A Chinese cave links climate change, social impacts, and human adaptation over the last 500 years

The collapse of some pre-historical and historical cultures, including Chinese dynasties were presumably linked to widespread droughts, on the basis of synchronicities of societal crises and proxy-based climate events. Here, we present a comparison of ancient inscriptions in Dayu Cave from Qinling Mountains, central China, which described accurate times and detailed impacts of seven drought events during the period of 1520–1920 CE, with high-resolution speleothem records from the same cave. The comparable results provide unique and robust tests on relationships among speleothem δ18O changes, drought events, and societal unrest. With direct historical evidences, our results suggest that droughts and even modest events interrupting otherwise wet intervals can cause serious social crises. Modeling results of speleothem δ18O series suggest that future precipitation in central China may be below the average of the past 500 years. As Qinling Mountain is the main recharge area of two large water transfer projects and habitats of many endangered species, it is imperative to explore an adaptive strategy for the decline in precipitation and/or drought events.

In recent years, increasing attention is paid to the impact of climate change and adaptation strategies 1,2 . It is evident that climate change could pose critical impact on ecosystems and society. For instance, rainfall amount deviations directly affect agricultural crop yields, forest advance and retreat, and human health [3][4][5] . In particular, drought events have widely occurred on various timescales and in turn played an important role in changing social stability and human welfare during pre-historical and historical times [6][7][8][9] . Collapses of many ancient civilizations, such as the Neolithic culture in north central China 10 , the Akkadian Empire 6 , pyramid-constructing Old Kingdom civilization of Egypt 9 and classic Maya 11,12 , have all been linked to intense droughts during mid-and late Holocene, on the basis of apparent synchronicities between proxy-inferred drought events and historically documented societal crises.
In the past decade, stalagmite δ 18 O records from China have characterized many aspects of the Asian monsoon variability over the past 500 ka years on centennial to orbital scales [13][14][15] . These records also show a possible linkage between climate change and the demise of several Chinese dynasties during the last 1800 years, such as Tang, Yuan and Ming Dynasties 16,17 . However, the relationship between Chinese stalagmite δ 18 O variations, monsoon climate change, and social crises is still in dispute 18 , and more evidence is required to evaluate the impact of past climate change.
Dayu Cave (33°08′ N, 106°18′ E, 870 m a. s. l.) is located on the southern slope of the Qinling Mountains, central China, and is more than 2 km long 19 (Fig. 1). The cave has a high relative humidity

