A 200-year annually laminated stalagmite record of precipitation seasonality in southeastern China and its linkages to ENSO and PDO

In southeastern China (SEC), the precipitation amount produced by the East Asian summer monsoon (EASM) is almost equivalent to that during the non-summer monsoon (NSM) period, both of them significantly affecting agriculture and socioeconomy. Here, we present a seasonally-resolved stalagmite δ18O record (δ18Os) for the interval 1810–2009 AD from E’mei cave, Jiangxi Province, SEC. The comparison between δ18Os and instrumental data indicates that the δ18Os variability is primarily controlled by the precipitation seasonality (i.e., the ratio of EASM/NSM precipitation) modulated by the El Niño/Southern Oscillation (ENSO) on interannual to interdecadal timescales. Higher (lower) δ18Os values thereby correspond to lower (higher) EASM/NSM ratios associated with El Niño (La Niña) events. Significant correlations with ENSO and the Pacific Decadal Oscillation (PDO) indicate that the precipitation seasonality in SEC is remarkably influenced by ocean-atmosphere interactions, with lower (higher) EASM/NSM ratios during warm (cold) phases of ENSO/PDO. The progressive increase in δ18Os since 2005 AD may reflect a strengthening of the central Pacific El Niño under continued anthropogenic global warming. The relationship between seasonal precipitation and δ18Os with ENSO/PDO requires further studies.

. Therefore, the variation in δ 18 O w during 1988-1992 is probably influenced by the ratio of EASM/NSM precipitation but not by the amount of annual precipitation which lacks a significant variation ( Supplementary Fig. S3).

Results
Chronology. Elevated Table S1 and Fig. S4) and hinder precise age models. Although the dating errors are large, the 230 Th dates still indicate that stalagmite EM1 grew during about last 200 years, which is consistent with the 200 ± 3 well-developed laminae within errors (see next paragraph). The 230 Th dates at 6 and 18.5 mm from the top are 1997 ± 267 and 1932 ± 158 AD, respectively, and the inferred age of the top of EM1 is 2028 ± 267 AD. This is consistent with the observation of active dripping when this stalagmite was collected in December 2009. The top age of EM1 was therefore assumed to be late 2009 AD.
The cross-section of EM1 shows continuous, well-developed laminae consisting of alternating dense, translucent sub-layers (TDSL) and white, porous sub-layers (WPSL) 21,22 (Supplementary Fig. S1). The visible WPSL are opaque under the transmitted-light microscope, but brightly luminescent under mercury light source UV reflected light and in the confocal laser fluorescent microscope (CLFM) ( Supplementary Fig. S1), which is consistent with observations from other caves 21,23 . As suggested by previous studies, the fluorescent sub-layer forms from organic-rich dripwater during the wet season, while dripwater depleted in organics during the dry season results in the non-fluorescent sub-layer 23 . We identified a total of 159 δ 18 O s cycles between 0 and 45.6 mm from the top of EM1, consistent with 159 bright-dark couplets, confirming the annual nature of these highly regular laminae ( Supplementary Fig. S5b and c). 200 ± 3 bright-dark couplets were counted using CLFM between 0 and 53.6 mm from the top of EM1. The number of annual laminae and δ 18 O s cycles was used to establish an age model for EM1 (Supplementary Figs S4 and S5), considering that EM1 grew continuously from 1810 to 2009 AD, which is consistent with the 230 Th dates within errors.
In order to constrain the chronology of EM1, we also compared the δ 18 O and δ 13 C records of EM1 with stalagmite YQ15-1, which was obtained from Yongquan Cave, 50 km northeast of E'mei Cave. The dating errors of YQ15-1 (12-15 years, Supplementary Fig. S5a) are much smaller than those of EM1. The δ 18 O and δ 13 C records of YQ15-1 and EM-1 replicate well ( Supplementary Fig. S5a) Fig. S4). We therefore interpolated this interval by using the corresponding relationship between the micromilling track of stable isotope analyses and layer counting ( Supplementary Fig. S1b). Together with the 159 annual δ 18 O s cycles between 0 and 45.6 mm, δ 18 O and δ 13 C records of EM1 were established for the period 1810-2009 AD (Fig. 2a). The good replication of the stable isotope records between EM1 and YQ15-1 indicates that EM1 was deposited close to isotope equilibrium conditions 24,25 , although "Hendy tests" 26 yielded a significant correlation (r = 0.31, p < 0.01, n = 780) between δ 18 O and δ 13 C values along the growth axis. The co-variation of δ 18 O and δ 13 C can also result from the climatic and environmental change 24 . The inverse relationship between δ 18 O and δ 13 C values on seasonal timescales ( Supplementary Fig. S5b) further indicates that kinetic isotope fractionation has a negligible effect only and the δ 18 O s signal recorded in EM1 is primarily of climatic origin.

