Co-evolutions of terrestrial temperature and seasonal precipitation from the latest Pleistocene to the mid-Holocene in Japan: carbonate clumped isotope record of a stalagmite

Quantitative paleotemperature reconstruction is a challenging and important issue in terrestrial paleoenvironmental studies, for which carbonate clumped isotope (Δ 47 ) thermometry is a promising approach. Here we analyzed Δ 47 values from 68 layers of OT02 stalagmite from Ohtaki Cave in central Japan, covering two separate time intervals (2.6–8.8 and 34.8–63.5 ka) to reconstruct temperature and meteoric d 18 O records. The average Δ 47 temperature of the Holocene portion of this stalagmite was 16.3 ℃ ± 5.6 ℃ , 6.6 ℃ ± 7.2 ℃ higher than the average of the latest Pleistocene portion, which was 9.7 ℃ ± 4.6 ℃ . Δ 47 thermometry also revealed that the coldest intervals (5 ℃ –10 ℃ ) correspond to the Heinrich cooling events H4–6, and the warmest interval (up to 19.9 ℃ ± 6.0 ℃ ) in middle Holocene (approximately 6–5 ka) accompanied by the Hypsithermal climate optimum. We also reconstructed past meteoric δ 18 O by subtracting the temperature effect from stalagmite δ 18 O. Average meteoric δ 18 O was less negative in the Holocene (8.22 ‰ ± 0.99 ‰ VSMOW) than in the latest Pleistocene (8.81 ‰ ± 0.84 ‰ ). Over centennial timescales, meteoric δ 18 O was more negative during colder periods, such as Heinrich cooling events and the cooling event around 7 ka, and less negative in warmer periods, such as Hypsithermal warming. These relations indicated co-evolution of terrestrial paleotemperature and paleoprecipitation. A temperature dependency of 18 O fractionation from water to vapor is a likely reason for the negative correlation between temperature and meteoric δ 18 O. Additionally, it is possible that increasing lower δ 18 O precipitation from East Asian winter monsoon (EAWM) has decreased the averaged meteoric δ 18 O in colder periods. These temperature effects on meteoric δ 18 O occur in opposite directions to fractionation between water and the stalagmite δ 18 O, explaining the small amplitudes of changes observed in the δ 18 O of Japanese stalagmites.


