Alkenone-sea surface temperatures in the Japan Sea
over the past 36 kyr: warm temperatures at the
last glacial maximum
R. Ishiwatari
a,*, M. Houtatsu
a, H. Okada
baDepartment of Chemistry, Graduate School of Science, Tokyo Metropolitan University, Hachioji, Tokyo 192-03, Japan bDepartment of Earth and Planetary Science, Graduate School of Science, Hokkaido University, Sapporo 060, Japan
Received 4 January 2000; accepted 11 October 2000 (returned to author for revision 27 April 2000)
Abstract
The Japan Sea experienced bottom water anoxia at the last glacial maximum (LGM) since it is surrounded by four shallow straits, the sill depths of which are close to, or shallower than, the drop in sea level (120 m) that occurred then. A distinctive negatived18O excursion of planktonic foraminifera also took place during the LGM. This excursion
has been interpreted from foraminiferal data as recording a drop in the paleosalinity of surface waters on the assumption of a constant low sea surface temperatures between 34 and 11 ka. We present here a pro®le of alkenone-based sea surface temperatures (alkenone-SSTs) over the past 36 kyr. Our results suggest that SSTs during the LGM were much higher than those previously assumed. After considering the factors that might aect estimation of alke-none-SSTs and comparisons of core-top alkealke-none-SSTs values with values for modern seawater we conclude that the higher alkenone-SSTs during the LGM are reliable and reasonable. These warm SSTs were probably caused by radiative equilibrium associated with the development of stable water strati®cation in the Japan Sea during the LGM.
#2001 Elsevier Science Ltd. All rights reserved.
Keywords:Sea surface temperature; Alkenones; Uk0
37index; Paleoceanography; Japan Sea; Marine sediments
1. Introduction
During the late Quaternary, the Japan Sea, which is a semi-isolated marginal sea located between Japan and the Asian continent and connected to the Paci®c Ocean through shallow straits with sill depths of less than 130 m, experienced signi®cant changes in sea-level (e.g. Oba et al., 1991, 1995; Ishiwatari et al., 1994, 1999; Tada et al., 1999). During the last glacial interval, the Japan Sea became almost isolated from the open ocean as sea-level dropped by approximately 120 m at the last glacial maximum (LGM).
Oba and his coworkers discovered a negative excur-sion in oxygen isotopic composition of planktonic for-aminifera at site L-3 (Oki ridge) in the Japan Sea, where thed18O became gradually lighter by2.7%between 33
and 17.5 ka, the latter corresponding to the end of the LGM [19±18 ka (calendar age) is used for the LGM around the Japan Sea (Oba et al., 1995)], and then shifted to heavier values by3.3%until 1.2 ka (Oba et al., 1991;
1995). Oba et al. (1991) assumed that the negatived18O
excursion of planktonic foraminifera during the 33±17.5 ka period was caused only by the decrease in paleosali-nity of surface seawater. They estimated that the paleo-salinity dropped from 33 to 20 during this interval, using a temperature-d18O relationship of foraminifera
(cited in Section 4.1.2.) and assuming that the paleo-SST remained constant at 4.41C, together with
assump-tions about thed18O values of seawater and freshwater for
0146-6380/01/$ - see front matter#2001 Elsevier Science Ltd. All rights reserved. P I I : S 0 1 4 6 - 6 3 8 0 ( 0 0 ) 0 0 1 5 1 - 0
www.elsevier.nl/locate/orggeochem
* Corresponding author at present address: 3-16-11 Takaido-nishi, Suginami, Tokyo 168-0071, Japan. Tel.: +81-3-5841-7707; fax: +81-3-3334-7035.
mixing (Matsui et al., 1998). A paleosalinity of 20 is14 less than the modern level.
However, their conclusion was based on an interpreta-tion of oxygen isotope analysis of planktonic and benthic foraminifera. No paleo-SSTs have been measured to date. Nevertheless, the model of depleted paleosalinity and low-SSTs during the LGM seems accepted in the scienti®c community (e.g. Keigwin and Gorbarenko, 1992; Crusius et al., 1999), although CLIMAP estimates give higher SSTs than4C (16±20C for August and 4-6C for
Feb-ruary) around site L-3 at the LGM (CLIMAP Project Members, 1981). Therefore, it was important to evaluate the scenario of depleted paleosalinity in the Japan Sea during the LGM by measuring paleo-SSTs. We felt that theUK0
37 method might provide a useful contribution for estimating paleo-SSTs in the Japan Sea. In Ishiwatari et al. (1999) we reported alkenone-based SSTs of up to 18C in the Japan Sea at approximately 17.5 ka. The SST
of18C is much higher than SST previously assumed by
Oba et al. (1991) and Matsui et al. (1998). However, we failed to measure alkenone-based SSTs in sediments older than 17.5 ka, because the abundance of alkenones in the sediment core (core L-3: described later) was too low to be measured, and hence we could not discuss the changes in paleo-SSTs during the LGM in the Japan Sea.
