Introduction

Empirical orthogonal function (EOF) analysis of instrumental climate data indicates that the spatiotemporal pattern of rainfall variability in East China exhibits a tripole or a dipole structure1,2,3,4,5,6,7,8. The tripole structure demonstrates a pattern of ‘−/+/−’ (rainfall decrease/increase/decrease) in South–Central–North China1,2,9, while the dipole structure demonstrates a ‘+/−’ rainfall pattern from the south to the north of the Yangtze River5,10 (Fig. 1). These structures have been previously linked to the coupled ocean–atmosphere modes in the Pacific, including the El Niño southern oscillation (ENSO) and the pacific decadal oscillation (PDO), and North Atlantic, including the North Atlantic oscillation (NAO) and Atlantic Multidecadal Oscillation (AMO), and their interaction with the East Asian monsoon3,4,5,8,11,12. During the warm phase of ENSO and PDO, there is a rainfall decrease in South and North China, but a rainfall increase around the Yangtze River area, showing a tripole pattern of ‘−/+/−’11,13. The rainfall distribution pattern is also influenced by the North Atlantic, though, with a dipole pattern of ‘+/−’, with a rainfall increase in South China and a rainfall decrease in North China associated with a positive NAO and AMO5,14,15. However, these patterns are based on meteorological data from recent decades, when the climate has been significantly influenced by human activities. Whether these patterns were the same before industrialization is still not well understood. Climate model simulations and paleoclimate reconstructions have suggested that similar rainfall patterns were found in East China during the little ice age (LIA), the medieval climate anomaly (MCA)16,17, the Holocene18, and the last deglaciation19. However, due to the paucity of continuous high-resolution records, how the ocean-atmosphere coupling affects the spatial pattern of paleo-precipitation in East China is still unclear.

Fig. 1: The rainfall spatial distribution structures in eastern China from 1951 to 2012 CE.
figure 1

The tripole rainfall pattern presents anti-phased changes between central China and northern and southern China (a). The dipole rainfall pattern reflects the anti-phased change between southern and northern China (b). The stars mark the cave locations of Dongge29, Heshang (this study), and Wanxiang28, where three high-resolution stalagmite δ18O records come from. We added the research sites on the base map from ref. 35.

Full size image

Chinese stalagmite oxygen isotope (δ18O) records with precise absolute chronology have revealed Asian monsoon variations from orbital to seasonal timescales20,21,22,23,24,25,26,27,28,29. In recent years, with more published speleothem δ18O records from different caves in East China, there is now potential to explore spatiotemporal climate variability in the region over longer timescales. However, Chinese speleothem δ18O is influenced by multiple factors, such as moisture source region, precipitation amount, evaporation, etc., making the interpretation of speleothem δ18O complicated. Some studies suggest that precipitation in the upstream areas is the main factor controlling stalagmite δ18O, which response to the variations of the Asian monsoon intensity27,30,31. However, local precipitation and evaporation should not be ignored32,33,34, as local rainfall amount and evaporation in the soil zone could also influence the δ18O of rainfall and infiltrating cave water, thus modulating the stalagmite δ18O. Increased local and regional rainfall likely leads to a strong Rayleigh distillation and more negative δ18O values, e.g., precipitation amount effect. At the same time, increased rainfall results in high relative humidity and weak evaporation, which also leads to more negative δ18O values. That is, any single stalagmite δ18O record likely contains both large-scale atmospheric circulation and local precipitation information, which are driven by complex ocean-atmosphere dynamics linked to the monsoon and coupled modes such as ENSO, PDO, NAO, and AMO. Separating these processes from spatially separated cave records could help to understand the ocean-atmosphere interaction over eastern China.

Principal component analysis (PCA) is a statistical procedure using orthogonal transformation to convert a set of observations of possibly correlated variables into a set of values of linearly uncorrelated variables, which could separate signals from different sources. For example, Chinese stalagmite oxygen isotope signals likely contain both Asian monsoon (isotopic fractionation by the variation of upper stream precipitation) and local rainfall amount (amount effect and evaporation) information. So, different cave locations in the monsoon area contain the common monsoon signal but different local rainfall amount information. That means that PCA of the different cave δ18O records from eastern China should separate the monsoon and local rainfall signals, and the past spatial rainfall distribution pattern in the past could be reconstructed.