Introduction
The oxygen isotopic composition of stalagmite calcite (δ 18 O C ) is a key source of information for terrestrial paleoclimates. Nevertheless, paleoclimatic interpretation of stalagmite δ 18 O C should consider two main controlling factors: the temperature of calcite formation and the oxygen isotopic composition of cave drip water (δ 18 O W ). The latter factor also involves changes in some meteorological conditions, such as monsoon intensity, moisture trajectory, and precipitation seasonality. Although previous stalagmite studies in East Asia often emphasized the so-called "amount effect" or the change in moisture sources (e.g., Wang et al., 2001) as a cause of changes in δ 18 O C , limited attention has been paid to the role of temperature changes on the stalagmite δ 18 O C records (e.g., Tremaine et al., 2011;Meckler et al., 2015).
The importance of temperature signals was rea rmed for late Pleistocene to middle Holocene stalagmites from Mie and Gifu Prefectures, central Japan, located at the eastern margin of the East Asian monsoon regime (Mori et al., 2018;Fig. 1). These stalagmites exhibit notably smaller amplitudes of variation in δ 18 O C than comparable Chinese cave records; hence, the glacial/interglacial contrast within these stalagmites can be accounted for by a typical warming temperature of 9℃ from the last glacial maximum (LGM) to the mid-Holocene (Mori et al., 2018). This estimated temperature change is comparable with other paleoclimate studies using stalagmites and lake and marine deposits around the Japanese Islands (e.g., Nakagawa et al., 2002;Kawahata et al., 2011;Kigoshi et al., 2014;Uemura et al., 2016). Mori et al. (2018) concluded that climate control on δ 18 O W values was insigni cant for Japanese caves, perhaps because of their proximity to moisture sources in the Paci c.
Changes in temperature and precipitation are principal factors in uencing terrestrial climate and are intimately connected. To reconstruct their co-evolution, their signals should be divided quantitatively.
Regrettably, it is challenging to determine the relative importance of the two fundamental controls on stalagmite δ 18 O C , temperature and δ 18 O W , using conventional isotopic compositions alone. An independent approach to estimate paleotemperature is carbonate clumped isotope (Δ 47 ) thermometry (e.g., Ghosh et al., 2006;Eiler et al., 2007), which does not require assumptions regarding δ 18 O W values.
In our previous study , we measured Δ 47 values from 50 layers of a well-dated stalagmite Hiro-1 (4.5-18.1 ka) from Maboroshi Cave in Hiroshima Prefecture, southwestern Japan (Shen et al., 2010;Hori et al., 2013Hori et al., , 2014Fig. 1). We revealed terrestrial temperature changes after the LGM to the mid-Holocene, except for two hiatuses around 12.8-11.4 and 10.8-7.7 ka and some layers in which Δ 47 and δ 18 O C were signi cantly disturbed by kinetic isotope effect (KIE) because of dry cave conditions. The difference between the average Δ 47 temperatures in the pre-Holocene (18.0-16.0 ka) and mid-Holocene (7.7-4.9 ka) was 8.3℃ ± 4.2℃, which is comparable with other estimates for the difference between the LGM and the Holocene from the two Japanese caves, Gyokusendo (8.2℃, 0 versus 26 ka; Uemura et al., 2016) and Kiriana (9℃, 7 versus 24 ka; Mori et al., 2018). The Δ 47 values of Hiro-1 exhibit an abrupt warming around 6.3-4.9 ka, which likely corresponds to the Hypsithermal event (e.g., Wanner et al., 2008), but the record during the corresponding Heinrich cooling event (HS1) was unavailable because of strong in uences of KIE related to dry cave conditions . We also reconstructed changes in the meteoric water δ 18 O MW from the stalagmite δ 18 O and the Δ 47 temperature, nding a correlation between δ 18 O MW and Δ 47 temperature. We show that the δ 18 O MW largely shifted positively because of changes in vapor trajectory to the region due to the Holocene transgression of the Seto Inland Sea (SIS; Fig. 1b; Kato et al., 2021).
In this study, we analyzed the Δ 47 values of the OT02 stalagmite (2.6-8.8 and 34.8-63.5 ka; Mori et al., 2018) from Ohtaki Cave (Fig. 1) to obtain a longer climatic record in Japan (back to 63 ka) including four periods of Heinrich cooling that have not yet been elucidated in Japan. We discuss the co-evolution of terrestrial temperature and precipitation from the Late Pleistocene to Holocene, making comparisons with the results from the Hiro-1 stalagmite .
Ohtaki Cave has a total length of more than 1000 m (Yura, 2011) into the Permian limestone of the Mino Terrane (Kajita et al., 1971). It comprises three cave levels; stalagmites are best developed in the middle level, which is located 110 m underground. The cave has been modi ed by arti cial tunnels for tourists, but the temperature in the cave remains stable at approximately 13.0℃, which is higher than the annual mean temperature observed in the nearest weather observatory at Nagataki (11.5℃). A larger amount of water seepage during summer likely elevates the cave water temperature throughout the year.
We analyzed A 140-mm-long stalagmite (OT02) collected 200 m from the cave entrance. U-Th ages and δ 18 O values from OT02 have been reported by Mori et al. (2018). The age model of OT02 was established on the basis of U-Th ages from 11 horizons using a Bayesian statistical model (StalAge; Scholz and Hoffman, 2011). A hiatus of 26 kyr (8.8-34.8 ka) was recognized at a discontinuous surface at the 55 mm horizon (Mori et al., 2018). The age ranges of the upper and lower parts are 2. 6-8.8 and 34.8-63.5 ka, respectively; however, the age model of the upper portion of OT02 includes a relatively large uncertainty (up to ±2.5 kyr; Mori et al., 2018). The lower part of OT02 stalagmite records four Heinrich cooling events (H4-6 including H5.2).

