更小的粒度 27,29,31,33 。根据黄土（古沉积）层与冰期（间冰期）的相关性 27,29,31,33,43 ，我们将中位粒度和 χ 与海洋底栖生物 δ 18 O 堆栈 47 相关联，构建了洛川和朝那黄土-古沉积剖面的年龄模型。我们反复进行关联，直到获得令人满意的周期与周期之间的关联。洛川剖面和朝那剖面的最终年龄模型分别采用了 37 个和 38 个年龄相关点。这些年龄相关点大多邻近黄土-古沉积边界（补图 1），这有助于在整个断面上选择一致的相关点。年龄模型得到了每个剖面已建立的磁层地层学的支持 77,78 ，该地层学为包括 Jaramillo 亚时相在内的 Brunhes 和 Matuyama 极性谱系提供了年龄相关点，这些年龄相关点与地磁极性时间尺度（GPTS）相一致 47 （补充图 1），同时考虑了中电黄土-古沉积物中沉积后天然剩磁锁定深度（一般小于 50 厘米）的不确定性 44 。 年龄模型的结果表明，两个剖面的年龄与深度之间存在平滑的线性关系，没有突然的变化，这为我们建立中电普年表提供了更多的支持（补充图 2）。

Other CLP loess-palaeosol sections, including Lingtai29,33, Jingchuan33,45, Baicaoyuan44, Zhaojiachuan29, and Lantian46, were synchronized to this chronology by cycle-to-cycle correlation of median grain size or χ data to their counterparts at Luochuan and Chaona. Thus, all sections considered here were placed on the same, newly refined chronology (Figs. 2, 3). After synchronization, we used the interpolating function in the Acycle software79 to conservatively resample the grain size and χ records at 0.5-kyr intervals to obtain an evenly-spaced data series. We constructed the CLP median grain size stack by averaging the evenly-spaced median grain size time series of the Chaona, Luochuan, Lingtai, Jingchuan, and Baicaoyuan sections with equal weight. Similarly, we constructed the CLP χ stack by averaging the evenly-spaced χ time series of Luochuan, Chaona, Jingchuan, Zhaojiachuan, Lantian, and Lingtai sections with equal weight. The χ, median grain size, and LR04 benthic δ18O stacks47 were subjected to spectral analysis to evaluate the robustness of their potential orbital signature. Evolutionary power spectra were calculated using the Acycle software79 with a 320-kyr sliding window and 2.6-kyr step. To improve the expression of the orbital transition across the MPT from 41-kyr to ~100-kyr cycles in the χ and median grain size records, their longer trends were removed with a low-band-pass filter; the residual records were used for evolutionary power spectral analysis.
其他中电黄土-古沉积剖面，包括灵台 29,33 、泾川 33,45 、百草园 44 、赵家川 29 和蓝田 46 ，通过中值粒度或χ数据与它们在洛川和朝那的对应剖面的周期-周期相关性，与这一年代学同步。因此，本文考虑的所有断面都被置于同一新改进的年代学上（图 2、图 3）。在同步化之后，我们使用 Acycle 软件 79 中的内插函数对粒度和 χ 记录进行保守的重采样，采样间隔为 0.5-kyr，以获得均匀的数据序列。我们将朝那、洛川、灵台、泾川和百草园剖面的中值粒度时间序列以等权重平均，构建了中电中值粒度堆栈。同样，我们将洛川、朝那、泾川、赵家川、蓝田和灵台断面的χ时间序列等间距平均加权，构建了CLP χ堆。对χ、中位数粒度和 LR04 底栖生物δ 18 O 叠加值 47 进行谱分析，以评估其潜在轨道特征的稳健性。使用 Acycle 软件 79 计算了演化功率谱，滑动窗口为 320kyr，步长为 2.6kyr。为了更好地表达 χ 和中位粒度记录中从 41kyr 到 ~100kyr 周期的 MPT 轨道转变，使用低带通滤波器去除了它们的较长趋势；残余记录用于演化功率谱分析。

