The Global Paleomonsoon as seen through speleothem records from Asia and the Americas

Abstract

The regional monsoons of the world have long been viewed as seasonal atmospheric circulation reversal—analogous to a thermally-driven land-sea breeze on a continental scale. This conventional view of monsoons is now being integrated at a global scale and accordingly, a new paradigm has emerged which considers regional monsoons to be manifestations of global-scale seasonal changes in response to overturning of atmospheric circulation in the tropics and subtropics, and henceforth, interactive components of a singular Global Monsoon (GM) system. The paleoclimate community, however, tends to view ‘paleomonsoon’ (PM), largely in terms of regional circulation phenomena. In the past decade, many high-quality speleothem oxygen isotope (δ¹⁸O) records have been established from the Asian Monsoon and the South American Monsoon regions that primarily reflect changes in the integrated intensities of monsoons on orbital-to-decadal timescales. With the emergence of these high-resolution and absolute-dated records from both sides of the Equator, it is now possible to test a concept of the ‘Global-Paleo-Monsoon’ (GPM) on a wide-range of timescales. Here we present a comprehensive synthesis of globally-distributed speleothem δ¹⁸O records and highlight three aspects of the GPM that are comparable to the modern GM: (1) the GPM intensity swings on different timescales; (2) their global extent; and (3) an anti-phased inter-hemispheric relationship between the Asian and South American monsoon systems on a wide range of timescales.

1 Introduction

The word ‘monsoon’ is derived from an Arabic word ‘Mausim’, which means ‘season’. Although the term monsoon was originally used to describe the large-scale reversal of surface winds over the Indian Ocean during the boreal summer, in the modern parlance, it is now used to describe significant seasonal atmospheric circulation change associated with changes in wind direction and precipitation in low latitude regions. The term ‘Global Monsoon’ (GM) was coined to refer global-scale seasonal changes in the atmospheric circulation, particularly emphasizing the connection among major regional monsoon systems in the tropics and subtropics on both sides of the Equator (e.g., Trenberth et al. 2000). The GM embodies a system characterized by a persistent global-scale overturning of the atmosphere that varies according to the time of year. These changes are associated with seasonal reversal in surface winds and shifts in precipitation tied to the seasonal migration of the Intertropical Convergence Zone (ITCZ) and reversal in the northern and southern hemispheric heating and temperature gradients between continent and ocean in response to the annual solar cycle (e.g., Trenberth et al. 2000; Qian et al. 2002; Wang and Ding 2006; Wang and Ding 2008).

Although the concept of the GM was proposed more than a decade ago (e.g., Trenberth et al. 2000), the paleoclimate community has yet to fully test this concept on longer timescales. In most cases, the proxy-inferred site-specific paleomonsoon (PM) variability have been viewed as one resulting from the regional monsoon variability. In 2007, a Past Global Changes (PAGES) Working Group on ‘Global-Monsoon’ was established in an effort to bring this holistic approach to PM research. The PAGES sponsored Global Monsoon Symposiums in 2008 and 2010 brought together paleo- and modern climatologists for in-depth discussions on the concept of the GM and its potential applicability to the ‘Global-Paleo-Monsoon’ (GPM) on longer timescales (Wang et al. 2009). A timely review by Wang (2009) further provides a profound analysis and description of GPM scenarios from tectonic (~10⁸–10⁶ years) to millennial-centennial (10³–10² years) scales.

The dominant monsoon systems in the world include the Asian–Australian, African, and the American Monsoons (e.g., Wang and Ding 2008; Zhou et al. 2008). While each of these modern regional monsoon systems has specific characteristics, all are coordinated by the annual cycle of solar radiation. In the past decade, speleothem oxygen isotope (δ¹⁸O) based studies have been widely used to reconstruct patterns of PM changes, most notably from Asian and South American Monsoon regions. These speleothem δ¹⁸O records have proven remarkably valuable because of their high-resolution and absolute-dated features that not only allow comprehensive reconstructions of regional PM patterns but also permit robust correlations among them.

Recently, Caley et al. (2011a) using a variety of proxy records with varying chronological resolutions and uncertainties from different monsoon systems, disputes the concept of the GPM. Here we argue that a thorough test of the GPM scenario on a wide range of timescales is best accomplished through the use of well-dated and precipitation-sensitive speleothem records from various monsoon systems from both hemispheres. Accordingly, we examine a number of existing speleothem δ¹⁸O records from Asia, the Americas, and the Eastern Mediterranean to highlight the major characteristics of the GPM on different timescales and provide possible dynamical mechanisms explaining linkages among different monsoon systems.

2 Speleothem δ¹⁸O records of regional monsoons

2.1 The Asian Monsoon domain

Nearly two-thirds of the global population lives under the direct influence of the seasonal rainfall associated with the Asian summer monsoon (ASM). Spring-time solar heating of the Asian continent overturns the atmosphere circulation and initiates the ASM, which transports large amounts of moisture and heat northward from northern Australia across the Indian Ocean, into India, southeastern China, and as far as northeastern China and Japan. Following the annual cycle of solar heating, during the boreal winter, cold-dry Asian winter monsoon flows from Siberia southward across eastern China, India, and the Indian Ocean, ultimately contributing to the Australian summer monsoon (An 2000).

The ASM regime includes two interactive sub-components: the East Asian summer monsoon (EASM) and the Indian summer monsoon (ISM) (Fig. 1). Although one may divide these two sub-monsoon systems geographically, it is unfeasible to clearly distinguish them mechanistically, as modern observations show that the ISM circulation and associated moisture penetrate northeastward, deep into East Asia (Figs. 1 and 2). As a result, it has been argued that a significant component of EASM precipitation (for example, in southeastern China) is derived essentially from the ISM domain (e.g., Ding 2004; Clemens et al. 2010; Pausata et al. 2011). EASM variations on a variety of timescales have been well characterized by speleothem δ¹⁸O records from southeastern China, such as from Sanbao (Wang et al. 2008; Cheng et al. 2009b), Hulu (Wang et al. 2001; Cheng et al. 2006), Dongge (Yuan et al. 2004; Dykoski et al. 2005; Wang et al. 2005; Kelly et al. 2006), Wanxiang (Zhang et al. 2008) and Linzhu (Cheng et al. 2009b) Caves (Fig. 1). Variations in the ISM have been documented from speleothem records from Hoti, Qunf, Defore, and Mukalla Caves in the Arabian Peninsula (Fleitmann et al. 2003, 2007, 2011); Moomi and Dimarshim Caves in Socotra Island (Burns et al. 2003; Fleitmann et al. 2007; Shakun et al. 2007); Timta Cave in northwestern India (Sinha et al. 2005); Xiaobailong Cave in Southwestern China (Cai et al. 2011) and Tianmen Cave in the Tibetan Plateau, China (Cai et al. 2010). The aforementioned ISM records, though limited in temporal duration, sketch out major features of the PM variability in the region and together with speleothem records from the EASM domain, illustrate ASM variations on orbital-millennial-centennial-decadal timescales and provide an important test of the GPM.

