Environmental interpretation of isotopic records

Changes in speleothem δ¹⁸O are mainly dictated by the isotopic composition of meteoric precipitation. Along the moisture trajectory, the precipitation δ¹⁸O signal is further modulated by various processes, including the source effect, altitude and latitude effects, rainfall amount, surface and in-cave temperature changes, and water–rock interactions (McDermott, Reference McDermott2004; Fairchild et al., Reference Fairchild, Smith, Baker, Fuller, Spötl, Mattey and McDermott2006; Lachniet, Reference Lachniet2009). At the millennial scale, it has been suggested that speleothem δ¹⁸O records from the ASM area, especially from southeastern China, reflect local rainfall amounts (Cai et al., Reference Cai, Fung, Edwards, An, Cheng, Lee and Tan2015), precipitation seasonality (Wang et al., Reference Wang, Cheng, Edwards, An, Wu, Shen and Dorale2001; Zhang et al., Reference Zhang, Cheng, Spötl, Cai, Sinha, Tan and Yi2018), or ASM intensity (Cheng et al., Reference Cheng, Edwards, Wang, Kong, Ming, Kelly, Wang, Gallup and Liu2006; Wang et al., Reference Wang, Cheng, Edwards, Kong, Shao, Chen, Wu, Jiang, Wang and An2008). The rainfall amount assumption, however, is believed to be inconsistent with other precipitation proxies from loess/paleosol magnetic properties (Maher, Reference Maher2008). It is most likely that changes in the upstream circulation regime are more important (Pausata et al., Reference Pausata, Battisti, Nisancioglu and Bitz2011; Cheng et al., Reference Cheng, Sinha, Wang, Cruz and Edwards2012; Maher and Thompson, Reference Maher and Thompson2012;Tan, Reference Tan2014). Consequently, a strong southeasterly flow developing over the upstream region likely induces a strong ASM (with ¹⁸O-depleted moisture). At our study site, where the summer rainfall (June through September) contributes 64% (on average) to the annual mean, it is likely that summer moisture and its isotopic composition are most important. Following this reasoning, we interpret our speleothem δ¹⁸O signal as an indicator of average ASM intensity, with minimum values corresponding to a strong ASM. This explanation does not imply a reflection of rainfall amount or rainfall seasonality, but rather of changes in the average intensity of circulation regime, which transfers ¹⁸O-bearing moisture from tropical oceans to the Asian continent and results in the widely observed replication in spatially separated cave records (Cheng et al., Reference Cheng, Sinha, Wang, Cruz and Edwards2012; Maher and Thompson, Reference Maher and Thompson2012; Tan, Reference Tan2014).

Carbon isotopes (δ¹³C) in speleothems are primarily (60%–90%) derived from soil CO₂ by plant respiration and microbial decomposition (Genty et al., Reference Genty, Blamart, Ouahdi, Gilmour, Baker, Jouzel and Van-Exter2003; Meyer et al., Reference Meyer, Feng, Breecker, Banner and Guifoyle2014). At the Wulu site, the modern vegetation type above the cave is mainly composed of deciduous herbs. Our previous work on speleothem records from the same cave revealed a consistent coupling of centennial-scale changes in δ¹⁸O and δ¹³C (Liu et al., Reference Liu, Wang, Cheng, Edwards, Kong and Li2016). This suggests that the soil moisture is important for δ¹³C variability at the centennial scale and is further supported by a recent study on δ¹³C values of dissolved inorganic carbon in cave dripwater in southern China (Li and Li, Reference Li and Li2018). At 72.7 ka (before the hiatus), sharp enrichments in both isotopic values (Fig. 2C) are likely a response to water deficiency at this time.