The Community Earth System Model (CESM 1.2) was used to test the sensitivity and underlying Asian climate response dynamics to Northern Hemisphere ice sheet expansion across the MPT. The CESM consists of coupled dynamic atmosphere, ocean, land, sea-ice, and land-ice components80. We used the Community Atmosphere Model (CAM), Community Land Model (CLM), Parallel Ocean Program (POP), Community Sea-Ice Component (CICE), and Coupler modules in the CESM. The CAM has a Bulk Aerosol Model (BAM) parameterization of dust emission, transport, and deposition81, which has been shown to simulate well dust emission flux and loading in Asia82. The atmosphere has 26 vertical layers and ~0.9° (latitude) × 1.25° (longitude) horizontal resolution. The land model has 15 soil layers and the same horizontal resolution as the atmosphere. Ocean and sea-ice components have 60 vertical layers with 0.5° horizontal resolution. To evaluate Asian (hydro-) climate response to MPT Northern Hemisphere ice sheet expansion, we performed two sensitivity experiments: the SIS and LIS experiments (see main text for boundary condition details). In contrast to LGM values (185 ppm CO2 and 350 ppb CH4), which are often used in simulations82,83,84,85,86, we used 220 ppm CO2 and 450 ppb CH4 concentrations that are similar to their values during glacial MIS 50 (~1.5 Ma) as reconstructed from Antarctic ice cores, which may better reflect early Pleistocene glacial greenhouse gas concentrations during the MPT72. In our experiments, only Northern Hemisphere ice sheet distributions were variable while keeping other boundary conditions fixed. This model design enables an understanding of how Asian (hydro-) climate responds exactly to Northern Hemisphere ice sheet expansion. The ice volume difference between SIS and LIS experiments is broadly in agreement with the amplitude of change across the MPT according to sea level and benthic foraminiferal δ18O records6,8,9,10,47. Both experiments were integrated for 150 model years from the equilibrated LGM initial conditions based on the Palaeoclimate Modeling Intercomparison Project (PMIP3) LGM experiment (http://pmip3.lsce.ipsl.fr). Climatological means of the last 50 model years were used here.
利用群落地球系统模式（CESM 1.2）测试了北半球冰盖扩张对整个 MPT 的敏感性和亚洲气候的基本响应动态。CESM 由耦合的动态大气、海洋、陆地、海冰和陆冰部分组成 80 。我们使用了 CESM 中的共同体大气模式（CAM）、共同体陆地模式（CLM）、平行海洋计划（POP）、共同体海冰组件（CICE）和耦合器模块。CAM 对沙尘的排放、传输和沉积进行了大量气溶胶模型参数化 81 ，该模型已被证明能很好地模拟亚洲的沙尘排放通量和负荷 82 。大气模型有 26 个垂直层，水平分辨率为 ~0.9°（纬度）×1.25°（经度）。陆地模型有 15 个土壤层，水平分辨率与大气层相同。海洋和海冰部分有 60 个垂直层，水平分辨率为 0.5°。为了评估亚洲（水文）气候对 MPT 北半球冰盖扩张的响应，我们进行了两个敏感性实验：SIS 和 LIS 实验（边界条件详见正文）。与模拟中经常使用的 LGM 值（185 ppm CO 2 和 350 ppb CH 4 ）相比 82,83,84,85,86 ，我们使用了 220 ppm CO 2 和 450 ppb CH 4 的浓度，与冰川 MIS 50 期间（约 1.在我们的实验中，只有北半球冰盖的分布是可变的，而其他边界条件是固定的。这种模型设计有助于了解亚洲（水文）气候如何对北半球冰盖扩张做出精确响应。 根据海平面和底栖有孔虫δ 18 O 记录，SIS 和 LIS 实验之间的冰量差异与整个 MPT 的变化幅度基本一致 6,8,9,10,47 。这两项实验都是根据古气候模拟相互比较项目（PMIP3）LGM 实验（http://pmip3.lsce.ipsl.fr）的平衡 LGM 初始条件整合了 150 个模型年。这里使用的是过去 50 个模式年的气候学平均值。

