The Asian Summer Monsoon: Teleconnections and Forcing Mechanisms—A Review from Chinese Speleothem δ 18 O Records

: Asian summer monsoon (ASM) variability signiﬁcantly a ﬀ ects hydro-climate, and thus socio-economics, in the East Asian region, where nearly one-third of the global population resides. Over the last two decades, speleothem δ 18 O records from China have been utilized to reconstruct ASM variability and its underlying forcing mechanisms on orbital to seasonal timescales. Here, we use the Speleothem Isotopes Synthesis and Analysis database (SISAL_v1) to present an overview of hydro-climate variability related to the ASM during three periods: the late Pleistocene, the Holocene, and the last two millennia. We highlight the possible global teleconnections and forcing mechanisms of the ASM on di ﬀ erent timescales. The longest composite stalagmite δ 18 O record over the past 640 kyr BP from the region demonstrates that ASM variability on orbital timescales is dominated by the 23 kyr precessional cycles, which are in phase with Northern Hemisphere summer insolation (NHSI). During the last glacial, millennial changes in the intensity of the ASM appear to be controlled by North Atlantic climate and oceanic feedbacks. During the Holocene, changes in ASM intensity were primarily controlled by NHSI. However, the spatio-temporal distribution of monsoon rain belts may vary with changes in ASM intensity on decadal to millennial timescales.


Introduction
The Asian summer monsoon (ASM) transports heat and moisture during boreal summer (JJA) across the Indian Ocean and the tropical western Pacific into the Indian subcontinent and southeastern Asia, extending as far as northeast China and Japan [1].As the strength of the ASM directly influences the hydro-climate over these regions, its variability is of profound relevance to a large fraction of the world's population.The ASM system includes two interacting subsystems: the East Asian summer monsoon (EASM) and the Indian summer monsoon (ISM).Modern observations show that the ISM circulation and its associated moisture penetrate northeastwards, deep into East Asia and, as a result, the two subsystems cannot be clearly distinguished mechanistically [2].China is influenced by both the EASM and ISM and, therefore, it is essential to study monsoon variability on various timescales to understand their global teleconnections, underlying forcing mechanisms, and, in turn, improve our ability to predict long-term trends of hydroclimatic change.
The Speleothem Isotopes Synthesis and Analysis (SISAL) Working Group supported by Past Global Changes (PAGES) is an international effort to compile and synthesize stalagmite δ 18 O and δ 13 C records to explore past climate changes and enable climate model evaluation [16,17].The first version of the database, SISAL_v1 [18] and database structure and content were described by Atsawawaranunt et al. [19].An overview of the ISM evolution, as interpreted from speleothem records, has been published in this special issue by Kaushal et al. [20].In this paper, we drew on SISAL_v1 to investigate climatic changes across China during the Late Quaternary.The database contains δ 18 O and δ 13 C data of stalagmites along with relevant age and cave information.Because the interpretation of Chinese δ 13 C records is underdeveloped due to its complexity, we focused this review on available δ 18 O records.

Cave Locations and Climatic Characteristics in China
The SISAL_v1 contains 64 speleothem records from 17 caves in China (Figure 1).These cave sites are mostly distributed in southern and central China, with fewer sites in northeastern and southeastern China.Although northwestern China and the Tibetan Plateau have well-developed karst systems (Figure 1), very few records have been published from these regions, such as the Kesang [21,22] and Baluk [23] caves in northwest China and the Tianmen [24,25] and Bengle [26] caves in Tibet.Difficulties with cave access, partly for religious concerns, have mainly contributed to the paucity of speleothem records from these areas.While significantly more stalagmite datasets have been published from southwestern China (i.e., Yunnan and Guangxi provinces), relatively low uranium concentrations tend to preclude their use for precise climate reconstructions.The most relevant records from this limestone-dominated region are from the Xiaobailong [13] and Dongge [8,12] caves in the southern part of southwestern China (Figure 1).
The EASM variability significantly affects hydro-climate across most of China, with two notable exceptions: (1) in NW China and the adjacent northern part of the Tibetan Plateau, where the westerly climate prevails; and (2) in SW China and the southern part of the Tibetan Plateau, where hydro-climate variability is mainly controlled by the ISM (Figure 2).Regions located around 110 • E are influenced by interactions between ISM and EASM (Figure 2).The EASM undergoes a meridional transition during its seasonal evolution, characterized by three quasi-stationary stages separated by two abrupt northward shifts in the frontal rainfall belt.Ding and Chan [31] summarized the meridional advance of the EASM as follows: (1) the first onset starts in early May, following the "spring persistent rainfall", is designated as the "pre-mei-yu" season; (2) in mid-June, the frontal rainfall abruptly shifts northwards, forming the Chinese "mei-yu"; and (3) the second abrupt shift occurs around mid-July, enhancing monsoon-associated rainfall over northern China.The conventional interpretation of the EASM seasonal evolution underlines the land-sea moisture-bearing thermal contrast caused by the differential heat capacities of the Asian continent and the Pacific Ocean, which induce a baroclinic contrast that drives a low-level monsoonal flow from South China Sea.However, Wang and Lin [32] suggested that the persistent spring rainfall over SE China is not related to the EASM, because the large-scale atmospheric circulation and rain-bearing systems differ from those associated with the typical summer monsoon frontal rainfall.Changes in EASM intensity are essentially reflected by the advance or retreat of the monsoon frontal rainfall belt, more northerly the penetration of the frontal rainfall belt, the greater the intensity of the EASM [33].Recent publications suggest that changes in the timing and duration of the transition between EASM seasonal stages are linked to the north-south migration of the westerly jet relative to the Tibetan Plateau, with an early seasonal transition being associated to an early transition of the westerly jet [34].The increase in insolation over the boreal summer reduces the latitudinal temperature gradient, leading to a weakened, northwardly shifted westerly jet.The meridional position of the westerlies relative to the Tibetan Plateau determines the onset of mei-yu and possibly the onset of the mid-summer stage.An earlier northward shift in the westerly jet triggers earlier seasonal rainfall transitions, and thus a shorter "mei-yu" and longer mid-summer stage [35].