Therefore, as an extension of the previous study, this study was initiated to: (i) check the reproducibility of alkenone-SSTs in the Japan Sea by measuring alkenones at another site, (ii) present data on alkenone-derived SSTs of the Japan Sea over the last 36 kyr with the discussion of the reliability of the alkenone-method in the light of pre-vious studies, and (iii) present some new ideas on the history of oceanographic conditions of the Japan Sea.
2. Materials and methods
2.1. Sediment samples
Two sediment cores Ð GH93, KI-5 (core length 302 cm; core bottom age 18 ka; ``KI-5'' is used hereafter) which was taken at 39 34.070N, 139 26.280E (water
depth 754 m) in 1993, and KT94-15, PC-9 (core length 500 cm; core bottom age 38 ka; ``PC-9'' is used here-after) which was taken at 3934.360N, 13924.410E (o
Akita city: water depth 807 m) in 1994 Ð were used for this study. Analytical data from a sediment core KH-79-3, L-3 (core length 235 cm) taken from the top of the Oki Ridge (3703.50N, 13442.60E; water depth 935 m;
``L-3'' is used hereafter) in 1979 (Oba et al., 1991) was used for reference. The sampling locations of cores KI-5 and PC-9 and core L-3 are shown in Fig. 1.
The lithologies of the KI-5 and PC-9 sediment core have been described by Okumura et al. (1996). Brie¯y, the upper part of these cores consists of olive gray to greenish clay-silt, indicating that the sediments were deposited under
oxic conditions. A thinly laminated layer (TL-1) is present at 192±201 cm depth in core KI-5, and at 129±137 cm in core PC-9. The top of a second thick laminated layer called TL-2 is observed at 272 cm depth in KI-5, and at 215 cm in PC-9. These two laminated layers (TL-1 and TL-2) cor-respond well to those commonly observed in cores taken from the Japan Sea (Ikehara et al., 1994; Ishiwatari et al., 1994,1999; Oba et al., 1991). In the case of core L-3, TL-1 is observed at 115 cm and the top of TL-2 at 160 cm depth, respectively. These two laminated layers indicate highly anoxic conditions in the bottom sediments.
The age model for core L-3 has been described by Ishiwatari et al. (1999). Brie¯y, it was established by using well-known volcanic ash layers (K-Ah, U-Oki, and AT) which occur in the core. Their ages are 7300 calendar (cal) years B.P. for K-Ah (80±76 cm depth in core L-3), 10,750 cal years B.P. for U-Oki (94±86 cm depth in core L-3) and 24,50014C years B.P. for AT. AMS-14C
ages of eight sections (planktonic foraminifera) were determined by Oba et al. (1995). Calendar ages were cal-culated from radiocarbon ages using the data set obtained for Lake Suigetsu (Kitagawa and van der Plicht, 1998). The age model for cores KI-5 and PC-9 was established by correlating the volcanic ash layers as well as the two laminated layers (TL-1: 11.3 ka and the top of TL-2: 17.5 ka) with those in core L-3.
Subsamples were taken from these cores. Subsamples were 50 mm thick (corresponding to550 yr) with some exceptions at volcanic layers for core L-3, 60 mm (500 yr) for core PC-9, and 20 mm for core KI-5 (125 yr), respectively. Wet subsamples were used for lipid analy-sis for cores PC-9 and KI-5, while freeze-dried sub-samples were used for core L-3, since the latter core has been stored at room temperatures for a long time after
collection and was almost in an air-dried state. No appreciable eect of sample storage at room tempera-tures on Uk37' values has been reported (Sikes et al., 1991), nor is it seen in core L-3 (described later).
2.2. Total organic carbon and total nitrogen analysis
Subsamples taken from these cores were freeze-dried and ground to powder for total organic carbon (TOC) and total nitrogen (TN: total organic nitrogen plus ammonium nitrogen) analysis. Duplicate determination used HCl-treated sediment samples (100 mg) analysed by dry combustion using a Fison NA1500-NCS elemental ana-lyzer. The relative standard deviation of these measure-ments was better than 10% for both TOC and TN.