Discussion
Significance of δ 18 O s proxy. Although δ 18 O s variability in the monsoon region of China on orbital to millennial timescales generally reflects the variation of EASM intensity, its significance on short timescales remains to be understood. Previous studies demonstrated that the δ 18 O s variability in the marginal zone of the EASM reflects the variation of monsoon intensity or precipitation amount 22,27,28 . In SEC, however, this relationship is more complex, because of the various factors influencing precipitation δ 18 O (δ 18 O p ), e.g., precipitation amount, moisture sources, moisture pathways and rainfall seasonality. Dayem et al. 29 suggested that several processes associated with precipitation seasonality might jointly influence the fluctuation of Chinese δ 18 O s 29 . Based on the analyses of modern precipitation δ 18 O data 30 , Tan 31 suggested that δ 18 O s in the monsoon region of China records the moisture source changes related to ENSO circulation but not the monsoon intensity or precipitation amount. The higher δ 18 O s values reflect more moisture originating from the adjacent Pacific Ocean relative to the remote Indian Ocean during El Niño phases, and vice versa 31 . However, some studies suggested that EASM precipitation is primarily derived from the Indian Ocean, while the moisture from the West Pacific Ocean is a minor contributor to monsoonal rainfall only 32,33 , and the δ 18 O p variability is caused by both precipitation amount and moisture transport history 32 . Cheng et al. 34 and Yang et al. 35 suggested that δ 18 O s is primarily controlled by large-scale monsoon intensity and upstream precipitation. As introduced above, δ 18 O w in the study area is influenced by both the EASM and the NSM associated with ENSO circulation (Supplementary Fig. S2). Thus, the EM1 δ 18 O s record was compared to the instrumental seasonal precipitation and the Southern Oscillation Index (SOI) in order to explore its significance on interannual to interdecadal timescales.
On an interannual timescale, the comparison between δ 18 O s and the SOI shows a significantly negative correlation (Fig. 3a, r = −0.50, p < 0.01 using a 5-year moving average for 1951-2004 AD), with higher (lower) δ 18 O s during El Niño (La Niña) events. This is consistent with previous studies based on instrumental data 30,33,35 and δ 18 O records of tree-ring cellulose [18][19][20]36 and stalagmites 37,38 in SEC. δ 18 O s is significantly and negatively correlated with the ratio of EASM/NSM precipitation (r = −0.32, p < 0.05) using a 5-year moving average for 1951-2009 AD (Fig. 3a). There is no significant correlation between δ 18 O s and the EASM, NSM and annual precipitation, although a good coherent variation can be observed (grey bars in Fig. 3a). We also compared the interannual variation between EM1 δ 18 O s and the local Drought/Flood (D/F) index for 1810-2000 AD. The annual D/F index, based on drought or flood events in spring, summer and autumn recorded in historical documentaries 39 , can reflect local precipitation variations in May-September, with a decreased (increased) D/F index representing more (less) precipitation. The good coherence indicates that lower (higher) δ 18 O s values are associated with increased (decreased) EASM precipitation in the study area (Fig. 2b). PDO and ENSO are strongly correlated on decadal timescales 11 , with warm PDO phases (positive PDO index) corresponding to warm ENSO phases (El Niño phases, negative SOI index). On an interdecadal timescale, previous studies demonstrated that the PDO/ENSO circulation shifted from a cold phase during 1951-1976 AD to a warm phase during 1977-2000 AD 7,11 . During the cold phase of PDO/ENSO, EM1 δ 18 O s values are generally lower than during the subsequent warm phase (green arrows in Fig. 3a), whereas precipitation amount does not show a significantly systematic shift (Fig. 3a). Actually, studies demonstrated that the intensity and duration of the EASM have a strong variability on interdecadal timescales 40 . The EASM starts late (early) and tends to be weaker (stronger) during El Niño (La Niña) years 41 , and the onset varies from late April to early June 40,41 . Thus, May-September precipitation cannot be always used to represent the EASM precipitation, especially in SEC characterized by a strong precipitation seasonality. This is also supported by instrumental δ 18 O p data from the Changsha GNIP station (Supplementary Fig. S6). They show that the δ 18 O p values in May during La Niña years (1988( -1989 are lower than those during El Niño years (1991-1992 AD), indicating that the EASM starts early (late) during La Niña events (El Niño events). Therefore, EASM precipitation during 1951-1976 AD likely represented May-September precipitation, while it shifted to June-September precipitation during 1977-2010 AD; the NSM precipitation should be adjusted accordingly. Hence, less EASM and more NSM precipitation results in higher δ 18 O s values during long-term El Niño phases (Fig. 3b) and vice versa.