During the summer monsoon season, precipitation occurs along a long trajectory (i.e., from more distal water sources).As a result, precipitation δ 18 O values in eastern China become progressively lower, reaching their minimum during July-August [2].This δ 18 O minimum also coincides with the maximum land-sea temperature contrast.During autumn, precipitation δ 18 O values become progressively higher with the retreat of the summer monsoon.In contrast, boreal winter (DJF) precipitation δ 18 O in eastern China exhibits a large spatial range in δ 18 O, with the lowest values in the north and highest values in the south.This meridional gradient likely results from the reversal of the predominant wind direction in association with the East Asian Winter Monsoon, driven rather by the eastern flank of the Siberian High [36], for which the "temperature effect" strongly determines regional meteoric δ 18 O values [37].Because winter precipitation contributes relatively little volume to the annual mean [2], spatial patterns in DJF precipitation δ 18 O should generally have a negligible impact on speleothem δ 18 O and, therefore, are less important to the interpretation of Chinese cave records.Overall, these observations demonstrate that seasonal precipitation δ 18 O values are principally (and inversely) correlated with summer monsoon intensity and/or distance from moisture sources, which, in turn, is closely related to atmospheric circulation.gridded gauge-analysis data from the GPCC (Global Precipitation Climatology Center [38]).Wind data at 850 hPa (June-September) are from MERRA (Modern-Era Retrospective analysis for Research and Applications [39]) database.Red and black dots are the locations of speleothem δ 18 O records discussed herein (details in Table 1).Cave locations numbered 1-17 (red dots) are the same as in Figure 1."EASM", "ISM", and "Westerly" denote regions mainly influenced by the East Asian summer monsoon, Indian summer monsoon, and Westerly climates, respectively.The regions located around 110 • E represent the interaction area between ISM and EASM.The modern ASM limit is shown by the black, dashed line, as in Chen et al. [40].Numbers indicate cave sites of records in SISAL_v1 (1: Kulishu; 2: Hulu; 3: Sanbao; 4: Heshang; 5: Dongge; 6: Yangkou; 7: Furong; 8: Xiaobailong; 9: Zhuliuping; 10: Huangye; 11: Dayu; 12: Suozi; 13: Jiuxian; 14: Xinya; 15: Xinglong; 16: Tianmen; 17: Kesang).See Table 1 for details of each record.

Chinese δ 18 O Records in SISAL_v1
More than 200 speleothem δ 18 O records from China have been published in the past two decades and SISAL_v1 contains 64 of these records from 17 caves.As discussed in Comas-Bru and Harrison (2109) [16], a range of challenges were faced at the time of compiling data for SISAL_v1 (e.g., large amounts of metadata missing) and efforts are being made towards incorporating the missing records in subsequent versions of the SISAL database.Nonetheless, most of the monsoon regions in eastern China are represented in SISAL_v1 with sufficient spatial representation to support our initial assessment (Figures 1 and 2).Future updates of the SISAL database will have a suitable spatial density to enable the investigation of spatio-temporal variations of the EASM and its interactions with the ISM and the westerly climate.Details on each speleothem δ 18 O record and their status in the SISAL database are summarized in Table 1.The temporal distribution of the Chinese stalagmite δ 18 O records in SISAL_v1 (Figure 3) spans a wide range of timescales.Multiple records provide information on glacial-interglacial and orbital timescales, such as the composite speleothem δ 18 O record from Hulu [5], Sanbao [9], and Linzhu [11] caves in central-south China.This is the longest speleothem-based record published (it covers the full U-Th dating range [1]) and provides a 640 kyr record of EASM variability on orbital to millennial timescales [1].The Kesang cave record is sensitive to changes in the westerly climate from NW China and spans from the Late Holocene to 500 kyr BP [21].In SW China, the composite δ 18 O record from the Xiaobailong Cave presents a 252 kyr record that has been linked to ISM variability and its interaction with the EASM [13].The sampling resolution of these longer records varies from ~5 to 1300 years (Figure 3).More than ten δ 18 O records capture the last interglacial period, though with large variance of temporal ranges and age constraints (Figure 3).Six of these have a temporal resolution between 6 and 240 years and a precise age control (Figure 3).This enabled an in-depth characterization of the last glacial EASM variability, especially at the millennial time-scale, and of abrupt changes such as Dansgaard/Oeschger and Heinrich events [5,9,43,120,146,173].The highest temporal resolutions are found in the Holocene, with some records having a resolution between 1 to

Significance of Speleothem δ 18 O as a Climate Proxy in the EASM Area