2.3. Alkenone and alkenoate determination
Free lipids were extracted by ultrasonication of wet samples (ca. 4 g) in 50 ml Pyrex glass centrifuge tubes containing benzene/methanol (6:4 v/v). The extracts were collected after centrifugation (10 min, 3000 rpm), and saponi®ed with 5% NaOH-methanol. Distilled water was added to the solvent extracts after their trans-fer to separatory funnels and the neutrals were extracted withn-hexane/diethyl ether (9:1 v/v). The neutrals were fractionated on a silica gel column (silica gel: Mal-linckrodt Inc. 100 mesh; column size: 505 mm i.d.). Long-chain alkenones were obtained as the third (ben-zene) eluate after hexane (1st eluate containing satu-rated hydrocarbons) and hexane/benzene (2nd eluate containing unsaturated and/or aromatic hydrocarbons) elutions. The 1st and 2nd eluates were stored for later study. The third eluates were evaporated to small volumes and analyzed by gas chromatography (GC) or gas chromatography/mass spectrometry (GC/MS).
GC analysis of alkenones used an HP 5890 instrument ®tted with a splitless injector and a DB-5 fused-silica capillary column (60 m0.32 mm i.d.). The oven tem-perature was programmed from 50 to 150C at 30C
minÿ1and then to 325C at 3C minÿ1. Quantitation was
achieved by integration of FID peak areas; 1-docosene was used as a reference (injection) standard. GC/MS ana-lysis used a Finnigan INCOS 50 MS/Varian 3400 GC, with analytical conditions similar to those for GC analysis. Alkenone-SSTs were calculated in this study using the empirical equation given by Prahl et al. (1988):Uk0
37=0.034 T(C)+0.039, whereUk0
37=[C37:2]/[C37:2+C37:3].
Alkenoates were analyzed by a similar procedure used for the alkenones except for elimination of the NaOH± methanol saponi®cation.
2.4. Micropaleontological analysis
Coccolith assemblages were analyzed on smear slides using a polarizing microscope with 1250 times
magni®cation. The abundance of coccoliths was mea-sured only qualitatively, and ranked into four categories: very rare, rare, few and common. For the samples con-taining rare to common coccoliths, 200 specimens were identi®ed to species or genus level, and their numbers were recorded. Occurrences ofFlorisphaera profundaand reworked specimens were counted separately.
3. Results
3.1. TOC and alkenone/alkenoate analytical data
The vertical pro®les for TOC concentrations were useful for correlating the sediment cores L-3, PC-9 and KI-5 (Fig. 2). The TOC pro®les for cores PC-5 and KI-5 were very similar, even though the two cores were taken in dierent years (1994 and 1993). Since the sampling location for core PC-9 is close to that for core KI-5, this result indicates the good reproducibility of TOC pro®les in this region. TOC concentrations in core PC-9 slightly decreased upward from 500 cm (37 ka) to 357 cm (28 ka), became uniform until 214 cm (17.5 ka), and then increased upward. At 128.5 cm (11.3 ka: TL-1 layer), TOC concentrations exhibited a maximum and then decreased again slightly followed by nearly constant values. This vertical pro®le matches well that in core L-3. In both cores L-3 and PC-9, a considerable amount of C37alkenones was detected in the Holocene section (C37
conc.: 1.70.3mg gÿ1for L-3, 1.20.7mg gÿ1for PC-9).
For the sections below the top of TL-2 (17.5 ka), how-ever, the abundance of alkenones in core L-3 was extre-mely low (C37conc.: <0.1mg gÿ1), compared with that
in core PC-9 (0.40.2mg gÿ1). This result suggests that
coccolithophore production was higher in the region where core PC-9 was taken compared to the region of core L-3 during the glacial period.
Crossplots of EE/K37 (a ratio of C36:2ethyl alkenoate
to C37 alkenones) versus U
k'
37 and EE/K38 (a ratio of C36:2ethyl alkenoate to C38alkenones) versusU
k0
37 have been proposed by Sawada et al. (1996) from culture experiments to estimate the extent of production of
Emiliania huxleyi relative to Gephyrocapsa oceanica in seawater. This is based on their result that the regression curves for EE/K37 vs.Uk0
37 and EE/K38 vs.U k0 37 in the two species are dierent. Although the results of Sawa-da's culture experiment are not necessarily supported by their own and other microscopic observations of sedi-ment samples (e.g. Herbert et al., 1998), we carried out both alkenone analyses and microscopic observation to check the coincidence of the two kinds of results.
We measured the concentrations of C36:2 ethyl
alkenoate, C37 and C38 alkenones in selected samples
is in con¯ict with the results of microscopic analysis, as described below.
3.2. Coccolith data
Out of 20 sections examined for core PC-9, 14 sec-tions yielded sucient concentration of coccoliths for analysis. In most samples, the assemblage was
domi-nated by E. huxleyi and G. oceanica, but Braarudo-sphaera bigelowiiwas the dominant species in the 339.0 and 363.0 cm sections (26.4 and 28.1 ka, respectively).