Muted ASM variability during interstadials in contrast to stadial instabilities

In Figure 4A and C, changes of Wu87 δ¹⁸O and δ¹³C during CIS 19 are characterized by weaker fluctuations (less than 0.3‰ and 0.5‰, respectively) than those in stadials. To further explore the structure of these minor oscillations, millennial-scale variability was removed using the software package ChangePoint (Killick et al., Reference Killick, Fearnhead and Eckley2012). Two detrended sequences (δ¹⁸O_d, δ¹³C_d) reveal similar variations on a centennial time scale (r = 0.51, P < 0.01), especially during isotopic enrichment events (Fig. 4B, green bar). At the decadal scale, however, significant divergence can be observed. During stadials surrounding CIS 19, variations in δ¹⁸O_d and δ¹³C_d are about 0.4‰ and 0.8‰, respectively, almost double the changes in CIS 19. Cross-wavelet analysis between δ¹⁸O_d and δ¹³C_d records reveals that millennial to centennial variability is almost absent between 71.2 and 68.8 ka (Fig. 4, lower panel), supporting our visual observations.

During (peak) interstadials, weak ASM variability appears as a pervasive phenomenon in Asian speleothem records (Fig. 5) (Liu et al., Reference Liu, Wang, Cheng, Edwards, Kong, Wang and Hardt2010, Reference Liu, Wang, Cheng, Edwards, Kong and Li2016; Cui et al., Reference Cui, Wang, Cheng, Zhao and Kong2012; Zhang et al., Reference Zhang, Liu, Liu, Kong and Fang2019). In these cave records, δ¹⁸O_d records are characterized by minor (less than 0.3‰ in amplitude) and high-frequency (centennial to decadal) variability during CIS 5.1 (Fig. 5B) (Liu et al., Reference Liu, Wang, Cheng, Edwards, Kong and Li2016), CIS 16 (Fig. 5C) (Liu et al., Reference Liu, Wang, Cheng, Edwards, Kong, Wang and Hardt2010), and CIS 19 (Fig. 5D) (Zhang et al., Reference Zhang, Liu, Liu, Kong and Fang2019). Similarly, such muted variability is also reflected in coeval δ¹³C_d records. During the stadials, changes in the δ¹⁸O_d and δ¹³C_d records increase in amplitude and decrease in frequency. Cross-wavelet analysis between these δ¹⁸O_d and δ¹³C_d sequences supports weak variability during interstadials (Supplementary Fig. 1). This suggests that limited perturbations were introduced into the ASM system and regional environments during peak interstadials. At the Medieval Warm Period (MWP)/Little Ice Age (LIA) boundary, however, a prominent change is reflected in the frequency domain of the two isotopic data sets, but not in the amplitude (Fig. 5A) (Cui et al., Reference Cui, Wang, Cheng, Zhao and Kong2012). Especially during the MWP, large oscillations are evident in the δ¹³C record as compared with the LIA. An increase of summer precipitation during the LIA caused by the Mei-Yu front was proposed by Cui et al. (Reference Cui, Wang, Cheng, Zhao and Kong2012). This effect of excess humidity in the LIA probably did not favor the dissolution of ¹³C-depleted soil CO₂ in the infiltration water and led to elevated speleothem δ¹³C values (Bar-Matthews et al., Reference Bar-Matthews, Ayalon and Kaufman2000; Plagnes et al., Reference Plagnes, Causse, Genty, Paterne and Blamart2002).

