1 Introduction

High-resolution isotopic records of well-dated stalagmites have been used as paleoclimatic archives in a terrestrial setting (e.g., Cheng et al. 2016). More specifically, the oxygen isotope (δ18O) records of stalagmites can provide important insight into the precipitation dynamics of the Late Pleistocene and Holocene epochs and have become a standard proxy for terrestrial climatic changes at the global scale. As a prominent example, Chinese cave records demonstrate that changes in stalagmite calcite δ18O (δ18OC) are strongly correlated with δ18O records from Greenland ice cores (Wang et al. 2001) and shifts in Northern Hemisphere summer insolation on orbital timescales (Cheng et al. 2016; Wang et al. 2008).

The link between climate and δ18OC has been mostly attributed to variability in the δ18O values of local meteoric water. A popular interpretation is that this results from the amount effect on meteoric water δ18O (δ18OW) from the intensified East Asian summer monsoon (EASM), which controls the climate in Asia (Cheng et al. 2009; Wang et al. 2001, 2008). Furthermore, δ18OC is seemingly synchronized with climatic changes in the North Atlantic Ocean during the late Quaternary (Cheng et al. 2016; Sun et al. 2012; Wang et al. 2001, 2008; Zhao et al. 2018). Regarding the warming transition associated with deglaciation from the last glacial period to the mid-Holocene, the stalagmites generally exhibit distinct reductions in δ18OC values, again ascribed to the intensification of the EASM.

The East Asian winter monsoon (EAWM) is another meteorological system in East Asia (Fig. 1a). The EAWM is driven by the thermal contrast between the Asian continent and the North Pacific during winter, which blows dry and cold northerly winds from the Siberian High (Tada et al. 2016). The EAWM affects the climate of East Asia (Porter and An 1995) and transports heat from the Northern Hemisphere to the Southern Hemisphere as it crosses the equator (Chu et al. 2017; Yamamoto et al. 2013). EAWM variability during the late Quaternary has been reconstructed using marine sediment cores from the Sulu Sea (de Garidel-Thoron et al. 2001), South China Sea (Huang et al. 2011; Steinke et al. 2010, 2011; Yamamoto et al. 2013), Northwestern Pacific Ocean (Sagawa et al. 2011), and Japan Sea (Nagashima et al. 2007, 2011), as well as from terrestrial archives (e.g., aeolian sediments from the northwestern Chinese Loess Plateau (Sun et al. 2012) and lake sediments in south China (Chu et al. 2017; Yancheva et al. 2007). The δ18OC rarely reflects the intensity of the EAWM because meteoric water from a dry EAWM makes only a small contribution to total precipitation in East Asia (e.g., An 2000). Exceptional findings were reported in Fukugaguchi Cave at Itoigawa in central Japan, on the coast of the Japan Sea (Sone et al. 2013, 2015), where water vapor from the Japan Sea warmed by the Tsushima Warm Current is carried by a northwesterly of the EAWM from Siberia (Fig. 1b; Hirose and Fukudome 2006). This area is wet in winter, and the snow/rainfall of the winter months (December–February) constitutes ~ 40% of annual rainfall (Fig. 1c). Considering that the EAWM usually starts in November and ends in March, the EAWM snow/rainfall contributes a larger proportion (it falls 57% of annual rainfall during five months from November to March). Sone et al. (2013) analyzed a Holocene stalagmite (i.e., FG01) from Fukugaguchi Cave (Fig. 1d) and found that δ18OC variability at the top of FG01 was strongly correlated with that of winter precipitation observed from 1924 to 2010 at Takada (Fig. 1d). On the other hand, the δ18OC variability did not show relevance to precipitation of other seasons (Sone et al. 2013 in their Fig. 5). The amount effect apparent in δ18O values of winter precipitation (collected from Toyama, Fig. 1d) also implied that the δ18OC reflected winter precipitation variability. In addition, the δ18OC profile of the entire FG01, covering the past 10 kyr, demonstrated a trend similar to that of EAWM variability reported by other studies on loess and lake sediments (Sone et al. 2013). However, the variability of the EAWM beyond the pre-Holocene period has never been demonstrated based on a stalagmite record. Such data may provide insight regarding the hydroclimate during glacial periods.