In this study, we present a continuous annually resolved δ18O record spanning the past 1150 years (850–2000 CE) from Heshang Cave, Central China. Then combined with the published speleothem δ18O records from Wanxiang Cave, Gansu28, and Dongge Cave, Guizhou29, we identify three principal components which are linked to monsoon and local rainfall distribution patterns. Finally, we investigate the influence of the Asian monsoon, PDO, and AMO on the spatiotemporal rainfall patterns in East China over the last millennium. Dongge, Heshang, and Wanxiang caves are located in South, Central, and North China, respectively, and are ideally situated to test whether the modern rainfall tripole and dipole structures35 existed in the past (Fig. 1).

Results

New annual resolution δ18O record from Heshang cave, Central China

Early studies confirmed that distinguishable laminations in stalagmite HS4 are annual23,36,37. By sampling along each lamina (Fig. 2), a continuous annual resolution δ18O record covering 850–2000 CE is established (see Supplementary data 1). The variation of HS4 annual δ18O record has been shown to be similar to other high-resolution speleothem records from nearby caves38,39 in recent centuries (Fig. 3), indicating it likely represents a robust record of regional δ18O variation in Central China. Stalagmite DX1 and DX2, from Dongge cave, South China, are also annually laminated and an annual resolution δ18O record is composited29. Though the layers of the stalagmite from Wanxiang cave, North China, are not clear, the extra high concentrations of uranium with low 230Th values provide a chronologic uncertainty of fewer than 2 years, and high-resolution sampling of stalagmite WX42B provides a δ18O record with a resolution of 2.5 years28. With these three precisely-dated high-resolution stalagmite δ18O records covering the past 1150 years, we aim to investigate spatiotemporal rainfall patterns in East China during different climate stages over the last millennium, such as the current warm period (CWP), the LIA, and MCA.

Fig. 2: The sedimentary characteristics and laminations of stalagmite HS4 from Heshang cave in the recent 1150 years.
figure 2

The section consists of two broken parts (the top 7 cm and the rest 40 cm). The numbers along the right and left edges mark the depth and chronology of HS4. Annual-resolved carbonate powder samples were scraped year by year for δ18O measurement.

Full size image
Fig. 3: High-resolved δ18O comparisons between HS4 and other stalagmites from Central China.
figure 3

a HS4 in this study and LH from Lianhua cave, Hunan38 (blue curve). b HS4 and XN from Xiuniu cave, Hubei39 (orange curve). The HS4 annual δ18O record duplicates other high-resolution speleothem records from nearby caves in recent centuries, suggesting the HS4 record could represent the stalagmite δ18O variation in Central China.

Full size image

Unlike the similar orbital-scale trends revealed from stalagmite δ18O across the broad Asian monsoon region20,22,24,25, the three highly-resolved stalagmite δ18O records (Fig. 4) from Dongge, Heshang, and Wanxiang caves (DG, HS, and WX hereafter) show different variations over the last millennium. The differences likely indicate that though all the records are from eastern China and share a common Asian monsoon signal, the interannual to centennial-scale stalagmite δ18O variability at each site is likely modulated by different local climate signals (e.g., precipitation and evaporation) and karst hydrologic processes32,33. Wavelet analysis reveals distinct inter-annual, inter-decadal, and centennial cycles from the three stalagmites δ18O records (Fig. 4) which may indicate the differing impacts of internal climate variability on rainfall at each site37,40,41,42.

Fig. 4: Comparison of the stalagmite δ18O records from DG, HS, and WX cave.
figure 4

The δ18O records from DG (a), HS (b), and WX (c), as well as their wavelet analysis (d–f) demonstrate their variations on different time scales with different frequencies and amplitudes, indicating that the contributions of the driving forces to the stalagmite δ18O are different from site to site.