Figure 5. Correlation of intrainterstadial and intrastadial Asian summer monsoon (ASM) changes. Speleothem records from Heilong Cave, central China (A) (Cui et al., Reference Cui, Wang, Cheng, Zhao and Kong2012), Wulu Cave, southern China (B, C) (Liu et al., Reference Liu, Wang, Cheng, Edwards, Kong, Wang and Hardt2010; Reference Liu, Wang, Cheng, Edwards, Kong and Li2016), and Tian'e Cave, central China (D) (Zhang et al., Reference Zhang, Liu, Liu, Kong and Fang2019). Gray bars in these panels indicate interstadial or peak interstadial conditions, with interstadials/stadials labeled at the top. Detrended δ¹⁸O (green line) and δ¹³C (orange line) records are shown in the middle of each panel. CIS, Chinese Interstadial; CS, Chinese Stadial; HS, Heinrich Stadial; MWP, Medieval Warm Period; LIA, Little Ice Age. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Stable interstadial climate conditions have been identified in North Atlantic sediment records (Vautravers et al., Reference Vautravers, Shackleton, Lopez-Martinez and Grimalt2004; Goñi et al., Reference Goñi, Landais, Cacho, Duprat and Rossignol2009) and paleolake records from western Europe (Ampel et al., Reference Ampel, Bigler, Wohlfarth, Risberg, Lotter and Veres2010). Typically, summer-related indicators in these studies exhibit a pattern similar to the ASM variability observed in our study These observations possibly shed some light on climate variability in the context of current global warming. In the last millennium, severe drought events seldom occurred during warming episodes (Zhang, Reference Zhang2005; Hou et al., Reference Hou, Yang and Feng2008), which is partly consistent with our results for interstadial conditions. Under the current warming, if the increased melting of the Greenland and Antarctic Ice Sheets (Bjørk et al., Reference Bjørk, Kjær, Korsgaard, Khan, Kjeldsen, Andresen, Box, Larsen and Funder2012; Shen et al., Reference Shen, Wang, Shum, Jiang, Hsu and Dong2018) induces a reduction of the Atlantic meridional overturning circulation (AMOC) (Mikolajewicz et al., Reference Mikolajewicz, Gröger, Maier-Reimer, Schurgers, Vizcáno and Winguth2007), a stadial-like climate and significant climate extremes are likely.

Detailed correlation of high- and low-latitude climates in MIS 4

In the last glaciation, abrupt reduction in the ASM intensity at the millennial scale was thought to be in line with Greenland temperature variability (Wang et al., Reference Wang, Cheng, Edwards, An, Wu, Shen and Dorale2001), possibly via the westerly winds (Porter and An, Reference Porter and An1995). Increasing evidence, however, shows that significant discrepancies can be observed between high- and low-latitude climate records. First, relatively gradual speleothem-based ASM variability was found during the late MIS 4 (Zhang et al., Reference Zhang, Li, Cheng, Edwards, Shen, Spötl, Li, Han, Li, Huang and Zhao2017), at the onset of CIS 12 (Cai et al., Reference Cai, An, Cheng, Edwards, Kelly, Liu, Wang and Shen2006), during the Bølling interstadial (Wang et al., Reference Wang, Cheng, Edwards, An, Wu, Shen and Dorale2001), and during the Younger Dryas (Liu et al., Reference Liu, Wang, Cheng, Edwards, Kong, Wang, Wu and Chen2008). During these periods, the rise/fall of ASM intensity was frequently accomplished in a longer time compared with Greenland temperature variability. Second, a decoupling of the ASM and Greenland temperature variability is clear during HSs (Zhang et al., Reference Zhang, Griffiths, Huang, Cai, Wang, Zhang, Cheng, Ning, Hu and Xie2016; Liu et al., Reference Liu, Wang, Cheng, Edwards, Kong, Chen and Liu2018), when the ASM is highly unstable. This suggests that the speleothem δ¹⁸O signal from the Asian continent exhibits an appreciably regional feature, which is not exclusive to cave records, as ASM-type changes are widely reflected in various marine and terrestrial records (Vautravers et al., Reference Vautravers, Shackleton, Lopez-Martinez and Grimalt2004; Schmidt et al., Reference Schmidt, Vautravers and Spero2006; Goñi et al., Reference Goñi, Landais, Fletcher, Naughton, Desprat and Duprat2008).