Coccolithus pelagicusthat dwells in cold waters was rare in the upper sections down to section 131.5 cm (11.5 ka) and became abundant below section 151.0 cm (12.9 ka). This observation re¯ects the increased in¯uence of the Tsusima Warm Current over the last 12 kyr.
Fig. 3. Plots of (a) EE/K37 [C36:2ethyl alkenoate /C37alkenones] and (b) EE/K38 (C36:2ethyl alkenoate/C38alkenones) againstU
k0
37in
the Japan Sea. The curves drawn in the ®gure are derived from Sawada et al. (1996). The samples from core PC-9 marked by a±h correspond to the sections (a-h) shown in Fig. 4.
The relative abundances ofE. huxleyiandG. oceanica
coccoliths are shown in Fig. 4. The highest concentra-tion of coccoliths was observed in the secconcentra-tion at 131.5 cm (12.9 ka), which corresponds to the maximum TOC concentration (Fig. 2). Although the alkenone/alkeno-ate-type analysis suggests that the major contributor to the sedimentary alkenones is E. huxleyi, the coccolith data indicate thatG. oceanicacomprised the dominant part (93%) of the haptophyte species in the sections at 19.3 and 20.6 ka (Fig. 4). The discrepancy between mole-cular analysis and microscopic observation is discussed in Section 4.2.1.
3.3. Sea surface temperatures (SSTs) obtained from alkenone analyses
Fig. 5 displays the temporal pro®les of alkenone-SSTs calculated from the empirical equation given by Prahl et al. (1988) for cores L-3, PC-9, and KI-5. The pro®le of
alkenone-SST for core PC-9 [Fig. 5(b)] is similar to that for core KI-5 [Fig. 5(c)], demonstrating that the results for alkenone-SSTs through time are reproducible.
The pro®les for cores PC-9 and KI-5 indicate that SSTs were12C at36 ka [Fig. 5 (b)], increased from
12C to 18±19C between 36 and 30 ka, and remained
almost constant until17.5 ka. The water column became strati®ed at this time due to restricted vertical mixing as a result of sea-level drop during the LGM. With the degla-cial rise in sea level, SST decreased from 18±19 to12C
at site PC-9 at 17-12 ka. Between 12 and 9 ka, SSTs increased from12 to 18C, remained almost constant
until 7 ka and then decreased to16C and ¯uctuated
around 16±17C thereafter.
The time resolution for the SST pro®le of core L-3 [Fig. 5(a)] is low compared to those for cores PC-9 and KI-5, but the following common features can be identi-®ed among those SST pro®les: (i) relatively high SSTs at 17.5 ka, (ii) an SST drop between 17 and 12 ka, and (iii) an SST rise after 12 ka. This suggests that the SST variations described above occurred over the whole area of the Japan Sea.
During the 12±11 ka interval, the SST at site L-3 shows a sharp rise from 15 to 19C, probably as the warm
sea-water from the south began to ¯ow into the Japan Sea through the Tsushima Strait (Oba et al., 1991). In the same interval, the rise in SST at site PC-9 was more gra-dual than that at site L-3, probably re¯ecting changes in the hydrodynamics in the Japan Sea.
4. Discussion
As described above, we obtained relatively high alke-none-SSTs in the Japan Sea during the LGM, which is consistent with our previous results (Ishiwatari et al., 1999). However these estimated alkenone-SSTs at the LGM are much higher than SST values previously assumed by Oba et al. (1991) and Matsui et al. (1998). Before going into a discussion of the possible reasons for this, we need to establish the reliability of the alkenone method. We have done this from a comparison with the other data in modern and late Holocene (0±7 ka) sedi-ments when the Japan Sea was probably under relatively stable oceanographic conditions. We also discuss some uncertainties associated with the alkenone method.
4.1. Assessment of core-top and late Holocene alkenone-SSTs
4.1.1. Core-top alkenone-SSTs
The shallowest samples we analyzed were the 5±10 cm section (0.7 ka) of core L-3, the 0±6 cm Section (0.3 ka) from core PC-9 and the 2±4 cm section (0.1 ka) from KI-5. The sample from core L-3 is clearly not a core-top sample. The temperature estimated from Uk370 was
19.1C for L-3, 16.2C for PC-9 and 15.2C for KI-5.