Fig. 1
figure 1

Geographic and climatological setting of the study site, Itoigawa. a Map showing the general wind directions in the winter, the southern limit of the EAWM, and the locations of the Itoigawa and Chinese caves. b Warm and cold currents around Japan. The Tsushima Warm Current enters the Japan Sea through the Tsushima strait and outflows to the Northwestern Pacific Ocean through the Tsugaru and Soya straits. c Monthly mean temperature (°C) at Itoigawa (upper) and rainfall (mm) at the Hiraiwa Precipitation Observatory (lower) during the period 1990–2009. d Locations of caves (Maboroshi, Ohtaki, and Kiriana), lakes (Suigetsu and Biwa), a coring site (U1427/PC3), and observatory sites (Toyama and Takada) discussed in this study

Full size image

Here, we present a new δ18OC profile spanning the last glacial period from 32.2 to 22.3 ka, obtained from another stalagmite (FG02) in Fukugaguchi Cave. A drastic shift in oceanographic conditions occurred during the last glacial period in the Japan Sea, the major moisture source in the study area. The Japan Sea is currently connected to the open ocean via four shallow straits (Fig. 1b). However, during glacial periods, the lowered sea level largely restricted entry of the Tsushima Warm Current. Freshwater influx from the surrounding land area reduced surface water salinity in the Japan Sea, as indicated by unusually low δ18O in foraminifers (Oba et al. 1991; Sagawa et al. 2018). Closing and opening of the Japan Sea led to glacial–interglacial contrast in the compositions and structures of core sediments recovered from the Japan Sea, which are generally laminated during glacial and bioturbated during interglacials (e.g., Irino et al. 2018; Tada et al. 2018; Seki et al. 2019). The closure of the Japan Sea may also have affected the hydrodynamics of vapor generation during glacial periods. Such changes, as well as the intensity of the EAWM and the EASM, may lead to differences in the δ18OC in FG02 from other cave records in Japan (Maboroshi, Ohtaki, and Kiriana; Fig. 1d), where precipitation from the EAWM is minor. Moreover, the δ18OC profile of FG02 exhibits millennial-scale changes that might correspond to climatic changes in the North Pacific Ocean. In this paper, the δ18OC records from Fukugaguchi Cave are discussed as new stalagmite evidence for the development of 18O-depleted surface water in the Japan Sea during the last glacial period.

2 Materials and methods

2.1 Study site and stalagmite sample

Fukugaguchi Cave (36° 96.5′ N, 137° 80.0′ E) is located in Itoigawa City, Niigata Prefecture, on the coasts of Japanese islands along the Japan Sea (Fig. 1). Itoigawa currently experiences some of the heaviest snowfall in Japan because of its geographic and altitudinal position, i.e., on a steep slope behind high mountains (e.g., Mt. Asahidake, 2418 m asl). A northwesterly of the EAWM reaches this location across the widest breadth of the Japan Sea, where the Tsushima Warm Current supplies water vapor to the initially dry northwesterly (Fig. 1a, b). The wet air mass then interacts with the mountain slope, causing heavy snowfalls. Therefore, the 5-month period (November–March) under the influence of the EAWM is generally wet in this area, accounting for 59% of the annual precipitation in Itoigawa (Sone et al. 2013).

Our sample analyzed in this study was a 22-cm-long stalagmite (FG02; Fig. 2a) collected from Fukugaguchi Cave in 2010. The sampling point was approximately 700 m from the cave entrance and near the location of the Holocene sample (FG01; Sone et al. 2013, 2015), where relative humidity (RH) is nearly 100%. FG02 has two discontinuous surfaces at 13.0 and 198.0 mm from the top (hiatus in Fig. 2a). The stalagmite lacks the regular laminae (Fig. 2b–e) that may constitute annual bands. It mainly consists of transparent calcite mass (Fig. 2b, d), although some distal parts of the stalagmite exhibit mm-scale layering (Fig. 2c, e).

Fig. 2
figure 2

Textures and age model of FG02. a The half-cut specimen of FG-02 showing sampling transect of isotopic analysis and positions of thin section image. Two hiatus horizons are indicated by dotted lines. b Transparent calcite at 3.5 cm from the top. c Millimeter-scale layering at 6.5 cm from the top. d Transparent calcite at 9.7 cm from the top. e Vague-layered structure at 13.5 cm from the top. f Age model: the length of the stalagmite, starting from the top, is plotted against age. Measured ages are presented with error bars of 2σ. The black line represents the smoothed age data, while the gray-shaded region represents the error range calculated using StalAge (Scholz and Hoffmann 2011)

Full size image

2.2 U–Th age measurement

The ages of FG02 were determined by U–Th dating carried out at the National Taiwan University. The methods were described in detail by Shen et al. (2002, 2003, 2012). The stalagmite was cut along its growth axis for polishing and then drilled at nine horizons along the growth lines. Nine 0.2–0.3-g subsamples were obtained for each analysis. Because of the low uranium concentration in FG02, the amount of each sample was more than twofold greater than the amount used in ordinary U–Th dating (approximately 0.1 g; Shen et al. 2012). These powdered samples were dissolved with 5% nitric acid and spiked with an artificial radiometric tracer (229Th–233U–236U). An Fe3+ solution was added to this solution to remove Ca2+ by iron co-precipitation. U and Th were purified by anion exchange chromatography. The isotopic signature of each purified fraction was measured with a multicollector inductively coupled plasma mass spectrometer (Neptune; Thermo). An age–depth model was constructed using StalAge software, by means of a statistical algorithm based on Bayesian Monte Carlo simulation (Scholz and Hoffmann 2011).