Chinese speleothem δ 18 O records have substantially improved our understanding of the EASM variability on different timescales.However, the paleoclimatic interpretation of Chinese speleothem δ 18 O records remains an issue of debate [16,[181][182][183][184][185][186][187][188][189][190][191][192][193].For example, there is a strong correlation between speleothem δ 18 O series in the EASM and ISM regions, but these data contrast with some other Chinese rainfall records reconstructed by loess and lake sediments [181,182,188].A few studies suggest that Chinese speleothem δ 18 O variability is not controlled by rainfall amount but by moisture source [181,182] or pathway [192], while others conclude that they are not only influenced by the EASM but also to a significant extent by winter temperature [184,185].Instrumental data and paleoclimatic reconstructions based on other natural archives imply a spatial distribution of rainfall over eastern China that is not consistent with the observations from speleothem δ 18 O records.In this regard, Liu et al. [189] argued that speleothem δ 18 O records in the region cannot be used as an indicator of EASM intensity.These disagreements partially arose from differing definitions of "ASM intensity".Chinese speleothem δ 18 O records, particularly on orbital to millennial scales, reflect first order changes in the fraction of water vapor rained out between tropical sources and cave sites (e.g., [1,12,50]).By this definition, a strong ASM implies a more remote moisture source, with larger rainout along the moisture trajectory, and thus lighter speleothem δ 18 O values (e.g., [12]).On orbital scales, this observation is reproducible because Chinese speleothem records follow the NHSI, but their relation to regional rainfall amount remains in dispute [181,182,188,189,193].Regarding large millennial-scale events, speleothem δ 18 O records show consistent positive excursions (i.e., weak EASM) corresponding to cold events that occurred in NH high latitudes.However, the associated spatial changes in rainfall in the EASM domain could have some regional differences, as suggested by model simulations (e.g., [186,194,195]), and as of yet, very limited and difficult to interpret observational data (e.g., [14,196]).The spatial distribution of rainfall in monsoonal China is primarily controlled by changes in the position of the monsoonal rainfall belt in response to changes in EASM intensity coupled with some geographic considerations.It has been suggested that a weak EASM (i.e., consistent positive δ 18 O values in the monsoonal China) leads to a southward shift of the rain belt that in turn results in either a dipole (i.e., dry northern China and a wet southern China [44,122,151]) or a tripole pattern (i.e., dry northern and southern China but wet in Central China [14]).Therefore, the speleothem δ 18 O records ultimately show the integrated monsoonal signal from the moisture source to the cave sites on orbital to millennial timescales [1], in line with modeling results (e.g., [191,194,197,198]).In this sense, the longstanding disparity between interpretations of Chinese cave δ 18 O records may be resolvable through a thorough investigation of site-specific controls on speleothem δ 18 O and/or independently corroborating records of rainfall intensity.
The paleoclimatic interpretation and synthesis of speleothem δ 18 O records (and their controlling mechanisms) on annual to centennial timescales are further subject to debate, mainly because precipitation δ 18 O can be influenced by a broad range of factors (e.g., rainfall amount, moisture source and pathway, storm trajectory, and the seasonality of precipitation).On centennial to decadal timescales, several δ 18 O records from the transitional monsoon regions in China show a significant correlation with regional drought/flood indices, suggesting that speleothem δ 18 O is controlled by rainfall amount in North China, with lower δ 18 O reflecting enhanced precipitation (e.g., [10,44,128,150,151]). On annual to decadal timescales, many modern stalagmite δ 18 O records show significant negative correlations with local instrumental EASM or annual precipitation amount for the last 200 years [44,62,63,128,150,178,199,200].However, there exists a mismatch between speleothem δ 18 O and mean annual rainfall in the south region of the EASM area [54,201].It has been argued that speleothem δ 18 O changes over the EASM region reflect variations in the ratio of water vapor sourced from the 18 O-depleted Indian Ocean to the nearby 18 O-enriched Western Pacific Ocean-the so-called "circulation effect"-which appears to be controlled by the El Niño-Southern Oscillation (ENSO) cycle [202][203][204].Baker et al. (2015) [192] suggested that the atmospheric moisture pathway might significantly influence precipitation and speleothem δ 18 O in the monsoonal China.In SE China, especially in the region of "spring persistent rain", speleothem δ 18 O variability seems to be controlled primarily by the seasonality of rainfall (i.e., the ratio of EASM precipitation versus that from the remaining seasons), which is modulated by ENSO on annual to decadal timescales [54,205].Furthermore, the "circulation effect" may play an important role in Central China, where the ISM and the EASM dominate alternatively on annual to multidecadal scales, depending on the position/strength of the Western Pacific Subtropical High (WPSH) [50].Some speleothem δ 18 O records also show close links with the Pacific Decadal Oscillation (PDO) [54,128,199], WPSH [50,54], and ENSO [54,201].A range of climate dynamics thus modify speleothem δ 18 O values on annual to centennial scales, further complicating the assessment of its relationship with EASM variability.Nevertheless, it can be generalized that when the ENSO is in the cold mode (i.e., La Niña), the WPSH intensifies and/or shifts westward, whereas during the cold mode of the PDO, the EASM tends to transport more moisture from more distal regions into northern regions in China, resulting in lighter speleothem δ 18 O values and vice versa [54,203,205].
