Cyclic precipitation variation on the western Loess Plateau of China during the past four centuries

Precipitation variation on the Loess Plateau (LP) of China is not only important for rain-fed agriculture in this environmentally sensitive region, but also critical for the water and life securities over the whole Yellow River basin. Here we reconstruct high resolution precipitation variation on the western LP during the past 370 years by using two replicated, annually-laminated stalagmites. Spatial analysis suggests that the reconstruction can be also representative for the whole LP region. The precipitation variations show a significant quasi-50 year periodicity during the last 370 years, and have an important role in determining the runoff of the middle Yellow River. The main factor controlling the decadal scale variations and long-term trend in precipitation over this region is southerly water vapour transport associated with the Asian summer monsoon. The Pacific Decadal Oscillation is also an important influence on precipitation variation in this region, as it can affect the East Asian summer monsoon and the West Pacific Subtropical High.


Results
Lamina counting results suggest that there are 370 6 15 and 262 6 10 (64% error applied to the lamina counts) DCL-WPL couplets in WY27 and WY33, respectively. High detrital 232 Th levels of 7-28 3 10 3 ppt result in large dating uncertainty of 113-874 years for 230 Th dates (Supplementary Table S1) and hinder precise age models.
To further constrain the chronology, we use 210 Pb dating. The 210 Pb dating results show exponential decrease of excess 210 Pb activities with growth depth in the two stalagmites (Supplementary Figs. S4,  S5 and Table S2), indicating the upper sections of WY27 and WY33 are younger than ,130 years 16,17 . The average growth rate calculated from the depth profile of excess 210 Pb for the top 23.5 mm of WY27 (0.171 mm/yr) is in good agreement with that estimated by lamina counting (0.156 mm/yr). The growth rate for the top 39 mm of WY33 determined by 210 Pb (0.289 mm/yr) is slightly lower than that calculated by lamina counting (0.361 mm/yr), which may be caused by the high sampling width (3-5 mm) and the errors of 210 Pb dating (,10%) and lamina counting (64%). The comparison suggests the DCL-WPL couplets in both stalagmites are annual. According to the layer counting chronologies, WY27 grew from 1641 to 2010 CE, and WY33 from 1749 to 2010 CE ( Supplementary Fig. S6).
Stalagmite d 18 O variation can be affected by different factors, including temperature, rainfall amount, moisture recycling and circulation, change in the precipitation-to-evaporation ratio, and changes in the moisture source and transport pathway [20][21][22][23] . Previous studies 11,12 suggested there was an anti-correlation between speleothem d 18 O and rainfall amount in this monsoonal front region during the last 1800 years on decadal to centennial scales. Recently, by combining simulation records and paleo-moisture records, Liu et al. 24  During wetter conditions, the decreased residence time of the seepage water may cause less bedrock to be dissolved, resulting in lower stalagmite d 13 C values. In addition, a wet climate favors vegetation growth and biological productivity, which may also contribute to the relatively lower d 13 C values on a decadal scale. In contrast, the increased residence time of the seepage water during drier condition may allow more dissolution of the bedrock, resulting in the heavier d 13 C values 25 . Decreased precipitation may also reduce vegetation cover, and favor prior carbonate precipitation in the unsaturated zone above the cave, and enhanced degassing of CO 2 within the cave 26,27 , also resulting in higher d 13 C values. Thus, the co-variations of the D 13 C and D 18 O time series lead further support to the anti-phase relationship between our stalagmite d 18 O and local rainfall amount on annual-to decadal-timescales.
Spatial analysis shows significant positive correlations between precipitation changes around Wuya Cave area and the LP, especially the western LP during the period 1950-2011 CE ( Supplementary Fig.  S12). Therefore, we interpret our stalagmite d 18 O record as a reliable indicator of precipitation changes on the western LP and even the whole LP region, with higher stalagmite d 18 O values representing lower precipitation and vice versa.