Similarly, the poor correlation between δ 18 O s and the D/F index during 1930-1970 AD on an interdecadal timescale (Fig. 2b) might be caused by the variation of precipitation seasonality associated with shifts in oceanic-atmospheric circulation. We calculated the correlation coefficients between δ 18 O s and the adjusted EASM and NSM precipitation as well as EASM/NSM ratios using a 5-year moving average for 1951-2009 AD. The data show that δ 18 O s is significantly negatively correlated with the ratio of EASM/NSM precipitation (r = −0.67, p < 0.01) and the EASM precipitation (r = −0.54, p < 0.01), and is significantly positively correlated with the NSM precipitation (r = 0.58, p < 0.01), suggesting that higher (lower) δ 18 O s values are associated with lower (higher) EASM/NSM ratios during El Niño (La Niña) phases on interannual to interdecadal timescales.
Considering the strong precipitation seasonality (including precipitation amount and moisture sources/ pathways) in the study area, we suggest that the EASM/NSM ratio better explains the δ 18 O s variation on interannual-interdecadal timescales. Studies demonstrated that EASM precipitation is mainly sourced from the remote Indian Ocean while NSM precipitation primarily originates from the adjacent Pacific Ocean 32,33 . Thus, the EASM/NSM ratio may also represent the ratio of moisture sources/pathways during the EASM season to that of the NSM season, partly similar to the "circulation effect" proposed by Tan 31 . He suggested that the δ 18 O p and δ 18 O s values are affected by the variation of moisture source in the summer monsoon season. We also emphasize the influence of NSM precipitation, which accounts for 45~65% of the annual precipitation in the study area. As a consequence, we suggest that the variation in precipitation seasonality (e.g., EASM/NSM ratio) modulated by ENSO plays a key factor in controlling the δ 18 O s variability in the study area on interannual to interdecadal timescales, with higher (lower) δ 18 O s values reflecting lower (higher) EASM/NSM ratios associated with El Niño (La Niña) phases.  Fig. S7).

Links between
A comparison of available high-resolution δ 18 O s records from annually laminated or high-precisely 230 Th-dated stalagmites in the monsoon region of China shows a consistent long-term increasing trend (Fig. 5), indicating a gradual weakening of the EASM intensity associated with the large-scale weakening of the Asian summer monsoon 28 . However, these δ 18 O s records show differences in long-term trends and amplitude. For example, δ 18 O s records from Wanxiang, Dayu and Dongge caves do not show the long-term increasing trend during last 200 years, and their interdecadal amplitudes are smaller than in other records (Fig. 5). The dating errors of the records from Wanxiang, Dayu, Xiaobailong and Dongge caves are less than 5 years, except for Dongge record. The records from Shihua, Wuya, Heshang and E'mei caves are based on annual layer counting combined with 230 Th dates. Therefore, these differences should not result from the chronology uncertainties. If δ 18 O s is influenced by the "circulation effect", δ 18 O s in SEC, a region more sensitive to circulation changes, should show larger amplitudes. Actually, the long-term δ 18 O s amplitudes of stalagmites from Heshang, Dongge and E'mei caves are smaller than those from Shihua and Wuya caves in North China and Xiaobailong cave in Northwest China (green dotted lines in Fig. 5). In addition, the interdecadal variations are not coherent in these δ 18 O s records from the monsoon region of China, and similarly inconsistent changes are also seen in other studies on decadal-centennial timescales 25,47 . This further demonstrates that δ 18 O s is not only affected by circulation but also by precipitation amount, because the migration of the EASM rainbelt results in different precipitation patterns in the monsoon region of China. Except for the variations in EASM precipitation, changes in NSM precipitation also contribute to the interdecadal divergences seen in the EM1 δ 18 O s record.
