1 Introduction

Frequent occurrences of droughts in southwest China in recent years have raised increasing concerns among public and scientific communities (Wu 2011; Yin 2010). Due to limited storage capacity of soil water over the widely distributed karst landform in southwest China along with impacts of human activities, droughts have induced severe economic and social problems in recent years. For example, droughts in Yunnan province, southwest China during 2009–2012 AD have directly caused more than 7.3 billion US dollar of financial loss and threatened the life of many people (Tao et al. 2014).

Meteorologists (Barriopedro et al. 2012; Huang et al. 2011, 2012; Li et al. 2009; Zhang et al. 2011) suggested that droughts during rainy seasons were caused by the westward shift of enhanced Western Pacific Subtropical High (WPSH), which blocked the transfer of warm and humid moistures from the Bay of Bengal to southwest China. Meanwhile, the meridional water vapor flux from the Bay of Bengal was weak. However, these analyses were based on short-term observations for the past 60 years and mainly focused on synoptic scale drought events. Only limited studies investigate decadal scale drought events based on tree ring records from the western margin of southwest China, i.e. southeastern Tibetan plateau (Fan et al. 2008; An et al. 2013; Bi et al. 2015). Decadal scale drought can cause much more damage to society, and even played an important role on the collapse of some Chinese dynasties (Zhang et al. 2008; Tan et al. 2011) and classic Maya civilization (Kennett et al. 2012; Medina-Elizalde and Rohling 2012).

A declining trend in monsoon precipitation was thought to exist in southwest China during the last 60 years (Zhang and Li 2014) along with frequent occurrences of extreme droughts (He et al. 2011). What’s the driving force of the declined monsoon precipitation in the last 60 years? Was the recent decadal drought unprecedented? Will the reduced monsoon precipitation trend continue in the future? Answers to these important questions are crucial for water and resource management in southwest China.

In this paper, we reconstructed the monsoon precipitation variation history in southwest China during the past 240 years based on δ18O records of a high-resolution, absolutely dated stalagmite from Yunnan province, southwest China. Variations and driving forces of monsoon precipitation in this region were further discussed.

2 Cave location and climatology

Xiaobailong Cave (N24°12′, E103°21′, 1500 m above sea level), located 20 km south of Mile, Yunnan, southwest China (Fig. 1), formed in Middle Triassic dolomitic limestone and calcite dolomite of the Gejiu Group. The cave, with a small entrance, is about 500 m long and the cave roof ranges from 20 to 60 m thick. The relative humidity is >90 % inside the cave (Cai et al. 2006), and cave air temperature is 17.2 °C, consistent with local mean annual temperature (17.3 °C, Cai et al. 2015). Annual precipitation in this region is 960 mm, and ~80 % occurs during summer monsoon months (June–September). Most of the water surplus occurs between July and October as calculated by the Thornthwaite evapotranspiration model (Thornthwaite 1948; McCabe and Markstrom 2007; Fig. 2). As a result, recharge of the aquifer in this area occurs mainly from summer monsoon rainfall. Previous work suggested that the regional precipitation is positively correlated with that in the northern Indo-China Peninsula (Cai et al. 2015). Water vapor of this region in rainy season mainly comes from the Bay of Bengal (Fig. 1) and is dominated by Indian summer monsoon.

Fig. 1
figure 1

Location of Xiaobailong Cave (red star) and moisture transport pathway. Rainy season (Jun.–Sep.) water vapor transport (kg m−1 s−1) averaged for the period 1961–2012 at low level (700 hPa) based on the NCEP/NCAR reanalysis datasets (Kalnay et al. 1996), using the Grid Analysis and Display System (GrADS). The shading shows absolute value of water vapor transport at each grid. The green solid circle and black solid square represent Jhumar Cave from India (Sinha et al. 2011) and tree ring site from Nepal (Sano et al. 2012), respectively

Full size image
Fig. 2
figure 2

Monthly mean precipitation (gray bars), temperature (blue squares), and water surplus (red squares) at the Kunming station, ~100 km northwest of Xiaobailong Cave (1957–2009 AD)

Full size image

By using various stalagmite records from Xiaobailong Cave, Cai et al. (2006, 2015) reconstructed millennial- to orbital-scales Indian monsoon precipitation variability over the past 252,000 years. Over orbital scale, monsoon precipitation in this region is dominated by 23-kyr precessional cycle. However, there are also clear glacial–interglacial variations, which are different from cave records from East China. On millennial scale, there is a clear link between monsoon precipitation and climate in the North Atlantic (Cai et al. 2015). In addition, climate of the high southern latitudes may also influence some millennial scale monsoon precipitation variations in this region (Cai et al. 2006).

