E’mei Cave (115°29′44″E, 29°33′18″N, 53 m a.s.l.) is located in Henggang Town of Ruichang City, northern Jiangxi Province (Figure 1A). The thickness of the cave roof is approximately 10–30 m.

The study area is dominated by subtropical humid and monsoon climate with strong rainfall seasonality. Spring persistent rain (SPR) lasts from March to May, and the Meiyu period with high temperature and heavy rainfall occurs in June. In contrast, precipitation significantly declines from July to August under the control of the western Pacific subtropical high, which is reflected as high temperature and little rainfall. Precipitation is the lowest during the period from autumn to winter. According to instrumental data (1951–2010 AD) from the Jiujiang meteorological station near E’mei Cave (∼50 km from the cave), the region has an annual mean precipitation of 1426.9 mm and an annual mean temperature of 17.5°C. The spring (March to May) and summer (June to August) precipitation account for 35.8% and 34.6% of the annual precipitation, respectively. According to data from the nearest Global Network for Isotopes in Precipitation (GNIP) station (Changsha station, ∼100 km west of E’mei Cave) during 1988–1992 AD, March-May precipitation (Spring rain) δ¹⁸O values are higher and June-September precipitation (East Asian summer monsoon precipitation) δ¹⁸O values are lower (Figure 1B).

A rapidly growing calcitic stalagmite (EM1) of 60 mm in height was collected in December 2009 from an active drip site located ∼150 m behind the entrance of E’mei Cave. The top of EM1 is smooth.

The sample was halved along its growth axis and then polished. The polished profile shows continuous laminations; the annual growth laminae are all clearly visible to the naked eye, alternating between translucent, dense sub-layers (TDSL) and white, porous sub-layers (WPSL) (Genty and Quinif, 1996; Tan et al., 2014b) (Supplementary Figure S1). X-ray diffraction (XRD) analyses suggest that the mineral composition of EM1 is pure calcite.

Four sub-samples were drilled from polished sections of EM1 using a carbide dental drill for ²³⁰Th dating. These subsamples were dated on a Multi-Collector Inductively Coupled Plasma Mass Spectrometry at the Institute of Global Environmental Change, Xi’an Jiaotong University. The laminae of EM1 were counted according to the scan image using a confocal laser fluorescent microscope at the State Key Laboratory for Manufacturing Systems Engineering, Xi’an Jiaotong University. The age model of EM1 was established by combining the ²³⁰Th dating results, δ¹⁸O cycle counting, and annual lamina counting. Further details on the age model are provided by Zhang et al. (2018).

In order to minimize the sampling error, a 20.4 mm-long section (3.5–23.9 mm from the top) of EM1 with very clear annual growth laminae was selected (Supplementary Figure S1). Each annual lamina of this section consists of a translucent, dense and white, porous sub-layer couplet. The growth laminae are quite parallel with less visible pores. A thin section (0–28.5 mm from the top) was also prepared and examined using the Zeiss Axioscope5 microscope.

A total of 204 sub-samples were milled using a high-resolution Micromill device at an interval of 0.1 mm at 3.5–23.9 mm from the top of the polished EM1 profile. The correspondence relationship between each subsample and translucent-white couplet was recorded during the sampling process. Each subsample was divided into two parts for analyses: one part was measured for δ¹⁸O and δ¹³C using a Delta V isotope ratio mass spectrometer connected with a Kiel-IV system at Xi’an Jiaotong University and the other part was analyzed for Mg/Ca, Sr/Ca, and Ba/Ca using the Agilent 5110 ICP-OES at the Institute of Earth Environment, Chinese Academy of Sciences. The laboratory standard TTB1 was analyzed after every 10 subsamples to ensure the stability of the stable isotope data. The analytical precision of the δ¹⁸O and δ¹³C analyses was 0.07‰ and 0.05‰ (1σ), respectively. All isotopic values are reported in the δ notation relative to the Vienna Pee Dee Belemnite (VPDB) standards. Throughout the trace element measurements, the subsample was dissolved in 5% nitric acid in cleaned centrifuge tubes. A standard sample was inserted after every 5 subsamples and a set of standard samples was inserted after every 15 subsamples for calibration. The relative standard deviations (RSD) were less than 5% for Mg/Ca, Sr/Ca, and Ba/Ca.