The modern mean annual SSTs obtained from the compiled data of JODC (Japan Oceanographic Data Center: on a 1 square grid) give 17.65.3C for the
region near L-3 site and 16.35.8C for that near PC-9
site, respectively. Therefore, the alkenone-SST value for L-3 is 1.5C higher than the modern mean annual SST,
while the alkenone-SST values for PC-9 and KI-5 are close to the modern mean annual SST. The coincidence between the alkenone-SSTs and modern mean annual SSTs at two sites implies that theUk0
37data provide reliable estimates of SST in this region.
4.1.2. Comparison between alkenone-SSTs andd18
O-derived SSTs in the late Holocene
Oba and his coworkers determined oxygen isotopic compositions of planktonic (mainly Neogloboquadrina pachyderma and Globigerina umblicata) and benthic for-aminifera (mainlyCassidulina japonica) in core L-3 (Oba et al., 1991; 1995). Later they reported vertical variations in d18O of planktonic foraminifera in cores PC-9 and
KI-5 (Okumura et al., 1996). These data are shown in Fig. 5(d).
N. pachyderma(right-coiling), which is a major spe-cies in the Holocene, is thought to live at depths shal-lower than 100 m, while the life style of G. umblicata, which is the most abundant species in the glacial inter-val, is unknown since it is an extinct species (Oba, per. comm. 1999). Oba and his coworkers estimated SSTs (d18O-SST is used hereafter) over the past 60 kyr except
for the interval 16±30 ka when benthic foraminifera
were absent in the sediments, using those oxygen iso-topic compositions. For this estimation they ignored dierence in the species of planktonic foraminifera pre-sent and assumed that: (i) bottom water temperatures were constant at 0.2C which is the same as the present
temperature over the periods studied, and (ii) vertical pro®les of oxygen isotopic composition of seawater as a function of water depth were the same as those in the present time; i.e. O2 in shallow sea water was assumed
to be isotopically heavier by 0.2%than that in bottom
water (below 1000 m in depth) (Oba et al., 1980). Thus, they calculated d18O-SSTs to be 16.6±17.7C for 0±10
ka, 8.2±11.5C for 10±17 ka and 8.0±9.7C for the
period before2.5 ka.
Although uncertainties are present in estimating SSTs using oxygen isotopic compositions of benthic and planktonic foraminifera, we calculatedd18O-SSTs using
the same assumptions as cited above, and compared these with alkenone-SSTs using new microfossil ana-lyses (Oba et al., 1995). The sediment sections which we examined are limited to those before 7 ka (0±75 cm in depth), because we excluded extinct species such asG. umbilicata. For the calculation ofd18O-SST, we used the
same equations of temperature-d18O relationship as
those used by Oba et al. (1980):
T C :0:2 16:9ÿ6:17 dfÿdw
0:52 dfÿdw2 1
SST C 21:4ÿ4:19 sfÿsw 0:05 sfÿsw2 2
where (dfÿdw) is the oxygen isotopic dierence
between calcite of benthic foraminifera and deep water, and (sfÿsw) is the oxygen isotopic dierence between
calcite of planktonic foraminifera and surface water. Eq. (1) is used for benthic foraminifera (C. japonica) and Eq. (2) is used for planktonic foraminifera (N. pachy-derma). We did not use the temperatureÿd18O
relation-ship [T (C)=16.9±4.38 (dfÿsw)+0.10 (ddfÿdsw)2 ] of
Matsui et al. (1998) for estimating paleosalinity varia-tions, because thed18O-SST obtained from this equation
for the core-top sample of core L-3 (12.7C for 0±1 cm
section) was about 5C lower than the modern mean
annual SST at site L-3 (17.6C).
It is clear from Table 1 that the agreement between d18O-SSTs and alkenone-SSTs is good, although the
former is slightly lower by aboutÿ0.4C0.4C (n=5).
This result seems in favor of the reliability of the alke-none-SSTs at least in late Holocene. A possible expla-nation for the slight dierence in estimated SSTs is a dierence in the depth of water inhabited by cocco-lithophores and planktonic foraminifera. From a study of long-chain alkenones in suspended materials during well de®ned thermocline conditions, Bentaleb et al. (1999) concluded that the C37 alkenone record re¯ects
the temperature of highest primary production (max-imum algal growth) which occurred at 30 m depth in their study. Herbert et al. (1998) found that alkenone-temperatures agree closely with mean annual SSTs in surface waters (0 m) and nearly as well with the warmest month (September) temperatures at 30 m depth along the California margin. Coccolithophores generally live in the upper photic zone, but planktonic foraminifera generally migrate vertically in the water column through their life cycle and thus re¯ect temperature changes through the surface water mass (Weaver et al., 1999). Oba et al. (1980) assumed that the planktonic foraminifera lived at depths shallower than 200 m. These facts suggest that tempera-tures estimated from d18O of planktonic foraminifera
re¯ect temperatures in waters deeper than those in which coccolithophores live. This might be particularly so at the time of development of strongly strati®ed sur-face waters in the Japan Sea at the LGM.