What is the relationship between δ 18 O s and seasonal precipitation amount in the study area during different PDO/ENSO phases? On an interannual timescale lower δ 18 O s values are associated with more EASM precipitation during La Niña phases (Figs 2b and 3a) and vice versa. By analyzing the precipitation variability in East China during 14 El Niño events since 1950s Kong and Tu 48 found that less EASM precipitation in May-September in the lower reaches of YRV are associated with El Niño events. Xu et al. 20 showed that the interannual variability in tree-ring cellulose δ 18 O data in SEC reveals the variability in May-October precipitation or relative humidity associated with ENSO, with higher (lower) δ 18 O values corresponding to less (more) precipitation during El Niño (La Niña) events. Studies based on instrumental and reanalysis data suggest increased spring 5,13 and EAWM precipitation 12,14,15 in SEC associated with El Niño events, indicating increased NSM precipitation during El Niño events on interannual timescales. Therefore, we suggest that the interannual variability recorded by the EM1 δ 18 O s data might also reflect the variability in seasonal precipitation amount, with more (less) EASM and less (more) NSM precipitation during La Niña (El Niño) phases. On interdecadal timescales, many studies demonstrate that more summer monsoon precipitation in June-August in the YRV region results from warm PDO/ENSO phases, and vice versa 7,8,49,50 . These studies indicate that decreased summer rainfall in North China and increased summer rainfall in South China (called "north drought-south flood") 49 result from a strong EASM during cold phases of PDO/ENSO, and vice versa. However, Chan and Zhou 11 found that less (more) early summer monsoon precipitation in May-June in South China results from the warm (cold) phases of PDO/ENSO. This indicates that the pattern of interdecadal variability of June-August precipitation is opposite to that of May-June precipitation associated with PDO/ENSO. Yang et al. 9 also demonstrated that the effect of the PDO on precipitation anomalies in East China is depends on the month: May-June precipitation patterns in East China are opposite July-August or annual mean patterns. In the study area, the interdecadal variation of precipitation in May also shows an opposite trend to that in June-September for 1951-2009 AD ( Supplementary  Fig. S8). This further supports our hypothesis that either May-September or June-August precipitation should be used to represent EASM precipitation under different phases of PDO/ENSO (Fig. 3b). Recent observations and simulations indicate that increased (decreased) spring 6 and EAWM 17 precipitation in SEC occurred during warm (cold) phases of ENSO/PDO on interdecadal timescales, indicating more (less) NSM precipitation during warm (cold) phases of ENSO/PDO. This is consistent with our observations (Fig. 3a,b). Therefore, on interdecadal timescales, we suggest that more EASM and less NSM precipitation leading to higher EASM/NSM ratios results in lower δ 18 O s values during the cold phases of ENSO/PDO, and vice versa (Fig. 3b).
The δ 18 O s values gradually increase during the late 20th century reaching a maximum of the entire 200-year record. A similar trend was observed in a tree-ring δ 18 O record from Taiwan 36 and a tree-ring δ 13 C record from SEC 51 . This trend might be caused by an increased frequency of the central Pacific El Niño (CP El Niño) under continued anthropogenic climate warming during recent decades 36,52 . The anomalous low-level anticyclone and Walker circulation associated with CP El Niño events contributes to anomalously dry conditions over southeastern Asia and an intensified winter monsoon with a deeper East Asian trough at 500 hPa resulting in lower temperatures in South China 53 . CP El Niño events can also lead to decreased spring precipitation in South China and increased spring precipitation in the lower reaches of Yangtze River 13 . More NSM precipitation might also contribute to increased δ 18 O s values. We suggest that the highest δ 18 O s values during the late 20th century result from dry conditions in summer, consistent with the results based on tree-ring δ 13 C data from SEC 51 . The synchronously increased δ 13 C values in the EM1 and YQ-15 stalagmites (Supplementary Fig. S5) probably indicate degraded vegetation due to these dry conditions 54 , further supporting our conclusion.