In summary, Chinese speleothem δ 18 O records seem to be inversely related to EASM intensity in orbital to millennial timescales.Thus, we interpret low and high values of Chinese speleothem δ 18 O records on these timescales to reflect a "strong" and a "weak" monsoon, respectively.This is broadly consistent with both theoretical and empirical studies (e.g., [1,2,5,8,9,14,191,194,[196][197][198].The interpretation of Chinese speleothem δ 18 O records on annual to centennial timescales is more complex.At these short time-scales, the speleothem signal is understood to reflect summer rainfall, the location of the moisture source and storm trajectory, or the seasonality of precipitation as associated with changes in ocean-atmosphere circulations.More empirical and theoretical studies remain critical in order to further understand Chinese speleothem δ 18   3a and 4a), for example, the records from Hulu [5], Dongge [12], Sanbao [1,9], Xiaobailong [13], Kesang [21], Yangkou [166], and Yongxing [173] caves.All these records, except Yongxing, are in SISAL_v1.The composite stalagmite δ 18 O record from the Hulu [5], Sanbao [9], and Linzhu caves [1,9] (Figure 5a) is dominated by the 23 kyr precessional cycles that are in phase with NHSI (21 July, 65 • N) (Figure 5b) [1].On the basis of this composite, Cheng et al. [1] demonstrated that glacial terminations are separated by either four or five precession cycles, supporting the idea that the "100 kyr" ice-age cycle is an average of discrete numbers of precession cycles [209].The timings of the past six glacial terminations precisely coincide with the timing and sequence of the termination events [1] (gray bars in Figure 5) as established from ice cores (Figure 5e) and marine sediments (Figure 5f).Changes in NHSI caused by the Earth's precession appear to be the main driver of the last seven ice-age terminations, as well as of the millennia-long intervals of reduced monsoon intensity associated with each of the terminations [1].These observations are consistent with a classic NHSI trigger: an initial retreat of the northern ice sheets releases meltwater and icebergs into the North Atlantic Ocean, altering oceanic and atmospheric circulation patterns and associated heat and carbon fluxes, which causes an increase in atmospheric CO 2 (Figure 5e) and Antarctic temperatures and finally drives the termination in the Southern Hemisphere [11].Increasing CO 2 and summer insolation further amplifies the recession of northern ice sheets, accelerating a rise in sea level and CO 2 through a positive feedback cycle [11] (Figure 5).
Kesang Cave is located in the eastern side of Central Asia, currently a semiarid-arid region dominated by the westerly climate (Figure 1).High-resolution (~200 year) and high-precision δ 18 O records from Kesang Cave cover most of the last 500 kyr [21] (Figure 5c).This record shows that climate change in this region exhibits a precessional rhythm with abrupt inceptions of low δ 18 O at times of high NHSI, followed by gradual δ 18 O increases that track the decline in insolation.While it is unclear whether the Kesang record suggests a possible incursion of the ASM rainfall or related moisture into the Kesang site and/or adjacent areas during the high NHSI times, it shows that orbital changes in the westerly hydro-climate are in phase with ASM variations (Figure 5b,c), which is supported by model simulations [34,197,210].
A speleothem δ 18 O record from Xiaobailong Cave in SW China characterizes changes in the ISM over the last 252 kyrs [13] (Figure 5d).This record is dominated by 23 kyr precessional cycles (Figure 5d) and, to some extent, exhibits glacial-interglacial changes (Figure 2 in Cat et al. (2015) [13]) that are in agreement with marine and other terrestrial proxies, but contrast with other speleothem δ 18 O records from eastern China [13].It has been corroborated by isotope-enabled global circulation modeling that the different responses of South and East Asian speleothem δ 18 O records (Figure 5) might be caused by the exposure of the land bridge in the western equatorial Pacific during glacial periods, which results in more depleted precipitation/stalagmite δ 18 O over eastern China [13].However, it remains an open question whether the observed glacial-interglacial variations are a manifestation of the large amplitude of ISM variations on the orbital-scale.Indeed, the ISM amplitude observed in Xiaobailong δ 18 O records at this temporal scale is ~7-8% [13], which is much larger than the typical EASM amplitude (~3-4% , [1]).Absolute speleothem δ 18 O values reflecting ISM precipitation are also lower (Figure 5b,d, Figure 2 in Reference [13]).In addition, similar ranges in modern precipitation δ 18 O and glacial-interglacial speleothem δ 18 O values have been observed between SW and SE China [211].A possible explanation might involve differences in moisture sources and trajectories in the ISM and EASM regimes.While, the ISM moisture is mainly derived from the remote Indian Ocean, the source of EASM moisture is apparently more complex, with moisture originating from the nearby Pacific Ocean, the South China Sea, and/or the Bay of Bengal (Figure 2) [203,212].This may partially explain the low δ 18 O values observed in speleothems and precipitation in SW China compared to SE China.
Overall, EASM speleothem δ 18 O records (e.g., Sanbao and Hulu [1]) show a temporal variability that is coherent with ISM records (e.g., Xiaobailong [13] and Bittoo [213]) on both orbital and millennial timescales, indicating similar forcing factors.For example, the EASM and ISM exhibit a coupled response to changes in NHSI without significant temporal lags [213].Additionally, two speleothem δ 18 O records from the Dongge [12] and Yangkou [166] caves in Guizhou and Chongqing (SW China), where the relative impact of EASM/ISM cannot be disentangled, track changes in NHSI on orbital timescales.As such, we suggest that on orbital timescales, speleothem records from both EASM and ISM regions are dominated by the Earth's precessional cycle that is in phase with NHSI.This coherence supports the idea that tropical/subtropical monsoons predominantly and directly respond to changes in NHSI [213].
