LETTER

doi:10.1038/nature13076

Mid-latitude interhemispheric hydrologic seesaw over the past 550,000 years Kyoung-nam Jo1, Kyung Sik Woo2, Sangheon Yi1, Dong Yoon Yang1, Hyoun Soo Lim3, Yongjin Wang4, Hai Cheng5,6 & R. Lawrence Edwards6

from 24 separate speleothems, but still allow the identification of periods of peak speleothem growth and, thus, precipitation. The clear hemispheric antiphasing indicates that the sphere of influence of the interhemispheric hydrologic seesaw over the past 550,000 years extended at least to the mid-latitudes, such as northeast Asia, and that orbital-timescale ITCZ shifts can have serious effects on temperate climate systems. Furthermore, our result implies that insolationdriven ITCZ dynamics may provoke water vapour and vegetation feedbacks in northern mid-latitude regions and could have regulated global climate conditions throughout the late Quaternary ice age cycles. The Korean peninsula in northeast Asia is situated on the boundary between the world’s largest continent and largest ocean5 and has a typical mid-latitude climate that is strongly affected by the East Asian monsoon (EAM) system (Fig. 1). The peninsula has experienced dramatic palaeogeographic changes and eustatic sea-level changes during the late Quaternary glacial–interglacial cycles. For example, during the last glacial maximum (LGM), vast areas of the Yellow Sea (about 500,000 km2) to the west of the peninsula were completely exposed subaerially, preventing

An interhemispheric hydrologic seesaw—in which latitudinal migrations of the Intertropical Convergence Zone (ITCZ) produce simultaneous wetting (increased precipitation) in one hemisphere and drying in the other—has been discovered in some tropical and subtropical regions1–3. For instance, Chinese and Brazilian subtropical speleothem (cave formations such as stalactites and stalagmites) records show opposite trends in time series of oxygen isotopes (a proxy for precipitation variability) at millennial to orbital timescales2,3, suggesting that hydrologic cycles were antiphased in the northerly versus southerly subtropics. This tropical to subtropical hydrologic phenomenon is likely to be an initial and important climatic response to orbital forcing3. The impacts of such an interhemispheric hydrologic seesaw on higher-latitude regions and the global climate system, however, are unknown. Here we show that the antiphasing seen in the tropical records is also present in both hemispheres of the mid-latitude western Pacific Ocean. Our results are based on a new 550,000-year record of the growth frequency of speleothems from the Korean peninsula, which we compare to Southern Hemisphere equivalents4. The Korean data are discontinuous and derived a

b

42° N

Russia

128° E 200 1,000

China

N

Summer rainfall belt (Early to middle July)

n

soo

on ter m

a

Win

East Sea

Korea

50 38° N

Asian continent

Siberian High

100

3,000 2,000 100 200 2,000 1,000

Study area

Middle to late June

Korea 3,000 2,000

1,000

WKWC

124° E

1,000

1,000

1,000

Seoul

EKWC

Late May

China

2,000

Yellow Sea 50

KC

1,000

Summer monsoon mo

an

100 200

Ta iw

Japan 50 34° N

10

0

Jeju Island

100

LGM

100

Japan

TC

LGM 0

Pacific Ocean

150 km

Figure 1 | Location of the study area and current atmospheric and oceanographic conditions. Also shown are coastlines during the last glacial maximum (LGM; dashed lines). a, Black-shaded areas represent the location of the Cambro–Ordovician Joseon Supergroup, where we collected all the speleothem samples. The black lines indicate the current coastlines, and grey lines with numbers (depths in metres) show current bathymetric features. A vast area of the Yellow Sea was subaerially exposed during the LGM6

North Pacific High

(shaded areas and dashed lines). b, The arrows describe atmospheric and oceanographic conditions. KC, Kuroshio current; TC, Tsushima current; EKWC, East Korean warm current; WKWC, West Korean warm current. Thick dashed lines indicate the current summer rainfall belt, called Jangma in Korea. The background image is from NASA’s Visible Earth Web Site (http://visibleearth.nasa.gov/).

1

Korea Institute of Geoscience and Mineral Resources, Daejeon 305-350, South Korea. 2Department of Geology, Kangwon National University, Gangwondo 200-701, South Korea. 3Department of Geological Sciences, Pusan National University, Busan 609-735, South Korea. 4College of Geography Science, Nanjing Normal University, Nanjing 210097, China. 5Institute of Global Environmental Change, Xi’an Jiaotong University, Xi’an 710049, China. 6Department of Geology and Geophysics, University of Minnesota, Minneapolis, Minnesota 55455, USA. 0 0 M O N T H 2 0 1 4 | VO L 0 0 0 | N AT U R E | 1

©2014 Macmillan Publishers Limited. All rights reserved

RESEARCH LETTER warm current inflow6 (Fig. 1). These palaeoenvironmental settings produced major changes in terrestrial ecosystems during the late Quaternary glacial–interglacial cycles7,8, and the changes on the Korean peninsula can be considered to reflect global changes in vegetation composition and density. Proxy records from the Korean peninsula may provide invaluable information on continent–ocean climate linkages and the impacts of global changes on ice sheets and sea level. However, only a few orbital-scale proxy records representing more than one glacial–interglacial cycle have been reported, possibly owing to the lack of directly datable materials. In this study, we have applied the 238U–234U–230Th dating technique to describe the first record of speleothem growth frequency from the northeast Asian sector over the past 550,000 years (550 kyr). The modern annual mean temperature (AMT) of the study area is approximately 10–13 uC, with strong seasonal variations of .30 uC between January and August9. The winter temperature is generally below freezing, with monthly mean temperatures of 26 uC to 27 uC in January. Annual mean precipitation (AMP) is approximately 1,100–1,400 mm, with high seasonality. Three quarters of the total annual precipitation falls during the humid summer season from June to September. All the caves in this study were formed in Cambro–Ordovician carbonate rocks of the Joseon Supergroup, which is composed mainly of limestone, with a minor contribution of dolomite and argillaceous limestone (Extended Data Table 1). The caves we studied were mostly formed along numerous joint and fault planes created by Jurassic tectonic movements10 (Fig. 1). The atmospheric environment in Korean limestone caves is characterized by stable temperature and humidity (see Methods for the detailed discussion) (Extended Data Fig. 1). Typical limestone caves in the Korean peninsula contain large numbers of actively growing speleothems11, indicating that current climatic and environmental conditions are suitable for the formation of various types of speleothem. We obtained 148 230Th ages from 24 speleothems (22 stalagmites and two flowstones from 15 typical limestone caves) using multi-collector inductively coupled plasma mass spectroscopic techniques (Extended Data Fig. 2 and Supplementary Table 1). Filtering to avoid Holocene T2