2.3 Stable isotope analysis

Stable isotope analysis was performed with an isotope ratio mass spectrometer (DeltaPlus; Thermo Finnigan) connected to an online gas separation and introduction system (GasBench II) at Kyushu University. Subsampling was conducted down the middle of the growth band along the growth axis with a dental microdrill (Tas-35LX; Shofu) at 0.2-mm intervals. In addition, a Hendy test to examine the stability of δ18O values along specific growth bands (Hendy 1971) was performed for 8–10 subsamples at three horizons. Each ~ 0.15-mg sample was enclosed in a 12-mL vial. Following replacement with He gas, these subsamples were reacted with phosphoric acid for > 5 h in a 50 °C thermostat chamber. Generated CO2 was introduced to the analysis system. δ18O values were normalized by using an in-house standard, which corresponded to the Vienna Pee Dee Belemnite standard. The reproducibility of the measurements of in-house standard (N = 190) was ±0.14‰ (2 SD). A typical measuring error is ±0.2‰ (2 σ). Additional details of δ18O measurement were described by Hori et al. (2009).

2.4 Evaluation of the amount effect of winter precipitation

For the Holocene stalagmite (FG01), Sone et al. (2013) concluded that the variability of δ18OC largely reflected the intensity of winter precipitation, consistent with the negative correlation between δ18OW and the amount of rain containing meteoric water during winter. We quantified this amount effect by using a bootstrap method and δ18OW data collected at Toyama, approximately 60 km southwest of the cave (Fig. 1d). The δ18OW data included 78 meteoric rain/snow events in winter months (December–February) from 2010 to 2012, which were collected by Sone et al. (2013). A virtual set of winter precipitation data consisted of a given number (N) of rainfall events, which were randomly selected from among the 78 events. First, the range of N was set at 24–30, which reproduced the distribution of total winter precipitation for the last 50 years at Toyama. Then, the total amount and weighted mean of δ18OW were calculated for each virtual set. For each of seven cases of N (24–30), the selection of virtual sets was repeated 100 times, such that the total number in the virtual set was 700.

3 Results

3.1 Dating result and age–depth model

U–Th dating for FG02 yielded suitable ages from nine horizons, which ranged from ca. 13 to ca. 29 ka (Table 1). Although each age from the nine horizons exhibited relatively large uncertainty, due to the low uranium concentration (typically 6.0 ppb) and relatively high 232Th (e.g., 3820 ppt at 9.5 mm horizon), they were generally in the correct stratigraphic order (Table 1). A distinctly younger age (ca. 13 ka) was obtained from the uppermost dated horizon (10 mm from the top; Table 1, and above the upper hiatus; Fig. 2a). An approximately 10-kyr age difference between this horizon and the next dated horizon (ca. 23 ka) at 30 mm likely signified a hiatus at an upper discontinuous surface (13.0 mm). Another hiatus was suspected at a lower discontinuous surface (198.0 mm), although no reliable age was obtained below this surface. The age model generated with StalAge (Scholz and Hoffmann 2011) indicated that the middle section between the two hiatuses was formed between ca. 32.2 ka and ca. 22.3 ka. The age–depth relation shown in Fig. 2 implies that the growth rate was fast above 130 mm (approximately 45 mm/kyr for 22–24 ka) and slow below 130 mm (approximately 7 mm/kyr for 24–32 ka).

Table 1 Uranium and thorium isotopic compositions and 230Th ages (before 1950 AD) for subsamples of stalagmite FG02 by MC-ICP-MS
Full size table

3.2 Stable isotope analysis

The δ18OC values of the middle section of FG02 ranged from − 7.48‰ to − 10.68‰, with a mean value of − 8.87‰. This was clearly lower than the range observed in Holocene FG01 (mainly − 8.5‰ to − 7.0‰; Sone et al. 2013, 2015). The δ18OC values of FG02 indicated millennial-scale changes resembling the changes in the paleoclimatic records of the last glacial period. From 32.2 ka (i.e., the bottom) to ca. 26 ka, the δ18OC variability showed sawtooth-shaped fluctuation on a millennial scale (Fig. 3). The fluctuation amplitude was 0.5‰–1.0‰, clearly greater than the δ18OC measurement deviation (± 0.14‰). The upper δ18OC profile in the interval from 26 to 24 ka was nearly flat around − 9‰; this was followed by a positive shift to − 8‰ at 23.1 ka, which rapidly recovered at ca. 22.8 ka (Fig. 3).