Millennial-scale events during the last glacial were reconstructed using three stalagmites from the Hulu Cave [5].The Hulu record shows that the EASM intensity changed in agreement with Greenland's temperature between 75 and 11 kyr BP (Figure 6a,b), indicating a close link between the EASM and North Atlantic climate.Recently, the Hulu record has been significantly improved in terms of both sample resolution and dating (Figure 6b) [64].A close comparison between the Sanbao, Hulu, and Greenland ice core records suggests that the monsoonal interstadials between the last and penultimate glacial periods are similar in their duration and frequency, implying that the millennial fluctuations may be synchronous under glacial conditions [9].
The millennial weak EASM events are generally considered to be caused by the decay of the northern ice sheets, resulting in the flux of ice and meltwater into the North Atlantic.The ensuing slowdown in Atlantic Meridional Overturning Circulation (AMOC) generated a cold anomaly over the North Atlantic, resulting in reorganizations of oceanic and atmospheric circulations, and, in turn, the weak monsoon events [1,5].By removing orbital-scale components from the 640 kyr composite speleothem δ 18 O record and the Antarctic δD record (Figure 3 in Reference [1]), we obtained the residual ∆δ 18 O and ∆δD records, respectively [1].The ∆δ 18 O and ∆δD records are remarkably similar and negatively correlated [1].A high-resolution reconstruction of EASM variability between 88 and 22 kyr BP from the Yongxing Cave in central China ( [173]; Figure 6c) is strikingly similar to the Hulu records, suggesting a regionally coherent pattern of speleothem δ 18 O on millennial timescales (Figure 6b,c).After removing the 65 • N insolation signal from the Yongxing δ 18 O record, the residual ∆δ 18 O was also strongly anti-phased with Antarctic temperature variability on sub-orbital timescales during the Marine Isotope Stage (MIS) 3 (Figure 2 in Reference [173]).It seems evident that during North Atlantic Heinrich events, Antarctica became warmer and the ASM weakened (Figure 6).These results provide a robust linkage between northern and southern high-latitude and low-latitude monsoon climates that likely operated via the bi-polar seesaw mechanism [217,218].The strong coupling between EASM circulation and millennial-scale climate at high latitudes indicates that atmospheric circulation changes are important in transmitting abrupt climate signals globally.On the other hand, the more gradual changes observed in the EASM and in Antarctica compared to Greenland's temperature during Heinrich events (Figure 6) implies that oceanic circulation and/or sea surface temperatures (SSTs) also play an important role in the propagation of the climatic signal on millennial-centennial scales.
The Xiaobailong δ 18 O record (Figure 6d) reveals that the millennial variability of the ISM was synchronous with the EASM, as recorded by the Hulu and Yongxing records (Figure 6b,c), but with systematically lower δ 18 O values.In addition, some ISM millennial-scale features (Figure 6d) seem to resemble the temperature changes recorded in the Antarctic ice core records (Figure 6f), particularly during the Heinrich events, consistent with the mechanism previously described.However, a number of authors have emphasized the potential role of Antarctic glacial and sea-ice retreat to influence the ISM region through perturbations to oceanic overturning circulation that originated with freshwater discharge from the southern ice sheets (e.g., [1,2,58,160,169,173]).Some differences between the EASM and ISM are also notable, for example, in the Shizi δ 18 O record, which exhibits a significant negative excursion around 47.5-46.6 kyr BP that was not clearly documented in the other two δ 18 O records from SW China [132].The Sanxing δ 18 O record shows a weakening ISM trend from 22 to 17 kyr BP, while the Hulu and Qingtian records express a 3 kyr period with an intensified EASM event during that period [120].The decoupling of the EASM and ISM may be due to the different sensitivities of the two ASM sub-systems in response to internal feedback mechanisms associated with the complex geographical or land-ocean configuration, as well as SST differences between the Indian and Pacific oceans [75,120].
A large number of Chinese speleothem records cover the Younger Dryas (YD) and Bølling-Allerød (BA) events during the last deglaciation, including samples from the Hulu [5], Dongge [12,45], Yamen [165], Songjia [139], Qingtian [99,100], Kulishu [77], Haozhuzi [14,58], Longfugong [85], Xianglong [152], Linyi [83], and Lianhua [82] caves.The timing, structure, and mechanism of the YD event have been discussed in detail in the context of the Hulu, Dongge, Yamen, Qingtian, and Kulishu cave records.These δ 18 O time series are fixed by precise chronologies and, therefore, provide detailed information on the structure of the YD cooling, which is interpreted in Asia as a weakened monsoon event.Based on the Hulu records, the BA event lasted from 14,645 ± 60 to 12,823 ± 60 yr BP and the YD event lasted from 12,823 ± 60 to 11,473 ± 100 yr BP [5].The Dongge record shows striking similarity with the Hulu record during both periods despite the fact that ~1200 km separates the caves [12].A record from Yamen Cave indicates that the onset and termination of the YD monsoon event are 12,850 ± 50 and 11,500 ± 40 yr BP, respectively [165].The QT δ 18 O record from Qingtian Cave [99], which is based on annual-layer counting and 230 Th dates, documents with ultra-high precision the transition into the YD.A new δ 18 O record (QT16) from the same cave reveals a gradual shift into the YD from 12,970 ± 30 to 12,290 ± 80 yr BP and a rapid termination of YD within ~11 yrs [100].Based on annual-layer counting, the record from Kulishu Cave [77] indicates that the shift into the YD began at 12,850 ± 40 yr BP and lasted for ~340 yr, while the end of the event began at 11,560 ± 40 yr BP as an abrupt positive δ 18 O shift that lasted less than 38 yrs (or a best estimate of ~20 yrs).These results are broadly similar to other speleothem records from Hulu, Dongge, Yamen, and Qingtian Caves.During the mid-YD, three centennial wetting peaks in the Qingtian record, consistent with those in the Kulishu record, show similar structures to Greenland's temperature variations (i.e., three warming peaks) [77,100].