Our stalagmite d 18 O records show good coherence with the treering based annual rainfall reconstruction from Mt. Xinglong 10 (Fig. 3), and drought reconstruction from Mt. Guiqing 9 on the western LP. They are also consistent with a previously published relative low resolution stalagmite d 18 O record from the Huangye Cave 11 in this area (Fig. 3). Based on the reconstructions, a series of decadalscale drought events on the western LP during the last 370 years can be identified: 1670s-1680s, 1760s-early 1770s, 1810s-1820s, 1860s-1870s, late 1910s-1920s, 1970s, and late 1990s. A remarkable decreasing trend in the precipitation since the end of the 19 th century is shown in both speleothem time series and the tree ring record (Fig. 3). In contrast, notable negative d 18 O peaks, which indicate humid climate, can be observed in both series during the periods of 1690-1730 CE, 1780-1800 CE, 1830-1850 CE, and 1880-1900 CE.
The identified drought intervals on the western LP during the last 370 years show a periodicity of about half a century (Fig. 3). Power spectrum analysis indicates that the most significant periodicity of the WY33 d 18 O series is 52 years, which passes the 99% significant level ( Supplementary Fig. S13), further supporting the quasi-50-year periodicity.

Discussion
Impacts. There were two large-scale land reclamations on the western LP during the past four centuries. The first one happened in the early 18 th century during the Qing Dynasty 28,29 , and the other one happened in the late 1930s and 1940s 30 . Both land reclamations corresponded to humid periods in our reconstruction (Fig. 3), suggesting that humid climate, in addition to the orientation of policy [28][29][30] , provided favorable conditions for agricultural developments on the western LP during these periods.
Decadal scale droughts on the western LP during historical times caused serious damage to the society in this environmental sensitive region. For example, historical documents recorded the Weihe River on the western LP dried up in the summer of 1862 CE when a serious drought occurred 31 . The drought in the late half of the 1870s was much more serious and affected a vast region from LP to North China Plain, and even from South and Southeast Asia to the Great Basin of North America 32 . Chinese local officers at that time regarded this drought as the severest one since the establishment of the Qing Dynasty (1644-1911 CE), and referred to it as ''Dingwu Disaster'' 33 . The drought had caused catastrophic collapse of the society in northern China. About half of the population (160-200 million) was affected, and at least 10 million people were killed by famine and pestilence caused by the severe drought 33,34 . The affected areas in China of the drought in the 1920s were similar to that in the 1870s 35 . Descriptions like ''not even a blade of grass grows'', ''river dried up'', ''big cannibalism'', ''terrible holocaust'' were reported in the newspapers at that time 33,36 .
The historical river runoff changes of the middle Yellow River from Sanmenxia hydrological station during the period of 1766-2004 CE 37 match the stalagmite-inferred precipitation data (Fig. 3). The synchroneity indicates their close relationship. For instance, when droughts occurred in the 1860s-1870s, late 1920s, 1970s, and the late 1990s on the LP, abnormally low river runoff of the middle Yellow River was observed (Fig. 3). In contrast, four megafloods of the Yellow River occurring in 1841-1843, 1855, 1887 and 1938 CE 33,38 coincided with the humid intervals in this region (Fig. 3).