3 Sample and methods

We collected a new stalagmite sample XBL41, 7.5 cm in length, in the inner chamber of the cave (about 400 m from the entrance) in November 2006. The top of the stalagmite was receiving drip water when collected, indicating possibly active growth. After halved and polished, the stalagmite section showed clear growth bandings. Two obvious growth hiatuses were observed at 0.5 and 6.0 cm of the stalagmite (Fig. 3). From the first hiatus to the topmost part, the lithology of the stalagmite changed from aragonite to calcite as identified by petrographic and X-ray Diffraction (XRD) analyses.

Fig. 3
figure 3

Cross-section of XBL41 from Xiaobailong Cave. All 230Th dating positions are marked on the stalagmite with 2σ errors. The red dot indicates an excluded dating subsample, which was contaminated. The dashed lines represent two growth hiatuses

Full size image

Ten subsamples (Fig. 3) were drilled and dated with U-series methods on a multi-collector inductively coupled plasma mass spectrometer (MC-ICPMS) at the University of Minnesota, USA and Xi’an Jiaotong University, China (Edwards et al. 1987; Shen et al. 2003; Cheng et al. 2013). An initial 230Th/232Th atomic ratio of 4.4 ± 2.2 × 10−6 were used to correct for initial 230Th amount. Subsamples for stable isotope analyses were drilled out at intervals of 0.5 mm. Seven coeval subsamples from two layers, respectively, were selected for the “Hendy test” (Hendy 1971) to evaluate the isotopic equilibrium during mineral precipitation. Stable isotope analyses were performed on a Finnigan MAT-252 mass spectrometer equipped with a Kiel III Carbonate Device at the Institute of Earth Environment, Chinese Academy of Sciences. The replicates showed that the precision of δ18O analysis is better than ±0.06 ‰ (2σ, VPDB).

4 Results

4.1 Chronology

Table 1 shows 230Th dating results. All the subsamples measured are in stratigraphic order, except for the date of XBL41-7. This subsample contains high detrital 232Th levels (2.1 × 104 ppt) and large dating uncertainties (±252 years, Table 1), which was contaminated by high detrital 232Th in the hiatus layer. The layer XBL41-7 was excluded from the established age model. A linear interpolation model was used to establish the chronology of XBL41 (Fig. 4). The result indicates that the top 6 cm part of the stalagmite grew from 1760 AD to the present (2006 AD, the time of sampling), with a short hiatus occurred after 1981 AD.

Table 1 U-Th isotopic data and 230Th dates of stalagmite XBL41
Full size table
Fig. 4
figure 4

Plot of age verses depth for XBL41. Error bars indicate 230Th dates with 2σ errors. Growth rates of the stalagmite range from 0.11 to 0.37 mm/year

Full size image

As we are unable to confirm the accurate duration of the hiatus, the period of 1760–1981 AD was selected for this study. The corrections for the initial 230Th were negligible because of high 238U concentrations (11–31 ppm) and the low 232Th levels (11–300 ppt) in the subsamples (Table 1). Most of the absolute dating errors were ±1–2 years. The growth rates of XBL41 range from 0.11 to 0.37 mm/year (Fig. 4).

4.2 δ18O record

The δ18O values of XBL41 vary between −8.8 and −11.0 ‰ with an average value of −9.7 ‰. During the last 240 years, the δ18O record shows progressively increasing trend. The δ18O values increased about 2.2 ‰ from 1760 to 1850 AD, followed by a decrease of 1.1 ‰ in the next 20 years. Since 1870 AD, the δ18O values increased again, with a trend of about 1.6 ‰ towards heavy δ18O (Fig. 5).