3 4 5a 5c 5e 5b 5d

1 2

3.0 3.5

T3 7a 7c 7e 7b 7d

6

T4 8

9

T5 10

11

T6 12

13

14

Sea-level/ deep-water temperature

a

4.0 5.0

Antarctic temperature

16.0 15.5 15.0 14.5 14.0 14.5 13.0

5 11 1

c

4

6 7

5

4

2

7

8

9

12

8 9

2 3

1d

–9.5 –8.5 –7.5 –6.5

Relative probability density

3

10

6

12 11

10

40

14 13

16 15

Northern Hemisphere summer insolation 19

17 18

Temperature and precipitation in Korean peninsula 80

120

160

200

240

280

560 540 520 500 480 460 440 420

13

e 0

Temperature in western North America

320

360

400

440

480

520

560

–12

δ18O (‰, VPDB) Chinese caves

Korean caves Devils Hole δ18C (‰, VPDB) δ18O (‰, VSMOW)

b

–350 –370 –390 –410 –430 –450

Insolation (W m–2) July at 65° N

4.5

δD (‰, VSMOW) EPICA Dome C

LR04 benthic stack δ18O (‰, VPDB)

Termination and MIS T1

bias removed 22 dates (Extended Data Fig. 3). We supplemented the plot of the number of speleothem depositions with textural examination results and confirmed hiatuses (Extended Data Figs 4 and 5). Our speleothem growth frequency data were augmented by 640 new and previous stable isotope data from three stalagmites (ED1, GE1 and DY1 from Eden cave, Gwaneum cave and Daeya cave, respectively). The growth frequency of Korean speleothems has varied widely over time (Fig. 2 and Extended Data Fig. 6), indicating the past fluctuations of speleothem growth in the limestone caves of the Korean peninsula. Although various high-resolution speleothem d18O records from tropical to subtropical China have been reported, this record is the first to show speleothem growth frequency from the northeast Asian temperate region. Each high peak provides the precise intervals for the active growth periods of the speleothems, and these data can be directly correlated with orbital-scale changes in global palaeoclimate records closely related to advances and retreats of massive continental ice sheets. High- and low-growth-frequency phases of Korean speleothems are well matched to interglacial/interstadial and glacial/stadial periods, respectively, based on marine isotope stages (MIS)12 (Fig. 2). Although the older part of the record was more or less unclear compared with the penultimate glacial period, it is notable that Korean speleothems grew markedly in interglacial/interstadial periods (MIS 1, MIS 3, MIS 5a, MIS 5c, MIS 5e, MIS 7a, MIS 7c, MIS 7e, MIS 9, MIS 11 and MIS 13), whereas their growth significantly declined during glacial/ stadial conditions (MIS 2, MIS 4, MIS 6, MIS 7d, MIS 8 and MIS 12). Moreover, almost all the peaks in our record are possibly linked to the second half of each peak of Northern Hemisphere summer insolation13 and the EAM intensity record14,15, as well as to interglacial/interstadial periods represented by isotopic records from an Antarctic ice core16 and Devils Hole17 (19 of 27 peaks in Northern Hemisphere summer insolation, 13 of 16 peaks in marine and terrestrial records), with a few exceptions mainly before 400 kyr ago (Fig. 2). These agreements between the growth frequency record of the Korean speleothems and the major palaeoclimatic sequences strongly suggest that Korean speleothem growth was primarily controlled by climatic and environmental changes. This

–10 –8 –6

600

Age (kyr)

Figure 2 | Comparison of the growth frequency record from Korean speleothems with major palaeoclimatic records over the last 600 kyr. a, d18O record of LR04 benthic stack12. The marine isotope stage (MIS) is indicated above the curve. Vertical dashed lines show the timing of glacial terminations. b, dD record of EPICA Dome C, Antarctica16. c, d18O record of vein calcite from Devils Hole, western USA17. The numbers above the curve represent high peak intervals and are matched to the Korean speleothem record. d, Summer insolation at 65u N (ref. 13) superimposed on the composite d18O record from the Hulu, Sanbao and Linzhu caves in China15. The numbers above the

curve represent high peaks matched to the Korean speleothem record. e, Growth frequency record of Korean speleothems with the simple plot of Th dates (open circles with horizontal lines). Error bars of each date depict 2s error. See the Methods for a detailed description of the sample collection. Periods of high growth frequency are shown in brown (interglacial/interstadial) and purple (glacial) vertical bars. The d13C records (red curves) closely overlap with Chinese d18O records and are from three stalagmites: GE-1, DY-1 and ED-1. The scale of the d13C values is that of ED-1. The GE-1 and DY-1 records were tuned to the Chinese d18O records. 230

2 | N AT U R E | VO L 0 0 0 | 0 0 M O N T H 2 0 1 4

©2014 Macmillan Publishers Limited. All rights reserved

LETTER RESEARCH in the mean position and vitality of the ITCZ (Fig. 1). Impacts of ITCZ migration on regions at latitudes higher than the tropics and subtropics can be ascertained by the linkage between the Korean speleothem record and Mediterranean sapropel events24 (Fig. 3). Our record shows excellent matches with the sapropel events, indicating that both regions have been affected by synchronous movement of the ITCZ, which resulted in atmospheric circulation changes in both regions. This observation in the mid-latitudes is also shown in speleothem evidence of millennial-scale fluctuations from the southwestern United States22 and from southwestern New Zealand23. Thus, our data support the idea that ITCZ shift causes latitudinal displacement of all climatic zones. To delineate the mid-latitudinal interhemispheric impacts of ITCZ migration, we combined the growth frequency record of Korean speleothems with the Southern Hemisphere geographical equivalent from Naracoorte, southeastern Australia4 (Fig. 3). The Korean speleothem record shows a distinct inverse correlation with the Southern Hemisphere mid-latitude speleothem record. For example, the Holocene and last interglacial periods are indicated as remarkable growth periods of Korean speleothems but as periods of little growth of Australian speleothems. Similarly, the southeastern Australian speleothems actively developed during cool interstadials to stadials, whereas Korean speleothems showed only moderate growth near interstadial peaks. Notably, no or very low growth frequency was shown in both hemispheres during the massive glacial periods including MIS 4 and MIS 7d (Fig. 3). Once again, this interhemispheric anti-correlation supports the theory that variations in the mean position of the ITCZ, directly affected by solar insolation, redistributed the global configuration of latitudinal climatic zones and air masses. During peak interglacial periods, the Korean peninsula had been influenced by a relatively warmer and wetter climate than during interstadials, owing to the enhanced effect of the North Pacific high-pressure system along with the northern position of the ITCZ (Figs 1 and 3)5. In contrast, during cool interstadials to stadials, southeastern Australia has experienced a wetter and cooler climate by southward displacement of the ITCZ and/or the strong effect of Walker circulation4. Alternatively, a steeper temperature gradient between the southern polar front and ITCZ could have induced stronger Southern Hemisphere westerly winds and prevailing southern frontal systems along with early Equatorward advance of the polar front and