4.2. Problems related to the estimation of alkenone-SSTs
4.2.1. Eects of haptophyte species on alkenone-SST E. huxleyiand the closely related speciesG. oceanica
are known to synthesize C37±C39methyl and ethyl
alke-nones and related alkenoates (Marlowe et al., 1990). The dierence in theUk0
37 vs. SST relationship between the two species is controversial. Volkman et al. (1995) and Sawada et al. (1996) claimed from culture experi-ments that a linear equation expressing theUk370 vs. SST relationship is dierent between the two species; SST estimated forG. oceanicabeing higher than that forE. huxleyiat the sameUk370 value.
Recently, however, Conte et al. (1998) showed from culture experiments using a variety of strains ofE. hux-leyiand G. oceanicathat the Uk370 vs. SST relationship forG. oceanicafalls within the range of those observed for cultured strains ofE. huxleyi. In addition, Conte et al. (1998) argued against the idea of Sawada et al. (1996) of distinguishingG. oceanicaproduction fromE. huxleyi
from their alkenoate/alkenone ratio relative toUk370, by presenting evidence that the ratio re¯ects a dierent cell physiology rather than genetics. Alkenone/alkenoate analysis for core PC-9 would indicate that the major contributor of alkenoates and alkenones isE. huxleyiif we adopt the ideas of Sawada et al. (1996). But this result does not agree with the microscopic observations of haptophyte species in our samples. A similar obser-vation was reported by Herbert et al. (1998) for Cali-fornia margin sediments. Their core top data fall closest to the Prahl et al. (1988) temperature calibration derived from laboratory cultures of E. huxleyi, although the smear slide data for selected samples show that G. oceanica comprises an important part of haptophyte ¯ora. They claimed that theG. oceanicaculture calibra-tion of Volkman et al. (1995) does not seem to accu-rately re¯ect natural conditions. Moreover, if we adopt the Uk0
37 vs. SST relationship obtained forG. oceanica reported by Volkman et al. (1995) or Sawada et al. (1996), alkenone-SSTs should be higher than those obtained from Prahl et al. (1988) equation. Therefore, SSTs obtained after correction for the presence of G. oceanicawould be higher than the uncorrected SSTs.
4.2.2. Eects of postdepositional conditions on alkenone-SST
Thick laminated sections called TL-2 are observed between 26 and 17.5 ka. This fact together with other evidence, such as the disappearance of most benthic fauna (at least for site L-3: Oba et al., 1991), presence of highly 13C-depleted diploptene (Yamada et al., 1997),
and high sulfur contents (Ishiwatari et al. 1999 unpub-lished data), collectively indicate the prevalence of anoxic bottom water conditions during this interval. Therefore, if strongly sulfate-reducing conditions in bottom water and sediments play a role in the reduction of alkenones, then the abundance of C37:2 and C37:3
alkenones might be modi®ed by reduction or sulfuriza-tion. However, since no distinct change in the distribu-tion of C37:2and C37:3alkenones has been observed in
the anoxic Black Sea sediments (Sun and Wakeham, 1994), anoxic bottom water condition does not seem to produce abnormalUk0
37 values.
Some studies have shown thatUk370 tends to be higher in oxic sediments as a result of preferential degradation of C37:3 alkenone relative to C37:2 alkenone and
oxic and anoxic conditions in an incubation experiment of bacterial degradation of E. huxleyi (Teece et al., 1998). Alkenone-SSTs at the core top and subsurface sections of our cores (Section 4.1.1.) are in agreement with (or at least not far from) the modern mean annual SSTs. Therefore, oxic condition of the sediments do not seem to exert serious eects on alkenone-SST in the Japan Sea sediments.
4.2.3. Salinity eects on alkenone-SST
According to Rosell-Mele (1998),Uk0
37 is not a good proxy for SST below 10C in the Atlantic Ocean and the
Nordic seas, whereasUk0
37does correlate with SST down to about 7C in the Southern Ocean (Sikes et al., 1997).
Rosell-Mele (1998) also reported that a lower salinity appears to produce a higher abundance of C37:4relative
to the total C37alkenones (%37:4 ). A similar
relation-ship between low salinity and unusually high SSTs and high%37:4 is observed for cores in the North Atlantic (Rosell-MeleÂ, 1998). That is, alkenone-SSTs give values of 12±13C instead of1C from diatom transfer
func-tion-SST, or 5±6C from foraminifera transfer
func-tions. The eect of salinity on the abundance of the C37:4
alkenone is quite marked at SSTs below 6C in the
Nordic seas (Rosell-MeleÂ, 1998). The percentage of C37:4alkenone in the total C37alkenones (%37:4) is high
(>5) below6C, and moreover, the increase of %37:4
corresponds to the decrease in salinity (1.5) for the Nordic sea core.