Fig. 1

Map of cave locations. Red dots indicate cave sites: 1-Hulu (32°30′N, 119°10′E), 2-Sanbao (31°40′N, 110°26′E), 3-Dongge (25°17′N, 108°5′E), 4-Xiaobailong (24°12′N, 103°21′E), 5-Wanxiang (33°19′N, 105°00′E), 6-Tianmen (30°55′N, 90°4′E), 7-Timta (29°50′N, 80°02′E), 8-Hoti (23°5′N, 57°21′E), 9-Qunf (17°10′ N, 54°18′ E), 10-Moomi (12°30′N, 54°E), 11-Borneo (4°N, 114°E), 12-Liang Luar (8°32′S, 120°26′E), 13-Kesang Cave (42°52′N, 81°45′E), 14-Caves in southern Israel, 15-Soreq Cave (31.45°N, 35.03°E), 16-Peqiin Cave (32.58°N, 35.19°E), 17-Sofular (41°25′N, 31°56′E,), 18-Corchia (44°2′N, 10°17′E), 19-Rio Grande do Norte (5°36′S, 37°44′W), 20-Paixão (12°39′S, 41°3′W), 21-Padre (13°13′S, 44°3′W), 22-Lapa Grande (14°25′S, 44°22′W), 23-Santana (24°32′S, 48°44′W), 24-Botuverá (27°13′S, 49°9′W), 25-Cueva del Tigre Perdido (5°56′S, 77°18′W), 26-Pacupahuain Cave (11°14′S, 75°49′W), 27-Fort Stanton Cave (33°18′N, 105°18′W), 28-Bells Cave (31°45′N, 110°45′W). Arrows depict wind directions of the ISM, EASM, NASM, SASM and the Westerlies. Dashed lines show the modern position of the ITCZ in July and January, respectively

Fig. 2

Moisture transport patterns averaged for 1990–1999 (adapted from Ding 2004; Ding et al. 2004) during Pre-onset of the ASM (the 6th pentad of March-the 3rd pentad of May) (a) and Post-onset of the ASM (the 5th pentad of May-the 2nd pentad of July) (b). The strong moisture transport zone is shaded (in kg m⁻¹ s⁻¹). Atmospheric circulation and moisture transport patterns are entirely different between pre- and post-onset of the ASM

2.2 The South American Monsoon domain

The South American summer monsoon (SASM) is a large-scale atmospheric circulation system over most of the tropical and subtropical South American continent (Zhou and Lau 1998) and exhibit many similar features of its ASM counterpart. When the SASM dominates during the austral summer season, equatorial trade winds transport moisture westward from the tropical Atlantic across the Amazon Basin to as far east as the Andes. Blocked by the high Andes mountains to the west, the warm and humid air is then forced to turn southeastward in a form of a low-level jet (LLJ) which develops deep convection over central and southeastern Brazil (Fig. 1). During the austral winter, this atmospheric circulation is weakened or even virtually comes to a halt (Zhou and Lau 1998). Speleothem δ¹⁸O records from the region shed new light on the hydroclimate history of tropical and subtropical South America. The speleothem records discussed here are mainly from Rio Grande do Norte, Paixão, Padre, Santana, Botuverá, Cueva del Tigre Perdido, Lapa Grande, and Pacupahuain Caves (Fig. 1) which not only characterize SASM variations on different timescales but also describe changes in its spatial patterns (e.g., Cruz et al. 2005, 2006, 2007, 2009a, b; Wang et al. 2004, 2006, 2007a, b; van Breukelen et al. 2008; Cheng et al. 2009a; Stríkis et al. 2011; Kanner et al. 2012).

2.3 The North American Monsoon domain

The North American Monsoon prevails mainly over the southwestern US and northwestern Mexico. In the southwestern US, during boreal summer, easterly and southeasterly winds transport moisture into the area from the Gulf of Mexico and the Gulf of California. On the other hand, Pacific westerly storms during the boreal winter provide about one-half to one-third of the annual precipitation. Recently, two speleothem δ¹⁸O records from southwestern US characterized the millennial scale fluctuations in precipitation history in the region over most portions of the past 60 ka (Asmerom et al. 2010; Wagner et al. 2010) (Fig. 1).

2.4 The Eastern Mediterranean domain

The modern climate in the Eastern Mediterranean region is semiarid with rainfall occurring mainly during the boreal-winter season (December–May), associated with air masses originating from the northeastern Europe and the Atlantic Ocean through the Mediterranean Sea in the west. A number of speleothem δ¹⁸O records from the region have been developed mainly from Israeli caves. The long-term speleothem δ¹⁸O variations in the Eastern Mediterranean climate region have been interpreted to reflect the δ¹⁸O variability of the Mediterranean Sea surface water (Bar-Matthews et al. 2003). The growth of cave calcite deposits in southern Israel appears to have occurred at the times of the strongest North African summer monsoon (NASM) (Vaks et al. 2006, 2007, 2010; Waldmann et al. 2010).

3 Speleothem δ¹⁸O proxy

Speleothem calcite δ¹⁸O has been widely used as a PM proxy. Although the speleothem δ¹⁸O signatures can be influenced by numerous and complex factors (e.g., Fairchild et al. 2006; Lachniet 2009), robust replications among speleothem δ¹⁸O records on orbital-centennial timescales from different caves in ASM and SASM regions demonstrate that these speleothems grew under conditions close to isotopic equilibrium, and therefore their δ¹⁸O records mainly represent changes in precipitation δ¹⁸O values with regional extent (e.g., Cheng et al. 2006). However, the climatic interpretation of δ¹⁸O records remains a subject of considerable debate, particularly in the EASM domain (e.g., Maher 2008; Dayem et al. 2010; Clemens et al. 2010; Pausata et al. 2011; Caley et al. 2011b).

In the ISM domain, precipitation δ¹⁸O changes have generally been interpreted to primarily indicate variations in local rainfall amount (e.g., Burns et al. 1998, 2001, 2003; Sinha et al. 2005, 2011; Shakun et al. 2007) and this interpretation is supported by recent climate modeling studies (e.g., Vuille and Werner 2005; LeGrande and Schmidt 2009; Pausata et al. 2011). However, this explanation may not be applicable in the EASM domain. In the EASM domain, the onset of summer monsoon rainfall occurs in late May in the South China Sea and southeastern coastal areas of China, with monsoon rainfall advancing to central-eastern China by June and eventually reaching northern-northeastern China in July. During this seasonal change in precipitation, the land-sea temperature contrast increases progressively, driving stronger and larger atmospheric circulation, thus, transporting more moisture from distal ocean sources into China (Fig. 2). As a larger fraction of water vapor is removed during the transfer trajectory from more distal water sources, δ¹⁸O values in precipitation become progressively lower (e.g., Yuan et al. 2004) and reach the minima during July–August (Fig. 3), coinciding with the maxima in the land-sea temperature contrast. During autumn, the δ¹⁸O value becomes progressively higher with the retreat of the summer monsoon (Fig. 3). The winter (December–February) precipitation in eastern China however, shows a relatively large δ¹⁸O variation from low δ¹⁸O values in the north to high values in the south (Fig. 3), possibly in response to contrasting winter temperatures. In general, the aforementioned observations demonstrate that the seasonal δ¹⁸O precipitation cycle correlates inversely with the summer monsoon intensity and/or the distance of moisture sources, except the lower δ¹⁸O values associated with winter precipitation in North China. The latter, however, is less important because of its small contribution to the weighted mean annual precipitation in the area.