The weak EASM during the YD event may be tied to a weakened AMOC, which affects North Atlantic climate and, in turn, the mean latitudinal position of the ITCZ, resulting in a decrease in northern low-latitude precipitation [5,77,100].Cross-spectral analysis of the NGRIP and QT16 records shows a coherent power of ~200 yr that is prominent during the YD event, partly supporting the hypothesis that centennial variability in mid-YD is associated with solar activity [100].The consistent age control of this ASM YD structure indicates that the ASM region experienced a longer transition into the YD than the corresponding Greenland temperature shift by at least 130 yr [221], implying that, apart from the direct link between Greenland and the ASM via atmospheric circulation, oceanic circulation changes may have been important.Recently published trace-element ratios (Sr/Ca, Mg/Ca, Ba/Ca) and δ 13 C data from the Haozhuzi Cave in the middle Yangtze region indicate a wetter central eastern China during the last deglaciation, when the North Atlantic was in the cold episodes (the Heinrich 1 and YD events), even though the speleothem δ 18 O record suggests a weaker monsoon state [14].In accordance with the "jet transition hypothesis" [34], some studies suggest that a cold North Atlantic climate leads to a southward shift of the ITCZ, which, in turn, results in a lengthening of the mei-yu rains and a shortening post-mei-yu stage [14].
Climate dynamic factors influencing Holocene speleothem δ 18 O changes in Kesang [21,22] and Baluk [23] caves in NW China are more complex.A synthesis of speleothem δ 18 O records from Central Asia (Uzbekistan to western China) reflects a supra-regional pattern of climate variability on orbital to millennial timescales [196], but local hydroclimatic changes during the Holocene are not in phase with the orbitally paced, regional signal of the Asian monsoons.For example, carbon-isotope and trace-element proxies of hydroclimate are shown to lag the supra-regional climate variability by up to several thousand years.Cai et al. (2017) [22] proposed that both moisture source and the amount of precipitation contributed to δ 18 O variability in the Kesang Cave during the early and middle Holocene.On the other hand, the Baluk δ 18 O record strongly suggests a link between hydroclimatic variability in northwestern China and solar activity on centennial-to-multi-decadal timescales [23].In the following discussion, we focused on the Holocene records from the monsoonal China.

Spatio-Temporal Distribution of δ 18 O Records During the Holocene
In the lower-latitude monsoonal region, the increasing δ 18 O trend observed (waning of the Holocene Climatic Optimum; hereafter HCO) commenced as early as ~7.5 kyr BP, while at higher latitudes this excursion started progressively later: ~7.0 kyr BP in the Dongge record, ~5.3 kyr BP in the Heshang record, ~4.7 kyr BP in the Sanbao record and ~4.5 kyr BP in the Jiuxian record [75].This meridional pattern of HCO waning suggests an asynchronous change in EASM intensity during the Holocene, which could be explained by the response of a coupled tropical and subtropical monsoon system to changes in insolation gradients and, in turn, variable thermal forcing associated with the regional geographical configuration [75].SST variability in western tropical Pacific may also have an important influence on EASM variability in central and northern China via its impact on WPSH, which regulates monsoon front migration.Higher SST in the western tropical Pacific can force a northward migration of the subtropical high over East Asia, which implies a northward shift of the monsoonal front and associated rain band.This hypothesis was reinforced by a new Holocene record from Shigao Cave, SW China [125], that shows an increasing trend starting at ~6 kyr BP and that is in agreement with the spatial pattern proposed by Cai et al. (2010) [75].Some studies also suggested that the spatially asynchronous ending of the HCO in Asia may be attributed to SST changes in the western tropical Pacific [75,125], which is an important moisture source region of the East Asian monsoon.
To estimate the timing and duration of the HCO in Chinese stalagmite δ 18 O records, we analyzed 10 Holocene records using RAMPFIT [225]: Lianhua (Shanxi) [82], Jiuxian [75], Xianglong [151], Sanbao [113], Heshang [15], Lianhua (Hunan) [81], Dongge [8,45], Shigao [125], and Tianmen [25] cave records.RAMPFIT is a weighted-square method commonly used to determine a ramp between different states of a variable in a time series [225].Our results show that the HCO, as identified by low δ 18 O values, is spatially synchronous from South to North China (Figure 7).This disagreement with previous findings is partially attributable to dating uncertainties and different δ 18 O temporal resolutions.However, the actual spatio-temporal pattern and underlying mechanism may be more complex than has previously been recognized [75].The results presented here call for more high-quality Holocene records to reconcile these contradicting observations and to gain insights into the processes controlling the HCO signature in SE Asia.