Fig. 3
figure 3

δ18OC age profile of FG02. Note the inverted vertical axis. Dating results are shown at the top; open circles for the corrected 230Th ages (Table 1), solid circles, and bars for the smoothed ages and the error ranges (Fig. 2), respectively

Full size image

The results of Hendy tests conducted at 26-, 60-, and 120-mm horizons are shown in Fig. 4a. The mean value was − 8.43‰ at 26 mm, − 8.79‰ at 60 mm, and − 9.28‰ at 120 mm. The three horizons exhibited perturbation within ±0.25‰, which was slightly larger than the measurement error of δ18OC (2σ < ± 0.2‰; Fig. 4a). The covariance between δ18OC and δ13C was not confirmed (R = 0.04; Fig. 4b), presumably supporting the absence of significant isotopic non-equilibrium.

Fig. 4
figure 4

Isotopic behaviors of FG02. a Hendy test results for δ18OC at three growth lines, 26 mm (red), 60 mm (black), and 120 mm (blue) from the top. Subsamples were collected at 1-mm intervals from the central growth axis. b Covariance between δ18O and δ13C

Full size image

3.3 Relationship between δ18O and winter precipitation at Toyama

A weak negative correlation between the total winter precipitation and weighted mean of δ18O was observed upon analysis of 78 meteoric water samples during winter months (December–February) at Toyama (Fig. 5) (Sone et al. 2013). In the past 50 years (1969–2018), 395.5–994.5 mm of rain and snow (mean, 696.4 ± 135 mm) fell each winter (Fig. 5a). First, we reproduced this observed distribution of total winter precipitation by random selection of a range of N (number of rain/snow events) from among the 78 actual events. The closest distribution was obtained with the range of N from 24 to 30 (mean, 687.9 ± 132 mm) (Fig. 5a). The calculation was repeated 100 times for each case of N (24–30; 7 cases), and the 700 virtual sets of winter precipitation data yielded weighted δ18OC values ranging from − 10.74‰ to − 7.81‰ (mean, − 9.34‰ ± 0.49‰). The precipitation amount and weighted mean (Fig. 5b) indicated a weak but significant negative correlation (R = − 0.33, p = 6 × 10–19), with a slope of − 1.2‰/1,000 mm. We use this slope as the amount effect of modern winter precipitation in the following section.

Fig. 5
figure 5

Model calculation of the amount effect of winter precipitation based on the δ18OW data of 78 meteoric water samples collected during winter months (December to February) at Toyama (Sone et al. 2013). a Comparison of the frequency of winter precipitation based on 50 years of data from the Toyama weather observatory (1969 to 2018; blue bar) and 700 virtual model calculation sets (red plots). Resampling 24–30 rain/snow events best reproduced the observed frequency (mean and standard deviation) of cumulative winter precipitation. b Cross plot of winter precipitation and δ18OW in meteoric water. A weak but clear correlation is evident between these two variances

Full size image

4 Discussion

4.1 Millennial-scale changes in FG02

The FG02 stalagmite recorded millennial-scale changes according to the characteristic features in the paleoclimatic records during the last glacial period. Figure 6 compares the FG02 δ18OC with the δ18O and Ca2+ concentration of a Greenland ice core (Rasmussen et al. 2014; Seierstad et al. 2014), as well as a reconstructed EAWM derived from grain size variation in the Chinese Loess Plateau (Sun et al. 2012), and a reconstructed EASM derived from stalagmites in Sanbao and Hulu Caves in China (Wang et al. 2001) and stalagmites in Kiriana Cave on the Pacific Ocean side of Japan (Mori et al. 2018). The FG02 stalagmite shows millennial-scale changes in δ18OC. Because the low uranium content of FG02 yields U–Th ages with relatively large uncertainty (Table 1), two possible relationships exist between the δ18OC and Dansgaard–Oeschger (D–O) events (cases 1 and 2; Fig. 6). In order to determine which case is true, stalagmite samples with small uncertainty of U-Th ages are needed.

Fig. 6
figure 6

Comparison of FG02 δ18O data and climate interpretation with other paleoclimatic studies in the period 20–25 ka. a Greenland ice core δ18O (Seierstad et al. 2014) and b Ca2+ (Rasmussen et al. 2014). c FG02 δ18O data (this study). Climate interpretation is according to case 1. d Grain size variation of quartz in loess sediments from the Chinese Loess Plateau (Sun et al. 2012). e δ18O variability in Hulu Cave (south China; Wang et al. 2001). f δ18O variability in Kiriana Cave (Mie Prefecture, Japan; Mori et al. 2018)