Fig. 3

Weighted monthly δ¹⁸O values of precipitation obtained from meteorological observatories in eastern China (data from IAEA, accessible at: http://isohis.iaea.org). Green curves are from southern China (approximate 30–20°N, i.e., Hong Kong, Guangzhou, Chengdu, Guilin, Guiyang, Changsha, Haikou, Kunming, Liuzhou and Zunyi observatories). Blue from central eastern China (approximate 30–35°N, i.e., Nanjing, Wuhan and zhengzhou observatories), and red from northern China (approximate 35–40°N, i.e., Shijiazhuang, Yantai, Tianjing, Xi’an, and Taiyuan observatories). The wide vertical arrow indicates the onset of the summer monsoon in May at China’s southeast coast, June in central eastern China and July in northern China. The vertical grey bar illustrates approximate time period of the summer monsoon. The summer monsoon precipitation is characterized by distinctly lower δ¹⁸O than precipitation during the rest of the year. The altitude effect may have some bearing on the δ¹⁸O values for the high altitude sites (e.g., Kunming). It also appears that the atmospheric temperature effect may affect the winter δ¹⁸O values, because these values display a decrease trend from southern to northern China, possibly linked to the winter temperature decrease northward

The seasonal cycle of EASM precipitation δ¹⁸O can serve as an analogue for interpreting δ¹⁸O variations in the speleothem records. When the Northern Hemisphere summer insolation (NHSI) is high on precessional scale, the summer (winter) insolation becomes higher (lower), while the spring–autumn insolation remains approximately the same (Berger 1978). As a result, it is plausible that both summer and winter monsoons become more intense. However, the summer monsoon enhancement may play a more important role in modulating the mean annual δ¹⁸O values in precipitation (and by extension, δ¹⁸O values in speleothem) because increase in the summer insolation would lead to an overall higher summer monsoon rainfall whereas decrease in winter insolation would lead to stronger winter monsoon circulation or colder-drier conditions during winter. The speleothem δ¹⁸O variance on precession cycles has, thus, been explained by a change in the ratio of the amount of summer (lower δ¹⁸O) to non-summer precipitation (higher δ¹⁸O) (Wang et al. 2001).

In eastern China, amount of local precipitation could be decoupled, to some extent, from the EASM intensity. For instance, strong summer monsoon in July–August may drive more frontal rainfall up to northern China and less in central-eastern China than in June. Modern meteorological observations on interannual-scale also show that the stronger EASM can penetrate farther into North China and bring more frontal rainfall into the region and, conversely, a weaker EASM may deliver more rainfall in central-eastern China and less up to the north (Ding 1992). A recent simulation study shows that Holocene precipitation δ¹⁸O values correlate well with summer monsoon rainfall amount in the ISM domain, but not in the EASM domain, and instead, the EASM precipitation δ¹⁸O indicates the mean state of monsoon intensity or integrated summer water vapor transported into the EASM region (LeGrande and Schmidt 2009). On millennial timescale, simulation study also suggests a possible scenario that excursions of the precipitation δ¹⁸O depletion (enrichment) may mainly reflect strong (weak) ASM events as a whole, rather than amount of local rainfall at an individual site in southeastern China (Pausata et al. 2011). Based on aforementioned observations, it is apparent that variations in the δ¹⁸O of EASM precipitation indicate a mean state of summer monsoon intensity or integrated moisture transport rather than amount of local precipitations (Johnson and Ingram 2004; Cheng et al. 2006).

In the SASM domain, speleothem δ¹⁸O records are also tightly linked to the SASM intensity (e.g., Wang et al. 2006; van Breukelen et al. 2008; Cruz et al. 2005, 2007, 2009b; Kanner et al. 2012). In southeastern Brazil, for instance, precipitation δ¹⁸O has a significant negative correlation with the SASM intensity as indicated by both instrumental data and Atmospheric General Circulation Model simulations (e.g., Vuille and Werner 2005; Lee et al. 2009). There are two moisture sources that contribute to rainfalls in the region. The LLJ transports moisture eastwardly with lower δ¹⁸O value due to its long trajectory (Hoffmann 2003), and the water vapor from nearby Atlantic is characterized by distinctly higher δ¹⁸O (Fig. 1). As a result, the speleothem δ¹⁸O variations mainly represent the change in ratio between the precipitation from the two sources (Cruz et al. 2005; Wang et al. 2006). Broadly, an enhanced SASM is associated with a stronger LLJ, hence, transporting more moisture into downstream southeastern Brazil (Liebmann et al. 2004), resulting in the lower δ¹⁸O of precipitation (Cruz et al. 2006; Wang et al. 2006). On the other hand, the influence of moisture from local Atlantic source is less important on the δ¹⁸O variability (Wang et al. 2006; Cruz et al. 2007). Thus, cave δ¹⁸O values in the region primarily indicate the strength of the LLJ along the moisture trajectory from the Amazon Basin to southeastern Brazil (i.e., the intensity of the SASM) with lower (higher) calcite δ¹⁸O in speleothem records indicating an intensified (weakened) SASM with a more (less) Amazon moisture contribution.

In the North American Monsoon region, the speleothem δ¹⁸O records mainly document the change in the winter (low δ¹⁸O)/summer (high δ¹⁸O) precipitation ratio or the amount of winter precipitation which is characterized by relatively depleted δ¹⁸O (Asmerom et al. 2010; Wagner et al. 2010).

4 Correlations among speleothem records

4.1 Orbital timescales

The EASM δ¹⁸O record shows a broadly sinusoidal pattern with a dominant cycle of ~23-ka that closely follows the NHSI on precessional timescale (Yuan et al. 2004; Wang et al. 2008; Cheng et al. 2009b). Spectral analysis shows no significant phase difference between mid-July insolation and the mean EASM δ¹⁸O signal (Wang et al. 2008). Recently, two hypotheses were proposed to explain the observation that the EASM follows the mid-July NHSI. (1) The first hypothesis (Wang et al. 2008) is based on the analogy that in the EASM domain, seasonal precipitation δ¹⁸O reaches the minima around mid-July (Fig. 3)—when the land-sea temperature contrast becomes largest and the summer monsoon is strongest (Fig. 3). (2) EASM speleothem δ¹⁸O records show an apparent phase lag of 30° (or ~2–3 ka) to the mid-June NHSI and this lag (but in phase with the mid-July insolation) has been attributed to the influence of high-latitude millennial-scale abrupt climate events on the EASM, which presumably delayed the monsoon response to the rising NHSI (Ziegler et al. 2010).

Although the two hypotheses are different in details, both emphasize variations in insolation as a major driving force of the EASM. Unfortunately, the speleothem records from the ISM domain are rather discontinuous on orbital timescale without a sufficient temporal coverage or resolution to fully test the insolation hypothesis. Nevertheless, existing speleothem δ¹⁸O records from Qunf Cave in southern Oman (Fleitmann et al. 2003) and Timta Cave in northwestern India (Sinha et al. 2005) indeed demonstrate that variations in the ISM occurred in concert with the EASM at least over the past 15 ka (Fig. 4). The possible synchronicity between the EASM and ISM on orbital timescales is also supported by a recent speleothem record from Tianmen Cave, Tibet- a region within the ISM domain (Cai et al. 2010). The Tianmen record shows that the ISM variability during portions of the Marine Isotope Stage (MIS) 5 (mainly 5a, 5c, and 5e) is also broadly similar to Chinese EASM records (Fig. 5).