Fig. 5
figure 5

XBL41 δ18O record during 1760–1981 AD (solid circle) and its comparison with annual precipitation (red line, 3-point moving average) record from Kunming meteorological station

Full size image

5 Discussion

5.1 Climate significance of stalagmite δ18O

“Hendy test” results show that the δ18O values remain constant along the two growth layers of XBL41 (Fig. 6), suggesting it was most likely deposited at isotopic equilibrium fractionation conditions (Hendy 1971). Both cave temperature (close to surface annual mean temperature) and the δ18O in precipitation can affect the δ18O of stalagmite under isotopic equilibrium fractionation conditions. Temperature-dependent fractionation change [oxygen isotope fractionation of aragonite-water is 103 lnα = (17.88 ± 0.13) × 103/T − (31.14 ± 0.46); (Kim et al. 2007)] should not be the dominant factor controlling the stalagmite δ18O variations in this study. Otherwise, low temperatures in the cold Little Ice Age (15th–19th centuries (Ge et al. 2013) would enhance δ18O values, which is opposite to our δ18O record (Fig. 5).

Fig. 6
figure 6

Hendy test on two horizons of stalagmite XBL41

Full size image

Rainfall amount, moisture recycling and circulation, and changes in moisture source and transport pathway may affect the δ18O in precipitation (Breitenbach et al. 2010; Cai et al. 2010; Tan 2014). Recently, Cai et al. (2015) suggested that there were similar 850-hPa wind trajectories for both present day and Last Glacial Maximum in Xiaobailong Cave site, indicating the moisture paths from the Bay of Bengal to Xiaobailong Cave were relatively stable even on millennial- to orbital-timescales. In addition, the impact of seawater δ18O of the Indian Ocean and the Pacific Ocean on the precipitation δ18O is negligible compared with the rather large magnitude of inter-annual variations in precipitation δ18O (Tan 2014). The distance from Xiaobailong Cave to the Bay of Bengal is similar to the distance from the other potential moisture source-South China Sea in modern condition, which reduces the impacts of changes in moisture source and transport pathway (Tan et al. 2015).

Recently, a simulation result indicated a significant negative correlation between the precipitation δ18O and monsoon rainfall amount in southwest China (Liu et al. 2014). To further confirm the climate significance of our stalagmite δ18O, we directly compare the XBL41 δ18O record with the observed annual rainfall (mainly contributed by monsoonal rainfall) record. Because of the short overlapping period (~30 years, 1951–1981 AD), different resolution and the age uncertainties of XBL41 record, it is hard to get a significant correlation coefficient between the two series. However, the two series are consistent from sub-decadal- to decadal-timescales. For example, the relative dry periods in the early 1950s and the late 1960s and 1970s corresponded to higher δ18O values. In contrast, the relative wet periods in the early 1970s and middle 1960s corresponded to lower δ18O within age errors (Fig. 5). The correspondence confirms the robust anti-phase relationship between the speleothem/precipitation δ18O and monsoon rainfall amount in southwest China (Liu et al. 2014). As a result, the XBL41 δ18O record reflects Indian monsoon precipitation changes in southwest China during the last 240 years.

5.2 Indian monsoon precipitation variations in southwest China during the last 240 years

Our stalagmite δ18O record also corresponds well with a moisture record (Drought/Flood index) reconstructed from historical documents in Kunming station (CAMS 1981). Because Drought/Flood (D/F) index was defined by the method of 5-level classification, it can hardly reflect the long-term trend of rainfall variations. As a result, we compare the detrended δ18O record with the D/F index record. As shown in Fig. 7, the decadal scale variations of the stalagmite record were also recorded in the D/F index series. For example, both series recorded the wet periods in the 1770s, 1800s, 1870s and the early 1900 and 1920s.