Full size image

In case 1, the δ18OC values of FG02 are presumed to be heavier during the D–O interstadials. Three positive δ18OC excursions, at 23.1, 27.3, and 28.7 ka, are correlated with D–O events 2, 3, and 4, respectively (Fig. 6c). Another positive excursion around 30.8 ka is also recognized in the Greenland ice core and Chinese loess (Fig. 6) and could correspond to the Greenland Interstadial 5.1 (at 30.84 ka) described by Rasmussen et al. (2014). Furthermore, Sone et al. (2013, 2015) concluded that the variability of δ18OC in Holocene FG01 reflects the amount of winter precipitation. According to this interpretation, the millennial-scale positive excursions of δ18OC in FG02 (with an amplitude of 0.5‰–1‰) should have been associated with less winter precipitation and the EAWM during short-term interstadials in Greenland and the North Atlantic Ocean. There might be a linkage between the D–O events and the weakened EAWM, which enriched 18O of meteoric water and FG02.

The δ18OC values of FG02 are presumed to be lighter during the D–O interstadials in case 2. In this scenario, four low-δ18OC intervals at 22.4, 27.6, 29.0, and 31.4 ka correspond to D–O events 2, 3, 4, and 5.1, respectively (Fig. 6c). This scenario fully accords with the stalagmites from China (e.g., Wang et al. 2008), whereby the amount effect due to the intensified EASM reduced δ18OC values during D–O interstadials. However, this is inconsistent with the findings of Sone et al. (2013), suggesting that the variability of the EASM did not strongly affect the stalagmite δ18OC. In addition, the δ18OC reduction during these D–O interstadials is unclear even in Kiriana Cave on the Pacific Ocean side of the Japanese islands (Fig. 6f; Mori et al. 2018), where the majority of meteoric water originated from the EASM.

The relationship in case 1 is likely to be more plausible than the relationship in case 2; the δ18OC values of FG02 increased during the D–O events, which contrasts with the findings in Chinese caves. Stalagmite FG02 exhibits an overall decreasing trend of δ18OC (Fig. 6c), which also contrasts with the Chinese stalagmites (Fig. 6e) and findings in Kiriana Cave (Fig. 6f). The case 1 scenario is consistent with the overall trend of δ18OC, as well as the conclusions of Sone et al. (2013, 2015), whereby the variability of δ18OC reflects the amount of winter precipitation. However, considering the calculated amount effect of the modern meteoric water (− 1.2‰/1000 mm; Fig. 5), a 0.5‰ to 1‰ enrichment of 18OC was only partly compensated by the weakened winter precipitation. A more robust amount effect during the glacial period could be the cause of high δ18OC in FG02 during the D–O events. Alternatively, high δ18OC values during the D–O events were also associated with other factors, such as the evaporation of water in the soil or cave, which increases both δ18OW and δ18OC. Our data do not rule out an evaporation effect during water infiltration from the soil. However, any such evaporation effect in the cave was presumably unimportant at the sampling site deep in the cave (700 m from the entrance), because it only appears to be robust where RH is unstable (Deininger et al. 2012).

4.2 Factors responsible for low δ18OC in glacial FG02

Figure 7 compares the δ18OC records of the last glacial period (FG02) and the Holocene (FG01) (Sone et al. 2013) with those from three other Japanese stalagmites in Maboroshi Cave in Hiroshima Prefecture (Hiro-1; Hori et al. 2014; Shen et al. 2010; Kato et al. 2021), and Kiriana and Ohtaki Caves in Mie and Gifu Prefectures, respectively (KA03 and OT02; Mori et al. 2018) (see Fig. 1d for locations). The stalagmite records from Fukugaguchi Cave indicate lower δ18OC values during the last glacial period (FG02) compared with Holocene values (FG01; Sone et al. 2013). This is a unique feature of Fukugaguchi Cave; such variability is not seen in other Japanese and Chinese caves where the EASM is the major source of meteoric water. The Hiro-1, KA03, and OT02 stalagmites exhibit similar δ18OC variability, with higher values during the last glacial period and lower values in the Holocene, consistent with the Chinese records (e.g., Sanbao and Hulu Caves; Wang et al. 2001, 2008). Comparing our results from Fukugaguchi Cave with the KA03 records from Kiriana Cave that span the entire FG02, a substantial difference (> 3‰) was apparent in the mid-Holocene value minus the glacial period value (Δ18OH-G; + 1.23‰ in Fukugaguchi and − 1.81‰ in Kiriana Cave; Table 2). The negative Δ18OH-G was used to explain the weakened EASM during the glacial period in Chinese records (e.g., Wang et al. 2001). Mori et al. (2018) challenged the notion of an effect of EASM intensity on δ18OC, suggesting that the variability of δ18OC was largely controlled by air temperature and δ18O in seawater, which constituted the major moisture source on the Pacific Ocean side of the Japanese islands. Thus, the Δ18OH-G value of − 1.81‰ at Kiriana Cave can be explained by the lower glacial period temperature that caused an increase in fractionation δ18O and by 18O-enriched glacial period seawater that caused an increase in moisture δ18O. The importance of the temperature effect on the stalagmite δ18OC was also demonstrated for Hiro-1 in Maboroshi Cave by co-variation of δ18OC and carbonate clumped isotope (Δ47; Kato et al. 2021).