Climatic and hydrological settings of Ohtaki and Maboroshi Caves
Two major moisture sources to Japan are the Paci c Ocean and the Japan Sea, from which the East Asian summer and winter monsoons (EASM and EAWM, respectively) bring moisture mainly by warm southerly winds and cold northerly winds, respectively. The climate of Japan is characterized by clear seasonality because the East Asian summer/winter monsoons migrate across the region. Consequently, the precipitation source switches seasonally between the Paci c Ocean in summer and the Japan Sea in winter (Fukui, 1977). On the Paci c side of the central mountains in Japan, including the sites of Ohtaki Cave (OT02) and Maboroshi Cave (Hiro-1), the EASM delivers the majority of the annual precipitation to the Paci c side of Japan. During the winter, the cold and dry EAWM winds acquire moisture from the Tsushima Warm Current, which blows into the Japan Sea. Most moisture from the Japan Sea side is then released as heavy snow and rainfall on the Japan Sea coast, whereas residual moisture generates comparatively short snow and rainfall on the Paci c side. Consequently, less negative δ 18 O MW values are observed during the warm season, and more negative δ 18 O MW values are observed during the cold season in several areas in Japan. Hori et al. (2009) collected meteoric water at monthly intervals at Nagaya (Okayama Prefecture), located 20 km east of the site of Hiro-1, in 2005Hiro-1, in -2007. δ 18 O MW exhibited less negative values from April to October (−7.1‰ ± 1.4‰ in amount-weighted average) and more negative values from November to March (−9.7‰ ± 1.4‰; Hori et al., 2009;Fig. 2d). Mori et al. (2018) (Asai et al., 2014). Although the altitudes of these sites (Nagaya, Ohgaki, and Ontake) differ signi cantly from that of Ohtaki Cave, there is a common trend in that the most negative δ 18 O MW values were observed in winter. Since Ohgaki is located upstream of Ohtaki along the major moisture trajectory of the region, the seasonal pattern of δ 18 O MW around Ohtaki Cave is likely most similar to that in Ohgaki.
A major difference between the climatic settings of Maboroshi Cave (Yuki) and Ohtaki Cave (Nagataki) is a seasonal bias of precipitation amount. In Yuki, precipitation during the warmer 7 months (April-October) accounts for 78% of the annual total, which is more than three times the precipitation in the other 5 months (November-March; 22%). Particularly, precipitation during the coldest 3 months (December-February) is limited to only 10% of the annual total and the yearly variation is also very limited in these months (Fig. 2d). In the region of Maboroshi Cave, the Chugoku Mountains (1729 m asl) obstruct the winter rain/snowfall from the Japan Sea side. Conversely, in Nagataki (near Ohtaki Cave), precipitation during the three winter months (December-February) accounts for 16.8% of annual precipitation and yearly variations in winter precipitation are far larger than those in Yuki ( Fig. 2c and d). Consequently, the amount ratio of winter and annual precipitation is also an important factor controlling the annual average δ 18 O MW values at Ohtaki Cave. Figure 3 shows the standard deviation of monthly average temperature (1σ; black bold line) and the correlation coe cients between annual and monthly average temperatures (R; dashed line) in Nagataki ( Fig. 3a) and Yuki (Fig. 3b) over the last 40 years. It also shows the impact of monthly temperatures on the annual average (℃/℃; gray line), i.e., the regression slope between monthly and annual temperature. In Nagataki (near Ohtaki Cave), temperature deviations tend to be large in the colder season (December-April) and small in the warmer season (May-October) (the largest deviation occurs in February and November and the smallest in June). The correlation coe cient (R) of annual and monthly temperatures is large in February-May and September-October and smallest in July-August, the warmest months ( Fig. 2). This indicates that summer temperature is the least important factor in uencing the annual average temperature in Nagataki. These relations will be discussed in section 5.4.  (Schrag et al., 1996(Schrag et al., , 2002Bintanja and van de Wal, 2008;Shakun et al., 2015). We follow Mori et al. (2018) in de ning the residual variation in δ 18 O OT02 by subtracting the in uence from δ 18 O SW change (Δ 18 O SW ) as follows: We then assume that residual variation (Δ 18 O OT02−SW ) re ects variations in air temperature (temperature dependency of isotopic fractionation) and hydrologic processes (e.g., fractionation from seawater to vapor, amount effect, and continental effect on meteoric δ 18 O).