Average monthly precipitation and temperature data from three near meteorological stations (Jiujiang, Lushan and Nanchang) for the period 1951–2010 AD were obtained from the National Climate Center (NCA, http://ncc.cma.gov.cn). The D/F index records of Jiujiang region was reconstructed from drought and flood events during spring, summer, and autumn recorded in historical literature, reflecting the local precipitation variation. Monthly precipitation δ¹⁸O data (1988–1992 AD) from the GNIP Changsha stations were obtained from the GNIP (http://www.iaea.org/). Both Changsha station and the study area are located in the spring persistent rain area, their precipitation δ¹⁸O variabilities are consistent (Zhang et al., 2020). Monthly simulated δ¹⁸O data of precipitation (1979–2015 AD) were acquired from a water isotope-permitting general circulation model (IsoGSM) (Yoshimura et al., 2008). IsoGSM has been proven to be capable of producing reliable data for eastern China (Zhang et al., 2020). A comparison between the precipitation δ¹⁸O data (1988–1992 AD) from the GNIP Changsha station and those from IsoGSM simulation showed consistent variations, with a significantly positive correlation [r=0.754; 95% confidence interval (0.614; 0.848)], although the amplitude of the IsoGSM simulated δ¹⁸O was slightly smaller than that of the GNIP observation data (Supplementary Figure S2). Multivariate El Nino-Southern Oscillation Index (MEI) and West Pacific Subtropical High (WPSH) data (1951–1990 AD) were obtained from the Earth System Research Laboratory, National Oceanic & Atmospheric Administration (NOAA) (http://www.cdc.noaa.gov/people/klaus.wolter/MEI). The Pacific Decadal Oscillation (PDO) Index (1951–1990 AD) was also obtained from NOAA (http://www.ncdc.noaa.gov/teleconnections/pdo).

Under isotope equilibrium fractionation, the δ¹⁸O values of calcite are mainly determined from the δ¹⁸O values of drip water and cave temperature, while the δ¹⁸O values of drip water are derived from the δ¹⁸O of precipitation. In order to calculate the fraction of stalagmite δ¹⁸O attributable to precipitation δ¹⁸O, the temperature dependent oxygen isotope fractionation factor between calcite and water established by Kim and O 'Neil (1997) was applied (Equation 1) (Kim and O ′Neil., 1997). δ 18 O c a l c i t e V P D B = δ 18 O d r i p   w a t e r V P D B + 18.03 10 3 T − 32.42 (1)

Where T is the absolute temperature in K (T=273+t°C), t is average annual temperature of the cave in °C. In this study, the IsoGSM simulated precipitation δ¹⁸O was used as drip water δ¹⁸O. It should be noted that the unit of precipitation δ¹⁸O is relative to the standard mean ocean water (VSMOW) and the unit of speleothem δ¹⁸O is relative to Pee Dee Belemnite (VPDB), and they need to be transformed into a consistent unit (Equation 2). δ 18 O V P D B = δ 18 O V S M O W − 30.864 / 1.030864 (2)

In order to better analyze the seasonal variations of trace element ratios and stable isotopes, the long-term trend of each record was removed using a data detrending method. PearsonT3 was used to calculate the linear Pearson correlation coefficient between different records, which provides the valid corrected 95% confidence intervals in the case of autocorrelation. Beyond that, the confidence interval can be used as a significance test by verifying whether it contains zero.

EM1 consists of couplets of alternating TDSL and WPSL under natural light (Figure 2). WPSL corresponds to the opaque layer under transmitted light and it is weak to the strong luminescence band under fluorescence (Figure 2). WPSL has a higher abundance of opaque particles under transmitted and cross-polarized light. These particles glow under fluorescence and may be clay-like impurities or organic matter (Figures 2B,C). Several wide fluorescent layers include multiple extremely thin fluorescent sublayers (Figure 2C). The thin-section image in Figure 2A shows that both sub-layers are composed of typical columnar and fibrous calcite crystalline structures under cross-polarized light.