If SSTs were steady at 4C and paleosalinity
decreased to 22 in the Japan Sea during the 33±17.5 ka interval as claimed by Matsui et al. (1998), and if the relationship between %37:4 and salinity found by Rosell-Mele (1998) is valid for the Japan Sea, then a high %37:4 content should be observed for the sediment samples in the 33±17.5 ka interval. The temporal varia-tion in %37:4 over the last 36 kyr is displayed in Fig. 6. Evidently, no high %37:4 values for core PC-9 are observed in the 33-17.5 ka interval, suggesting that the SST cannot be so low. In the light of the ®ndings of Rosell-Mele (1998), SSTs at site PC-9 must have been at least higher than6C.
4.2.4. Other uncertainties
There are still several possible uncertainties disturbing the estimation of accurate SSTs. It is known that long-chain alkenones preserved in sediments are synthesized essentially at the depths of highest primary production (Bentaleb et al., 1999; Weaver et al., 1999). However, there exists a possibility that the growth temperatures at which the C37 alkenones were produced might have
been in¯uenced by changes in the glacial±interglacial
Table 1
Data of benthic and planktonic foraminifera, their oxygen isotope composition and comparison betweend18O-SSTs and alkenone-SSTs in late Holocene for core L-3
Deep water Surface water
a Data from Oba et al. (1995),C. jap.;Cassidulina japonica;N.pac;Neogloboquadrina pachyderma.
b ddw was calculated fromT(C)=16.9ÿ6.17 (ddfÿddw)+0.52 (ddfÿddw)2(Oba et al., 1980) on the assumption that bottom water
temperature is 0.2C.
c dsw was assumed to be isotopically heavier by 0.2%than that of bottom water (Oba et al., 1980). d SST was calculated from SST (C)=21.4ÿ4.19 (dsfÿdsw)+0.05 (dsfÿdsw)2(Oba et al., 1980). e SST was calculated from Uk0
climatic regime or by changing oceanographic condi-tions caused by severely strati®ed surface seawaters. As discussed by Herbert et al. (1995), if the season of max-imal production of haptophyte microalgae were to shift with climatic regime, it would produce alkenone-SSTs anomalies. Although it is dicult to rule out such potential anomalies, we have no evidence to suggest that they are important here.
4.3. Environmental changes in the Japan Sea
Notwithstanding the possibility that Uk370 values are controlled by unknown factor(s), we conclude that the temporal variations in mean annual SST over the past 36 kyr, as shown in Fig. 5, are reasonable. Thus we conclude that the following environmental changes occurred in the Japan Sea over the past 36 kyr.
. 36±17.5 ka: During the interval 36-17.5 ka, the water column of the Japan Sea became strati®ed as sea-level dropped. As vertical mixing and upward transport of nutrients became restricted, deep waters [deeper than 400 m from the sea surface as suggested by Ikehara et al. (1994)]
became anoxic, leading to the formation of lami-nated sediment layers of TL-2. Evidence for anoxic deep waters was presented in Section 4.2.2. The high mean annual SSTs (18±19C) may have
occurred under heavily strati®ed surface seawater where thermal energy due to solar radiation is trapped in shallow waters. A similar phenomenon of radiative equilibrium is observed in brackish lakes (Yoshimura, 1976).
According to Oba et al. (1991),Thalassiosira lacustris, a diatom species common in the less saline waters of coastal areas, became abundant in the sedimentary sec-tions during the 32±17.5 ka interval. This suggests that paleosalinity was lower in this interval. Therefore, a realistic scenario for environmental changes in the Japan Sea is that both SST rose and paleosalinity fell during this interval. Further studies are needed to develop a method of extracting signals of paleosalinity from those preserved in the Japan Sea sediments.
. 17.5-8 ka: Between 17.5 and 12 ka, the SST
decreased and then increased between 12 and 9 ka at sites PC-9 and L-3. The decrease in SSTs is interpreted to have been caused by the onset of vertical mixing of seawater by the in¯ow of sea-water from the north (Oyashio Current: Oba et al., 1991), or the south (Keigwin and Gorbarenko, 1992). The vertical mixing of seawater led to the disappearance of anoxic deep water. This is shown by the positive shift in d13C values of diploptene
from ÿ57 to ÿ25% (Yamada et al., 1997), the
disappearance of laminae.