Fig. 4

Cave δ¹⁸O records over the past 20 ka from EASM, ISM, SASM and Eastern Mediterranean regions. a EASM records from Hulu (blue) and Dongge (dark blue) Caves (Wang et al. 2001; Dykoski et al. 2005). b ISM records from Qunf (red) and Timta (orange, relative scale) Caves (Fleitmann et al. 2003; Sinha et al. 2005). c SASM records from Botuverá Cave (Cruz et al. 2005; Wang et al. 2006, 2007a). Different colors denote different stalagmites. d The Eastern mediterranean record from Soreq Cave (Bar-Matthews et al. 2003). Summer insolation (grey curves) at 65°N (in a, b and d) and 30°S (in c) were plotted for comparison (Berger 1978)

Fig. 5

Comparison between the ISM and EASM δ¹⁸O records over the Marine Isotope Stage (MIS) 5 (adapted from Cai et al. 2010). a Hulu and Dongge records (Cheng et al. 2006; Kelly et al. 2006). b The Sanbao record (Wang et al. 2008). c The Tianmen record (Cai et al. 2010). Different colors denote different stalagmites. Summer insolation at 65°N (grey) was plotted for comparison (Berger 1978)

A number of speleothem δ¹⁸O records from tropical-subtropical South America have demonstrated that variations in the SASM track changes in Southern Hemisphere summer insolation on orbital timescale (Cruz et al. 2005, 2007; Wang et al. 2007a) and thus, exhibit an interhemispheric anti-phased relationship with ASM records at precession bands (Fig. 6). This is because Earth’s orbital precession affects the seasonal distribution of solar radiation resulting in precessional increase in northern tropical-subtropical summer insolation which is balanced by decreases in southern tropical-subtropical summer insolation and vice versa. The sea-saw oscillations in solar radiation, alternating between tropics-subtropics of two hemispheres, can thus explain the anti-phased relationship between the ASM and SASM, because the insolation forcing intensifies summer monsoon intensity due to enhanced land-sea thermal contrast (e.g., Kutzbach 1981; Kutzbach et al. 2008). Similar interhemispheric anti-phased relationship had also been noted in the African Monsoon domain. For example, South African Monsoon variations reconstructed from the Pretoria Saltpan sedimentary record, South Africa (~25.5°S) (Partridge et al. 1997) and the NASM index based on fossil faunal assemblage variations in deep-sea sediment core RC24-07 (20°N) (McIntyre et al. 1989), track southern and northern tropical-subtropical summer insolation changes, respectively, during the late Pleistocene.

Fig. 6

Comparison among the benthic, atmospheric and speleothem δ¹⁸O records, and marine records. a The stacked benthic δ¹⁸O record (Lisiecki and Raymo 2005). b The atmospheric δ¹⁸O record from Vostok ice core, Antarctica (Suwa and Bender 2008). c EASM records are composited from records of Dongge (dark blue), Hulu (blue), Sanbao (green) and Linzhu (black) Caves (Wang et al. 2001; Dykoski et al. 2005; Wang et al. 2008; Cheng et al. 2009b). For comparison, the Hulu and Dongge δ¹⁸O records are plotted 1.6 % more negative to account for their higher values than Sanbao cave. d The Eastern Mediterranean ODP 968 color reflectance record (540 nm, %) (Ziegler et al. 2010). e The stacked marine ISM record (Clemens and Prell 2003). f The SASM record from Botuverá Cave (Cruz et al. 2005). g The Eastern Mediterranean record from Soreq (brown) and Peqiin (blue) Caves (Bar-Matthews et al. 2003). h The Kesang Cave record from the Westerlies region. The vertical yellow bars depict the Terminations (T-I to T-IV, Cheng et al. 2009b). Red arrows show strong ASM events during glacial times as documented by terrestrial mollusk records from the Chinese Loess Plateau (Rousseau et al. 2009) and red dots depict their possible correlations to the atmospheric and ASM δ¹⁸O records. Summer insolation (mid-July, grey curves) at 65°N (in c, e, g and h) and 30°S (mid-January, in f) (Berger 1978) were plotted for comparison. While the EASM record (c) broadly tracks the mid-July insolation change at 65°N and proves its significant similarity to the atmospheric δ¹⁸O record (b), the marine ISM from the Arabian Sea (e) appears to lag to the same insolation change by ~8 ka at precession bands. The remarkable similarity between the Eastern Mediterranean ODP 968 (d) and the ASM (c) records demonstrates a close link between the NASM and ASM

In the eastern Mediterranean region, a composite δ¹⁸O record for the last 250 ka from Soreq and Peqiin Caves, Israel is characterized by precessional variations that are superimposed on strong 100-ka cycles (Bar-Matthews et al. 2003). The rainfall in this region is derived mainly from the adjacent Eastern Mediterranean Sea and, thus, the Soreq and Peqiin δ¹⁸O records broadly follow variations in the δ¹⁸O of the sea surface water, the latter being inferred from the marine Globigerinoides ruber δ¹⁸O record (e.g., Fontugne and Calvert 1992). The δ¹⁸O of surface water of the Mediterranean Sea is primarily controlled by river discharge, local precipitation/evaporation, and exchanges of the Mediterranean Sea water with the Atlantic Ocean through the Strait of Gibraltar, and with the Black Sea through the Dardanelles/Bosporus Straits (Badertscher et al. 2011).

Because the Atlantic seawater δ¹⁸O variance is dominated by the 100-ka cyclicity (e.g., Lisiecki and Raymo 2005), the water exchange between the Mediterranean Sea and Atlantic waters (except during the periods characterized by formation of Mediterranean sapropels) presumably modulated the δ¹⁸O variability of the Eastern Mediterranean Sea surface water on a 100-ka periodicity (e.g., Arz et al. 2003; Tzedakis 2007; Rohling et al. 2009; Göktürk et al. 2011) and thus, by extension, the δ¹⁸O variations in Soreq and Peqiin records. On the other hand, the Eastern Mediterranean Sea also receives large quantity of freshwater discharge from the northern Africa rivers (e.g., the Nile River). The precession cyclicity of the NASM, exerted on the fluctuations in freshwater discharge, can be imprinted in marine sediment and speleothem records in the region as well (e.g., Rossignol-Strick et al. 1982; Rossignol-Strick 1983; Pokras and Mix 1987; Weldeab et al. 2007; Ruddiman 2008; Ziegler et al. 2010). Recently, a remarkable correlation has been established between the Eastern Mediterranean ODP 968 color reflectance and ASM records (Ziegler et al. 2010), further suggesting an in-phased response of ASM and NASM systems to external forcing (Fig. 6).

The speleothem records from Soreq Cave, Israel (Bar-Matthews et al. 2003) and Antro del Corchia, Italy (Drysdale et al. 2009) are characterized by a gradual decrease in δ¹⁸O values during a Weak Monsoon Interval characterized in the EASM record around Termination II when the sea level was presumably rising (Cheng et al. 2009b) (Fig. 7). Following the gradual decreasing trend is an abrupt decrease in δ¹⁸O at ~129 ka in both Israeli and Italian records, which is correlated, within dating errors, with an abrupt increase in both ASM (Cheng et al. 2009b) and NASM (Weldeab et al. 2007) (Fig. 7) intensities. A plausible mechanism may involve a gradual rise in sea level from the continental ice meltdown followed by an abrupt increase of the Nile River discharge due to abrupt intensification of the NASM. This hypothesis is consistent with the observation that while the magnitude of δ¹⁸O change in both Israeli and Italian cave records, corresponding to the gradual rise of the sea level, is similar, the abrupt decline in δ¹⁸O values at Israeli cave sites is much larger (more than 2 ‰) than the Italian site (slightly less than 1 ‰) (Fig. 7). The disparity in the δ¹⁸O amplitude likely stems from the fact that the Israeli site being in a close proximity with the Eastern Mediterranean was more sensitive to changes in the Nile River discharge (Fig. 1).