Fig. 7
figure 7

Comparison between (a) historical Drought/Flood (D/F) index record from Kunming (CAMS 1981) and (b) detrended XBL41 δ18O record. The yellow bars highlight the periods of decreased stalagmite δ18O values and D/F index

Full size image

In general, our stalagmite δ18O result suggests that the Indian monsoon precipitation in southwest China gradually decreased in the last 240 years. The first decreasing period is from 1760 AD to 1850 AD, and the second remarkable decreasing period is from the late 19th century to present (1981 AD). The drying rate in the last century was slower than that in the period of 1760–1850 AD. Notable decadal- and sub-decadal scale droughts were identified in the 1790s, 1820s, 1850s, early 1910s and 1940s in historical time (Fig. 5).

The decreasing precipitation trend in the last 240 years is consistent with the declined lake level of Chenghai Lake in the northwest Yunnan province. Historical document recorded that the Chenghai Lake was an open lake before 1779 AD, and after that, it was closed (Xu et al. 2015). Some of the exposed lake terraces were dated in the Little Ice Age (Xu H. unpublished data), further supporting a higher than present lake level at that time. By combining the speleothem δ18O record with the observed rainfall record, our result indicates that the drought during 2009–2012 AD was the driest since 1760 AD (Fig. 5). The drought occurred along the long-term drying trend, which means it was extremely severe.

5.3 Dipolar structure of Indian monsoon precipitation during historical time

Meteorological observation records indicate that the trends of the Indian monsoon rainfall over different regions are different during the last hundred years (Guhathakurta and Rajeevan 2008; Mishra et al. 2012). The region south of 20°N, receiving moisture from the Arabian Sea shows an increasing rainfall trend in monsoon season. On the contrary, northeastern India, which receives moisture from the Bay of Bengal, shows a decreasing rainfall trend (Guhathakurta and Rajeevan 2008; Konwar et al. 2012; Mishra et al. 2012). Recently, a stalagmite, JHU-1, from central India with decreasing δ18O values since the late 17th century, demonstrates a shift towards enhanced monsoon precipitation in this region. In contrast, increasing δ18O values were observed in a stalagmite, WS-B, from northeastern India during this period, indicating declined regional monsoon precipitation (Sinha et al. 2011). Moreover, a 2000-years long stalagmite δ18O record from northwestern India, where both receive moistures from the Arabian Sea and the Bay of Bengal, shown both similarities and discrepancies with the JHU-1 record from central India (Sinha et al. 2015). Therefore, it is crucial to confirm the dipolar structure of the precipitation of the two branches of Indian monsoon, as well as its temporal and spatial extension during historical time. Because there is no age control in the last 200 years of the WS-B record from northeastern India, more evidences are needed.

In this paper, XBL41 record is compared with the JHU-1 record from central India. As shown in Fig. 8, an anti-phase relationship of the two records was observed during the last 240 years. The increasing trends of XBL41 δ18O values during the periods of 1760–1850 AD and from the late 19th century to present correspond to the decreasing trends of JHU-1 δ18O records during these periods (Fig. 8). From 1850 to 1870 AD, the XBL41 δ18O values abruptly decreased, while the JHU-1 δ18O values increased drastically. Tree ring δ18O reconstructed monsoon precipitation in Nepal Himalaya (Sano et al. 2012), with moisture source from the Bay of Bengal, shows similar variations with the XBL41 δ18O record (Fig. 8). Declining monsoon precipitation was observed in the tree ring δ18O records from southern (Griebinger et al. 2011) and southeastern part of Tibetan Plateau (Wernicke et al. 2015) too. In addition, the long-term trend of the XBL41 record is consistent with that of the South Asia summer monsoon (e.g. Indian monsoon) index record (Shi et al. 2014), since the monsoon rainfall variability, with moisture from the Bay of Bengal, can well represent the Indian summer monsoon intensity variability (Gadgil 2003).