Stalagmite Δ 47 value
The principle of clumped isotope thermometry is based on the temperature dependency of the observed abundance anomaly of the 13 C-18 O bond in carbonate relative to the stochastic abundance calculated using bulk δ 13 C and δ 18 O values. The abundance anomaly of 13 C 18 O 16 O (Δ 47 ) in CO 2 from acid digestion of carbonate minerals is negatively correlated with the temperature of carbonate precipitation Schauble et al., 2006). The analysis of carbonates synthesized at known temperatures provides an experimental (empirical) calibration between temperature and Δ 47 (e.g., Ghosh et al., 2006;Dennis and Schrag, 2010;Zaarur et al., 2013;Fernandez et al., 2014;Tang et al., 2014;De iese et al., 2015;Kluge et al., 2015;Bonifacie et al., 2017;Kelson et al., 2017;Kato et al., 2019). Nevertheless, estimating paleotemperature from stalagmite Δ 47 values is challenging. CO 2 degassing from cave water leads to isotopic disequilibrium in the dissolved inorganic carbon (DIC) pool and thus a smaller Δ 47 value of carbonate, which yields higher temperatures than predicted (e.g., Guo and Zhou., 2019). KIEs under these conditions have been reported from natural and laboratory speleothem Δ 47 results (Affek et al., 2008Daëron et al., 2011;Kluge and Affek, 2012;Affek, 2013;Kato et al., 2021).
Our previous study  measured the Δ 47 values of two sample sets: 1) calcites synthesized at different temperatures (2.9℃-61.0℃) and 2) Japanese natural tufa collected at monthly intervals, for which the precipitation temperature was carefully estimated (5.6℃-16.0℃). Synthetic calcites were precipitated according to the "mixed-solution method" to minimize the kinetic effect of CO 2 degassing by bulk solution, and tufa samples were collected monthly at two sites by Kawai et al. (2006). Similar to stalagmite settings, CO 2 degassing induces calcite precipitation in tufa by increasing its saturation state in stream water (Ford and Pedley, 1996;Kano et al., 2003Kano et al., , 2019Kawai et al., 2006). We found a clear difference in the Δ 47 values of synthetic calcites and tufa samples. The Δ 47 values of tufa were lower than those of synthetic calcites, which is in the same direction as reported in the stalagmite studies referred above . According to our results, the range of the Δ 47 offset between the tufa samples and synthetic calcite appears stable throughout the year, within a temperature range of 5.6℃-16.0℃ and a pCO 2 of 0.55-1.01 matm . Since these factors (temperature and pCO 2 ) control the rate of CO 2 degassing, our observations indicate that the degassing rate does not signi cantly in uence the kinetic effect on the Δ 47 offset . Besides that for the synthetic calcite, Kato et al. (2019) developed an empirical temperature calibration for the natural tufa samples (Eq. 3) as follows: Δ 47 = (0.0336 ± 0.0036) * 10 6 / T 2 + (0.301 ± 0.048) (3) where T is absolute temperatures in Kelvins. The evaluated temperature offset between the two calibrations was ~4℃ , which broadly corresponds to the temperature offset from equilibrium conditions observed in modern-day speleothems, such as Soreq Cave, Israel . The Δ 47 temperature calibration of tufa (Eq. 3; Kato et al., 2019) was applied to the Δ 47 values of Hiro-1 stalagmite and successfully yielded the paleotemperature . We therefore applied the same calibration (Eq. 3) to stalagmite Δ 47 to estimate paleotemperature in this study.

Δ 47 measurement and calculation
We collected subsamples for carbonate clumped isotope (Δ 47 ) measurement from 68 layers throughout OT02. To avoid strong disequilibrium effects as much as possible, subsamples were collected from clear layers excluding opaque and muddy horizons presumed to have formed under depressed and/or discontinuous calcite precipitation. The signi cantly large prior calcite precipitation (PCP) associated with dry conditions in Maboroshi Cave likely caused smaller Δ 63 of drip water DIC and smaller Δ 47 of Hiro-1 stalagmite . Δ 47 measurements were performed using a dual inlet mass spectrometer (Finnigan MAT-253) at Kyushu University for four discrete periods from 2018 to 2019. Each subsample was measured once or twice, and the stability of Δ 47 measurements was con rmed by repeated measurements of an in-house calcite standard (Hiroshima standard; δ 13 C = −0.47‰, δ 18 O = −5.04‰ and Δ 47 = 0.671‰), which yielded 1 SD of ±0.006‰-0.009‰ (n = 5-13 for each period of measurement) corresponding to ±1.8℃-3.7℃ in the temperature range of OT02. Five milligrams of powdered carbonate were digested by phosphoric acid at 90℃ for ~5 min and the generated gas was immediately trapped by a stainless-steel tube cooled by liquid nitrogen. Moisture was separated from CO 2 gas using a liquid nitrogen and ethanol slush, after which CO 2 was introduced to a 30-m-long capillary column (Supel-Q PLOT) with a helium carrier gas. The column was cooled to −10℃ to remove organic contaminants. Puri ed CO 2 was analyzed by MAT-253 con gured for measurements of masses 44-49 with pressure adjustment to produce an m/z 44 signal of 16 V. The integration time was 30 s. We applied the Pressure Baseline correction of He et al. (2012) with the off-peak measurement of the background intensities of masses 45-49. Each analysis comprised 5-8 acquisitions, with 4.5 off-peak (nine alternate detections; ve times for the reference side and four times for the sample side), 8 on-peak, and 4 off-peak cycles per acquisition.

Paleometeoric δ 18 O reconstruction
As mentioned above, stalagmite δ 18 O C value is controlled by the temperature of calcite formation and the δ 18 O W value of water. Using the paleotemperature determined by carbonate clumped isotope and the temperature dependency of isotopic fractionation between calcite and water, past δ 18 O W values can be calculated from the calcite δ 18 O C value. We applied the temperature dependency of δ 18 O of Eq. 4) as follows: 1000lnα water−calcite = 16.1 (10 3 T −1 ) − 24.6 (4) Eq. 4 was established for cave deposits  and is also consistent with δ 18 O variability in Japanese stalagmites studied by Mori et al. (2018) and Kato et al. (2021).
For the present study, we de ned two types of the index of past meteoric isotopes, δ 18 O MW and ε 18 O MW−SW , which were calculated from δ 18 O OT02 and Δ 18 O OT02−SW , respectively (de ned in section 3.1). δ 18 O MW principally refers to cave drip water, which is assumed to be almost the average value of meteoric water δ 18 O. ε 18 O MW−SW refers to the difference between δ 18 O MW and δ 18 O SW ; in other words, the total fractionation in the hydrologic circulation process (from seawater to meteoric water) involving changes in climatic factors including the monsoon intensity, moisture trajectory, and the seasonality of precipitation.