Millennial-Scale Events During the Holocene
At millennial timescales, the δ 18 O time series from Dongge is punctuated by eight weak monsoon events each lasting ~1 to 5 centuries [8].Significant multi-centennial variability is also evident in the Heshang record, with notably dry periods during the 8.2 kyr BP event and at 4.8-4.1 kyr BP, 3.7-3.1 kyr BP, 1.4-1.0kyr BP and the Little Ice Age (LIA) [15].These weak EASM events are causally linked to North Atlantic ice-rafting debris (IRD) events or Bond events through southward displacement of westerly and ITCZ circulation, in particular, the 8.2 kyr [61,226] and 4.2 kyr [122] events.The 4.2 kyr event in the EASM region may have contributed to the collapses of the Chinese Neolithic culture [227].An antiphase pattern was observed between the EASM and the South American summer monsoon (SASM) during the 8.2 kyr event [228], highlighting the global extent of this event.The teleconnection is explained by a slowdown in AMOC (triggered by a glacial lake draining event) that led to a cooling of the North Atlantic climate and a southward migration of the ITCZ, in turn reducing EASM and increasing SASM intensities [8,61,226,228].
During the 4.2 kyr BP event, wet conditions are reconstructed from sites in central and southern China (Xianglong, Jiuxian, Sanbao and Heshang δ 18 O records), while dry conditions are reconstructed from one site in northern China (Lianhua (Shanxi) δ 18 O record) (Figure 8) [151].A new high-resolution (6~30-year) δ 18 O record from Shennong Cave (Figure 8f) in SE China provides a critical evidence supporting the similar "north dry, south wet" pattern during the 4.2 kyr event [122].The Dongge and Mawmluh δ 18 O records also suggest a dry hydro-climate in SW China and the ISM region, respectively (Figure 8g,h).Thus, it appears that during this event, the monsoonal rain belt may have stayed longer in the south, and shorter in the north [122,151].These weakened EASM and ISM may have been triggered by the reduced AMOC as a result of the melting icebergs in the North Atlantic (Figure 8i) [122,151].However, a recent paleoclimate data synthesis with climate-model support for western Eurasia proposed that expansion of the Siberian High is a more plausible explanation for the geographic distribution of climate perturbations near 4.2 kyr BP [229].

Climate Variability During the Last 2000 Years
There are at least 15 published speleothem δ 18 O records that fully or partially cover the last 2000 years (Figures 3b and 4c, see Table 1 for details).Unfortunately, the records from Longquan [232], Xiniu [162], Buddha [233], Qingtian [234] and Qixing [109] caves are published in Chinese journals and their data have not been made publicly available.The following discussion will focus on the three Chinese speleothem δ 18 O records in SISAL_v1 as well as the records available from Table 1: WX42B from Wanxiang Cave [10], HY-1, HY-2 and HY-3 from Huangye Cave [63], DA from Dongge Cave [8], HS4 from Heshang Cave [15] and SQ1 from Shenqi Cave [235].
Strong links between the EASM and the NH temperature [236], the temperature of the warm season in northern China [127] and solar variability [237,238] have been suggested from Huangye and Wanxiang δ 18 O records [10,63] (Figure 9).However, such a strong link is unclear in Shenqi, Jiuxian, Sanbao, Dongge and Heshang records (Figure 9g-k).A comparison of speleothem δ 18 O from Dayu, Dongge, Wanxiang, Buddha, Heshang and Lianhua caves in the EASM region over the last 750 years revealed a large variability of monsoon precipitation on decadal to centennial scales with spatial differences between southern and northern regions, likely due to changes in rain belt dynamics that ultimately relate to variations in ASM intensity [44,235].
The Medieval Climate Anomaly (MCA, 950-1250 CE) and the LIA (1400-1700 CE) [239] also manifest as δ 18 O anomalies in Chinese cave records.A warm Northern Hemisphere MCA could drive the ITCZ northward thereby intensifying the EASM.This situation is consistent with observations of increased precipitation in northern China (Wanxiang and Huangye δ 18 O records) and SW China (Shenqi δ 18 O record) but decreased precipitation in southern China (Dongge and Heshang δ 18 O records) (Figure 9).In contrast, cold conditions in the Northern Hemisphere during the LIA drive the ITCZ southward, thereby weakening the EASM.This results in drier conditions in northern China (Wanxiang and Huangye δ 18 O records) and wetter conditions in southern China (Dongge and Heshang δ 18 O records), respectively (Figure 9).In general, precipitation in the EASM region shows a "north wet/south dry" pattern during the MCA and a "north dry/south wet" pattern during the LIA [235], similar to earlier Holocene events.Studies of Chinese speleothem records also provide unique and robust tests of the relationships between speleothem δ 18 O, the occurrence of droughts, and societal unrest [27,28,150].Besides the influence of Northern Hemisphere, some studies suggest that variations in low-latitude monsoon precipitation are significantly influenced by shifts in the mean position of the ITCZ and WPSH, which is further mediated by solar activity [235] and tropical SSTs [46,47].A number of studies suggest that a decreased gradient in the tropical Pacific SST and the associated cold phase of ENSO (La Niña phase) would result in a northeastward extension of the Western Pacific Subtropical High (WPSH), which would lead to more moisture being taken up from the remote Indian Ocean compared to the Pacific [203,204,240].This signal would be captured by negative excursions in the speleothem δ 18 O records, as seen in Wanxiang, Huangye and Shenqi δ 18 O records.A more La Niña-like phase in the tropical Pacific during the MCA and more El Niño-like phase during the LIA have been noticed as major driving factors [193,239] that impose profound influences on the EASM system.However, the opposite mechanism linking the dominant El Niño-like and La Niña-like conditions to the MCA and LIA, respectively, has also been proposed [241].The phase status of ENSO during both the MCA and LIA remain a debate issue [193], and further studies are clearly needed.