Fig. 8
figure 8

Comparisons of (a) tree ring δ18O record from Nepal Himalaya (Sano et al. 2012), (b) stalagmite XBL41 δ18O record from southwest China, and (c) stalagmite JHU-1 δ18O record from central India (Sinha et al. 2011). The dark yellow arrows represent the linear trends of different periods in the three records, respectively. The gray lines in panels b and c are observed monsoon rainfall records (3-point moving average) of central northeast India and peninsular India during 1880–2000 AD, respectively (data comes from Indian Institute of Tropical Meteorology). Observed records also show long-term regional differences in Indian monsoon rainfalls during the last century

Full size image

In summary, comparisons of the XBL41 record with other paleoclimate records in the region confirm the observed dipolar structure of Indian monsoon precipitation variability (Konwar et al. 2012; Mishra et al. 2012) on multi-decadal- to centennial- scales during historical time.

5.4 Controlling factors of Indian monsoon precipitation in southwest China

Many studies suggested that the tropical sea surface temperature (SST) variations have important influences on the spatial–temporal variations of Asian monsoon rainfall (Wang 2006). In particular, the ENSO’s impact on the Indian monsoon circulation has been studied extensively (Berkelhammer et al. 2014; Cherchi and Navarra 2013; Mishra et al. 2012; Park et al. 2010; Webster et al. 1998). It was evident that the teleconnection between ISM precipitation and ENSO has undergone a protracted weakening since the late 1980s (Kumar et al. 1999). As a notable example, the 1997/1998 El Nińo event, despite its intensity, produced only marginal rainfall anomalies over India (Kucharski et al. 2007). However, Chen and Yoon (2000) demonstrated that ENSO could significantly influence the moisture conditions over Indo-china on inter-annual timescale. Buckley et al. (2007) also suggested that ENSO might be the major contributing factor to decadal scale drought over northwestern Thailand. Ummenhofer et al. (2013) stated that the drought patterns across monsoon and temperate Asia over the period 1877–2005 were linked to Indo-Pacific climate variability associated with the ENSO and the Indian Ocean Dipole (IOD). Recently, Zhang et al. (2013) suggested that the drought in southwest China during autumn 2009 was caused by a nonconventional El Nińo, with the maximum SST anomalies confined to the central equatorial Pacific Ocean. The XBL41 record also demonstrated the influence of tropical ocean SST and ENSO on the monsoon precipitation in southwest China during the last 240 years.

As show in Fig. 9, there is a long-term increasing trend of tropical Indo-pacific Ocean SST during the last two century (Wilson et al. 2006). Meanwhile, the increase of temperature in continent where the Indian low pressure develops may be smaller than that over the sea surface, which was resulted from increasing aerosol cooling effect and shrinking Himalayan snow cover (Ramanathan et al. 2005; Zhao and Moore 2006; Xu et al. 2012). The decreasing land-sea thermal gradient may reduce the moisture transport from the Bay of Bengal and cause the long-term decreasing of rainfall in southwest China.

Fig. 9
figure 9

Comparison between (b) stalagmite XBL41 δ18O record and (a) tropical Indo-Pacific ocean SST anomaly (Wilson et al. 2006). The blue arrows in a and b represent the linear trends of the two records

Full size image

The low-level cross-equatorial flow in the western Indian Ocean is an essential component of the Indian summer monsoon (Wang 2006). Compared to that of the Bay of Bengal, the major contribution for the moisture fluxes to the Arabian Sea comes from the southern Indian Ocean brought by the cross-equatorial flow (Kumar et al. 1999). A marine record from northwest Arabian Sea reveals intensified upwelling in the last 240 years (Anderson et al. 2002), indicating strengthening northward cross-equatorial moisture transport. The intensified northward cross-equatorial moisture transport could enhance the monsoon rainfall in peninsular and west central India (Puranik et al. 2014; Sinha et al. 2011). Modern observation data indicate the low level wind speed and moisture transport over the Arabian Sea increased, which caused increased rainfall in peninsular and west central India during the last 30 years. In contrast, decreasing trends of low level wind speed and moisture transport from the Bay of Bengal were observed (Konwar et al. 2012). The observation further supports our explanation.