result is in accordance with the growth of a speleothem in northeastern China (Shihua cave, 39u 479 N, 115u 569 E) which shows that much faster growth occurred during warmer periods18. Indeed, the most basic and simplest palaeoclimatic clue provided by speleothems is whether the speleothems grew under the prevailing climatic conditions existing outside the caves19. In theory, speleothem growth requires both a sufficient supply of cave water and suitable temperature, that is, at least positive effective precipitation2,4 and ice- and permafrost-free conditions20,21. Additionally, for speleothem growth in the temperate climatic regime, productive vegetation and soil depths thick enough to produce and store soil CO2 and moisture should be present4,19. Thus, we attributed the high speleothem growth frequency during the interglacial (and/or pluvial) periods to reinforcement of surficial vegetation productivity and soil depth under clearly favourable climatic conditions. This interpretation is confirmed by noticeably lower d13C values (about 210.7% to 29.0% in the Eden cave scale) from three stalagmites during interglacial and interstadial periods than during glacial conditions (about 28.5% to 26.5%) (Fig. 2), suggesting that terrestrial productivity in the study area was highly intensified during interstadial and interglacial periods. For the case of glacial periods with low speleothem growth, our interpretation is also supported by pollen and vertebrate evidence7,8 from the Korean peninsula as well as numerical model simulations. Considering both the significantly reduced growth frequency of the Korean speleothems and the estimated amount of precipitation during rainy seasons, the peninsula probably experienced a further southward shift or even failure of the Jangma rain belt (also called the Meiyu in Japan and Bai-u in China), which is the strong frontal system that develops each year in early summer between the maritime Okhotsk and North Pacific air masses (see Methods for a detailed discussion). The timing of palaeoclimate changes in our record is correlated with boreal monsoonal changes (Fig. 2). Movement of the ITCZ—which is the main controlling factor of interhemispheric monsoon intensity— has climatic impacts in middle and higher latitudes, as successive displacements in zonal flow and related air masses compensate for the ITCZ motion, as in modern seasonal movements22,23. The Korean peninsula, as a mid-latitude temperate region and one of the northernmost parts of the EAM sector, should be highly sensitive to even small shifts T1

T2

T4

T3

T5

T6

8

489

407 357 385 123

Southward migration of ITCZ

296 281

210 157 202

b 174

0.000

79 109

76

168

c 0

70

0.025

0.010

50

44

40

58

85

128

99 93 ~96

80

205 192 217 241

310 289

328

359 411

120 160 200

428

440

240 280 320 360 400 440 480 Age (kyr)

Figure 3 | Comparison of the Korean speleothem record with the Southern Hemisphere counterpart record from southeastern Australia. a, Growth frequency record of southeastern Australian speleothems4. The black numbers around each peak show the age. b, Summer insolation at 65u N (ref. 13). c, Growth frequency record of Korean speleothems. The red numbers around each peak show the age of growth periods. The light grey vertical bars represent the periods of high growth frequency indicated in the Korean speleothem record during interglacial/interstadial periods. We note the distinctive

519

Northward migration of ITCZ

520 560

560 540 520 500 480 460 440

Insolation (W m–2) July at 65° N

0.019

112

0.015

Relative probability density Korean caves

24

Relative probability density Southeastern Australian caves

a

0.005

0.029

Termination and full glaciation Permafrost thawing Sapropel sequences

420 400

600

anti-correlations between the northern and southern mid-latitude speleothem records. The vertical dashed lines show the timings of glacial terminations. The bars in the uppermost part of the figure indicate periods of massive glaciation, including MIS 4 and MIS 7d. The middle bar shows periods of permafrost thawing from the eastern Siberian speleothem record21. Mediterranean sapropel events reported by ref. 24 and other literature are shown in the lower bar.

0 0 M O N T H 2 0 1 4 | VO L 0 0 0 | N AT U R E | 3

©2014 Macmillan Publishers Limited. All rights reserved

RESEARCH LETTER expansion of the sea ice extent23. This inference describes the midlatitude interhemispheric hydrologic seesaw from peak interglacial to mid-sized glacial periods. Because the southeastern Australian record reflects effective precipitation (higher precipitation with cooler temperature in Naracoorte)4, this is also likely to indicate synchronous icedriven forcing in both the northern and southern mid-latitudes25 in the western Pacific sector. During full glacial conditions, very limited growth frequencies were found for speleothems in both hemispheres (Fig. 3). This suggests that both regions experienced maximal cold and dry conditions as the northern continental ice sheet and southern sea ice reached their maximum extents. In other words, this indicates the equatorward push of the ITCZ by strong ice-driven forcing and weakening of the global rainfall system in both hemispheres26 (Extended Data Fig. 7). The interconnecting atmospheric processes described above are likely to be concurrent with the climatic feedbacks in the northern mid-latitude regions (Extended Data Fig. 7). Insolation-driven northward displacement of the ITCZ can bring about sensible heating and positive feedback by the potent effect of water vapour as a major greenhouse gas27 in the northern mid- to high-latitude regions. It is possible that this type of positive feedback was combined with the precipitation–vegetation– albedo feedback28,29, leading to acceleration in the increase in surface air temperature and rapid ice melting in the Northern Hemisphere, especially during glacial terminations (see Methods for the detailed discussion). Our record strongly supports this explanation because the higher speleothem growth frequency with lower d13C values in the Korean peninsula should have been accompanied by a significant increase in vegetation density and/or soil depth during the terminations (Fig. 2). According to pollen records of the Korean peninsula8, the enhanced monsoon precipitation and increasing temperature led to shifts in vegetation types from glacial grassland to interglacial forest. This explanation for the vegetation-related feedback mechanism in the northern midlatitude regions is also in line with a recent numerical model simulation30. Given that the northernmost region of the Korean peninsula is thought to have been one of the southern limits of discontinuous permafrost in the Asian continent during glacial maxima, our hypothesis could be closely related to thawing of the Northern Hemispheric permafrost region, a massive reservoir of greenhouse gases. Recently, the speleothem data from Eastern Siberia reveals that permafrost thawing is highly sensitive to slight increases in global temperature and can be potentially coupled with substantial release of greenhouse gases trapped in it into the atmosphere21. But this greenhouse effect caused by largescale permafrost thawing is likely to be most effective during glacial terminations because the magnitude of the thawing during other periods, such as the ends of stadials, must be significantly reduced owing to the inadequate extension of permafrost areas at those time periods.

and d13C are 0.15% and 0.2%, respectively. Detailed age models for the timeseries isotopic records of the stalagmites were derived by linear interpolation between all dates. The final age models for the three stalagmites were tuned to fit Chinese cave d18O records within age errors. Online Content Any additional Methods, Extended Data display items and Source Data are available in the online version of the paper; references unique to these sections appear only in the online paper. Received 27 June 2013; accepted 22 January 2014. Published online 30 March 2014. 1. 2. 3. 4. 5. 6.

7.

8. 9. 10. 11.

12. 13. 14. 15. 16. 17. 18.

19.

20.

21.

METHODS SUMMARY We obtained 148 available uranium-series dates using a multi-collector inductively coupled plasma mass spectrometer (MC-ICP-MS; Thermo-Finnigan Neptune) at the Minnesota Isotope Laboratory, University of Minnesota, USA. All age error bars in this paper depict 2s errors. Seventeen of the dates were previously published31,32. The relative probability density curve for the total dates of the Korean speleothems was plotted using Isoplot/Ex 3.0 software (http://bgc.org/isoplot_etc/isoplot.html). This plot has an inherent defect that may be artificially caused by the sampling strategy (closer sub-sampling intervals in a particular part). To solve this problem, we supplemented the plot of the number of speleothem depositions with textural examination results and confirmed hiatuses. Another weakness of the plot was that Holocene speleothem samples were more common than older samples. To address this, we primarily focused on the timings of peaks in the probability density curve, which were regarded as the periods of high-frequency speleothem growth in the Korean peninsula. Oxygen and carbon isotope ratios were analysed at the Isotope Laboratory at Nanjing Normal University, China. In total, 640 measurements of d18O and d13C were made on the three speleothems using a Finnigan MAT-253 ratio mass spectrometer with an on-line, automated, carbonate preparation system (Kiel III). Isotopic data of 346 measurements from GE1 and DY1 were published. Measured C- and O-isotope ratios in per mil (%) notation are reported relative to the belemnite from the Pee Dee Belemnite formation standard (PDB). Analytical errors for d18O

22.