Carbonate clumped isotope (Δ 47 )
The Δ 47 values of 68 layers in OT02 range from 0.670‰ to 0.747‰ (Fig. 4d) (Guo, 2008;Daëron et al., 2011;Kluge and Affek, 2012;Guo and Zhou, 2019) and can generate isotopic offsets (differences between actual and equilibrium values). Some speleothems show a negative correlation between δ 18 O offset and Δ 47 offset due to the KIE (Daëron et al., 2011;Wainer et al., 2011;Kluge and Affek, 2012;Kluge et al., 2013;. This type of covariation (negative correlation) between δ 18 O and Δ 47 offsets develops when KIE fractionation signi cantly affects δ 18 O and Δ 47 . In such cases, Δ 47 temperature is unusable for paleoclimatic reconstruction. Kato et al. (2021) examined the covariation between Δ 18 O Hiro−1−SW and Δ 47 values of the Hiro-1 stalagmite, which were calculated using the same method as in the present study, to distinguish the in uence from KIE. The positive covariation expected for equilibrium conditions was found in most parts of the Hiro-1 stalagmite, supporting the interpretation that both Δ 47 and δ 18 O Hiro−1 changed mainly as a function of temperature. Several layers of low growth rate and high δ 13 C in Hiro-1 which imply dry conditions and high PCP (Hori et al., 2013), however exhibited features of the strong in uence of KIE disequilibrium: less negative δ 18 O values and small Δ 47 values. In these layers of the Hiro-1 stalagmite, according to our estimate , signi cantly large PCP likely caused small Δ 63 of DIC and small stalagmite Δ 47 , and the KIE was likely further induced by rapid supersaturation and precipitation in the evaporating water lm, in which both δ 18 O and δ 13 C increased.
In the OT02 stalagmite, a weak but positive covariation between Δ 18 O OT02−SW and Δ 47 values was noted throughout (Fig. 5a, r = 0.36, n = 78). This supports the interpretation that temperature change is the principal factor controlling both Δ 47 and Δ 18 O OT02−SW and that the signals (Δ 47 , δ 18 O OT02 , and Δ 18 O OT02−SW ) seem not to be disturbed by strong KIE disequilibrium. The growth rate of the OT02 stalagmite is higher and more stable than that of the Hiro-1 stalagmite. Additionally, subsamples for Δ 47 measurements were collected from clear layers of OT02 excluding opaque/muddy horizons to avoid the effect of KIE as much as possible, as mentioned in section 3.2.2.