Fig. 7
figure 7

δ18O variability in Japanese caves during the Holocene and last glacial period. a Fukugaguchi Cave in Niigata Prefecture. The Holocene profile is based on FG01 data (Sone et al. 2013). b Maboroshi Cave in Hiroshima Prefecture (Hori et al. 2014; Shen et al. 2010). c Ohtaki Cave in Gifu Prefecture (Mori et al. 2018). d Kiriana Cave in Mie Prefecture (Mori et al. 2018)

Full size image
Table 2 Mean δ18O (vs VPDB) during mid-Holocene and the last glacial period recorded in Japanese and Chinese caves
Full size table

The EASM intensity, air temperature, and seawater δ18O from the Pacific Ocean are important factors controlling δ18OC in East Asia. However, none of these factors are considered responsible for the positive (+ 1.23‰) Δ18OH-G observed at Fukugaguchi Cave. The EASM is a secondary moisture source for Fukugaguchi Cave (Sone et al. 2013, 2015), and EASM weakening is likely to cause negative Δ18OH-G. Alkenone-based reconstruction studies for nearshore sediments from the Japan Sea suggested that the sea surface temperature during 17–32 ka was 1–2 °C higher than in the mid-Holocene (Fujine et al. 2006, 2009; Ishiwatari et al. 2001; Xing et al. 2011). Ishiwatari et al. (2001) ascribed these enigmatic results in the Japan Sea to the retention of solar energy on the surface water layer that was stratified with heavier, deeper layers. Other studies have suggested that diminished salinity during the last glacial period might have led to a change in alkenone production, resulting in a higher reconstructed temperature (Fujine et al. 2006, 2009). Although this interpretation cannot be ruled out, the air temperature on the land was generally lower during the glacial period than the interglacial period, as indicated by pollen assemblages from Lakes Suigetsu and Biwa (Nakagawa et al. 2006, 2008). Similar to other Japanese cave sites, the lower air temperature during the last glacial period would increase δ18OC, thus making Δ18OH-G negative.

With respect to the Holocene δ18OC variability at Fukugaguchi Cave (FG01), Sone et al. (2013, 2015) considered the intensity of the EAWM and winter precipitation to be dominant factors. The correlation between δ18OC and winter (December–February) precipitation at Takada (Fig. 1d) is evident in high-resolution δ18OC analysis of the uppermost FG01, with a 1‰ difference between the dry (1000 mm/3 months) winters in the late 2000s and wet (1500 mm/3 months) winters in the early 1940s at Itoigawa (Sone et al. 2013). Additionally, our analysis of data from Toyama, which is a drier area, revealed a negative correlation of δ18OW with the amount of precipitation (slope of − 1.2‰/1,000 mm (Fig. 5b)). In this scenario, a stronger EAWM during the last glacial period (Huang et al. 2011; Steinke et al. 2010; Tian et al. 2010) substantially increased winter precipitation (thus decreasing δ18OW) and generated an approximately 3‰ difference in Δ18OH-G between the Japan Sea side (+ 1.23‰ at Fukugaguchi Cave) and the Pacific Ocean side of the islands (− 1.81‰ at Kiriana Cave; Table 2). When the slope of − 1.2‰/1000 mm is regarded as the amount effect of the winter precipitation, the ~ 3‰ difference requires 2500 mm of winter precipitation during the glacial period, which is more than fourfold greater than the present amount of winter precipitation at Toyama. However, such a large amount of winter precipitation is considered unrealistic because the pollen records from Lakes Suigetsu (Nakagawa et al. 2006) and Biwa (Hayashi et al. 2010) suggest drier winter conditions during the same period. Furthermore, Schlolaut et al. (2014) suggested that the winter monsoon on the Japan Sea coast had less moisture during the glacial period because of the limited inflow of the Tsushima Warm Current.

If conditions on the sea surface were drier, the enhanced fractionation from water to vapor could generate lower δ18O in both water vapor and stalagmites (Lachniet 2009). The effect of RH on the δ18O in water vapor has been estimated in several studies; a 10% reduction in RH was posited to decrease the vapor δ18O by 1.3‰ (Gonfiantini 1986) and, in another study, by 0.6‰ (Merlivat and Jouzel 1979). Considering these contrasting estimates, the ~ 3‰ difference corresponds to a 25–50% decrease in RH on the Japan Sea side, whereas the RH has remained stable on the Pacific Ocean side. However, this scenario might reduce winter precipitation, causing the meteoric water at Itoigawa to be dominated by summer precipitation that is more 18O-rich than winter precipitation (Sone et al. 2013). Therefore, the RH effect is unlikely to be equivalent to an approximately 3‰ difference.