EM1 δ¹⁸O and δ¹³C records showed clear annual cycles, with δ¹⁸O values varying from −8.87‰ to −5.67‰ and δ¹³C values from −9.46‰ to −6.13‰ (Figure 3). Our previous study shows a good replication of the δ¹⁸O and δ¹³C records from E’mei and Yongquan caves, suggesting EM1 is less influenced by the kinetic fractionation during the growth (Zhang et al., 2018). Variations of growth rate caused changes in lamina thickness, resulting in different resolutions of EM1 δ¹⁸O and δ¹³C records at different time periods. For example, between 1951 and 1991 AD, δ¹⁸O and δ¹³C records exhibited distinct seasonal cycles. Between 1928 and 1950 AD, δ¹⁸O and δ¹³C records consisted of seasonal cycles and annual cycles. On the long-term scale, the δ¹⁸O values exhibited more positive trends during 1928–1950 and 1977–1991 AD than in the interval of 1951–1976 AD (Figure 3), which might be related to the stronger monsoon intensity in 1951–1976 AD (Zhang et al., 2018). Previous studies based on instrumental data show a strong-to-weak shift of the summer monsoon intensity from 1951–1976 AD to 1977–1991 AD (Gong and Ho, 2002; Zhou et al., 2009).

The seasonal amplitude of EM1 δ¹³C was smaller than that of δ¹⁸O. The EM1 δ¹⁸O and δ¹³C records showed an offset of ∼4 months in most years except for a few years (1965, 1971, 1983) that were in phase. δ¹³C values were lower in WPSL and higher in TDSL (Figure 3). The oscillation of δ¹⁸O exhibited clear seasonal characteristics, and δ¹⁸O showed the minimum or minimum to maximum values in TDSL. The annual seasonal variation of EM1 δ¹⁸O is basically consistent with that of precipitation δ¹⁸O from the GNIP Changsha station during 1988–1991 AD and IsoGSM simulated precipitation δ¹⁸O during 1979–1991 AD, showing a positive correlation with simulated precipitation δ¹⁸O [r=0.453; 95% confidence interval (0.31; 0.575)] (Figure 4). Lower δ¹⁸O values were observed during July-September and higher values during December-April. According to the water-calcite fractionation Equation 1, simulated calcite δ¹⁸O values were also calculated using IsoGSM simulated precipitation δ¹⁸O. The result showed that the seasonal variation trends of simulated calcite δ¹⁸O are consistent with those of EM1 δ¹⁸O, but the variation range of the simulated calcite δ¹⁸O was higher (−11.50 to −1.78‰, VPDB) than that of EM1 δ¹⁸O (−8.25 to −6.32‰, VPDB) (Figure 4).

The seasonal and interdecadal variations of Sr/Ca were always consistent with Ba/Ca in EM1, but the relationship of Mg/Ca with Sr/Ca and Ba/Ca showed some changes at different timescales (Figure 3, Supplementary Figure S3). In the long term (1928–1991 AD), Mg/Ca, Sr/Ca and Ba/Ca showed coherent variations, exhibiting an antiphase relationship with δ¹⁸O (Figure 3).

During 1951–1976 AD, seasonal Mg/Ca, Sr/Ca and Ba/Ca showed variations in phase, which are also consistent with δ¹³C. However, Mg/Ca showed an antiphase relationship with Sr/Ca and Ba/Ca in a few years, such as 1953, 1972, and 1973. The annual cycles of Mg/Ca showed an antiphase relationship with Sr/Ca and Ba/Ca in most years during 1977–1991 AD, with coherent variations between Mg/Ca, Sr/Ca, and Ba/Ca in few years (e.g., 1979, 1985–1986). Remarkably, when the seasonal variations of Mg/Ca are in phase with Sr/Ca and Ba/Ca, δ¹³C also varies in phase with them (Supplementary Figure S3). During 1951–1991 AD, an in-phase relationship between Mg/Ca and δ¹³C and an antiphase relationship between Mg/Ca and δ¹⁸O were observed. The relationships between Mg/Ca ratios and sublayers are clearly visible, with lower Mg/Ca ratios corresponding to WPSL and higher Mg/Ca ratios within TDSL (Figure 3).