Fig. 7

Monsoon records around Termination-II. a Three curves show the Sanbao δ¹⁸O record (green) (Wang et al. 2008), the Soreq δ¹⁸O record (brown) (Bar-Matthews et al. 2003) and the inferred sea level (blue) (Cheng et al. 2009b), respectively. b The green and blue curves are as same as those in (a). The purple cure is the Corchia δ¹⁸O record (Drysdale et al. 2009). c The NASM proxy record: the δ¹⁸O record of planktonic foraminifer Globigerinoides ruber in a marine sediment core from the Gulf of Guinea (Weldeab et al. 2007). The grey bar denotes the Weak Monsoon Interval (WMI) recorded in ASM records (Cheng et al. 2006, 2009b). The red arrow indicates a synchronous jump among the ASM, NASM and δ¹⁸O in Mediterranean cave records

Today, the Saharan-Arabian Desert is the largest and the most arid region in the world. However, this modern hyper-arid region appears to have many humid periods in its Quaternary climate history, when the summer monsoon was stronger during times when the NHSI was much higher than today, leading to formations of lakes and rivers (e.g., Szabo et al. 1995; Crombie et al. 1997). Similarly, intervals of enhanced speleothem formation in the Arabian Peninsula (e.g., Hoti Cave, northern Oman, Burns et al. 1998, 2001; Fleitmann et al. 2011, Mukalla Cave, southern Yemen Fleitmann et al. 2011) and in the Negev desert, Israel (Vaks et al. 2006, 2007, 2010) indicate episodes of wide-spread humid conditions that can be linked to intensifications of the ISM (Burns et al. 1998, 2001; Fleitmann et al. 2011) and perhaps the NASM (Waldmann et al. 2010; Battisti et al. 2012) during times of high NHSI, including MIS 1, 5.1, 5.3, 5.5, 7.1, and 9. In contrast, speleothem formation in these areas was virtually absent during glacial periods (Burns et al. 2001; Vaks et al. 2010; Fleitmann et al. 2011), implying significantly weak summer monsoon associated with glacial periods. Recently, a speleothem record from Kesang Cave in western China documents a precipitation history over the past 500 ka in the current semiarid-arid Westerlies region (Cheng et al. 2012a). The Kesang record also shows a precession rhythm that is possibly linked to incursions of ASM rainfall during the insolation peaks when the ASM was much stronger than at the present time (Fig. 6).

In summary, the aforementioned orbital-scale changes in monsoons are dominated by variation between two modes with low (high) δ¹⁸O values corresponding to strong (weak) PM state. These patterns of PM variations are primarily related to summer insolation changes and, to some extent, the ice volume and possibly other factors as well. The transitions between the two modes were however, generally abrupt or non-linear in comparison to insolation forcing, possibly related to abrupt climate events at high northern latitudes (Cheng et al. 2009b; Ziegler et al. 2010) and/or due to some threshold mechanisms, such as the atmospheric humidity level over oceans around the monsoon regions and its feedbacks (Levermann et al. 2009; Schewe et al. 2011).

4.2 Millennial-centennial timescales

The ASM δ¹⁸O records, most notably from China, contain prominent millennial-length events, including all the counterpart events recorded in Greenland ice core records, such as the Younger Dryas (YD), Dansgaard-Oeschger (D/O), and Heinrich (H) events (e.g., Wang et al. 2001, 2008; Ma et al. 2012). The so-called ‘Chinese interstadials’ (CIS) nomenclature has been introduced to indicate strong ASM events recorded in speleothem records (Cheng et al. 2006) that correlate with the D/O events during the last glacial period in Greenland ice cores. The complete series of CIS events are similar between the last and penultimate glacial periods in terms of the variability in their frequency and duration, indicating an ice volume effect on their characteristics and pacing (Cheng et al. 2006; Wang et al. 2008). The millennial events in ASM records (e.g., Wang et al. 2001, 2008; Burns et al. 2003; Cheng et al. 2006; Cai et al. 2006; Shakun et al. 2007) exhibit negative correlation with δ¹⁸O records from the SASM region including both equatorial and subtropical areas (Cruz et al. 2005, 2009a, b; Wang et al. 2004; 2007a, b; Cheng et al. 2012b; Kanner et al. 2012) (Fig. 8). As noted in earlier studies, this low-latitude interhemispheric precipitation see-saw pattern is likely related to shifts in positions of the ITCZ and associated asymmetry in Hadley circulation in two hemispheres, perhaps ultimately triggered by changes in deep water formation in the North Atlantic Ocean (e.g., Lindzen and Hou 1988; Clement et al. 2004; Wang et al. 2006, 2007a b). Model simulations show that large meltwater discharge into the North Atlantic can substantially weaken the Atlantic Meridional Overturning Circulation (AMOC), resulting in increasing sea ice coverage in the northern high latitudes and meridional Atlantic sea surface temperature (SST) gradients. This, in turn, can induce significant responses in the tropics-subtropics (e.g., Chiang et al. 2003; Chiang and Bitz 2005; Zhang and Delworth 2005), resulting in a southward ITCZ shift and subsequently an alteration of the Hadley circulation (Lindzen and Hou 1988). Such modification produces intensified uplift of moist air in the southern low latitudes but strong subsidence in the north (Lindzen and Hou 1988; Clement and Peterson 2008). As a result, the ASM is weakened while the SASM intensifies and vice versa. Indeed, similar to their Greenland counterparts, millennial paleomonsoon events in ASM and SASM records oscillate between two states along orbital long-term trend, demonstrating a causal linkage and their interhemisphere extent.

Fig. 8

Correlations among millennial events in ASM, SASM and North American Monsoon records. a The Greenland ice core δ¹⁸O record (NGRIP, Svensson et al. 2008). b Speleothem records from southwestern North America: the Fort Stanton cave record (brown, Asmerom et al. 2010) and Bells Cave record (red, Wagner et al. 2010). c The EASM record is compiled using the Hulu (green, Wang et al. 2001) and Sanbao (purple, Wang et al. 2008) records. d The ISM record from Moomi Cave (Burns et al. 2003; Shakun et al. 2007). e The ISM record from Xiaobailong Cave (Cai et al. 2006). f The SASM record from Pacupahuain Cave (Kanner et al. 2012). g Northeastern Brazil speleothem growth (wet) periods (green) (Wang et al. 2004). h The SASM record from Botuverá records (blue (Wang et al. 2006) and dark blue (Wang et al. 2007a)). Numbers indicate Greenland Interstadials. Vertical yellow bars denote H events (H2–H5) and grey bars indicate correlations among northeastern Brazil wet periods, strong SASM events and cold Greenland-weak ASM events. Summer insolation (grey curves) at 65°N (in c) and 30°S (in g) (Berger 1978) were plotted for comparison

In addition, speleothem δ¹⁸O records from southwestern North America (Asmerom et al. 2010; Wagner et al. 2010) are characterized by a series of millennial events during the last glacial period (Fig. 8). When the cold events documented in Greenland ice cores occurred, the polar jet stream shifted southward, modulating the position of the winter storm tracks and winter precipitation (with distinctly lower δ¹⁸O values) increased in the region (Asmerom et al. 2010). The millennial scale events in southwestern North America, both in terms of changes in precipitation δ¹⁸O and amount, are anti-phased with their counterparts in the ASM domain and in-phased with those in the SASM domain (Fig. 8).