A factor likely to be important in the positive Δ18OH-G at Fukugaguchi Cave is the change in δ18Osw on the surface of the Japan Sea (i.e., the dominant vapor source for the cave) (Sone et al. 2013). Currently, δ18OSW is comparable between the Japan Sea and the Pacific Ocean. The Japan Sea is a landlocked marginal sea that connects with the East China Sea and the Pacific Ocean through four narrow straits (Fig. 1b). Because all four straits are shallow, water exchange (e.g., inflow of the Tsushima Warm Current) was largely limited during the last glacial maximum when sea level was reduced by ~ 130 m (Matsui et al. 1998). Under this restricted condition, the surface seawater may have been diluted by riverine water from the Russian mainland and Japanese islands. Sediment cores from the Japan Sea indicate strong stratification during the last glacial period (Oba et al. 1991; Sagawa et al. 2018), which consisted of oxygen-deficient deep water and low-salinity surface water (Tada et al. 1992, 1999). 18O depletion of the Japan Sea surface was first proposed because of the unusually low δ18O in planktonic foraminifera during the last glacial period, recorded in sediment cores recovered from a depth of 935 m at Oki Ridge (Oba et al. 1991) and 800–750 m offshore of Akita (Okumura et al. 1996). By using δ18O records of planktonic foraminifera (Globigerina umbilicate; Oba et al. 1995) and assuming δ18OW in the modern freshwater inflow to the Japan Sea (–7.6‰), Matsui et al. (1998) estimated that the δ18OW in the Japan Sea surface water fell to 20‰ during the last glacial maximum. This drastic decrease in salinity is conceivable if the significant sea-level fall during the last glacial maximum led to reduced seawater inflow through the Tsushima Strait (Matsui et al. 1998). The development of the 18O-depleted water mass during the last glacial period was recently confirmed by higher-resolution data of foraminifera δ18O obtained from marine sediments drilled at a shallower depth (330 m at IODP site U1427/PC3; Sagawa et al. 2018). At this site, both benthic (Uvigerina spp. and Cassidulina spp.) and planktonic foraminifera (Globigerina bulloides) began to exhibit decreasing δ18O from ca. 32 ka to the last glacial maximum, then showed increasing δ18O at ca. 12 ka (Fig. 8; Sagawa et al. 2018). The amplitude of this negative anomaly, compared with the Pacific Ocean standard (Lisiecki and Stern 2016), is approximately 2.5‰ and has been attributed to decreased δ18OSW on the surface of the Japan Sea (Sagawa et al. 2018). Because the surface water of the Japan Sea is the major moisture source for the Fukugaguchi Cave, the anomalous 18O depletion could have reduced the δ18O in meteoric water and stalagmites. Considering the 2.5‰ depletion of δ18OSW at the coring site (330-m depth), greater depletion was suspected at the sea surface. This might be the dominant factor driving the 3‰ difference of Δ18OH-G. The δ18OC record from FG02 represents the first terrestrial data supporting seawater stratification and 18O depletion during the glacial period in the Japan Sea. Furthermore, it implies stronger stratification in an earlier stage of the last glacial period compared with previous findings. The FG02 δ18OC values were significantly lower at 32.2 ka, indicating that the 18O-depleted surface water mass had existed prior to 32.2 ka.

Fig. 8
figure 8

Isotopic evidence of 18O depletion in the Japan Sea during the last glacial period. a δ18O in stalagmites from Fukugaguchi Cave (FG01, Sone et al. 2013; FG02, this study). b Benthic foraminifera δ18O from a coring site at 330 m depth in the Japan Sea (Sagawa et al. 2018) and Northwestern Pacific Ocean (Lisiecki and Stern 2016). Both sets of isotopic data incorporate variability in water δ18OW and temperature

Full size image

The magnitude of the salinity reduction at the sea surface can be estimated using the δ18OW in freshwater entering the Japan Sea from the surrounding land area. There are only two Global Networks of Isotopes in Precipitation sites in coastal areas of the Japan Sea: Pohang in South Korea and Terney in Russia. These sites recorded weighted mean meteoric water δ18O (δ18OMW) values of − 7.78‰ in the period 1961–1976 and − 9.09‰ in the period 1996–2000, respectively (http://www-naweb.iaea.org/napc/ih/IHS_resources_gnip.html). Considering these data and the weighted mean δ18OMW at Toyama (− 9.71‰; Sone et al. 2013), the modern mean δ18OMW entering the Japan Sea is estimated to be approximately − 9‰. Here, we define the proportion of freshwater (fMW) in the Japan Sea surface water mass during the glacial period, as well as isotopic values in freshwater (δ18OMW), seawater (δ18OSW), and the Japan Sea surface (δ18OJS), as shown below.