During MIS 4, double peaks in CIS 19 (19a and 19b) and a rapid depletion of 0.8‰ in δ¹⁸O values at 63 ka (likely corresponding to CIS 18) in the Wu87 record (Fig. 6A) agree well, within dating uncertainties, with other cave records from China (Fig. 6B and C) (Chen et al., Reference Chen, Wang, Cheng, Edwards, Wang, Kong and Liu2016; Zhang et al., Reference Zhang, Liu, Liu, Kong and Fang2019). These millennial-scale ASM oscillations, including subcycles, are also mirrored in North Greenland Ice Core Project (NGRIP) ice-core Ca²⁺ and δ¹⁸O (Fig. 6D and E) (Rasmussen et al., Reference Rasmussen, Bigler, Blockley, Blunier, Buchardt, Clausen and Cvijanovic2014) and δ¹⁵N-inferred temperature records (Fig. 6F) (Kindler et al., Reference Kindler, Guillevic, Baumgartner, Schwander, Landais and Leuenberger2014). Among these events, the short-lived cold/aridity episode between CIS 19a and CIS 19b (Fig. 6A–F, dotted lines) corresponds to a weak Antarctic warming (Fig. 6I) (Veres et al., Reference Veres, Bazin, Landais, Kele, Lemieux-Dudon, Parrenin and Martinerie2013). Additionally, CIS 18 in speleothem records and its equivalents in Greenland ice-core records (Fig. 6, dotted line) register as a notable cooling in West Antarctica (Fig. 6H) (WAIS Divide Project Members, 2015), although they are weakly expressed at the EDML ice core (Fig. 6I) (Veres et al., Reference Veres, Bazin, Landais, Kele, Lemieux-Dudon, Parrenin and Martinerie2013). This relationship supports a conception of interhemispheric climate changes as a bipolar seesaw (Broecker, Reference Broecker1998; Blunier and Brook, Reference Blunier and Brook2001).

Figure 6. Climate events surrounding MIS 4. δ¹⁸O records from Wulu (A) (this study), Yongxing (B; Chen et al., Reference Chen, Wang, Cheng, Edwards, Wang, Kong and Liu2016), and Tian'e Caves (C; Zhang et al., Reference Zhang, Liu, Liu, Kong and Fang2019); Greenland ice-core Ca²⁺, δ¹⁸O (D, E; Rasmussen et al., Reference Rasmussen, Bigler, Blockley, Blunier, Buchardt, Clausen and Cvijanovic2014) and δ¹⁵N records (F; Kindler et al., Reference Kindler, Guillevic, Baumgartner, Schwander, Landais and Leuenberger2014); (G) Ice-rafted debris (IRD) is plotted as the number of detrital sediment grains (lithics) from core ODP980 (55°29′N, 14°42′W, depth of 2179 m) (gray line; McManus et al., Reference McManus, Oppo and Cullen1999) and ice-rafted detrital CaCO₃ grains g⁻¹ from core EW9302-2JPC (48°47.70′N, 45°05.09′W, depth of 1251 m), North Atlantic (pink line; Marcott et al., Reference Marcott, Clark, Padman, Klinkhammer, Springer, Liu and Otto-Bliesner2011); tetraunsaturated alkenone record from core ODP977A, western Mediterranean (36°1.907′N, 1°57.319′W, depth of 1984 m) is a tracer for low-salinity water masses (green line; Martrat et al., Reference Martrat, Grimalt, Lopez-Martinez, Cacho, Sierro, Flores, Zahn, Canals, Curtis and Hodell2004); and Antarctic δ¹⁸O records from the WDC (H; WAIS Divide Project Members, 2015) and EDML ice cores (I; Veres et al., Reference Veres, Bazin, Landais, Kele, Lemieux-Dudon, Parrenin and Martinerie2013). The dotted lines show the correlation of interhemispheric climate changes during Interstadial 18 and the stadial between Interstadials 19a and 19b, respectively. The orange bar denotes the initiation of Antarctic warming during MIS 4, and the gray bar indicates Stadial 19 (early MIS 4). Substages of Asian summer monsoon (ASM) variability are also shown at the bottom. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Regardless of this tight interhemispheric climate link, Greenland temperatures remained in a stable cold state during Stadial 19 (Fig. 6E and F, gray bar), a climate condition similar to that during late MIS 4. At Antarctica, a rise of surface temperature initiated at about 68.9 ka (Fig. 6I, orange line), immediately after the end of Interstadial 19a. This Antarctic warming persisted throughout MIS 4 and accelerated appreciably after Interstadial 18 (Fig. 6H and I). Similarly, a gradual ASM intensification is exceptionally clear in early MIS 4 (Fig. 6B and C, gray bar), especially in our newly retrieved Wu87 record (Fig. 6A). In these speleothem records, the ASM intensity during CS 19 appears stronger on average, or is at least not weaker than during CIS 19. This pattern of ASM variability generally follows Antarctic temperature changes, but is decoupled from Greenland temperatures. Consequently, these observations show that the climate of the Southern Hemisphere and the ASM were intimately related during early MIS 4.