Features in the correlation between Δ 47 vs Δ 18 O OT02−SW
The relationship between the two temperature-controlled factors (Δ 47 and δ 18 O C ) is not simple. In the case of the Hiro-1 stalagmite, different regressions of Δ 47 versus Δ 18 O Hiro−1−SW were obtained from three discrete periods, 18.0-16.0, 14.2-12.6, and 7.7-4.9 ka (Fig. 5a, Kato et al., 2021). A especially large difference was noted between regressions in the pre-Holocene and mid-Holocene. Kato et al. (2021) explained this discrepancy in regressions by differences in meteoric water δ 18 O MW (ε 18 O MW−SW ) due to the presence or absence of the SIS (Fig. 1b), which is an important vapor source for Maboroshi Cave in Hiroshima. The SIS is currently shallow (mostly <50 m deep) and was almost exposed during the glacial sea-level low. Because of the absence of this inland sea, moisture was predominantly transported over longer distances from the marine vapor source. Less negative δ 18 O MW from the SIS was established by the Holocene glacial retreat . By contrast, the in uence of the SIS is negligible for Ohtaki Cave in Gifu prefecture (Fig. 1b). This is likely the reason for the uni ed regression between Δ 47 and Δ 18 O OT02−SW for Ohtaki Cave (Fig. 5a).
Comparing to Hiro-1 values, another characteristic in the regression of OT02 Δ 47 vs Δ 18 O OT02−SW becomes apparent (Fig. 5a), which is expressed as follows: Δ 18 O OT02−SW = 5.05 Δ 47 − 11.15 (Eq. 5; n = 78, R 2 = 0.132) With a slope of 5.05, this is clearly shallower than that of the Hiro-1 stalagmite, for which the corresponding value is approximately 40 (37.71-44.15;Kato et al., 2021;Fig. 5a For Hiro-1, Kato et al. (2021) interpreted that a possible process for the shallower slopes was the temperature dependency of meteoric water δ 18 O MW ; thus, 18 O depletion during vapor generation decreases with increasing temperature (e.g., Horita and Wesolowski, 1994;Clark and Fritz, 1997) (7) where T is the temperature in ℃ and FT SW−MW is assumed to be a linear function of temperature. The temperature-dependent coe cient of FT SW−MW is expressed by a (‰/℃). The slope A in Eq. 6 increases with a of Eq. 7 and attains a theoretical slope of Δ 47 versus Δ 18 O stalagmtie−SW (approximately 70) when a takes an appropriate value.
In the Hiro-1 stalagmite, slope A in Eq. 6 attains the theoretical slope when a is at 0.077‰-0.098‰/℃ (black lines in Fig. 5b; Kato et al., 2021). The range of a Hiro−1 is comparable with the temperature dependency of 18 O fractionation between water and vapor, which is 0.094‰-0.102‰/℃ at 10℃, 0.085‰-0.090‰/℃ at 20℃, and 0.076‰-0.079‰/℃ at 30℃ (Horita and Wesolowski, 1994;Clark and Fritz, 1997). Figure 5b also shows the calculation results for the OT02 stalagmite (red line for the Holocene and blue line for latest Pleistocene). Slope A attains the theoretical slope (approximately 70) when a OT02 is at 0.18‰/℃ for both the Holocene and latest Pleistocene, which is clearly larger than a Hiro−1 .

Reason for the larger temperature-dependent coe cient of FT SW−MW (a) in OT02
We assume that the divergence in a values for OT02 and Hiro-1 are ascribed to regional differences in precipitation characteristics. As described in section 2.2, in both Nagataki (near Ohtaki Cave; Fig. 2c) and Yuki (near Maboroshi Cave; Fig. 2d), less negative δ 18 O MW values are observed during the warm season, whereas more negative δ 18 O MW values are observed in the cold season. However, in Nagataki, winter precipitation accounts for a larger proportion of the annual total, and the yearly variation of winter precipitation is far larger than in Yuki ( Fig. 2a and b). Hence, the amount ratio of winter/annual precipitations is an important factor in uencing the annual average δ 18 O MW values at Ohtaki Cave.
The major control of winter precipitation from the Japan Sea side is the strength of the EAWM. Hirose and Fukudome (2006) found a signi cant correlation (R 2 = 0.85) between the average winter precipitation on the Japan Sea side of Honshu and Hokkaido islands in 1907-2006 and the winter monsoon index of Hanawa et al. (1988), which re ects differences in the sea-level pressures between Irkutsk and Nemuro in winter and represents the strength of the EAWM. Hirose and Fukudome (2006) also found that the Japan Sea SST in winter (November-January) exhibited a signi cant correlation with winter precipitation on the Japan Sea side (R 2 = 0.82). This relation occurs because the high SST of the Japan Sea increases evaporation from seawater, which is the supply source of winter snow/rainfall to Japan. In the Northwest Paci c and East Asia, the land-sea thermal contrast is a driving force behind both the EASM and EAWM. As described above, the proportion of winter/summer precipitation is more variable in the Ohtaki region and the proportion is largely dependent on winter precipitation. Cooling of the land area results in increased winter precipitation and decreased annual average δ 18 O MW values in Ohtaki. This is likely the reason for the apparently larger temperature dependency of meteoric water δ 18 O MW (a; ‰/℃) in the Ohtaki region.