One of the most prominent and widespread abrupt climate events during the Holocene is the 8.2 ka event (Alley et al. 1997; Cheng et al. 2009a). The timing and structure of the 8.2 ka event has been well documented in ASM records (i.e., Hoti and Qunf records, Oman and the Dongge record, China) and SASM records (i.e., the Paixão, Padre and Lapa Grande records) (Wang et al. 2006; Cheng et al. 2009a; Stríkis et al. 2011). Similar to other millennial-scale climate events, the 8.2 ka event manifests as a weak (strong) monsoon event in the ASM (SASM) records. Interestingly, the magnitude of δ¹⁸O excursion in the SASM records is much larger (~2.5 ‰) than that in the ASM records (≤1 ‰) and tracks Greenland records closely in terms of structure (Cheng et al. 2009a). This observation is consistent with a North Atlantic origin of the event and a tight link between the Atlantic sector of the ITCZ and the North Atlantic climate (Wang et al. 2007a, b). Another notable centennial-scale event is the Little Ice Age (LIA) in human cultural history. The apparent weak ASM during the LIA (Zhang et al. 2008) coincides with the strong SASM at the same time period (Reuter et al. 2009). Furthermore, as Cheng et al. (2010) pointed out, a strong ASM period during the Chinese Northern Song Dynasty coincided with dry conditions that Tiwanaku endured in South America, which disrupted agricultural production (Ortloff and Kolata 1993). A recent high-resolution stalagmite record from central-eastern Brazil further revealed a general anti-phased relationship between the ASM and SASM during Bond events in the Holocene, similar to the general anti-phase relationship on longer timescales (Stríkis et al. 2011; Cheng et al. 2012b) (Fig. 9).

Fig. 9

Comparison of Bond events among the North Atlantic, ASM and SASM records (adapted from Stríkis et al. 2011). a Hematite stained quartz grain (HSG) record from the North Atlantic deep sea core VM 29-191 (Bond et al. 1997). b The SASM record from Lapa Grande Cave in central-eastern Brazil (Stríkis et al. 2011). c The ASM δ¹⁸O record from two stalagmites DA (green, detrended) and D4 (brown) from Dongge Cave in southeastern China (Wang et al. 2005; Dykoski et al. 2005). The bond events 0–6 are indicated by the grey bars

High-resolution and precise-dated speleothem records are required to reconstruct decadal-scale paleomonsoon variations. Such records are relatively few, in part because of the notion that on this timescale, the amplitude of monsoon related δ¹⁸O change in the speleothem is expected to be small and sometimes indiscernible from the ‘noise’ produced from karst and non-climate related processes. Notwithstanding, high-resolution speleothem δ¹⁸O records from Dandak and Jhumar Caves in central India (Sinha et al. 2011) and Wanxiang Cave in China bear remarkable similarity during the late Holocene (Berkelhammer et al. 2010). Records from both regions contain evidence for decadal-multidecadal length intervals of sustained reductions in monsoon rainfall (megadrought) during the last millennium. These megadroughts have been linked to discrete intervals of cooler conditions in the extratropical northern latitudes and southward displacement of the ITCZ (Sinha et al. 2011).

5 Equatorial-tropical records and possible links to the Walker circulation

While the ASM and SASM variations inferred from speleothem δ¹⁸O records clearly describe the anti-phased ~23 ka precessional cycles, speleothem δ¹⁸O records from sites near the equatorial-tropics (low latitudes (<10°)) tend to exhibit complex precipitation behavior at precessional bands with some of their δ¹⁸O changes tracking neither insolation nor δ¹⁸O changes from other low latitude sites within the same hemisphere. For example, speleothem δ¹⁸O variations spanning the last precession cycle from Liang Luar Cave in western Flores, Indonesia (~8.5°S, Griffiths et al. 2009), Borneo caves, Malaysia (~4°N, Partin et al. 2007), Cueva del Tigre Perdido Cave, northern Peru (~6°S, van Breukelen et al. 2008), and northeastern Brazil (~5.5°S, Cruz et al. 2009a) are quite different. While the δ¹⁸O orbital-scale changes in the above two Asian records bear some similarities, even though they are located in two hemispheres respectively, the two South American records within the same hemisphere are nearly anti-phased (Figs. 1 and 10). These disparities may have resulted from complex changes in the Walker circulation associated with deep monsoon convection over equatorial-tropical regions, perhaps tied to insolation and/or glacial boundary conditions (Clement et al. 2004; Cruz et al. 2009a). These complex phase relationships are, to some extent, expected, considering the dramatic changes observed in the zonal atmospheric circulations along the equatorial-tropics and associated large precipitation and temperature fluctuations between El Niño and La Niña years. Further investigations, particularly on longer timescales, are important to reach a comprehensive understanding about patterns of the paleoclimat change in the equatorial-tropical region and the possible cause of it.

Fig. 10

Cave δ¹⁸O records from equatorial-tropical areas. a The Borneo δ¹⁸O record (~4°N, Malaysia, Partin et al. 2007). b The Liang Luar record (~8.5°S, Indonesia, Griffiths et al. 2009). c The cave δ¹⁸O record from northeastern Brazil (~5.5°S, Cruz et al. 2009a). d The Cueva del Tigre Perdido record (~6°S, Peru, van Breukelen et al. 2008). Different colors denote different stalagmites. Summer insolation (grey curves) at 5°N (a) and 5°S (b–d) were plotted for comparison (Berger 1978)

6 Discussions

6.1 Orbital-scale variability

The speleothem δ¹⁸O records indicate that long-term histories of the ASM and SASM are dominated by precessional cyclicity (e.g., Wang et al. 2008; Cheng et al. 2009b; van Breukelen et al. 2008; Cruz et al. 2005, 2007, 2009b; Wang et al. 2007a), suggesting a dominating role of summer insolation forcing. This mechanism agrees well with a series of simulation studies which indicate that increased summer solar radiation appears to be most effective in strengthening monsoons (e.g., Prell and Kutzbach 1987, 1992; Masson et al. 2000; Kutzbach et al. 2008; Yin and Berger 2011; Battisti et al. 2012). In contrast, marine proxy records from the Arabian Sea suggest a significant time lag (~8 ka) of the ISM in response to NHSI change on the precessional cycle (e.g., Clemens et al. 1991, 2010; Clemens and Prell 2003; Caley et al. 2011b). The ~8 ka phase lag has been linked to mechanisms of latent heat transport from the southern Indian Ocean and global ice volume influence on the ASM that lags NHSI by ~11 and 5 ka at precession bands, respectively (Clemens et al. 2010).