Results
Many ancient inscriptions were disclosed in Dayu Cave, which indicate that local ancient people visited the cave frequently, at least 70 times during 1520-1920 CE. According to the inscriptions, seven major drought events were clearly described , occurring in 1528 CE, 1596 CE, 1707 CE, 1756 CE, 1839CE, 1891 CE and 1894 CE (Fig. 2), respectively. These inscriptions described many details of the droughts (Table 1). For example, one of them ( Fig. 2A) stated: "On May 24th, 17th year of the Emperor Guangxu period, Qing Dynasty [the traditional Chinese calendar, equivalent to June 30th, 1891 CE], the local mayor, Huaizong Zhu led more than 200 people into the cave to retrieve water. A fortuneteller named Zhenrong Ran prayed for rain during a ceremony". Three years later in 1894 CE (June 12th, 20th year of the Emperor Guangxu period, Qing Dynasty), another drought event occurred. The same mayor and fortuneteller again led more than 120 people into the cave to collect water (Fig. 2B). Another inscription indicated that "On June 8th, 46th year of the Emperor Kangxi period, Qing Dynasty [July 7th, 1707 CE], the governor of Ningqiang district came to the cave to pray for rain".  The seven drought events described in the inscriptions are notably reflected in the stable isotopic and trace elemental records of a stalagmite DY1 from the Dayu cave (Fig. 3). DY1 was collected about 1 km from the cave entrance, covering the period from ca. 1265 to 1982 CE continuously (Fig. S4, Table  S1). The initial low-resolution δ 18 O results from DY1 stalagmite were reported in 2009 19 . Here we built a more solid age model with additional six 230 Th dates and a higher resolved (~1.3 yrs) stable isotopic and trace elemental profiles capturing annual δ 18 O, δ 13 C (Table S2) and Sr/Ca ratio variations during the last 500 years. "Hendy test" 20 results show that both the δ 18 O and δ 13 C remain constant along growth layers of DY1 (Fig. S5). Some limitations of "Hendy test" were reported. For example, the isotopic equilibrium could theoretically occur in the center of the speleothem at the same time that kinetic fractionation occurs at the flanks 21 . However, the stalagmite was most likely deposited at isotopic equilibrium conditions, if the isotopic values remain constant along growth layers. In addition, the DY1 δ 18 O record is similar to a calcite stalagmite SF1 (r = 0.21, N = 393, p < 0.001) from the Buddha Cave in the southern Qinling Mountains 22 , 300 km northeast of the Dayu Cave, on decadal scale, with different mineral compositions and amplitudes of δ 18 O variations (Fig. S6). The correlation coefficient is not very high, mainly because of the uncertainties of the chronology of SF1, which was based on two average growth rates during the last 500 years. The "Hendy test" 20 and "Replication test" 21 indicate that DY1 deposited under conditions close to isotopic equilibrium, and its δ 18 O and δ 13 C variations can be interpreted as proxies primarily reflecting climate and environment variations.
There is a significant negative correlation between the DY1 δ 18 O and the annual rainfall during the period between ~1957 and 1982 CE (r = − 0.44, N = 24, p < 0.05). The correlation coefficient is not very high which may be ascribed to the "Smoothing effect" of the δ 18 O in drip water. The intra/inter-annual mixture of "fresh water" and "old water" may occurred in the karst aquifer 23 of Dayu Cave, because of the thick epikarst zone (~80m). The age uncertainties may also play a role. The age model between the section of 1950-1982 CE of DY1 was built by an average growth rate of 0.197 mm/yr based on two 230 Th dates at 1970 ± 1 and 1894 ± 0.4 CE. Because of possible variations in growth rate, the use of the average value may cause age uncertainties in given subsamples. Nevertheless, the δ 18 O sequence agrees with the observed annual and monsoon rainfall amount on long-term trend, with higher δ 18 O values corresponding to reduced precipitation, and vice versa (Fig. S7). During the last 500 years, the notable positive excursions in δ 18 O coincided with the droughts documented in the inscriptions, corroborating a generally inverse relationship between rainfall amount (mainly from summer monsoon) and speleothem δ 18 O in this region.
The δ 13 C values 24 are relatively higher during each drought event in the last 500 years too (Fig. 3). In fact, there is a significant positive correlation (r = 0.37, P < 0.01, N = 393) between the δ 13 C and the δ 18 O records of DY-1 during 1500-1982 CE (Fig. S8). Speleothem δ 13 C values have bedrock, atmospheric, and soil gas sources 25 (3) and (4) are related to the vegetation change and climatic conditions. On decadal-to annual-time scales, cold and dry climate could reduce the vegetation cover and microbial activity, and result in higher δ 13 C values in speleothems. Factors (2) and (5) are related to hydrogeochemical processes in the epikarst zone and affected by climatic conditions. The increased residence time of the seepage water during drier conditions may allow more bedrock to be dissolved, favor PCP in the unsaturated zone, resulting in higher δ 13 C values in speleothem [25][26][27] . In addition, dry condition may enhance the evaporation and CO 2 degassing of drip water 28 , and cause higher δ 13 C of speleothems in Dayu Cave.
As shown in Fig. 3, the droughts in 1596 CE, 1707 CE, 1756 CE, and 1839CE corresponded well with elevated Sr/Ca ratios. The other three droughts in 1528 CE , 1891 CE, and 1894 CE are also comparable with increased Sr/Ca ratios, considering age differences caused by different sampling intervals and paths. As discussed before, drier conditions could promote longer water residence times in the epikarst, decreased drip rates, and enhanced CO 2 degasing into air voids within the unwetted epikarst. These conditions lead to Sr/Ca ratios higher than congruent bedrock dissolution due to preferential removal of Ca during PCP and increase the Sr/Ca ratio in speleothem [29][30][31] . The positive correlation between δ 18 O and Sr/Ca records of DY1 (r = 0.22, P < 0.01, N = 393) further confirm the observed inverse relationship between speleothem δ 18 O and rainfall amount in this region.
In summary, Dayu Cave provides for the first time an in situ comparison between historical drought events and speleothem records from the same cave. The in-phase variations in speleothem δ 18 O, δ 13 C and Sr/Ca during droughts in the last 500 years demonstrate a convincing anti-correlation between rainfall amount and speleothem δ 18 O in this region 32 .