Terrestrial paleotemperature
We interpret temperature to be the dominant control on the OT02 Δ 47 record, on the basis of the discussion presented in section 5.1. The Holocene Δ 47 temperature appears consistent with the present cave temperature of 13℃, although the growth of OT02 was suspended in 2.6 ka and present Δ 47 data are absent. The Δ 47 temperature record from OT02 (Table 1) is also broadly consistent with known climatic stages, such as cooling during Heinrich events, warming during the Holocene, and a warming peak around the Hypsithermal event (9 to 6-5 ka; e.g., Wanner et al., 2008;Fig. 6). In the period common to the OT02 and Hiro-1 stalagmites (7.7-4.5 ka), the two stalagmites exhibit very similar patterns in Δ 47 temperature ( Fig. 4 and 6); a rapid warming in 7.0-6.0 ka after a temporal cooling of ~5℃ at 7.0 ka from around 15℃ at 8.0-7.5 ka and a gradual cooling to the cessation of stalagmite growth. Three intervals of distinct cooling by 3-5℃ are likely correlated to Heinrich events H6, H5.2, and H5 (Fig. 6). Although the averaged Δ 47 temperature of the latest Pleistocene portion of OT02 (63-35 ka) is lower than that of the Holocene portion, the Δ 47 temperature in warm periods between each Heinrich event reached 10℃-15℃, which is as high as the present temperature and the average of the middle Holocene.
Our Δ 47 temperature equation (Eq. 3) was calibrated using the Δ 47 values of natural tufa deposited at 5.6℃-16.0℃. The high Δ 47 temperatures over 20℃ around 5.4-2.9 ka therefore somewhat deviate from the temperature range covered by Eq. 3. Because of the nature of carbonate clumped isotope thermometry, the uncertainty of Δ 47 temperatures is greater at higher temperatures, because the Δ 47 value exhibits an inverse proportionality to the square of the temperature in Kelvins. Even after considering these larger errors and uncertainties, the average temperature during the period from 5.4 to 3.8 ka (considering the larger dating error for OT02 mentioned above, which likely corresponds to the warm maximum of 6.3-4.9 ka recorded in Hiro-1) is 19.9℃ ± 6.0℃ and is clearly higher than the present cave temperature of 13℃.
Seasonal temperature variation is currently very limited in the deeper part of Ohtaki Cave, and it is thus likely that the past cave temperature was also stable on seasonal timescales. We do not presume, however, that the high-average Δ 47 temperature during the Hypsithermal means uniform warming of all seasons. In Nagataki (near Ohtaki Cave), the temperature variation tends to be largest in the colder season (December-April) and smaller in the warmer season (May-October; Fig. 3a). The correlation coe cient (R) of annual and monthly temperatures is large in February-May and September-October and smallest in the warmest months (July-August; Fig. 3a). Together, this indicates that summer temperature is not an important control in the determination of annual average temperatures. Variability in annual average temperature is largely dependent on temperature variability in spring and autumn and is also somewhat dependent on winter temperature. This indicates that the annual average temperature in Nagataki is characterized by the length of the summer (warmer interval) and the winter (colder interval). We presume that the higher cave temperature of 19.9℃ ± 6.0℃ observed in the 5.4-3.8 ka period of OT02 (corresponding 6.3-4.9 ka of Hiro-1) was induced by longer high summers and shorter winters than at present, accompanied by the warm climate optimum of the Hypsithermal event.

Paleoprecipitation history
In section 5.4, we explained changes in cave temperature arising from interactions between the summer and winter seasons. In Japan, summer and winter climates are characterized by the EASM and EAWM winds, respectively, where rainfall from EASM winds has a less negative δ 18 O MW value and rain/snowfall from EAWM winds has a more negative δ 18 O MW value. Changes in summer and winter durations might therefore affect the precipitation balance from EASM and EAWM winds and the annual average of δ 18 O MW value. Based on this assumption, paleometeoric water should have a more negative δ 18 O MW value in colder periods and a less negative δ 18 O MW value in warmer periods. However, the relation between EASM and EAWM is not simple. Previous paleoclimatic studies have indicated numerous relationships between the intensities of EASM and EAWM: an inverse correlation (Xiao et al., 1995;Yancheva et al., 2007;Liu et al., 2009), a positive correlation (Zhang and Lu, 2007), and both positive and inverse correlations depending on the period and timescale concerned (Steinke et al., 2011;Ge et al., 2017). Recently, Yan et al. (2020) presented simulation results to investigate the relationship between time series changes in EASM and EAWM, showing that their intensities are positively correlated at orbital timescale due to seasonal insolation forcing but are negatively correlated over multidecadal to millennial timescales, primarily as a result of internal variability in the Atlantic Meridional Overturning Circulation and its subsequent teleconnection to East Asia via land-sea thermal contrasts.
Focusing on climatic changes at the centennial-millennial timescale, i.e., Heinrich cooling events and Hypsithermal warming, the result of our meteoric water δ 18 O MW reconstruction is consistent with the assumption of a more negative (less negative) δ 18 O MW value in colder (warmer) periods (Fig. 6). The ε 18 O MW−SW value (and accordingly the δ 18 O MW value) was more negative in colder periods such as Heinrich cooling events and the period of cooling around 7 ka and less negative in warmer periods. This is consistent with the assumption in section 5.4 that the longer summer durations and larger amounts of EASM rainfall caused the average δ 18 O MW value in the Hypsithermal warm period. By contrast, it is presumed that EAWM brought a larger amount of snow and rainfall during long winters to Ohtaki Cave during Heinrich cooling events. Very few previous studies have reported terrestrial climatic records in Heinrich events from Japan. Nakamura et al. (2013) performed stratigraphic analyses of acoustic records of sediments in Lake Nojiri, the Japan Sea side of central Japan (Fig. 1b) where winter snowfall accounts for approximately 30% of annual precipitation. They revealed lake level uctuation during the past 45,000 years and found eight sets of rising and falling levels, although their results contain large dating errors of ±1,000-15,000 years. They also found that peaks of lake level corresponded to cold climate stages such as Heinrich cooling events 1-4 and the Younger Dryas. They presumed that lake level rises were caused by increased snowfall from an enhanced winter monsoon (EAWM). Light blue vertical bands in Fig. 6 indicate the periods of peak lake level in Nojiri identi ed by Nakamura et al. (2013). Unfortunately, the record of Nakamura et al. (2013) does not include older Heinrich events H5-6, for which particularly negative ε 18 O MW−SW values were reconstructed from OT02 ( Fig. 6). However, the durations of high lake level periods in Nojiri seem to correspond to the periods of colder temperature and more negative ε 18 O MW−SW values. Increased precipitation from EAWM was likely also bought into the Ohtaki region and caused a decrease in the annual average of ε 18 O MW−SW during these cold stages.
As described here, changes in the amount ratio of summer and winter precipitations from EASM and EAWM caused the divergence in ε 18 O MW−SW values in the Ohtaki region, although the temperature dependency of the fractionation between seawater and meteoric water (as discussed in section 5.3) also accounts for the relationship between temperature and ε 18 O MW−SW .
As described in section 2.2, precipitation from EAWM is minor in the region of Maboroshi Cave. We suggest that a strong EAWM and weak EASM during cold periods caused the dry conditions of Maboroshi Cave, which might have resulted in the interruptions of the growth of Hiro-1 and KIE disturbance of Hiro-1 records (Fig. 6).