The apparent lag of the ISM, as inferred from the marine records, has recently been used to reinterpret Chinese speleothem δ¹⁸O records, because a major portion of summer monsoon precipitation in southeastern China is presumed to result from the ISM system (Clemens et al. 2010). This interpretation singles out three distinctive components relevant to precipitation seasonality in southeastern China: summer (JJA, 48 % of total annual precipitation with δ¹⁸O of ~−8.78 ‰ originating from the Indian Ocean source), winter (SONDJFM, 34 % of total annual precipitation with δ¹⁸O of ~−6.60 ‰ from continental sources) and spring (AM, 18 % of total annual precipitation with δ¹⁸O of ~−3.14 ‰ from the Pacific source). The summer δ¹⁸O signal in Chinese speleothem records (emanating from the ISM domain) has been deemed to exhibit a ~8 ka lag to NH June insolation. On the other hand, the residual δ¹⁸O signal (non-summer precipitation) has been argued to stem from winter temperature influence and is in phase with the NH June (summer solstice) insolation change at precession bands. In this sense, the Chinese speleothem records reflect a mixed signal produced by aforementioned three components, yielding an orbital phase that lags mid-June insolation by ~2.9 ka.

If the EASM δ¹⁸O records could be solely explained by mixing two distinct precipitation components, summer and non-summer, that have ~8 and 0 ka phase differences at precession bands respectively, it would be logical to presume that the δ¹⁸O records may vary accordingly from ~8 ka lag (100 % summer precipitation) to in-phase (100 % non-summer precipitation) relative to NHSI change on precessional cycles, depending on the ratio of local summer/non-summer precipitations. In other words, this interpretation indeed predicts possible phase differences on precessional scale among speleothem δ¹⁸O records in ASM regions, simply because seasonal distributions of precipitation and their moisture sources diverge significantly in the vast ASM domain. For instance, in eastern China, the summer monsoon onset is two months earlier in the south (in May) than in the north (in July). Furthermore, in southwestern China (e.g., Xiaobailong Cave, Yunnan (Fig. 1)), the spring Pacific moisture source is effectively absent, and thus the cave δ¹⁸O records from this region should be theoretically different from the records in eastern China. However, thus far, neither the Xiaobailong record (Cai et al. 2006, 2011), nor any other published or unpublished Chinese cave records denote any resolvable phase differences at precession bands.

A recent speleothem δ¹⁸O record from Tianmen Cave in the south-central Tibetan Plateau, located at considerable distance from southeastern China (Fig. 1), characterizes an ISM precipitation history over much of MIS 5 (Fig. 5) (Cai et al. 2010). The mean annual precipitation at this site is ~300 mm and ~86 % of it falls during summer months (June–September) when the ISM prevails in the area. The precipitation at this cave site virtually contains no moisture from the Pacific source and the winter precipitation component is also small and, therefore, this record is suitable for testing the orbital phase of the ISM. The Tianmen record is basically similar in both structure and timing to the cave records from southeastern China, suggesting no notable phase difference between Tianmen area and southeastern China (Fig. 5). In addition, two other cave records from the ISM region, the Timta record, northwestern India (~80 % of total annual precipitation from the ISM) and Qunf record, Oman (90 % of total annual precipitation from the ISM) (Fig. 1), also display a pattern during the past 15 ka which is essentially similar to those established from southeastern China (Fig. 4), even though modern precipitation patterns in the two cave sites are rather different from those in southeastern China (Sinha et al. 2005; Fleitmann et al. 2003).

The comparison between speleothem records from ISM and EASM domains provides critical evidences that patterns of the cave δ¹⁸O change in the vast ASM region are mostly coherent and their precession cyclicity appears to be in-phase with NHSI, notwithstanding substantial variance in local seasonal precipitation patterns and moisture sources. This observation suggests that the winter and summer monsoons may conceivably be in-phase at precession bands and, as a result, the overall phases are virtually similar over the vast ASM region regardless of the difference in seasonal precipitation distribution among different cave locations. A supporting line of evidence comes from the ice core δ¹⁸O record of atmosphere O₂ (δ¹⁸O_atm). The Antarctic Vostok ice core δ¹⁸O_atm record on an independent N₂/O₂ chronology (Suwa and Bender 2008) proves a striking similarity and strictly in-phase correlation with the EASM δ¹⁸O record (Wang et al. 2008; Cheng et al. 2009b) on precessional cyclicity (Fig. 6). Severinghaus et al. (2009) further demonstrated a definite and precise correlation between the Dongge-Hulu δ¹⁸O and high-resolution Siple Dome δ¹⁸O_atm records on orbital-millennial timescale. This correlation reveals an important impact of monsoon rainfall on the biosphere process which amends the “Dole Effect”. Moreover, because the ice core δ¹⁸O_atm changes pertain to be an integrated global signal, very likely linked to tropical-subtropical hydrological variations (e.g., Severinghaus et al. 2009), their in-phase correlation with the EASM δ¹⁸O record is consistent with the view that EASM variations are basically synchronous with those in the ISM and NASM domains, and incoherent with the idea that the EASM δ¹⁸O record from southeastern China has a random orbital phase shaped primarily by local seasonal precipitation distribution. Nevertheless, the disparity between the marine records from the Arabian Sea and the ASM speleothem records remain an important question. Various possibilities can be raised such as whether the Arabian Sea domain has unique precipitation variability which is dissimilar to the ASM in Indian subcontinent and East Asia (e.g., Conroy and Overpeck 2011); or whether the ISM, including the Arabian Sea region, is different from the East Asian Monsoon, in response to insolation change (Yin et al. 2009). And finally, whether the marine ISM records primarily reflect changes in monsoon wind strength (monsoonal upwelling offshore the Arabian coast) or monsoon precipitation (Ruddiman 2006)? The ongoing work on long and continuous speleothem records from typical ISM regions (Gayatri Kathayat and Sebastian Breitenbach, personal communications) is particularly critical for further testing the aforementioned phasing issue.

One of the notable features of ASM and SASM speleothem records is the absence of significant 100-ka cycle. Although some strong summer monsoon periods under glacial conditions were revealed from the Chinese Loess Plateau (Rousseau et al. 2009) which possibly occurred in the NASM domain as well (Mélières et al. 1997) (Fig. 6), the ASM reconstruction based on loess records from the Chinese Loess Plateau shows a strong 100-ka cycle, with strong (weak) summer monsoon occurring during interglacial (glacial) periods (e.g., Ding et al. 1995; Guo et al. 2011). The presence of dominant precessional cycles in Chinese speleothem records suggests a generally weak ice volume impact in tropical-subtropical monsoon regimes, and this appears to be consistent with a mechanism of ice volume threshold proposed in detail in Yin et al. (2009). We additionally offer here another possibility of threshold mechanism. If the speleothem δ¹⁸O variations are largely dependent on the ratio of summer versus winter precipitation, each with its characteristic δ¹⁸O values, rather than on absolute rainfall, a large increase in summer rainfall above certain threshold value could just result in a similar δ¹⁸O value virtually determined by the δ¹⁸O value of the summer monsoon component. In other words, once the summer monsoon dominates annual precipitations, speleothem δ¹⁸O value does not drop further although summer rainfall might continue to increase. The EASM speleothem records are indeed characterized by comparable absolute δ¹⁸O minima that coincide with NHSI peaks both during the glacial and interglacial periods (Fig. 6). At face value, this observation implies no significant weakening of the EASM at high summer insolation phases even during the full glacial periods in tropical-subtropical regions. On the other hand, speleothem growth in Yemen and Oman-locations that are at the current fringes of the ISM, occurred only during NHSI peaks in interglacial periods (Burns et al. 2001; Fleitmann et al. 2011), implying that the ISM was significantly weaker during glacial periods. The threshold hypothesis may explain this discrepancy, as the similar δ¹⁸O values in the EASM record reached at the insolation peaks during both glacial and interglacial periods could result from a same threshold value, likely the δ¹⁸O value of summer moisture component, which limits further decreases in cave δ¹⁸O value after the EASM dominants annual precipitations. This mechanism thus implies that the actual monsoon intensity at insolation peak in glacial periods could be weaker than that in interglacial periods and thus reconcilable at least partially with Chinese loess records and cave records from the Arabian Peninsula. Nevertheless, this hypothesis requires a further test and multiple-proxy records from both ISM and EASM regions are especially valuable for the purpose.