Methods 230 Th dating. Four subsamples (50-100 mg), drilled using a hand-held carbide dental drill, were analyzed at the Institute of Global Environmental Change, Xi'an Jiaotong University. The chemical procedure used to separate uranium and thorium was similar to that described in Edwards et al. 55 . Uranium and thorium isotopes were measured using a multicollector-inductively coupled plasma-mass spectrometer (MC-ICP-MS). Details about the instrumental setup are explained in Cheng et al. 56 .

Stable isotope measurements.
For stable isotope analyses (δ 18 O and δ 13 C), a total of 780 subsamples were micro-milled along the central axis of stalagmite EM1. 370 subsamples were micro-milled at a spatial resolution of 0.1 mm from 0 to 13 mm, and 0.05 mm from 13 to 25 mm distance from the top. These samples were measured using a MAT253 isotope ratio mass spectrometer equipped with a MultiPrep system at Xi'an Jiaotong University. Analytical precision of the δ 18 O and δ 13 C analyses was better than 0.06‰ and 0.03‰ (1σ), respectively. 410 subsamples were micromilled along the right side of central axis of stalagmite EM1 at a resolution of 0.07 mm from 25 to 53.7 mm distance from the top, and measured on a Thermo Fisher Delta plus XL isotope ratio mass spectrometer linked to a Gasbench II at the University of Innsbruck. Analytical precision was 0.08‰ and 0.06‰ for δ 18 O and δ 13 C, respectively (1σ) 57 . All isotopic values are reported in the δ notation relative to the Vienna Pee Dee Belemnite (VPDB) standards.
Layer counting and δ 18 O cycles counting. On the polished profile of stalagmite EM1, visible laminae consisting of a translucent, dense sub-layer (TDSL) and a white, porous sub-layer (WPSL) can be observed ( Supplementary Fig. S1). EM1 was imaged using a confocal laser fluorescent microscope (CLFM) at the State Key Laboratory for Manufacturing Systems Engineering, Xi'an Jiaotong University. The visible WPSL is brightly fluorescent while the visible TDSL is non-fluorescent ( Supplementary Fig. S1). We identified 200 ± 3 fluorescent laminae between 0 and 53.6 mm from the top of EM1. The 3-year error is a cumulative error from top to bottom. Because the mean growth rate of EM1 between 0 and 45.6 mm from the top is 0.29 mm/year, we obtained a seasonally-resolved δ 18 O record in this section. We counted 159 annual δ 18 O cycles between 0 and 45.6 mm from the top of EM1, which is consistent with the layer-counting result using the CLFM. Data sources. Monthly instrumental data from the Nanchang and Jiujiang meteorological station during the period 1951-2010 were obtained from the National Climate Center (NCA, http://ncc.cma.gov.cn). Monthly precipitation δ 18 O data (1988-1992) from Changsha stations were obtained from the Global Network for Isotopes in Precipitation (GNIP, http://www.iaea.org/). The Southern Oscillation Index (SOI) was obtained from http://www. bom.gov.au/climate/current/soihtm1.shtml. Positive SOI values represent El Niño events and negative values represent La Niña events. The Pacific Decadal Oscillation (PDO) index, defined as the leading standardized principal component of monthly sea-surface temperature (SST) anomalies in the North Pacific Ocean, was taken from http://www.ncdc.noaa.gov/teleconnections/pdo. The west Pacific subtropical high (WPSH) plays an important role in governing the variability over East China. The WPSH strength (WPSH s ) index 58 is used to represent the intensity of WPSH, with stronger WPSH s concurrent with a south-westward extension of WPSH and vice versa.

Spectral analyses.
Were computed using REDFIT software 59 on the PAST platform 60 . The 90% confidence level is shown in Supplementary Fig. S7.