Full size image

PCA of the spatially separated cave records

Here, we aim to separate the monsoon signal from these spatial precipitation patterns in East China through PCA analysis of the three δ18O records (Fig. 5). The PC1 explains 42.28% of the variability. The loadings for DX, HS, and WX δ18O are 0.69, 0.24, and 0.68, respectively, indicating a common signal is extracted from the three δ18O records (Fig. 5a). As all Chinese stalagmite oxygen isotopes are influenced by the upstream moisture source and rainout effect during the moisture transport, which is related to the Asian monsoon, we interpret PC1 as an indicator for monsoon intensity31,41,43. We identify three stages within PC1 which correspond well with the MCA (940–1340 CE), LIA (1340–1840 CE, and CWP (1840–1983 CE), indicating that the large-scale monsoon intensity increased during the MCA, weakened during the LIA, and strengthened again during the CWP. These trends are consistent with monsoon observations and modeling results42,44,45,46,47,48. Interestingly, PC1 follows the effective solar radiation robustly45 (Fig. 5a), with a correlation coefficient of 0.42 on a decadal scale (n = 113, P < 0.01), suggesting that solar irradiance may be the main controlling factor of the Asian monsoon over the last millennium.

Fig. 5: The principal components analysis of three stalagmite δ18O records from the eastern monsoon region of China.
figure 5

The PC1 (a), PC2 (b), and PC3 (c) explain 43.28%, 32.86%, and 24.87% of the variability, representing monsoon, tripole, and dipole rainfall distribution patterns, respectively. The variations of PCs are shown in black thin curves, with their 31-year smoothing average curves shown in orange. Three principal components are consistent with the effective solar radiation45 (ESR, purple curve), PDO49 (31-year smoothing of the data, dark blue curve), and AMO50 (31-year smoothing of data, light blue curve), which means internal climate dynamics persistently impact rainfall distribution in eastern China over the last millennium. The periods of MCA, LIA, and CWP are highlighted in PC1.

Full size image

PC2 explains 32.86% of the variability, with loadings from DG, HS, and WX δ18O records of −0.11, 0.97, and −0.22, respectively. Negative PC2 values imply drier conditions in North and South China, with wetter conditions in Central China, similar to the modern rainfall tripole pattern. In the past 1150 years, the PC2 variation (Fig. 5b) generally corresponds to the 3 stages of the monsoon shown in Fig. 5a. During the LIA, negative PC2 scores imply a tripole rainfall structure (−/+/−) following a weak monsoon, which coincides with the modern observation with wet conditions in Central China and drier conditions in South and North China. During medieval and current warm periods (stronger monsoon), the PC2 scores are positive, suggesting wet conditions in South and North China but a dry conditions in Central China (+/−/+, an anti-tripole structure). The decadal oscillations of PC2 are consistent with the PDO variation49 (Fig. 5b). Linear regression of 11-year average data shows that the PC2-PDO correlation exceeds 95% confidence level (r = 0.21, n = 113). Positive PDO corresponds to negative PC2, presenting a tripole structure, implying that Pacific dynamics likely dominate the tripole precipitation structure in East China. Our evidence clearly shows that even before industrialization, the tripole pattern of precipitation in East China existed and was likely driven by similar dynamics to the modern.

PC3 explains 24.87% of all variability, and the loadings of DG, HS, and WX δ18O records are −0.71, 0.08, and 0.70, respectively. This structure is similar to the observed dipole pattern with opposite precipitation variations between southern and northern China (Fig. 5c). Negative PC3 scores suggest less rainfall in South China but more rainfall in North China (an anti-dipole pattern), which is likely related to the strengthening Asian monsoon during the MCA. Subsequently, during the LIA, the PC3 scores become positive, indicating the rainfall belt moving southward, with a gradual rainfall increase in southern China and a decrease in northern China (dipole pattern). The variation of PC3 is consistent with that of the AMO50 (11-year average, r = 0.25, n = 113, p < 0.05), indicating the influence of the North Atlantic on the rainfall in East China (Fig. 5c).