$$ {\delta}^{18}{O}_{JS}={f}_{MW}{\delta}^{18}{O}_{MW}+\left(1-{f}_{MW}\right){\delta}^{18}{O}_{SW} $$

Assuming that δ18OSW and δ18OMW were + 1‰ (Schrag et al. 2002) and − 9‰, respectively, a 2.5–3.0‰ reduction in the Japan Sea surface water (δ18OJS = − 1.5 to − 2.0‰) indicates fMW of 0.25–0.3. Assuming glacial period ocean water salinity of 35, the Japan Sea surface salinity ranged from 24.5 to 26.3. This is in the lower part of the salinity range (24.0–32.5) estimated for 22–32 ka by Matsui et al. (1998); their estimate was based on δ18O in planktic foraminifers (Globigerina umbilicata), while ours was based on δ18O in the stalagmite. In the closed Japan Sea during the glacial period, the salinity at the sea surface, where vapor is generated, was probably lower than the salinity at the foraminifer habitat depth. In addition to freshwater inflow, sea ice production in the Northern Japan Sea may have contributed to 18O-depletion of the surface water, because frozen water becomes relatively enriched with 18O. Based on the occurrence of dropstones and ice-rafted debris in sediment cores, Ikehara (2003) reconstructed that the sea ice in the northern Japan Sea approached the southern end of Hokkaido during the last glacial maximum. However, ice melting during summer releases 18O-enriched water, and the effect of sea ice development on δ18OSW is not sustainable. Thus, seasonal waxing and waning of sea ice were unlikely to cause a marked reduction in δ18OSW. In addition, oxygen isotopic enrichment from water to ice is only ~ 3‰ (O’Neil 1968). Sustainable freshwater influx of − 9‰ was presumably a major factor in 18O depletion in the Japan Sea surface water.

A similar linkage between marine and δ18OC records was reported in the eastern Mediterranean (Bar-Matthews et al. 2003), where δ18OSW is sensitive to climatic conditions. Two stalagmite records in Israel extending to 250 ka are consistent with the δ18OC in Globigerinoides ruber in the eastern Mediterranean (Fontugne and Calvert 1992). Both marine and stalagmite δ18O showed minimum values during sapropel events (i.e., humid intervals in interglacial period marine isotope stages 5 and 7), which developed under enhanced low-latitude hydrological activity. Currently, the δ18OSW in the eastern Mediterranean is + 1.6‰ (Pierre 1999), although this has been reduced by additional meteoric water during humid periods (Kallel et al. 1997). While Bar-Matthews et al. (2003) partly ascribed the low δ18OC values of the stalagmites to the amount effect observed in the area, the reduction of δ18OSW was also an essential factor involved in the low δ18OC. Moisture source δ18OW is an essential factor governing stalagmite δ18OC.

5 Conclusions

Stalagmite FG02 during the last glacial period (32.3–22.3 ka) in Fukugaguchi Cave on the Japan Sea side of the Japanese islands shows unique δ18OC trends, which have not been previously described in other caves in East Asia. Due to the relatively large uncertainty of U–Th ages, we examined two possible relationships (Fig. 6). The scenario in case 2 contrasts with the findings of a previous Japanese cave study, suggesting that EASM variability did not generate clear δ18OC peaks during D–O interstadials. According to the scenario in case 1, four positive excursions of the δ18OC profile probably correspond to three D–O events and one interval of high δ18O in the Greenland ice sheet (Fig. 6). We considered that case 1 correlation was more likely. An important finding of this study was that the glacial FG02 had distinctively lower δ18OC than the Holocene stalagmite from the same cave (FG01). Its mean value (− 8.87‰) is 1.23‰ lower than the mid-Holocene mean value (4.2–8.2 ka, − 7.64‰; Sone et al. 2013). This feature is unique to the Fukugaguchi Cave, i.e., is inconsistent with the δ18OC records from other Japanese and Chinese caves (Fig. 7). The factor making the largest contribution to reduced δ18OC in FG02 was presumably the development of low-salinity water in the semi-isolated Japan Sea, consistent with foraminifera δ18O findings (Oba et al. 1991; Sagawa et al. 2018). The amount effect of the intensified EAWM was a potential factor in the seduction in δ18OC in FG02. However, our model calculation based on δ18OW at Toyama indicated that the amount effect of winter precipitation (1.2‰/1000 mm; Fig. 5) was insufficient to explain the observed 2.5–3.0‰ depletion. In addition, the generation of water vapor and winter precipitation likely decreased during the last glacial period due to the blocking of the Tsushima Warm Current. Our stalagmite record provides insight regarding 18O depletion and salinity in the Japan Sea surface water during the last glacial period.