6.2 Millennial-scale variability

ASM δ¹⁸O records contain prominent millennial-length events that exhibit negative correlations with the δ¹⁸O records from the SASM region. The observed anti-phased pattern between the ASM and SASM (Fig. 8) is consistent with the AMOC mechanism as mentioned previously.

Meteorological data over the last 50 years and simulations indicate a covariance between the SASM intensity and the SST in the neighboring southern tropical Atlantic (Barros et al. 2000; Liebmann et al. 2004; Vuille and Werner 2005). On millennial and even orbital timescales, the SST anomaly may also likely influence the SASM activity and associated rainfall in southern South America. During the last glacial period, for instance, millennial-scale changes in ocean heat circulation related to the bipolar see-saw mechanism may have changed meridional Atlantic SST gradients (Broecker 1998). Modeling efforts show that the weak ocean circulation may cause a positive SST anomaly in the South Atlantic (e.g., Crowley 1992), consistent with the result from ocean sediment cores off the Brazilian coast (e.g., Arz et al. 1998). A warm SST anomaly may in turn stimulate a persistent intense SASM and strong LLJ, which consequently supplies more isotopically depleted precipitation up to southern Brazil.

A recent oxygen isotope-enabled climate model simulation shows that isotope signatures of precipitation recorded in speleothems from eastern China during the H and/or YD events reflect reduced precipitation over the Indian subcontinent rather than a significant decrease over China (Pausata et al. 2011). The model simulates the reduced precipitation over India which produces isotopically enriched vapor transported eastwards into China. Consequently, increase in the δ¹⁸O of Chinese cave carbonates associated with these events has been argued to reflect changes in the intensity of the ISM rather than the EASM (Pausata et al. 2011). As noted earlier in this paper, the EASM is rather a geographic notion- a monsoonal phenomenon that occurs in East Asia. Modern observations show that the ISM and EASM can be viewed as the up- and low- stream portion of the large ASM circulation system, respectively (Figs. 1 and 2) (e.g., Ding 2004; Ding et al. 2004). In this sense, the H and YD events in speleothem δ¹⁸O records from southeastern China indicate a weakened ASM (Wang et al. 2001) (including the upstream ISM) (e.g., Yuan et al. 2004; Cheng et al. 2006) and thus, are reconcilable with the simulation result (Pausata et al. 2011). However, it is also noteworthy that other proxy records in eastern China indicate possible variances in precipitation associated with H and YD events (e.g., Zhou et al. 2001, 2005; Yancheva et al. 2007; Sun et al. 2010, 2011), in contrast with the simulation result that shows no precipitation changes in the region (Pausata et al. 2011).

6.3 Centennial-decadal-scale variability

A new speleothem record from central-eastern Brazil (Stríkis et al. 2011) revealed a series of strong monsoon events on centennial-decadal timescales during the Holocene that are correlated to Bond events recorded by marine sediments from the North Atlantic Ocean (Bond et al. 1997). These events are broadly anti-phase with weak ASM events (Wang et al. 2005; Fleitmann et al. 2007; Cheng et al. 2009a) (Fig. 9). Not unlike their millennial-scale counterparts, these abrupt events are primarily linked to anomalous cross-equatorial flow induced by negative AMOC phases and resulting SST positive anomalies in the southern tropical Atlantic (Barros et al. 2000; Liebmann et al. 2004), and this may have sequentially modulated not only the monsoon in South America (Stríkis et al. 2011) but also affected the ASM during the Holocene (Wang et al. 2005; Cheng et al. 2009a; Sinha et al. 2011). In fact, Bond events are correlated to solar radiation variations inferred from ¹⁴C production and ¹⁰Be flux changes (Bond et al. 2001). A set of high resolution and precisely dated speleothem records also show evidence for a possible connection between solar radiation and Holocene climate changes in the ISM (e.g., Neff et al. 2001; Fleitmann et al. 2003), EASM (Wang et al. 2005), SASM (Stríkis et al. 2011), and North American Monsoon (Asmerom et al. 2007). The variability of the solar radiation is however rather small in amplitude. Some amplify mechanisms must exist to trigger worldwide climate change and one of such mechanisms proposed by Emile-Geay et al. (2007) involves the SST and in turn ENSO as a mediator between the solar forcing and the resulting climate change.

7 Summary: the PM and GPM

Because it is yet difficult to extract detail paleo-seasonality information from proxy records, we consider the PM to describe alternation of a mean climatic condition between two (or more) states on various timescales (i.e., orbital, millennial, centennial, decadal and annual timescales). Although, mechanisms driving PM oscillations are quite distinct on different timescales, their alternation patterns and global scope, inferred from well-distributed speleothem records from both hemispheres, are indeed comparable to the modern annual monsoonal scenario. On orbital timescale, the ASM and SASM intensities have oscillated between strong (during high summer insolation times) and weak (during low summer insolation times) states, respectively. The anti-phased relationship between the ASM and SASM on orbital timescale (Fig. 6) demonstrates an important aspect of the GPM scenario with summer insolation as the major driving forcing. Orbital variations in ASM and SASM intensities were punctuated by a series of millennial-length events at least over the past two glacial-interglacial periods. Such variations may again be viewed as an oscillation between two states on millennial timescale with weak ASM events correlating with strong SASM events and vice versa. This millennial-scale pattern of climate variability conforms remarkably well to climate model projections, particularly with respect to the temporal synchrony of hemispherically asymmetric response of the monsoon system that accompanied the large-scale temperature fluctuations documented in the Greenland ice cores (Wang et al. 2008). These events show the global extent and may be casually linked to AMOC changes and the associated shifts in the mean latitudinal location of the ITCZ, asymmetry in the Hadley circulation in two hemispheres and SST anomalies (e.g., Lindzen and Hou 1988; Wang et al. 2006, 2007a, b). On centennial to decadal timescales, variability in the ASM and SASM inferred from speleothem records also point to a PM pattern with some features that are similar to their millennial counterparts (e.g., Wang et al. 2005; Cheng et al. 2009a; Stríkis et al. 2011; Cheng et al. 2012b). In summary, speleothem records demonstrate a GPM phenomenon with three key features that are analogous to modern observations of the GM scenario: (1) Paleomonsoonal oscillations on different timescales (Orbital-millennial and possibly even on centennial-decadal); (2) their global extent; and (3) interhemispheric anti-phased relationship between the ASM and SASM domains.