Discussion

The separation of monsoon (PC1) and precipitation signals (PCs 2 and 3) from speleothem δ18O allows us to investigate the relationship between tripole and dipole patterns and their link to the monsoon, PDO, and AMO. It is observed that two types of patterns alternate by the influence of the mutual interactions among monsoon, PDO, and AMO on centennial timescales. During the MCA, when a strong monsoon is coupled with a negative PDO and a positive AMO, a dipole pattern of precipitation in East China dominated (PC3 < 0), while the tripole was not significant (PC2 ≈ 0). During the LIA, the precipitation in East China was dominated by a tripole pattern (PC2 < 0) under the conditions of a weak monsoon, a positive PDO, and a negative AMO. In the CWP, the strengthening monsoon, combined with a negative PDO and a positive AMO, initially exhibits a dominant tripole which is gradually replaced by a dipole pattern, similar to the MCA status. This observation coincides with the earlier findings that there are different precipitation patterns during MCA, LIA, and CWP16,17.

On multidecadal timescales, the monsoon, PDO, and AMO also influence the precipitation distribution structure in East China. A bandpass filtering analysis between 20 and 80 years demonstrates that, during a strong monsoon, the main precipitation distributions are shown by a negative tripole pattern (a rainfall increase in South China with a rainfall decrease in Central China and a rainfall increase in North China) and a positive dipole pattern (a rainfall increase in South of China and a rainfall decrease in North China) (Fig. 6). On the other hand, during a weak monsoon, the rainfall distributions are shown as a positive tripole pattern and a negative dipole pattern. The correlations between monsoon (Fig. 6a) and precipitation distribution patterns (Fig. 6b, c) are consistent with modern observations2,3,16 and climate modeling7,17. In contrast to the tripole and dipole patterns of rainfall distribution on centennial timescales, and on multi-decadal timescales, the relationships of PC2-PDO and PC3-AMO are unstable. For example, after 1400 CE, the variations of the PDO and PC2 are in-phase, while before 1400 CE, they are anti-phase (Fig. 6b), which suggests that the different climate background states during the MCA and LIA may influence the interaction between PDO and the precipitation distribution pattern in East China. In addition, the interaction between AMO and the precipitation distribution pattern in East China is also unstable (Fig. 6c). During 880–1130, 1310–1500, and 1780–1980 CE, the variations of AMO and PC3 are in-phase, while during 1130–1310 and 1500–1780 CE, they are anti-phase, suggesting that it is not the AMO alone, but the coupling between AMO and PDO that affects the precipitation distribution in East China12,51.

Fig. 6: Interactions between the precipitation distribution pattern in East China and the monsoon with PDO and AMO.
figure 6

All the data are processed by bandpass filtering. a The monsoon variation indicated by PC1. b The tripole rainfall pattern revealed by PC2 and PDO (in the dark blue). The relationships between them are reversed in 1400 CE. Before 1400 CE, the variations of PC2 and PDO are in-phase, after that, they are anti-phase. c The dipole rainfall pattern suggested by PC3 and AMO (in light blue). During 880–1130, 1310–1500, and 1780–1980 CE, the variations of AMO and PC3 are generally in-phase, while during 1130–1310 and 1500–1780 CE, they are anti-phase. On multi-decadal scales, the unstable relationships of PC2-PDO and PC3-AMO suggest that it is the mutual interactions among the monsoon, PDO, and AMO, rather than any induvial factor, that influences the precipitation distribution pattern in East China.