Zhang et al. (2018) suggested that precipitation and stalagmite δ¹⁸O are influenced by precipitation seasonality and the large-scale atmosphere circulation modulated by the El Nino-Southern Oscillation (ENSO) in the study area. It was suggested that stalagmite δ¹⁸O is negatively correlated with the East Asian summer monsoon (EASM)/non-summer monsoon (NSM) precipitation ratio on interannual to decadal timescales because the EASM/NSM ratio are predominantly affected by ENSO (Zhang et al., 2018; Zhang et al., 2020). A good correlation was observed between EM1 δ¹⁸O, precipitation amount, and dry/flood (D/F) from 1970 to 1991, with lower (higher) EM1 δ¹⁸O corresponding to decreased (increased) D/F index and increased (decreased) precipitation in the study area (Figure 5). From 1928 to 1970 AD, the D/F index and EM1 δ¹⁸O showed weak correlation (Figure 5), which is considered to be driven by the variation of precipitation seasonality associated with shifts in oceanic-atmospheric circulation (Zhang et al., 2018). Previous studies have proved that the PDO/ENSO circulation shifted from a cold phase during 1951–1976 AD to a warm phase during 1977–1990 AD (Chan and Zhou, 2005) (Supplementary Figure S4), and different phases of ENSO or PDO have important effects on EASM precipitation in southeastern China by regulating the intensity and position of WPSH (Wang and Lin., 2002; Dong and Xue, 2016). The comparison between EM1 δ¹⁸O, PDO, MEI, and WPSH in this study suggest that lower (higher) δ¹⁸O values correspond to a weakened (strengthened) WPSH associated with cold (warm) PDO/ENSO phases on the inter-annual to inter-decadal timescales (Supplementary Figure S4), which is a reasonable explanation for the antiphase relationship between stalagmite δ¹⁸O and the EASM/NSM ratio modulated by ENSO during 1951–1991 (Figure 5).

Stalagmite δ¹³C can be influenced by the δ¹³C values of soil and atmospheric CO₂ (Baskaran and Krishnamurthy, 2013; Tan, 2013; Li and Liu, 2015). δ¹³C can reflect regional vegetation controlled by hydroclimate conditions and human activities on centennial-to-millennial scales (Zhang et al., 2015; Niu et al., 2022). Over annual to decadal scales, δ¹³C can be also largely influenced by the processes of PCP in epikarst systems, degassing speed of CO₂ within caves, bedrock dissolution and residence time of seepage water (Tan, 2013). Studies demonstrated that PCP usually induces a consistent increase of δ¹³C, Mg/Ca, Sr/Ca, and Ba/Ca (Mattey et al., 2010; Sinclair et al., 2012; Koltai et al., 2017; Vansteenberge et al., 2020). The δ¹³C values of EM1 from 1928 to 1991 did not exhibit any distinct change (Figure 5), probably due to the lack of apparent changes in the overlying vegetation in the past 70 years. EM1 δ¹³C exhibited several inconspicuous cycles on annual to decadal scales, with a strong covariation of increased δ¹⁸O, Mg/Ca, Sr/Ca and Ba/Ca values, indicating the occurrence of PCP during these dry periods on annual to decadal timescales (Figure 5).

On the interdecadal scale, Mg/Ca, Sr/Ca, and Ba/Ca records showed in-phase variation, which are in antiphase with annual precipitation amount (Figure 5). The relationship between trace element ratios and the D/F index of the study area during 1928–1991 AD was also investigated, with higher (lower) Mg/Ca, Sr/Ca, and Ba/Ca ratios corresponding to dry (wet) climate (Figure 5). The response rates of Mg, Sr, and Ba to hydroclimate change were different (Mg>Ba>Sr) (Figure 5), which might be attributable to the different distribution coefficients of these trace elements (Mg<<Ba<Sr) (Fairchild and Baker, 2012; Day and Henderson, 2013). This indicates that Mg is more sensitive to precipitation variations than Ba and Sr. On interdecadal timescales, with decreased precipitation, the extended residence time of seepage water in the epikarst system enhances WRI, which is conducive to the dissolution of Mg, Sr, and Ba in bedrock and finally lead to higher Mg/Ca, Sr/Ca, and Ba/Ca ratios. The fact that higher values of Sr/Ca and Ba/Ca corresponded to increased rainfall from 1972 to 1976 AD indicate that the incorporation of Sr and Ba into speleothem might have a more complex mechanism (Figure 5). On the interdecadal scale, Mg/Ca, Sr/Ca, and Ba/Ca in EM1 were found to respond to local hydroclimate conditions, with higher values in dry climates.