Driving forces. The positive teleconnection between summer precipitation over northern China and Indian monsoon have been reported according to modern instrumental data 39 . A recent study 40 suggested during the monsoonal season of the last 50 years (1961-2012CE) water vapor over the LP was transported mainly from the tropical Indian and Pacific Oceans by southerly monsoon flow, and that southerly water vapor transport controlled the interannual variability of monsoon precipitation on the LP. Our stalagmite d 18 O records show broad similarities with the stalagmite d 18 O record from the core monsoon region of India 41 on decadal scale during the last 370 years (Fig. 4). The drought events recorded in our records were also observed in the Nepal Himalaya, an Indian monsoon front region, as inferred from a 223 years tree ring d 18 O record 42 . The coincidence of the drought events in the Indian monsoon region and the western LP suggests the important influence of Indian monsoon on our study area. When the Indian summer monsoon declines, weakened southwesterly winds reduce the water vapor transport from the tropical ocean to the north, and induce a drought on the western LP. Conversely, strengthened Indian summer monsoon can enhance the southwesterly winds and the water vapor transport, resulting in increased monsoon precipitation on the western LP 43 (Fig. 4), i.e. warm PDO phases 44 . The correspondence is especially significant during the severe droughts in the 1860s-1870s and late 1910s-1920s. The only exception is the drought in the 1810s-1820s, which seems to correspond to a cold PDO phase in Fig. 4. However, in another two records of PDO reconstructed from the historical rainfall proxy data over eastern China 45 and the tree ring data over Asia 46 , the PDO was in its warm phase during this drought event. It has been suggest the East Asian monsoon was weak and the Western Pacific Subtropical High (WPSH) moved southward during warm PDO phases. These conditions could result in decreased precipitation in northern China, including the western LP 39,47 . In contrast, the strengthened East Asian monsoon and northward shifting of the WPSH in cold PDO phases could enhance the monsoon precipitation on the western LP 39,47 .
The significant 52-year cycle ( Supplementary Fig. S13) observed in the stalagmite d 18 O time series is generally consistent with a 55-60 years cycle in the observed Indian monsoon rainfall 48  Previous studies 39,43,50 reveal a stepwise shift of the EASM and its associated migrations of monsoon air masses and rain belt were closely related to seasonal changes in the westerly upper-level jet stream and the WPSH. As the western LP is located near the northern limit of the EASM, it has been suggested the westerly jet could be an important factor for the regional precipitation by affecting the northward shift of the monsoon rain belt. By comparing Wuya stalagmite records with the reconstructed strength of the Northern Hemisphere westerly in GISP2 ice core 51 , we find concomitant changes between decadal-scale droughts on the western LP in the 20 th century and the strengthened westerly jet (Fig. 4). However, the opposite is true before the 20 th century. Therefore, the relationship between the precipitation on the western LP and the Northern Hemisphere westerly on decadal scale is still uncertain.
Modern instrumental data show a declining trend of the EASM since the end of the 19 th century 52 . Although the trend of the Indian summer monsoon during the past century remains debatable 53 , the observed seasonality of wind field at 850 hPa within the South Asian domain suggests a decreasing monsoon trend since 1948 CE 54 . The reduced southerly water vapor transport caused by declined Asian summer monsoon may contribute to the decreasing trend in precipitation variations on the western LP since the end of the 19 th century.

Methods
We used a hand-held carbide dental drill, with a 0.9-mm diameter drill bit, to recover subsamples (50-100 mg powder) along the growth axes of WY27 and WY33 for 230 Th dating. Seven subsamples were dated with U-series methods at the Minnesota Isotope Laboratory on an inductively coupled plasma mass spectrometer (Thermo-Finnigan ELEMENT) using procedures described in ref. 55    A total of 20 subsamples were drilled along the growth axes of the stalagmites, and were dated with 210 Pb methods at the Institute of Earth Sciences, Academia Sinica. About 400-700 mg powders were dissolved in HCl. 210 Pb activities were determined via 210 Po by alpha spectrometry using 209 Po as a yield determinant 57 . The analytical error based on counting statistics was ,3%.
About 300 powdered subsamples (,50-80 mg) were drilled out by using a handheld carbide dental drill, with a 0.3-mm diameter drill bit, at an interval of 0.5 mm along the central growth axes of the stalagmites for stable isotope analyses. All isotopic compositions were measured on an IsoPrime100 gas source stable isotope ratio mass spectrometer equipped with a MultiPrep system at the Institute of Earth Environment, Chinese Academy of Sciences. The d 18 O values reported here are relative to the Vienna PeeDee Belemnite (VPDB) standard. Repeated measurements of one internal laboratory standard TTB1 showed that the long-term reproducibility of the d 18 O analyses was better than 60.06% (1s).