The lack of a precisely-dated, unequivocal climate proxy from northern China, where precipitation variability is traditionally considered as an East Asian summer monsoon (EASM) indicator, impedes our understanding of the behaviour and dynamics of the EASM. Here we present a well-dated, pollen-based, ~20-yr-resolution quantitative precipitation reconstruction (derived using a transfer function) from an alpine lake in North China, which provides for the first time a direct record of EASM evolution since 14.7 ka (ka = thousands of years before present, where the “present” is defined as the year AD 1950). Our record reveals a gradually intensifying monsoon from 14.7–7.0 ka, a maximum monsoon (30% higher precipitation than present) from ~7.8–5.3 ka, and a rapid decline since ~3.3 ka. These insolation-driven EASM trends were punctuated by two millennial-scale weakening events which occurred synchronously to the cold Younger Dryas and at ~9.5–8.5 ka, and by two centennial-scale intervals of enhanced (weakened) monsoon during the Medieval Warm Period (Little Ice Age). Our precipitation reconstruction, consistent with temperature changes but quite different from the prevailing view of EASM evolution, points to strong internal feedback processes driving the EASM, and may aid our understanding of future monsoon behaviour under ongoing anthropogenic climate change.

[1] Detailed rock magnetic studies show that susceptibility (mass-specific c) and anhysteretic remanent magnetization (ARM) of the Chinese loess/paleosol sequences are carried by almost identical magnetic carriers. Therefore the ratio Ác/c ARM (or equivalently c ARM /Ác, where Ác is defined as c À c 0 , and c 0 is the intercept susceptibility of the plot of c versus ARM, and c ARM is field-normalized ARM) can be used to quantify the grain size of c and ARM carriers. By determining this ratio for three Chinese loess/paleosol profiles (Jiuzhoutai, Yuanbao, and Yichuan) characterized by different degrees of environmentally controlled pedogenesis and sedimentation rates, we show that the lower grain-size limit of aeolian magnetic particles in the less pedogenically altered loess units is about 100-300 nm, in the finer-grained pseudosingle domain (PSD) grain-size range. In contrast, the grain sizes of pedogenically produced magnetic particles for mature paleosols dominantly cover both the superparamagnetic (SP) and single-domain (SD) ranges. On the basis of plots of Ác/c ARM against Ác, samples can be divided into four regions (I, II, III, and IV). Region I corresponds to the least pedogenically altered primary loess samples, with Ác/c ARM of 0.165-0.24. Samples in region II, a transition zone between the least altered loess and the onset of development of paleosols, have c values identical to those in region I but have lower Ác/c ARM of 0.09-0.165. With increasing susceptibility in zone III, Ác/c ARM is positively correlated with Ác, indicating the gradually increasing influence of SP particles. Finally, in zone IV with Ác higher than $6.5 Â 10 À7 m 3 kg À1 , Ác/c ARM is independent of the variations in Ác, suggesting that Ác/c ARM is totally controlled by the pedogenic finest-grained particles and the size distribution of these particles remains almost constant. The development of soils in the Chinese loess revealed by these three profiles from three sites can be clearly explained by a continuous process of pedogenesis, increasing from zone I to zone IV. The definition of the pedogenic zones can help to improve our understanding of the underlying mechanisms and variability of pedogenesis and thus could enable more successful and accurate separation of the authentic pedogenic signals from the background signal of the aeolian inputs at different loess sites worldwide.