Several studies have determined the seasonal variation of stalagmite δ¹⁸O through modern analysis techniques to explore the relationship between sub-annual climatic and cave environmental signals (Johnson et al., 2006; Baker et al., 2007; Orland et al., 2009; Orland et al., 2012; Koltai et al., 2017; Baker et al., 2019). Figure 4 and Figure 6 show a clear seasonal cycle of EM1 δ¹⁸O, EM1 retains seasonal climate responses probably because of abundant rainfall and the suitable cave environment. A comparison between EM1 δ¹⁸O records and meteorological data reveals that the seasonal variation of EM1 δ¹⁸O is significantly and positively correlated with precipitation δ¹⁸O but insignificantly correlation with precipitation amount (Figure 4). Cave monitoring studies from Shennong cave (around 200 km southeast of E’mei cave) and the study area show that there is more rainfall in spring with more positive δ¹⁸O values but less rainfall in summer with more negative δ¹⁸O values, this is not consistent with “amount effect” (Zhang et al., 2020; Tian et al., 2021). The relationship between seasonal rainfall amount and rainfall δ¹⁸O is consistent with that between seasonal rainfall amount and speleothem δ¹⁸O (Figure 1 and Figure 4). Therefore, we suggest that the “amount effect” does not fit the speleothem δ¹⁸O values in southeastern China on seasonal timescales.

We also find that the seasonal variation range δ¹⁸O in EM1 (−8.25 ∼ −6.32‰, VPDB) was much smaller compared to the simulated calcite δ¹⁸O (−11.50 ∼ −1.78‰, VPDB) (Figure 4). Some potential mechanisms can blur or alter seasonal δ¹⁸O signals from precipitation to stalagmites, such as selective recharging (Baker et al., 2007; Baker et al., 2019), mixing of old and fresh water (Baker et al., 2007; Baker and Bradley, 2010), seepage path (Duan et al., 2016; Markowska et al., 2016), and evaporative fractionation in the soil and epikarst zone (Zimmerman et al., 1967). Cai et al. (2010) found that δ¹⁸O in stalagmites from northwestern Thailand and 5-year moving average of the local rainfall data has a better correlation than that comparing with 1-year moving average of the rainfall data, indicating a retention and mixing of old and fresh water in karst system. On the basis of the correlation analysis between stalagmite δ¹⁸O and precipitation, Jex et al. (2010) concluded that the drip water is supplied by an aquifer recharged by waters from the current year and the previous 5 years. Regression correlation between original monthly rainfall and EM1 δ¹⁸O [r=0.453; 95% confidence interval (0.31; 0.575)] is worse than the correlation between 8-point smoothed rainfall data and EM1 δ¹⁸O [r=0.486; 95% confidence interval (0.348; 0.603)], which indicates a mixing of old and fresh water in the epikarst (Baker and Bradley, 2010). When the temperature is high, precipitation may have undergone evaporation in soil or near surface water during the formation of EM1 drips, resulting in heavier δ¹⁸O values of the drip water than the original precipitation δ¹⁸O (Baker et al., 2007; Markowska et al., 2016). Our monitoring work in Shennong cave also shows that the δ¹⁸O values of drip water inside the cave are significantly higher than the δ¹⁸O values of original precipitation, and the amplitude of drip water δ¹⁸O is smaller than original precipitation (Tian et al., 2021), indicating evaporation occurs in the process from precipitation to drip water. Unfortunately, monitoring work was not performed in the E’mei cave. However, we can find the most negative values of simulated calcite δ¹⁸O (−11.50‰, VPDB) are lower than those of the EM1 δ¹⁸O (−8.25‰, VPDB) during summer when the precipitation δ¹⁸O is negative but the temperature is high (Figure 4). The amplitude of simulated calcite δ¹⁸O (∼9.72‰, VPDB) is much larger than that of EM1 δ¹⁸O (∼1.93‰, VPDB). These observations indicate that evaporation and water mixing occur in the process from the precipitation, drip water to speleothem in the E’mei cave.