Discussion
Impacts of climate change on local society. Historical documents show that drought events recorded in Dayu Cave caused serious social problems. For example, the drought of 1528 CE led to "a big starvation and cannibalism" 33 around the Qinling Mountain region, from southern and central Shaanxi Province to eastern Gansu Province 34 . Droughts in the 1890s also caused severe starvation and triggered local social instability, which eventually resulted in a fierce conflict between government and civilians in 1900 CE 35 . A recent study suggested that the collapse of classic Maya civilization was caused by modest reduction in precipitation 11,12 . Similarly, comparisons between stalagmite δ 18 O changes and historical records inside Dayu Cave (Fig. 3) also provide robust evidence that even modest droughts during relatively wet periods had serious societal consequences. As shown in Fig. 3, the drought around 1596 CE was not very severe in comparison with others during the last 500 years in the Dayu δ 18 O record. However, in the context of the overall wet climate during 1530-1685 CE, the multi-year drought appears to be unusual and caused local societal unrest. The inscriptions (Fig. S9) describe the event as "mountains are crying due to drought", and local people "came to the cave to get water" in July and August when the summer monsoon is presumably the strongest. The δ 13 C value of the stalagmite also reaches the highest value in the last 500 years in the drought in 1596 CE 25 .
As described in the inscriptions, in an attempt to adapt to droughts, people in the area came to the cave to obtain water and pray for rain. Likewise, during recent droughts in southwestern China, karst groundwater became a very important water source for local people. This demonstrates a common human adaptation to such climatic changes under similar conditions.

Future climate change in southern Qinling Mountain region. A time domain combined model 36
was used here to evaluate potential future precipitation changes in the area on the basis of the Dayu δ 18 O series. The modeled δ 18 O series fits the original series before 1982 CE, and is then extrapolated for 60 years to 2042 CE (Fig. 4A). Strong coherence between the predicted δ 18 O series and the observed precipitation variations in the period of 1982-2012 CE further validate this approach (Fig. 4B). According to our predicted δ 18 O changes, precipitation between 1982 and 2042 CE will likely fall below the average over the past 500 years in central China. Two droughts, comparable with historical droughts, appear in the model: the 1990s and the late 2030s. Instrumental data confirm the first drought event in the 1990 s (Fig. 4B).
Spectral analysis of the Dayu record yields significant periodicities at 96, 6.3, 3.4 and 2.8 years, with the 2.8, 3.4, and 6.3 year periods corresponding to the El Nińo-Southern Oscillation (ENSO) cycle (Fig. 5A). Spatial correlation analysis also indicates that the precipitation in Dayu Cave region anti-correlates with the sea surface temperature (SST) of Niño 4 region during the period of 1960-2009 CE (Fig. 5B). Results of climate model simulations suggest that the tropical Pacific SST gradient decreases under conditions of global warming, resembling El Nińo-like SST patterns 37,38 . If the current warming continues, precipitation may decrease in the cave region, which is consistent with our prediction.
Precipitation in the southern Qinling Mountain region is the main recharge of Danjiangkou Reservoir, which supplies water to the middle route of South-to-North Water Transfer Project in China. Recently, another large Hanjiang-to-Weihe River Water Transfer Project is under construction, and its water source is similarly recharged by precipitation in the southern Qinling Mountain region. In addition, Qinling Mountains are important refugia for many rare and endangered species including giant pandas (Ailuropoda melanoleuca) 39 . It is therefore crucial to explore an adaptive strategy to prepare for the possible future decline in precipitation and/or drought events in the region. Th dating were obtained by drilling along the growth axis of the stalagmites with a hand-held carbide dental drill. The chemical procedure used to separate uranium and thorium followed those described in ref. S1. Measurements of uranium and thorium were performed on inductively coupled plasma mass spectrometers (ICPMS), Thermo-Finnigan ELEMENT and Thermo Fisher NEPTUNE, following procedures described in ref. S2 and ref. S3, respectively. Corrections for initial 230 Th were made assuming an initial 230 Th/ 232 Th atomic ratio of 4.4 ± 2.2 × 10 −6 . A total of 15 230 Th dates were obtained for DY1 (Table  S1). Two-sigma date errors are less than 4 years for 13 layers. Linear interpolations between 230 Th dates were used to establish the chronology (Fig. S4).
Stable isotope and trace elements analyses. We performed stable isotope analyses (δ 18 O and δ 13 C) for DY1 at intervals from 100 μ m to 250 μ m, depending on growth rate. All subsamples were analyzed on an on-line, automated carbonate preparation system (Kiel III) linked Finnigan MAT-252 gas source mass spectrometer (Table S2). International standard NBS19 and inter-laboratory standard TTB1 were measured every 10-15 samples. Arbitrarily selected duplicates were conducted to check the homogeneity and reproducibility. Stable isotopic values reported are relative to the Vienna PeeDee Belemnite (VPDB) standard. The standard results show that precisions of δ 18 O and δ 13 C analysis are better than  S4). Based on the cumulative variance contribution of ~90%, there were 82 significant harmonics. We then used an Auto Regressive And Moving Average [ARMA(p, q)] model to simulate the residual errors of the δ 18 O series, which were produced by subtracting period terms from the original series. According to the Bayesian information criterion, the degrees p and q of the ARMA model were selected as 2 and 3, respectively. By combining the periodic terms and the results of ARMA(p, q) model, the prediction model of the δ 18 O series was established 36 .