A southern and northern imprint on the ASM variability

The idea of an AMOC impact on the interhemispheric climate link appears valid during the interstadials, as shown in Figure 6, where an antiphase relationship is evident. In this regard, the northern displacement of the ITCZ during southern stadials (Broccoli et al., Reference Broccoli, Dahl and Stouffer2006; Stouffer et al., Reference Stouffer, Yin, Gregory, Dixon, Spelman, Hurlin and Weaver2006) that accompanies a strong AMOC can induce a strong ASM and more precipitation in southern China (Wang et al., Reference Wang, Auler, Edwards, Cheng, Ito and Solheid2006). Thus, a more stable ASM during CIS, as derived from our speleothem records (Fig. 5), likely points to an average northerly positioned ITCZ. In contrast, in northern stadials, the frequent intrusion of a northern high-pressure system, including a southerly positioned ITCZ, could have reduced ASM intensity and triggered ASM instabilities (Webster et al., Reference Webster, Magaña, Palmer, Shukla, Tomas, Yanai and Yasunari1998; Wang et al., Reference Wang, Wang, Cheng, Fasullo, Guo, Kiefer and Liu2014, Reference Wang, Wang, Cheng, Fasullo, Guo, Kiefer and Liu2017; An et al., Reference An, Wu, Li, Sun, Liu, Zhou and Cai2015) when the AMOC was weak and frigid cold returned to the northern high latitudes.

During MIS 4, peaks of ice-rafted debris (IRD) from limited-resolution oceanic sediment records suggest catastrophic iceberg and freshwater discharges into the North Atlantic occurred at about 66 ka (Fig. 6G, gray and orange lines) (McManus et al., Reference McManus, Oppo and Cullen1999; Marcott et al., Reference Marcott, Clark, Padman, Klinkhammer, Springer, Liu and Otto-Bliesner2011). Additionally, a high-resolution alkenone record from core ODP977A indicates that a large volume of low-salinity water masses intruded into the western Mediterranean at about 64 ka (Fig. 6G, green line) (Martrat et al., Reference Martrat, Grimalt, Lopez-Martinez, Cacho, Sierro, Flores, Zahn, Canals, Curtis and Hodell2004). Taking the dating uncertainties of these oceanic sediment records into consideration,, these lines of evidence show that a significant AMOC reduction most likely occurred in late MIS 4 (HS 6), as summarized in Hemming (Reference Hemming2004). Accordingly, Greenland temperature (Fig. 6E and F) (Kindler et al., Reference Kindler, Guillevic, Baumgartner, Schwander, Landais and Leuenberger2014; Rasmussen et al., Reference Rasmussen, Bigler, Blockley, Blunier, Buchardt, Clausen and Cvijanovic2014) and ASM intensity declined significantly after Interstadial 18 (Fig. 6C, arrowed line) (Chen et al., Reference Chen, Wang, Cheng, Edwards, Wang, Kong and Liu2016; Zhang et al., Reference Zhang, Liu, Liu, Kong and Fang2019). In response to the interhemispheric thermal gradient caused by freshwater perturbations, a southerly positioned ITCZ and an accelerated Antarctic warming would be expected (Fig. 6H and I) (Veres et al., Reference Veres, Bazin, Landais, Kele, Lemieux-Dudon, Parrenin and Martinerie2013; WAIS Divide Project Members, 2015).