23.

24.

25. 26.

27.

28.

29.

30.

Severinghaus, J. P. Monsoons and meltdowns. Science 326, 240–241 (2009). Wang, X. et al. Wet periods in northeastern Brazil over the past 210 kyr linked to distant climate anomalies. Nature 432, 740–743 (2004). Cruz, F. W. Jr et al. Insolation-driven changes in atmospheric circulation over the past 116,000 years in subtropical Brazil. Nature 434, 63–66 (2005). Ayliffe, L. K. et al. 500 ka precipitation record from southeastern Australia: evidence for interglacial relative aridity. Geology 26, 147–150 (1998). Nakagawa, T. et al. Regulation of the monsoon climate by two different orbital rhythms and forcing mechanisms. Geology 36, 491–494 (2008). Kim, J.-M. & Kucera, M. Benthic foraminifer record of environmental changes in the Yellow Sea (Hwanghae) during the last 15,000 years. Quat. Sci. Rev. 19, 1067–1085 (2000). Lim, H. S. et al. Vertebrate burrows in late Pleistocene paleosols at Korean Palaeolithic sites and their significance as a stratigraphic marker. Quat. Res. 68, 213–219 (2007). Yi, S. & Kim, S.-J. Vegetation changes in western central region of Korean Peninsula during the last glacial (ca. 21.1–26.1 cal kyr BP). Geosci. J. 14, 1–10 (2010). Korea Meteorological Administration. http://web.kma.go.kr/eng/biz/ climate_01.jsp (2011). Chough, S. K. et al. Tectonic and sedimentary evolution of the Korean peninsula: a review and new view. Earth Sci. Rev. 52, 175–235 (2000). Jo, K., Woo, K. S., Hong, G. H., Kim, S. H. & Suk, B. C. Rainfall and hydrological controls on speleothem geochemistry during climatic events (droughts and typhoons): an example from Seodong Cave, Republic of Korea. Earth Planet. Sci. Lett. 295, 441–450 (2010). Lisiecki, L. E. & Raymo, M. E. A. Pliocene-Pleistocene stack of 57 globally distributed benthic d18O records. Paleoceanography 20, PA1033 (2005). Berger, A. & Loutre, M. F. Insolation values for the climate of the last 10 million years. Quat. Sci. Rev. 10, 297–317 (1991). Wang, Y. et al. Millennial- and orbital-scale changes in the East Asian monsoon over the past 224,000 years. Nature 451, 1090–1093 (2008). Cheng, H. et al. The climatic cyclicity in semiarid-arid central Asia over the past 500,000 years. Geophys. Res. Lett. 39, L01705 (2012). Jouzel, J. et al. Orbital and millennial Antarctic climate variability over the past 800,000 years. Science 317, 793–796 (2007). Winograd, I. J. et al. Continuous 500,000-year climate record from vein calcite in Devils Hole, Nevada. Science 258, 255–260 (1992). Tan, M. et al. Cyclic rapid warming on centennial-scale revealed by a 2650-year stalagmite record of warm season temperature. Geophys. Res. Lett. 30, 1617 (2003). Baker, A., Smart, P. L. & Ford, D. C. Northwest European palaeoclimate as indicated by growth frequency variations of secondary calcite deposits. Palaeogeogr., Palaeoecol. Palaeoclimatol. 100, 291–301 (1993). Spo¨tl, C., Mangini, A., Frank, N., Eichsta¨dter, R. & Burns, S. J. Start of the last interglacial period at 135 ka: evidence from a high Alpine speleothem. Geology 30, 815–818 (2002). Vaks, A. et al. Speleothems reveal 500,000-year history of Siberian permafrost. Science 340, 183–186 (2013). Asmerom, Y., Polyak, V. J. & Burns, S. J. Variable winter moisture in the southwestern United States linked to rapid glacial climate shifts. Nature Geosci. 3, 114–117 (2010). Whittaker, T. E., Hendy, C. H. & Hellstrom, J. C. Abrupt millennial-scale changes in intensity of Southern Hemisphere westerly winds during marine isotope stages 2–4. Geology 39, 455–458 (2011). Ziegler, M., Tuenter, E. & Lourens, L. J. The precession phase of the boreal summer monsoon as viewed from the eastern Mediterranean (ODP Site 968). Quat. Sci. Rev. 29, 1481–1490 (2010). Weber, M. E. et al. Interhemispheric ice-sheet synchronicity during the Last Glacial Maximum. Science 334, 1265–1269 (2011). Stager, J. C., Ryves, D. B., Chase, B. M. & Pausata, F. S. R. Catastrophic drought in the Afro-Asian monsoon region during Heinrich Event 1. Science 331, 1299–1302 (2011). Soden, B. J., Wetherald, R. T., Stenchikov, G. L. & Robock, A. Global cooling after the eruption of Mount Pinatubo: a test of climate feedback by water vapor. Science 296, 727–730 (2002). Kutzbach, J., Bonan, G., Foley, J. & Harrison, S. P. Vegetation and soil feedbacks on the response of the African monsoon to orbital forcing in the early to middle Holocene. Nature 384, 623–626 (1996). Cheddadi, R. & Bar-Hen, A. Spatial gradient of temperature and potential vegetation feedback across Europe during the late Quaternary. Clim. Dyn. 32, 371–379 (2009). McCarthy, M. P. et al. The influence of vegetation on the ITCZ and South Asian monsoon in HadCM3. Earth Syst. Dyn. 3, 87–96 (2012).

4 | N AT U R E | VO L 0 0 0 | 0 0 M O N T H 2 0 1 4

©2014 Macmillan Publishers Limited. All rights reserved

LETTER RESEARCH 31. Jo, K. et al. Textural and carbon isotopic evidence of monsoonal changes recorded in a composite-type speleothem from Korea since MIS 5a. Quat. Res. 74, 100–112 (2010). 32. Jo, K. et al. Holocene and Eemian climatic optima in the Korean Peninsula based on textural and carbon isotopic records from the stalagmite of the Daeya Cave, South Korea. Quat. Sci. Rev. 30, 1218–1231 (2011). Supplementary Information is available in the online version of the paper. Acknowledgements We thank the Kangwon National University Cave Investigation Club (KNUCIC) for collecting some of the speleothem samples, S. S. Lee for the supplementary statistical test and K. R. Ludwig for providing the analytical software package. This research was a part of the project titled K-IODP (KIGAM; Korea Institute of Geoscience and Mineral Resources) funded by the Ministry of Oceans and Fisheries, Korea. This project was also partially supported by Basic Research Project

(GP2009-005) of KIGAM, grants NSFC 41230524 and NBRP 2013CB955902 (to H.C.) and US NSF grants 1103403 and 1337693 (to R.L.E. and H.C.). Author Contributions K.-n.J. collected most of the information and wrote the first draft of the manuscript. K.S.W. discussed the palaeoclimatic implications of the data and improved the manuscript. S.Y. commented on the palaeoclimatic interpretations. D.Y.Y. and H.S.L. provided research grants to K.S.W. and K.-n.J. for obtaining data on caves and speleothems. Y.W. provided the stable isotope data, and H.C. and R.L.E. supplied some of the 230Th/234U dating results. All authors approved the submitted form of the manuscript. Author Information Reprints and permissions information is available at www.nature.com/reprints. The authors declare no competing financial interests. Readers are welcome to comment on the online version of the paper. Correspondence and requests for materials should be addressed to K.S.W. ([email protected]).