Summary
We analyzed carbonate clumped isotopes (Δ 47 ) of the OT02 stalagmite from Ohtaki Cave for the age intervals of 2.6-8.8 and 34.8-63.5 ka and revealed changes in terrestrial temperature and meteoric δ 18 O. The average Δ 47 temperature of the Holocene portion (16.3℃, 8.8-2.6 ka) is 6.6℃ higher than that of the latest Pleistocene portion (9.7℃, 63.5-34.8 ka). Reductions in Δ 47 temperature correspond to Heinrich cooling events and are approximately 3℃-5℃. We presume that higher cave temperatures of 19.9℃ ± 6.0℃ observed during the middle Holocene (5.4-3.8 ka of OT02) were induced by longer high summers and shorter winters than at present, accompanied by the Hypsithermal warm climate optimum.
We also reconstructed two indices of meteoric oxygen isotopes, δ 18  Hence, we revealed a trend of more/less negative meteoric water δ 18 O MW in warmer/colder climate stages. This trend is the opposite of that assumed in conventional paleoclimate studies using Japanese stalagmites, which suggest that meteoric δ 18 O MW becomes more/less negative in warm-humid/cold-dry climates due to the "amount effect." Our results also do not follow previous interpretations that variation in meteoric δ 18 O MW values is the dominant controlling factor of stalagmite δ 18 O value. In our study regions, major factors determining the average δ 18 O MW value are 1) the proportion between summer and winter precipitation and 2) the temperature dependency of the fractionation from seawater to meteoric water. Additionally, 3) seawater δ 18 O SW change (Δ 18 O SW ) at orbital scales also in uences δ 18 O MW . These effects occur due to the hydrological setting in Japan in which moisture is brought from surrounding seas. Factor 1 is related to climatic stages on centennial scales, which is also deeply related to terrestrial temperature. Besides these three factors controlling δ 18 O MW , 4) temperature-dependent fractionation at the time of stalagmite deposition (FT water−stalagmite ) is an important control on stalagmite δ 18 O C . These temperature effects on δ 18 O MW and δ 18 O C are, however, in opposite directions and the negative in uence of temperature on δ 18 O C (FT water−stalagmite ) exceeds the information available for past δ 18 O MW . This explains the small amplitude of δ 18 O change in Japanese stalagmites (Mori et al., 2018;Kato et al., 2021) and has complicated the interpretation of stalagmite δ 18 O C records using only conventional methods of stalagmite climatology depending only on δ 18 O analysis.

Declarations
Availability of data and material Table 1 contains all data presented in this study.

Competing interests
The authors declare that they have no competing interest.

Funding
This study was supported by Grants-in-Aid from the Japan Society for the Promotion of Science

Authors' contributions
HK proposed the topic, conceived and designed the study. HK, TM, AK, CCW, and CCS carried out the experimental work. HK and AK wrote the manuscript of this study. AK and CCS supervised the study.The authors read and approved the nal manuscript.