Nevertheless, this idea of latitudinal shifts in the ITCZ (in a scenario of freshwater impact) lacks validity regarding synchronous increase in the ASM intensity and Antarctic temperature in early MIS 4 (CS 19) (Fig. 6), because IRD peaks for that time period were insignificant (Fig. 6G). One possibility for the coupling of ASM and Antarctic temperature is an expansion of the ITCZ (Collins et al., Reference Collins, Schefuß, Heslop, Mulitza, Prange, Zabel and Tjallingii2011; Yan et al., Reference Yan, Wei, Soon, An, Zhou, Liu, Wang and Carter2015; Denniston et al., Reference Denniston, Ummenhofer, Wanamaker, Lachniet, Villarini, Asmerom and Polyak2016). In early MIS 4, the Mg/Ca-based summertime sea-surface temperatures (SSTs), reconstructed from Globigerinoides ruber records from the western tropical Pacific (6.3°N, 125.83°E), were relatively stable (generally above 28°C) (Stott et al., Reference Stott, Poulsen, Lund and Thunell2002). This implies that thermal conditions in the Indo-Pacific Warm Pool (IPWP) were favorable for ITCZ expansion at this time, as a rise in SST to 27.5°C is required for initiation of deep convection (Graham and Barnett, Reference Graham and Barnett1987). ITCZ position is sensitive to slight shifts in the atmospheric energy balance. Currently, the hemispherically asymmetric contrast of mid-tropospheric temperatures is less than 3 Kelvin in the zonal mean between 20°N and 20°S (Schneider et al., Reference Schneider, Bischoff and Haug2014). It is believed that a gradient of 1 Kelvin between high- and low-latitude SSTs could trigger the displacement of the ITCZ (Hastenrath and Heller, Reference Hastenrath and Heller1977). Moreover, other sediment records have shown that the tropical Pacific SST in early MIS 4 reached its average modern value (22°C to 29°C), with centennial-scale changes falling within the seasonal range of 2°C (Fig. 7C) (Rosenthal et al., Reference Rosenthal, Oppo and Linsley2003; Lea et al., Reference Lea, Pak and Spero2000, Reference Lea, Pak, Belanger, Spero, Hall and Shackleton2006). A significant warming of about 2°C was also registered at the midlatitudes of the southern Pacific (Fig. 7B) (Pahnke and Sachs, Reference Pahnke and Sachs2006; Tachikawa et al., Reference Tachikawa, Vidal, Sonzogni and Bard2009) and the South China Sea (Fig. 7D) (Oppo and Sun, Reference Oppo and Sun2005). This SST variability is in line with the gradual increase in ASM intensity in our speleothem record (Fig. 7A). Consequently, the concurrent increase in ASM intensity and Antarctic temperature (Fig. 6, orange line) should reflect an expansion of ITCZ when freshwater discharges ceased.

Figure 7. (Color online) Dynamic link of Asian summer monsoon (ASM) changes and sea-surface temperature (SST) records in MIS 4. (A) Wu87 δ¹⁸O record (this study); (B) SST changes at subtropical southern Pacific (MD97-2120 [45°32.06′S, 174°55.85′E] and MD97-2121 [40°22.8′S, 177° 59.4′E]; Pahnke and Sachs, Reference Pahnke and Sachs2006; MD2125 [22°34′S, 161°44′E]; Tachikawa et al., Reference Tachikawa, Vidal, Sonzogni and Bard2009); (C) IPWP SST (ODP806B [0°19.04′N, 159°21.25′E]; Lea et al., Reference Lea, Pak and Spero2000; TR163 [0°30.9′N, 92°23.9′W]; Lea et al., Reference Lea, Pak, Belanger, Spero, Hall and Shackleton2006; MD97-2141 [8.8°N, 121.3°E]; Rosenthal et al., Reference Rosenthal, Oppo and Linsley2003); (D) SST changes at the South China Sea (ODP1145 [19°35′N, 117°38′E]; Oppo and Sun, Reference Oppo and Sun2005). Gray bar indicates early MIS 4, when SST in each site reached its modern level.