0 0 M O N T H 2 0 1 4 | VO L 0 0 0 | N AT U R E | 5

©2014 Macmillan Publishers Limited. All rights reserved

RESEARCH LETTER METHODS Sampling. Twenty-four speleothem specimens were collected from fifteen limestone caves from 2004 to 2009. All samplings followed the code of ethics for cave conservation. The caves were all located within an area of less than 5,500 km2. The collected samples consisted of 22 stalagmites and two flowstones. Each speleothem was halved along the growth axis, and one side of the sample was analysed for petrographic evidence such as calcite texture, changes in the growth axis and possible growth hiatuses. Each powder subsample of 200–500 mg for dating was extracted using a dental drill from top and bottom positions on the polished surface of each specimen. 238 U–234U–230Th disequilibrium dating. We produced 230Th dates using a multicollector inductively coupled plasma mass spectrometer (MC-ICP-MS; ThermoFinnigan Neptune) at Minnesota Isotope Lab, University of Minnesota, USA. Chemical separation and purification of the uranium and thorium for 230Th/234U dating was based on ref. 33. Mass spectroscopic techniques used in this study are identical to ref. 34. All uranium isotopes were measured on Faraday cups with tail corrections performed by measuring masses 234.5 and 237 on a MasCom multiplier. The smallest uranium isotope (234U) beam intensities are measured at ,30 mV with a 1,011 ohm resistor. Instrumental mass fractionations were determined by measuring the 233U–236U double spike simultaneously. The procedures of characterizing the multiplier are similar to those described in ref. 35. Instrumental mass fractionation was determined by bracketing measurements of a 233U–236U spike. 230Th dating techniques are described in refs 35, 36. Half-life values are those reported in ref. 36. After the first inspection of all samples for the growth period, old samples beyond the upper limitation of 230Th/234U techniques, young samples that only grew within the Holocene, and samples affected by diagenetic alterations were excluded from further analysis. High-resolution 230Th/234U dating was applied for the six textural hiatuses-bearing stalagmites grown over long-term periods. We finally obtained 148 available uranium-series dates. Seventeen of the dates were from a previously published paper31,32. All age error bars in this paper depict 2s errors. The dating systematics used in ref. 4 and its potential systematic difference from our method is difficult to evaluate. However, because of the nature of U-series dating equations, the systematic errors will increase with sample ages progressively and, thus, the systematic dating errors between different laboratories for samples less than 300 kyr old would be relatively small (less than a few thousand years). For samples older than 300 kyr, the dating precision would also most probably be worse than the accuracy and therefore, within quoted uncertainties, the comparison might not be significantly affected by the accuracy or systematic errors between different laboratories. Statistical analysis. Our sampling method of speleothems was entirely random and the samples were taken from several different caves at different altitudes. Also, we tried to determine the sub-sampling points for U-series dating along the whole section as evenly as possible considering the texture of speleothems to avoid subsampling bias. Filtering to avoid Holocene bias remained 126 dates (Extended Data Fig. 3). The relative probability density curve for the total dates of the Korean speleothems was plotted using Isoplot/Ex 3.0 software37. This plot has an inherent defect that may be artificially caused by the sampling strategy (closer subsampling intervals in a particular part). To solve this problem, we supplemented the plot of the number of speleothem depositions with textural examination results and confirmed hiatuses (Extended Data Figs 4 and 5). A remaining weakness of the plot was that Holocene speleothem samples were more common than older samples. To address this, we primarily focused on the timings of peaks in the probability density curve, which were regarded as the periods of high-frequency speleothem growth in the Korean peninsula. Stable isotope analysis. For d18O and d13C analyses for three stalagmites (GE1, DY1 and ED1), powder samples of 100 mg were drilled using a 0.3-mm-diameter drill bit along growth axis. Oxygen and carbon isotope ratios were analysed at the Isotope Laboratory at Nanjing Normal University, China. 640 measurements of d18O and d13C were made on three speleothems using a Finnigan MAT-253 ratio mass spectrometer with an on-line, automated, carbonate preparation system (Kiel III). Isotopic data of 346 from GE1 and DY1 were published by refs 31, 32. Measured C- and O-isotope ratios in the per mil (%) notation are reported relative to the PDB (the belemnite from the Pee Dee formation) standard. Analytical errors for d18O and d13C are 0.15% and 0.2%, respectively. A detailed age model for time-series isotopic records of the stalagmites was derived by linear interpolation between all dates. Final age models for three stalagmites were tuned to fit into Chinese cave d18O records within age errors31,32. Atmospheric conditions of Korean limestone caves. According to some previous reports31,32, the air temperature of limestone caves in South Korea ranges from 10.7 uC to 15.1 uC, although temperatures at locations near cave entrances are highly variable between summer and winter. For example, Daeya cave is a relatively deeply seated underground cave, located more than 200 m below the surface32. The temperature of the sampling location ranges from 13.7 uC to 14.1 uC