Our modeling differs from previous efforts in the following respects. Previous simulations generally used the full range of LGM boundary conditions82,83,84,85,86. Although our experiments employ LGM orography, vegetation, lakes, aerosol conditions, and orbital parameters, we adjusted CO2 and CH4 concentrations from LGM to early Pleistocene glacial values to better reflect MPT glacial greenhouse gas conditions72. Moreover, previous simulations examined Asian climate responses to the combined influences of global ice volume, orbital forcing, and greenhouse gas concentration changes from the LGM to Holocene, pre-industrial, or present day83,84,86,87,88, yet the response to ice volume change by itself was poorly constrained82,89. In our experiments, only the Northern Hemisphere ice sheet extent is varied and other parameters (e.g., greenhouse gas and orbital forcing) are kept constant, which helps to improve understanding of how Asian (hydro-) climate responds exactly to Northern Hemisphere ice sheet expansion across the MPT. Northern Hemisphere ice sheet expansions from the Holocene, pre-industrial, or present day to the LGM used in previous simulations are useful for understanding the impacts of large-amplitude ice sheet changes between full glacial and interglacial conditions. However, these ice volume changes overestimate the glacial changes across the MPT. The expansion of Northern Hemisphere ice sheets between our SIS and LIS experiments better represents the ice volume changes across the MPT. Despite these model differences, both previous82,84,86,89,90,91 and our new simulations indicate that Northern Hemisphere ice sheet expansion was associated with Asian cooling, winter monsoon strengthening, and aridification. Simulations suggest that large ice sheets can also drive cooling and aridification in North China under the warmer Pliocene boundary conditions92. However, our models do not simulate the wetter-than-present conditions in Central Asia when ice volume was larger, for example, under the LGM maximum Northern Hemisphere ice volume conditions that were observed in some previous simulations93; this response may have arisen from a boundary condition other than ice volume in previous simulations.
我们的模拟在以下几个方面不同于以往的工作。以前的模拟通常使用全部的 LGM 边界条件 82,83,84,85,86 。尽管我们的实验采用了 LGM 地形、植被、湖泊、气溶胶条件和轨道参数，但我们将 CO 2 和 CH 4 浓度从 LGM 调整为早更新世冰川时期的值，以更好地反映 MPT 冰川时期的温室气体条件 72 。此外，以前的模拟研究了亚洲气候对全球冰量、轨道强迫和温室气体浓度变化的综合影响的反应，包括从全新世到全新世、工业化前或现在的变化 83,84,86,87,88 ，但对冰量变化本身的反应却没有很好的约束 82,89 。 在我们的实验中，只改变了北半球冰盖的范围，而其他参数（如温室气体和轨道强迫）则没有改变 82,84,86,89,90,91 、在我们的实验中，只改变了北半球冰盖的范围，而其他参数（如温室气体和轨道强迫）保持不变，这有助于更好地理解亚洲（水文）气候是如何对北半球冰盖在整个 MPT 期间的扩张做出精确响应的。以往模拟中使用的北半球冰盖从全新世、工业化前或现在到全新世的扩张，有助于理解冰盖在全冰期和间冰期之间大振幅变化的影响。然而，这些冰量变化高估了整个大冰期的冰川变化。我们的 SIS 和 LIS 实验之间北半球冰盖的扩张更好地代表了整个 MPT 的冰量变化。尽管存在这些模型差异，但之前的82,84,86,89,90,91和我们的新模拟都表明，北半球冰盖的扩张与亚洲变冷、冬季季风加强和干旱化有关。 模拟结果表明，在较暖的上新世边界条件下，大冰盖也能驱动华北地区的降温和干旱化 92 。 然而，我们的模型并没有模拟出冰量较大时中亚地区比现在更湿润的条件，例如，在以往一些模拟中观测到的 LGM 北半球最大冰量条件下 93 ；这种反应可能是由以往模拟中冰量以外的边界条件引起的。