Hence, the thermal contrast between the two hemispheres is very important. We speculate that when the AMOC is exceptionally strong (or weak), a large interhemispheric thermal gradient would result in a latitudinal migration of the ITCZ, as demonstrated by modeling studies (Broccoli et al., Reference Broccoli, Dahl and Stouffer2006; Stouffer et al., Reference Stouffer, Yin, Gregory, Dixon, Spelman, Hurlin and Weaver2006). This could explain the antiphase relationship between the ASM and Antarctic temperature during Interstadials 19 and 18 and late MIS 4 (Fig. 6). In contrast, a small thermal gradient between the two hemispheres is favorable for an expansion/contraction of the ITCZ, which is partly reminiscent of the concurrent rise in the ASM strength and Antarctic temperature during early MIS 4. This simplified concept of ocean–atmosphere response is probably similar to the mechanism associated with triple AMOC states, that is, a glacial mode, a Heinrich mode, and a modern mode (Clark et al., Reference Clark, Pisias, Stocker and Weaver2002), which is further modulated by ice-shelf and sea-ice interactions (Boers et al., Reference Boers, Ghil and Rousseau2018). It is currently out of our scope to evaluate the northernmost boundary of the ITCZ. Our comparison suggests that the impact of the ITCZ is probably limited to mid- to low latitudes, as an increase in NGRIP ice-core Ca²⁺ concentration (indicating a strong Asian winter monsoon) can be observed before the onset of Interstadial 18 (Fig. 6D) (Rasmussen et al., Reference Rasmussen, Bigler, Blockley, Blunier, Buchardt, Clausen and Cvijanovic2014). In our speleothem records, the ASM intensification persisted throughout early MIS 4 (Fig. 6A–C, gray bar). A decoupling of ASM intensity and Greenland temperature was also observed during mid-HS (Liu et al., Reference Liu, Wang, Cheng, Edwards, Kong, Chen and Liu2018). In this study, speleothem records from central to southern China revealed a gradual ASM intensification in mid-HS, in stark contrast to a stable cold climate state in Greenland. It is probably a reflection of the impact of low-latitude hydrothermal processes.

These observations are of vital importance for understanding the dynamics of millennial-scale climate changes, because the frequency and amplitude of interstadials/stadials are controlled by North Atlantic processes, and their magnitude increases with increasing latitude (Masson-Delmotte et al., Reference Masson-Delmotte, Buiron, Ekaykin, Frezzotti, Gallée, Jouzel and Krinner2011; Shakun et al., Reference Shakun, Clark, Marcott, Mix, Liu, Otto-Bliesner, Schmittner and Bard2012; Cuffey et al., Reference Cuffey, Clow, Steig, Buizert, Fudge, Koutnik, Waddington, Alley and Severinghaus2016). However, the evolution and duration of millennial-scale climate events, especially within northern low latitudes, is probably determined by southern oceanic processes (Buizert and Schmittner, Reference Buizert and Schmittner2015). In Asian speleothem records, gradual δ¹⁸O variability can be expected at CIS/CS transitions in contrast to abrupt changes in Greenland temperature (Wang et al., Reference Wang, Cheng, Edwards, An, Wu, Shen and Dorale2001; Cai et al., Reference Cai, An, Cheng, Edwards, Kelly, Liu, Wang and Shen2006; Liu et al., Reference Liu, Wang, Cheng, Edwards, Kong, Wang, Wu and Chen2008, Reference Liu, Wang, Cheng, Edwards, Kong, Chen and Liu2018; Zhang et al., Reference Zhang, Li, Cheng, Edwards, Shen, Spötl, Li, Han, Li, Huang and Zhao2017). Sensitive to the interhemispheric temperature gradient, the ITCZ can act a bridge to link bipolar climates via expansion/contraction and “pull–push” mechanisms (Rohling et al., Reference Rohling, Liu, Roberts, Stanford, Rasmussen, Langen and Siddall2009). Therefore, one needs to precisely pinpoint the “zero-anomaly line” of the interhemispheric thermal gradient in the past (Stouffer et al., Reference Stouffer, Yin, Gregory, Dixon, Spelman, Hurlin and Weaver2006; Zarriess et al., Reference Zarriess, Johnstone, Prange, Steph, Groeneveld, Mulitza and Machensen2011). Within the IPWP, a sharp precipitation boundary between the maritime continents and Australia was located at about 8–10°S during the last termination (Lo et al., Reference Lo, Shen, Wei, Burr, Mii, Chen, Lee and Tsai2014); however, a geographic mismatch between meridional thermal and rainfall patterns (Linsley et al., Reference Linsley, Rosenthal and Oppo2010) suggests that the zero-anomaly line variability deserves further investigation, especially at contrasting climate boundaries.