throughout the year (Extended Data Fig. 1a). Eden cave provides another example. Eden cave is relatively small and is located close to the surface (less than 50 m below the surface) in the central part of the Korean peninsula. The temperature at the end of the cave passage from which the speleothem sample was collected is more or less stable throughout the year, ranging from 11.7 uC to 12.6 uC (Extended Data Fig. 1b). The measured data of cave temperature nearly coincide with the AMT of the study area. This suggests that the temperature range of most limestone caves in the Korean peninsula reflects long-term variations in surface air temperature near the caves. Humidity and partial pressure of carbon dioxide (pCO2 ) values measured in the Daeya and Eden caves also reflect atmospheric conditions typical of limestone caves in South Korea. Therefore, we believe that similar cave atmosphere conditions prevailed for the speleothems. Regional palaeoclimatic interpretation. The growth frequency of Korean speleothems has varied widely over time (Fig. 2 and Extended Data Fig. 6). Because our samples were randomly taken from several different caves, our result can be considered to reflect the general climatic trend (that is, not an arbitrary trend). Speleothem records from southwestern China, which is located only about 6 to 7 degrees of latitude south of the study area, do not show a similar trend, suggesting that major differences in climatic regimes existed between southwestern China and the Korean peninsula, mainly owing to dynamical responses to orbital-scale ITCZ migration38. Unlike the speleothem records from southwestern China, the growth of Korean speleothems corresponds with that of northeastern Chinese speleothems18. Our speleothem growth frequency record indicates unambiguous climatic fluctuations (interglacial and/or pluvial) without an isotopic equilibrium test (the so-called Hendy test and replicating examinations) as a high-resolution isotopic record39,40. The record reported here can help to solve the question of which climatic factors are critical and responsible for the growth of Korean speleothems. First, there is a clear difference between this Korean record and Chinese d18O records (Extended Data Fig. 6). Whereas it has been accepted that Chinese speleothem d18O records thoroughly reflect past variations in EAM intensity (mostly the amount of precipitation interpreted from the amount effect of d18O values and/or seasonal differences in d18O values of precipitation)41,42, the Korean speleothem record implies large temperature increases during the peak interglacials, MIS 5e and Holocene periods compared with the interstadial periods (MIS 5c, MIS 5a and MIS 3). This is because during the periods of middle-sized peaks in the Korean record, Chinese d18O records that reflect EAM intensity still show the high peaks of low values (Extended Data Fig. 6). Thus, the mid-sized peaks in Korean speleothem records cannot be explained by the amount of precipitation only. Temperature changes directly affect the soil respiration rate and in turn the pCO2 in the soil zone in the temperate climatic zones43. However, we cannot rule out the strong influence by precipitation changes on the growth of Korean speleothems because there is much evidence for changes in EAM intensity over the Korean peninsula at the orbital timescale31,32. Our record also shows good agreement with the timing of EAM changes (Fig. 2 and Extended Data Fig. 6). Therefore, large increases in both temperature and precipitation must have strongly promoted speleothem growth in the study area during peak interglacial periods. The mid-sized peaks in the Korean record are considered to have been affected by relatively highprecipitation and low-temperature conditions compared with conditions in peak interglacial periods. Second, the modern climate of the Korean peninsula is favourable for active speleothem growth11. In glacial periods, however, conditions in the Korean peninsula were not favourable for speleothem growth (Extended Data Fig. 6). A critical condition of the lowest limits of precipitation and temperature for speleothem growth is expected to exist. Speleothem growth in arid regions essentially requires at least 300 mm of precipitation for percolating seepage water44, and annual mean temperature should be higher than 21 uC to avoid the formation of permafrost in the soil zone45,46. Speleothem growth can be significantly limited near these thresholds. That is, even without the formation of permafrost and very low precipitation, speleothem growth may be reduced because of changes in precipitation and the melting dynamics of winter snow46. Numerical model and palynological results show that the study area suffered prominent palaeoclimatic changes of a 5–6u C decrease in AMT and an approximately 40% (.500 mm) decrease in precipitation during the LGM8,47. These changes corresponded to approximately 6u C of AMT and 700 mm of AMP, respectively. However, speleothem observations in cold and dry central Asia21 suggest that climatic conditions of ,6u C AMT and annual precipitation of ,700 mm—similar to modern Shenyang, China (,8 uC, 700 mm), Chunggang, Korea (,4 uC, 800 mm), and Vladivostok, Russia (,4 uC, 800 mm)—would not have been sufficient to largely inhibit speleothem growth during glacial periods. We noted that some of the Korean speleothems still grew, although very slightly, during some of the full glacial periods (Fig. 2 and Extended Data Fig. 6). This suggests that conditions in South Korea were near the threshold for speleothem growth but not below or far above the threshold during some full glacial periods. Climate conditions in Korea

©2014 Macmillan Publishers Limited. All rights reserved

LETTER RESEARCH during the glacial period may have been similar to those found in modern Mongolia or Manchuria (2–3 uC in AMT and 400 mm in AMP), a suggestion that agrees with Korean pollen8 and vertebrate evidence7. Taken together, these observations lead us to propose that the Korean peninsula experienced AMT of approximately 2–6 uC and AMP of approximately 300–700 mm during full glacial periods. The modern climate of the Korean peninsula is very sensitive to the seasonal switches in air masses through both atmospheric and oceanographic changes9. The warm, humid summer and cold, dry winter are largely governed by the North Pacific and Siberian high-pressure systems, respectively5. Speleothem growth is inhibited by significant decreases in temperature and precipitation2,46. In turn, the strong and persistent influences of the Siberian high-pressure system on the Korean peninsula during the full glacial periods were probably responsible for the very limited growth of Korean speleothems. It is also possible that further southward shift or even failure of the Jangma rain belt (called Meiyu in Japan and Bai-u in China) could have occurred under the glacial conditions. This is because the Jangma rainfall system is a strong frontal system developed between the Okhotsk and North Pacific air masses, and thus weakening and southward shifting of the North Pacific air mass could cause the Jangma system to weaken significantly9. In the extreme case, the Jangma rainfall belt may not have reached the Korean peninsula during full glacial periods. Because the amount of current Jangma rainfall is relatively constant every year (,300–400 mm), a reduction in AMP caused by changes in Jangma rainfall would be needed to explain the approximately 500–800 mm decrease in AMP, which is the value we assumed above. Scenario of ice age cycles linked with the interhemispheric hydrologic seesaw. Here we discuss (1) the peak interglacial periods, (2) the stadial periods, (3) the interstadial periods, (4) the full glacial periods and (5) the glacial termination periods linked with the interhemispheric hydrologic seesaw. (1) Seasonal migration of the Intertropical Convergence Zone (ITCZ) is a major consequence of interhemispheric adjustment for the thermal energy disproportion on the Earth’s surface caused by different distributions of solar radiation. Strong convective meridional overturning circulation transports equatorial excess heat energy by bringing enormous amounts of moisture and latent heat over higher latitudes and into the mid-continents. The dynamical response of the ITCZ to orbital forcing has been studied in various tropical and subtropical areas, including in African, Asian and South American monsoon regions at millennial to orbital timescales1–3. Under peak interglacial conditions, the ITCZ reaches its maximum extent of seasonal variation, and the mean position of the ITCZ should be in the northernmost part of its glacial–interglacial variations (Extended Data Fig. 7). This condition directly affects the strengthening and northward position of subtropical permanent high-pressure cells such as North Pacific air masses through large, deep convection of subtropical Hadley cells. The expansion of warm and wet atmospheric conditions to mid- to high-latitude regions promotes precipitation–vegetation feedbacks28,29 that, in turn, produce low-albedo effects (and thus high absorption of solar heat energy) in the mid-latitudes. Relatively warmer and wetter atmosphere in higher latitudes would permeate into poleward and landward regions and would shrink the ranges of continental ice sheets and permafrost regions on a global scale. Greenhouse gases released with the melting of ice sheets and permafrost would rapidly be injected into the atmosphere, causing a stronger greenhouse effect. Finally, sea level would reach the highest level of the glacial–interglacial variations because of melting of the continental ice sheets. Worldwide, vast areas of the continental margins would be submerged under sea water, and marine environments would reach into inner continental areas, further decreasing albedo effects. These phenomena would make the global shape of the ITCZ more irregular and landward (Extended Data Fig. 7). Additionally, active atmospheric and oceanographic circulations would distribute solar heat energy throughout the globe and highly amplify the initial solar radiation. Under these conditions, global karstification would be greatly amplified by the much higher precipitation and temperature in the mid- to high-latitude regions, even in Norway and Siberia (Extended Data Fig. 7). Karstification of Korean caves would reach maximum efficiency, including subsurface dissolution of limestone bedrock and active formation of speleothems (Fig. 2). (2) The mean position of the ITCZ migrated substantially south during the abrupt North Atlantic ice rafting events1,2. It is widely accepted that the southern positioning of the ITCZ inevitably leads to the weakening of the boreal summer monsoon intensity and that this resulted in severe drought episodes in northern subtropical to temperate regions during the late Quaternary glacial periods1. In contrast, southward movements of the ITCZ had the opposite hydrologic effect in the Southern Hemisphere2. With decreasing solar insolation, the mean position of the ITCZ quickly relocated to the southernmost position (Extended Data Fig. 7). This phenomenon caused very weak EAM intensity and significant reductions in vegetation cover in midlatitude regions of the Northern Hemisphere. The North Pacific air mass was probably weaker and shifted to a southern position5, resulting in much cooler and drier