Seasonal δ¹⁸O in EM1 peaks in spring (March-May) with high rainfall, decreases with the onset of the summer monsoon, and reaches the lowest level in late summer to early autumn with significantly reduced rainfall and strong evaporation. The high EM1 δ¹⁸O values in WPSL (luminescent layer) correspond to a large amount of precipitation with higher δ¹⁸O in the spring to early summer (Figure 6). The low EM1 δ¹⁸O values in TDSL (non-luminescent layer) are consistent with the low precipitation δ¹⁸O occurring during the late summer to early autumn when stalagmite sublayers are deficient in organic matter due to less precipitation and strong evaporation (low P-E), which are not conducive to soil organic matter production (Davidson et al., 2000; Fischer, 2009; Wang et al., 2016) and eluviation (Beynen et al., 2002). This further confirms the hypothesis of the seasonal variation of precipitation δ¹⁸O primarily controlling the seasonal variation of EM1 δ¹⁸O.

The seasonal variation of δ¹³C in cave sediments may be affected by various processes, such as bedrock dissolution (Riechelmann et al., 2011), soil biogenic CO₂ content (Fairchild et al., 2006), degassing of CO₂ (Sinclair et al., 2012), and PCP (Rampelbergh et al., 2014a). Seasonal cycles of δ¹³C are much less pronounced than those of δ¹⁸O, but an inverse correlation between seasonal δ¹⁸O and δ¹³C records could still be observed, with the minima of δ¹³C corresponding to the maxima (or maxima to minima) of δ¹⁸O (March-June) and the maxima of δ¹³C corresponding to the minima (or minima to maxima) of δ¹⁸O (July-February) (Figure 6). These can be attributed to the continuous spring precipitation from March to May and June being the Meiyu period in the region. In addition, the average temperature from March to June is suitable for plant growth. Therefore, soil contains more CO₂ rich in ¹²C because of flourishing plant roots and strong soil microbial activity (Amundson et al., 1998; Zhang et al., 2015). At the same time, the δ¹³C values of calcite tend to be lower because the faster drip rates are not conducive to CO₂ degassing (Banner et al., 2007; Fairchild et al., 2006). Rapid degassing of CO₂, which is mainly controlled by good ventilation or slow drip rate, leads to increases in calcite δ¹³C (Duan et al., 2013; Johnson et al., 2006). From July to August, the decrease of rainfall and the increase of temperature in this area not only contributed to the dissolution of cave bedrock (Asrat et al., 2018), but also weakened the activities of plant roots and soil microorganisms because of low P-E, and slower drip rates promoted CO₂ degassing (Banner et al., 2007). It is noteworthy that δ¹³C and δ¹⁸O displayed similar seasonal trends in individual years (e.g., 1965, 1971, 1983–1984). The δ¹⁸O, δ¹³C, Mg/Ca, Sr/Ca, Ba/Ca, of EM1 in the spring of 1965 and 1971 all showed high values. This may be attributable to piston flow. During the period of heavy spring rain, resident water in the surface karst in winter was flushed out, which may have been affected by PCP (Arbel et al., 2010). From 1983 to 1984, δ¹⁸O and δ¹³C showed in phase variation but antiphase variation with Mg/Ca, Sr/Ca, and Ba/Ca (Figure 7B), and the variation ranges of δ¹³C were 0.083‰ and 0.20‰, respectively (Figure 5). Therefore, rather than PCP, bedrock dissolution is suggested to be the cause of the δ¹³C increase under this scenario (Riechelmann et al., 2011).