On decadal timescale, the detrended XBL41 record shows similar patterns to the Nino3 SST reconstructions (D’Arrigo et al. 2005; Mann et al. 2000), as well as the Southern Oscillation index series (Stahle et al. 1998), despite some discrepancies among the different proxy reconstructions (Fig. 10). The Person correlation coefficient between XBL41 record and the Nino3 index record (10-point moving smooth) reconstructed by Mann et al. (2000) is 0.29 (p < 0.05). A significant negative correlation coefficient (r = −0.28, p < 0.05) was also observed between XBL41 record and the Southern Oscillation Index series (10-point moving smooth, Stahle et al. 1998). The comparisons suggest that the decadal variability of east Pacific SST significantly influences the monsoon rainfall in southwest China. In decadal El Nińo-like conditions, droughts occurred in southwest China. In contrast, enhanced monsoon rainfall in southwest China corresponded to decadal La Nińa-like conditions (Fig. 10). Tropical Pacific SST’s impact on the monsoon rainfall in southwest China was thought to be related to the east–west displacement of the ascending and descending branches of the Walker circulation (Ropelewski and Halpert 1987). The abnormal ocean warming during El Nińo-like conditions may cause the ascending branch of Walker circulation move eastward. As a result, the descending branch broadly distributed over northeast India and southwest China, suppressing monsoon rainfalls in these regions (Kumar et al. 2006; Ropelewski and Halpert 1987).

Fig. 10
figure 10

Comparisons of (a) detrended XBL41 δ18O record with (b) Nino3 index record (10-point moving smooth, Mann Nino3 index) reconstructed by Mann et al. (2000), (c) Nino3 index record (10-point moving smooth, Cook Nino3 index) reconstructed by D’Arrigo et al. (2005), and (d) Southern Oscillation index series (10-point moving smooth, Stahle et al. 1998). The shaded yellow bars indicate drought events in southwest China and their correlation with various series

Full size image

It was observed that at annual scale, the inverse relationship between the El Nińo events with the warmest SSTs in the eastern equatorial Pacific and the Indian monsoon rainfall broke down in recent decades (Kumar et al. 1999). In contrast, the El Nińo events with the warmest SST anomalies in the central equatorial Pacific are more effective in forcing drought and producing subsidence over India (Kumar et al. 2006; Ratnam et al. 2010; Wang et al. 2015). However, there are similar variations (r = 0.97, N = 707, p < 0.01) in SST of eastern and central equatorial Pacific in the long term (1300–2006 AD, Cook et al. 2008). Climate model simulation results suggest that the tropical Pacific SST gradient will decrease under conditions of global warming, resembling El Nińo-like SST patterns (An et al. 2012; Liu et al. 2013). Multi model ensembles of CMIP3 and CMIP5 also show consistently weak and eastward shift of the Walker Circulation along the equator under global warming condition (Bayr et al. 2014). Consequently, there might be a decreasing trend of monsoon precipitation in southwest China under the El Nińo-like conditions under the scenario of global warming in the future. Therefore, it is crucial to come up with adaptive strategies to prepare for potential declines of precipitation and/or drought events in southwest China in the future.

6 Conclusions

  1. 1.

    The high-resolution δ18O record of an absolutely dated stalagmite from Xiaobailong Cave in Yunnan province, China reveals an overall decreasing trend in monsoon precipitation in southwest China during the last 240 years. By combining the speleothem δ18O with the observed rainfall record, we suggested that the recent drought in 2009–2012 AD was the driest since 1760 AD. The drought occurred in a long-term drying trend, which means it is extremely severe.

  2. 2.

    Our speleothem record is consistent with monsoon precipitation records in Nepal Himalaya and southeastern Tibetan Plateau reconstructed from tree rings but anti-correlated with a speleothem record from central India. The comparisons confirm the observed anti-phase variations of Indian monsoon precipitation with moisture sources from the Bay of Bengal and Arabian Sea on multi-decadal to centennial- timescales during historical time.

  3. 3.

    On centennial scale, the warming of the tropical ocean during the last 240 years may have decreased the land-sea thermal gradient and reduced the moisture transport from the Bay of Bengal, therefore reduced the monsoon rainfalls in southwest China. On decadal scale, El Nińo-like conditions of east Pacific SST may also have significant influences on droughts in southwest China.