conditions in the Korean peninsula during the summer season. Furthermore, the Korean peninsula would have been under the direct influence of the stronger Siberian high-pressure system during autumn to spring. These conditions would have eventually led to decreased vegetation cover, the formation of permafrost, the advance of continental ice sheets, and the lowering of sea level. However, these processes should have been much slower than the opposite precipitation–vegetation– albedo feedbacks during peak interglacial periods because the feedbacks would have been broken; that is, although precipitation directly influences the expansion of permafrost and continental ice sheets, a decrease in precipitation in mid- to highlatitude regions of the Northern Hemisphere would only hamper the growth of permafrost and ice sheets because of the deficient moisture supply. In contrast to the Northern Hemisphere, the Southern Hemisphere experiences the opposite atmospheric conditions, including high precipitation by the Southern Hemisphere bias in incoming solar radiation1,2. However, temperature in the mid-latitude Southern Hemisphere would not be higher than in peak interglacial periods as a result of the large expansion of Northern Hemisphere ice sheets and subsequent global temperature decrease, which might have caused an advance in the Antarctic ice sheets and sea ice. Moreover, the Southern Hemisphere does not have large land areas for precipitation–vegetation–albedo feedbacks; thus, with only the effect of climatic processes in the Southern Hemisphere, it would be difficult to restore global climate to its former condition. In these periods, Northern Hemispheric speleothems, except in the tropics to subtropics, would grow little or not at all (Extended Data Fig. 7). Korean speleothems provide an example of the cessation of speleothem growth in lowermost latitude regions of a temperate climatic zone with dense vegetation. In contrast, in the Naracoorte cave sites in southeastern Australia, the stadial periods were the periods in which carbonate speleothems grew most rapidly, mainly controlled by wetter and cooler climate conditions4. (3) During interstadial periods, the ITCZ migrates northward and has high seasonal variability because of the recovery of high solar insolation (Extended Data Fig. 7). However, the relatively small areas of permafrost, ice sheets, vegetation cover and exposed continental margins compared with those during full glacial periods would confine the increasing degree of global temperature and precipitation to much lower levels than during the glacial terminations48. Thus, continental ice sheets would remain in both the northern and southern hemispheres and could continue to grow until the next stadial and full glacial periods. Speleothem growth probably resumed in the Northern Hemisphere mid-latitudes, but at a slower rate than during the peak interglacial periods (Extended Data Figs 6 and 7). Previous studies found that speleothem growth in Norway and Siberia stopped completely, and only limited speleothem growth took place at temperate sites in Europe and Korea during interstadial periods. In contrast, the growth of southeastern Australian speleothems was halted by northward shifts of the ITCZ and the reassignment of related atmospheric circulations. (4) The ITCZ shows its minimum extent of seasonal variations and southward movement of its mean position during full glacial periods (Extended Data Fig. 7). It has recently been suggested that systematic weakening of Hadley circulation with southward displacement of the ITCZ had occurred during a massive ice-rafting event, Heinrich event 126. This was linked to small and limited atmospheric and oceanographic circulations in low-latitude regions. Continental ice sheets that survived through interstadial periods would have contributed to the consistent and maximized expansion of ice sheets. Global temperatures would have significantly decreased with large expansion of continental ice sheets in the Northern Hemisphere. This would influence the maximum extent of the Antarctic ice sheet and sea ice by global reduction in precipitation and vegetation cover and highalbedo effects, followed by sea level changes. Ice sheets in both hemispheres would reach their maximum ranges, which would create strong ice-driven forcing in the mid-latitude regions by active and large-scale circulation of polar cells and would thrust the ITCZ towards equatorial regions (Extended Data Fig. 7). These processes would additionally contribute to the ITCZ having its minimum seasonal extent and flattened configuration. Global karstification would probably be at its minimum ranges because of the low precipitation, temperature, and terrestrial productivity in mid-latitude regions of both hemispheres. Growth of both Korean and southeastern Australian speleothems was nearly stopped by cold and dry palaeoclimatic conditions in the midlatitude regions during full glacial periods (Extended Data Figs 6 and 7). (5) The climatic processes of these periods are the same as those in the interstadial periods. However, the maximum capacity of permafrost and ice sheets in these periods would be enough to be linked to the highest rate of thawing of glacial ice sheets and permafrost and could heighten large-scale precipitation–vegetation– albedo feedbacks, turning full glacial conditions to peak interglacial conditions. 33. Edwards, R. L., Chen, J. H. & Wasserburg, G. J. 238U–234U–230Th–232Th systematics and the precise measurement of time over the past 500,000 years. Earth Planet. Sci. Lett. 81, 175–192 (1987).

©2014 Macmillan Publishers Limited. All rights reserved

RESEARCH LETTER 34. Cheng, H. et al. Ice age terminations. Science 326, 248–252 (2009). 35. Cheng, H. et al. The half-lives of uranium-234 and thorium-230. Chem. Geol. 169, 17–33 (2000). 36. Cheng, H. et al. A new generation of 230Th dating techniques: tests of precision and accuracy. Geochim. Cosmochim. Acta 72, A157 (2008). 37. Ludwig, K. R. Isoplot 3.00: a Geochronological Toolkit for Microsoft Excel 70 (Univ. of California, Special Publication 4, 2003). 38. He, X., Wang, J.-L., Li, H., Cheng, H. & Yuan, D. Growth pattern and depositional feature of stalagmites in the karst of Chongqing, SW China: climatic implications. Karst Waters I Spec. Pub. 10, 31–32 (2006). 39. Hendy, C. The isotopic geochemistry of speleothems—I. The calculation of the effects of different modes of formation on the isotopic composition of speleothems and their applicability as paleoclimatic indicators. Geochim. Cosmochim. Acta 35, 801–824 (1971). 40. Dorale, J. A. & Liu, Z. Limitations of Hendy Test criteria in judging the paleoclimatic suitability of speleothems and the need for replication. J. Caves Karst Stud. 71, 73–80 (2009). 41. Wang, Y. J. et al. A high-resolution absolute-dated late Pleistocene monsoon record from Hulu Cave, China. Science 294, 2345–2348 (2001). 42. Yuan, D. et al. Timing, duration, and transitions of the last interglacial Asian mosoon. Science 304, 575–578 (2004). 43. Chae, N. Annual variation of soil respiration and precipitation in a temperate forest (Quercus serrata and Carpinus laxiflora) under East Asian monsoon climate. J. Plant Biol. 54, 101–111 (2011). 44. Vaks, A. et al. Paleoclimate and location of the border between Mediterranean climate region and the Saharo-Arabian Desert as revealed by speleothems from the northern Negev Desert, Israel. Earth Planet. Sci. Lett. 249, 384–399 (2006). 45. Dobinski, W. et al. Permafrost. Earth Sci. Rev. 108, 158–169 (2011). 46. Ersek, V. et al. Environmental influences on speleothem growth in southwestern Oregon during the last 380,000 years. Earth Planet. Sci. Lett. 279, 316–325 (2009). 47. Kim, S.-J. et al. High-resolution climate simulation of the last glacial maximum. Clim. Dyn. 31, 1–16 (2008).