Full size image

The Western Pacific subtropical high (WPSH) is the bridge between the precipitation patterns in East China and the monsoon, PDO, and AMO7,52,53,54. The AMO controls the precipitation distribution in East China with the monsoon and PDO by modifying the middle and high latitude circulation and influencing the intensity and position of the WPSH7,54 (Fig. 7). A weak monsoon with positive PDO and negative AMO leads to a southward shift of the WPSH with increasing rainfall in Central China (HS) associated with the Meiyu front, while South China (DG) is covered by the WPSH with decreasing rainfall. At the same time, due to the weakening of the summer monsoon, the moisture transport to North China (WX) is reduced with less rainfall. Thus, the precipitation distribution in East China presents a tripole pattern of ‘-/+/-’ with less rainfall in both South and North China but more rainfall in Central China, a scenario that appeared in LIA (Fig. 5b). On the contrary, during a strong monsoon with a negative PDO and positive AMO, the position of the WPSH moves northward, and Central China (HS) is characterized by high temperatures and less rainfall. The northward shift of the WPSH increased rainfall in South China (DG) and North China (WX) associated with the monsoon trough and Meiyu front55,56. Thus the precipitation distribution in East China presents as an anti-tripole pattern (‘+/−/+’) with increasing rainfall in both South and North China but decreasing rainfall in Central China (Fig. 7b). During the MCA, Asian monsoon further strengthened, and AMO (PDO) was more positive (negative) than the CWP, then WPSH moved further to the northeast, and the monsoon trough leaves off DG site, leading to the precipitation decrease in South China, but increase in North China (WX) and Central China (HS), with a dipole structure shown in Fig.5c. However, in the past millennium, the monsoon, PDO and AMO were not always coupled (Fig. 6). When they are decoupled, the tripole and the dipole pattern alternate12,51, then the precipitation distribution in East China is more complex, as we have observed on multi-decadal scales (Fig. 6).

Fig. 7: Schematic diagram of the influence of the Asian monsoon, PDO, and AMO on the precipitation distribution patterns in East China.
figure 7

a A weak monsoon with positive PDO and negative AMO. During this period, the position of the West Pacific subtropic High (WPSH) was southward and Central China (HS) is controlled by Meiyu front with rich rainfall, while South China (DG) and North China (WX) with less rainfall, presenting as a tripole pattern (‘−/+/−’). b A strong monsoon with negative PDO and positive AMO. During this period, the position of the WPSH was northward, which covered Central China (HS) with less rainfall, while South China (DG) and North China (WX) were controlled by monsoon trough and Meiyu front with rich rainfall, presenting an anti-tropile rainfall pattern (‘+/−/+’). This diagram is drawn according to refs. 55,56.

Full size image

In summary, the monsoon and local precipitation signals are successfully extracted by the principal components analysis on three individual stalagmite δ18O records spanning North, Central, and South China, providing important pieces of evidence for understanding the link between local rainfall, the Asian monsoon, and Atlantic and Pacific climate modes. The shared monsoon signal (PC1) recorded at all three locations suggests an influence of solar irradiance on monsoon strength. The reconstructed rainfall records of the last millennium (PCs 2 and 3) suggest that the dipole and tripole structures, which reflect the combined influence of monsoon strength, PDO, and AMO, are persistent over multidecadal to centennial timescales, with potential implications for evaluating climate models and improving future hydroclimate projections.

Materials and methods

Layer identification and chronology

We selected the top section of stalagmite HS4 covering the last 1150 years for this study. The stalagmite is characterized by clear annual growth bands, and each band consists of a light–dark couplet23,36,37. After layer counting, year by year combined with the U–Th dating23, a precise chronology of stalagmite HS4 from the topmost to 47 cm in depth was established (Fig. 2).

Stable isotopic analysis

On the basis of the above-mentioned stratigraphic identification, we scratched the stalagmite layer by layer for stable isotopic analysis. A total of 1150 samples have been collected for δ18O measurements on a Finnigan MAT-253 isotope ratio mass spectrometer with an online, automated carbonate preparation system (GasBench II). The δ18O values are reported relative to VPDB (Vienna Pee Dee Belemnite) standard with an external error <0.20‰ (two sigmas).

Principal components analysis

The principal component analysis was carried on by using three stalagmite datasets of oxygen isotope from Dongge cave, Heshang cave, and Wanxiang cave, respectively. We linearly interpolated the WX record to an annual resolution before analysis, to couple the DG and HS datasets, then all of them are annual resolutions. Three principal components are extracted with variances of 43.28%, 32.86%, and 24.87%, respectively.