On a decadal scale, the Pacific Decadal Oscillation (PDO) and the ENSO are strongly correlated, with positive PDO indices (warm PDO phase) corresponding to El Niño events (warm ENSO phases) [54].During the cold phases of PDO and ENSO (La Niña phase), the WPSH weakens and/or shifts eastward, resulting in a tendency by the EASM system to transport moisture from more distal regions into northern China-as reflected by a lower δ 18 O in precipitation [54,203].This pattern is consistent with the observed positive correlation between spelethem δ 18 O from Wanxiang [10], Huangye [63], Shihua [128], E'mei [54] and the normalized multi-speleothem δ 18 O composite from monsoonal China and the PDO index on decadal scales (Figure 10), which suggests that speleothem δ 18 O is modulated by large-scale atmospheric-ocean circulation patterns.[238]; (E-K) Speleothem δ 18 O records from Huangye [63], Wanxiang [10], Shenqi [235], Jiuxian [75], Sanbao [113], Dongge [8] and Heshang caves [15].The black, solid line in panels A and B is the 100-year running mean.Black lines in panels C-K is the 50-year running mean.Yellow and gray bars show the intervals best defining the Medieval Climate Anomaly (MCA) and Little Ice age (LIA) based on the Northern Hemispheric temperature anomaly (A), respectively.High-resolution (1~3-year) records from SW China, SE China, central China, the eastern part of NW China and northern China [50, 54,128] show a trend towards higher speleothem δ 18 O during the 20th century (Figure 11).A similar trend is observed in solar irradiance [242] and global surface temperature anomaly [243,244] but drawing robust conclusions on the causal relationship between these variables is not straightforward.It is difficult, for example, to explain the observed EASM weakening trend during the 20th century in terms of solar irradiance (Figure 11a) and global surface temperature anomalies (Figure 11b) primarily because their increase would have enhanced the land-sea temperature gradient and thus strengthen (rather than weaken) the EASM, as shown by meteorological data.Anthropogenic forcing, such as that from aerosols, has been proposed to explain this trend [245], but further investigations are required to resolve the mechanisms underpinning these trends.[246]; (I) Heshang Cave [15]; (J) E'mei Cave [54] and (K) Yuhua Cave [178].Speleothem records in (D-K) have resolutions higher than 2 years.Dotted lines in (D-K) show the long-term trend of each speleothem record.Principal Component Analysis has been done on speleothem δ 18 O converted to standard Z-scores using the software Origin Pro 2016.All axes are reversed.

Conclusions
China has witnessed a rapid increase in the number of speleothem studies over the last 20 years, with more than 100 speleothem records from ~80 caves published in ~300 papers.Most studies attribute changes in speleothem δ 18 O to variable EASM intensities.The longest records from southern and central China show that the Northern Hemisphere Summer Insolation is the main driver of EASM variability on orbital-scales.On millennial timescales, weak EASM events are causally linked to cold events in the North Atlantic in the late Pleistocene and to North Atlantic ice-rafting events during the Holocene.On centennial to decadal timescales, however, changes in monsoonal precipitation appear to be spatially heterogeneous due to a variable spatio-temporal distribution of the monsoonal rainfall belt in response to a complex geographical configuration.EASM is also significantly affected by the PDO, WPSH and ENSO modes, as well as solar activity.More efforts are required to produce annual-decadal resolution δ 18 O records with small dating uncertainties in China in combination with other climate proxies (e.g., trace element ratios) in particular in SE, NE and SW China.This, coupled with an increased availability of isotope-enabled climate simulations, would strengthen our ability to understand the mechanisms underpinning changes in speleothem δ 18 O at various timescales.

Figure 3 .
Figure 3. Temporal distribution of stalagmite δ 18 O records from China in SISAL_v1 [19].Panel (A) shows all available records and panel (B) focuses on the last 12 kyr.The color bar represents the temporal difference among consecutive isotopic samples in years.Labels on the y-axis are the entity_names as inTable 1. Entities are sorted by latitude.

Figure 5 .
Figure 5. Orbital scale changes in speleothem δ 18 O records from China and comparison with other records.(A) Multi-speleothem record from China: three records from the Hulu Cave (green curves; green left y-axis), two records from the Linzhu Cave (blue curves; blue left y-axis), and 16 records from the Sanbao Cave (all other colors; black right y-axis).(B) The 640 kyr EASM composite record from (A) [1].(C) The composite δ 18 O record from the Kesang Cave (cyan curve) [21].(D) The composite δ 18 O record from the Xiaobailong Cave (brown curve) [13].Pink curves in panels B-D show 21 July insolation at 65 • N[206].(E) The composite CO 2 record[207].(F) The composite sea level record[208].Vertical gray bars mark the timing of seven glacial terminations, which correspond with weak monsoon intervals.

Table 1 .
[19]ary of Chinese speleothem records sorted alphabetically by site name.The records for which entity_ID is available are in SISAL_v1[19].Min/Max year BP correspond to the last/first stable isotope measurement of each record.BP is years before present, where present is 1950 CE. * indicates the papers published in Chinese journals.
O records.4.2.Late Pleistocene Variations of the ASM and Westerly Climates Recorded by Speleothem δ 18 O Records 4.2.1.Orbital-Scale Changes Seven published speleothem δ 18 O records in China have long temporal coverage for the orbital-scale ASM and westerly climate variations (Figures