48. Levavasseur, G. et al. Present and LGM permafrost from climate simulations: contribution of statistical downscaling. Clim. Past Discuss. 7, 1647–1692 (2011). 49. Hennig, G. J., Grun, R. & Brunnacker, K. Speleothems, travertines, and paleoclimates. Quat. Res. 20, 1–29 (1983). 50. Gordon, D. et al. Dating the late Pleistocene interglacial and interstadial periods in the United Kingdom from speleothem growth frequency. Quat. Res. 31, 14–26 (1989). 51. North Greenland Ice Core Project members. High-resolution record of Northern Hemisphere climate extending into the last interglacial period. Nature 431, 147–151 (2004). 52. Bar-Matthews, M. Ayalon, A. & Kaufman, A. Timing and hydrological conditions of sapropel events in the Eastern Mediterranean, as evident from speleothems, Soreq Cave, Israel. Chem. Geol. 169, 145–156 (2000). 53. Martinson, D. G. et al. Age dating and the orbital theory of the ice ages: development of a high-resolution 0 to 300,000 year chronostratigraphy. Quat. Res. 27, 1–29 (1987). 54. Linge, H., Lauritzen, S.-E. & Lundberg, J. Stable isotope stratigraphy of a late last interglacial speleothem from Rana, northern Norway. Quat. Res. 56, 155–164 (2001). 55. Burns, S. J., Fleitmann, D., Matter, A., Neff, U. & Mangini, A. Speleothem evidence from Oman for continental pluvial events during interglacial periods. Geology 29, 623–626 (2001). 56. Fleitmann, D. et al. Holocene and Pleistocene pluvial periods in Yemen, southern Arabia. Quat. Sci. Rev. 30, 783–787 (2011). 57. Partin, J. W., Cobb, K. M., Adkins, J. F., Clark, B. & Fernandez, D. P. Millennial-scale trends in west Pacific warm pool hydrology since the Last Glacial Maximum. Nature 449, 452–455 (2007). 58. Griffiths, M. L. et al. Increasing Australian-Indonesian monsoon rainfall linked to early Holocene sea-level rise. Nature Geosci. 2, 636–639 (2009). 59. Cai, Y. et al. Large variations of oxygen isotopes in precipitation over south-central Tibet during Marine Isotope Stage 5. Geology 38, 243–246 (2010).

©2014 Macmillan Publishers Limited. All rights reserved

LETTER RESEARCH

Extended Data Figure 1 | Temperature, humidity, and partial pressure of CO2 in the Daeya and Eden caves. a, b, The atmospheric conditions in Daeya cave (a) and Eden cave (b).

©2014 Macmillan Publishers Limited. All rights reserved

RESEARCH LETTER

Extended Data Figure 2 | Plot of 230Th dating results for all speleothem samples from the Korean peninsula. Error bars show the 2s analytical

uncertainty, including the error from the detrital 230Th. Grey horizontal bars divide each sample.

©2014 Macmillan Publishers Limited. All rights reserved

LETTER RESEARCH

Extended Data Figure 3 | Plot of relative probability constructed by summing the individual ages of all the dated samples. Panel a shows all dates,

and panel b presents the relative probability density, except for samples that grew only within the Holocene (Supplementary Table 1).

©2014 Macmillan Publishers Limited. All rights reserved

RESEARCH LETTER

Extended Data Figure 4 | The age data along the growth axes of three speleothem samples. The yellow vertical bars represent the minimum periods of major growth intervals. a, ED1 from Eden cave. b, JM3 from Joongmal cave.

c, JM2 from Joongmal cave. Error bars represent 2s analytical uncertainty with error from the detrital 230Th.

©2014 Macmillan Publishers Limited. All rights reserved

LETTER RESEARCH

Extended Data Figure 5 | Frequency distributions of speleothem growth in the Northern Hemisphere. a, North America and Europe49; b, United

Kingdom50; c, South Korea. Also shown is the oxygen isotope record (d) from Devils Hole in Nevada, western USA17 (red line with dots).

©2014 Macmillan Publishers Limited. All rights reserved

RESEARCH LETTER

Extended Data Figure 6 | Comparison of the growth frequency record from Korean speleothems with other high-resolution palaeoclimatic records from high-latitude polar to low-latitude tropical regions since the penultimate glacial termination. a, d18O record of Sanbao cave in central China15. b, Summer insolation at 65u N13. c, d18O record from the North Greenland Ice Core Project (NGRIP)51; VSMOW is Vienna standard mean ocean water. d, Growth frequency record of Korean speleothems. S numbers indicate

Mediterrean sapropel events. Periods of high growth frequency are shown in light brown. e, d18O record for Soreq cave in the eastern Mediterranean region52. f, Radiolarian temperature record from marine sediment core RC11-120 in the Southern Ocean53. The present values of each proxy record are also shown by horizontal dashed lines on each proxy curve. Palaeoclimatic signals from mid-latitude regions display both low- and high-latitude characteristics.

©2014 Macmillan Publishers Limited. All rights reserved

LETTER RESEARCH

Extended Data Figure 7 | Scenario of palaeoclimate changes linked with the interhemispheric hydrologic seesaw. Background image is from NASA Eclipse Web Site (http://eclipse.gsfc.nasa.gov/transit/TV2004.html). a, Peak interglacial periods are characterized by large geographic variations in the seasonal ITCZ and strong atmospheric meridional overturning circulations. b, Stadial periods show intermediate climatic conditions between those of peak interglacial and full glacial periods. c, Interstadial periods show intermediate climatic conditions between those of peak interglacial and full glacial periods.

d, Full glacial periods are characterized by small geographic variations in the seasonal ITCZ and weak atmospheric meridional overturning circulations. Conceptual geographic ranges of estimated permafrost and ice sheets during stadial periods are shown by grey- and white-coloured areas, respectively48. Also shown are the cave locations of active speleothem growth during each time period (red dots)4,15,21,42,49,52,54–59. See Methods for a detailed description of the scenario.

©2014 Macmillan Publishers Limited. All rights reserved

RESEARCH LETTER Extended Data Table 1 | Information about the caves

NS means not surveyed. For cave type, H 5 horizontal, V 5 vertical, and C 5 composite. For bedrock type, L 5 limestone, A 5 argillaceous limestone and D 5 dolomite.

©2014 Macmillan Publishers Limited. All rights reserved