The lowering of SST during 17±12 ka is qualitatively supported by the diatom fossil analysis where the ratio of warm-current species (Pseudoeunotia doliolus, Paralia sul-cata, Thalassiosira oestrupii) to total species [warm-cur-rent region sp.+cold-cur[warm-cur-rent region sp. (Denticulopsis seminae, Thalassiosira trifulta, Thalassiosira nordenskiol-dii)] dropped from about 0.8±0.9 at 17 ka to 0.4±0.5 at 12 ka for core C-3 taken at the site L-3 (Koizumi, 1984). This result also demonstrates that the constant4C model of
Matsui et al. (1998) at the LGM is not realistic.
The SST rise in the 12±11 ka interval were probably caused by the in¯ow of warm seawater from the south through the Tsushima Strait (Oba et al., 1991).
. 8±0 ka. The present oceanographic regime became established at about 8 ka, and conditions since then have been similar to what is observed today.
5. Conclusions
A marked negatived18O excursion of planktonic
for-aminifera during the LGM observed in the Japan Sea
sediment cores has been interpreted as recording a drop in paleosalinity of surface waters on the assumption of constant low sea surface temperatures between 34 and 11 ka. This assumption has now been checked and our results for alkenone-SSTs over the past 36 kyr at Sites L-3, PC-9 and KI-5 indicate that SSTs during the LGM were much higher than those previously assumed. Comparisons of core-top alkenone-SSTs values with values for modern seawater in the Japan Sea suggest that SSTs estimated from alkenones are reliable. The warm SSTs were probably the result of equilibrium of thermal energy due to solar radiation in strati®ed sur-face seawaters under conditions where the Japan Sea was largely isolated from the open ocean when sea level was lowest at the LGM. Physical modeling of the warm SSTs of the Japan Sea at the LGM, and magnitudes of paleosalinity change remain to be studied.
Acknowledgements
We are grateful for Drs. D. Brincat and J. Volkman who carefully read this manuscript for improvement. We thank Drs. T.D. Herbert and E. Gonzalez for their critical reviews of the manuscript and constructive comments. We also thank Dr. T. Oba for discussions during the revision of the manuscript. This work was conducted as an international collaborative study on paleoceanography of the Japan Sea (Chief scientist: T. Oba) partly supported by Japan Society for the Promo-tion of Science.
Associate EditorÐJ.K. Volkman
References
Bentaleb, I., Grimalt, J.O., Vidussi, F., Marty, J.-C., Martin, V., Denis, M., Hatte, C., Fontugne, M., 1999. The C37 alke-none record of seawater temperature during seasonal ther-mocline strati®cation. Marine Chemistry 64, 301±313. CLIMAP Project Members, 1981. Seasonal Reconstructions of
the Earth0s Surface at the Last Glacial Maximum, Map and Chart Series, MC-36, The Geological Society of America, Inc. Conte, M.H., Thompson, A., Lesley, D., Harris, R.P., 1998.
Genetic and physiological in¯uences on the alkenone/ alkenoate versus growth temperature relationship in Emilia-nia huxleyiandGephyrocapsa oceanica. Geochimica et Cos-mochimica Acta 62, 51±68.
Crusius, J., Pedersen, T.F., Calvert, S.E., Cowie, G.L., Oba, T., 1999. A 36 kyr geochemical record from the Sea of Japan of organic matter ¯ux variations and changes in intermediate water oxygen concentrations. Paleoceanography 14, 248±259. Gong, C., Hollander, D.J., 1999. Evidence for dierential
degradation of alkenones under contrasting bottom water oxygen conditions: Implication for paleotemperature recon-struction. Geochimica et Cosmochimica Acta 63, 405±411.
Herbert, T. D., Yasuda, M., Burnett, C., 1995. Glacial±inter-glacial sea-surface temperature record inferred from alke-none unsaturation indices, site 893, Santa Barbara basin. In: Kennett, J.P., Baldauf, J.G., Lyle, M. (Eds.), Proceedings of the Ocean Drilling Program, Scienti®c Results, Vol. 146 (Pt. 2), pp. 257±265.
Herbert, T.D., Schuert, J.D., Thomas, D., Lange, C., Wein-heimer, A., Peleo-Alampay, A., Hergiera, J.-C., 1998. Depth and seasonality of alkenone production along the California margin inferred from a core top transect. Paleoceanography 13, 263±271.
Hoefs, M.J.L., Versteegh, G.J.M., Rijpstra, W.I.C., de Leeuw, J.W., Sinninghe DamsteÂ, J.S., 1998. Postdepositional oxic degradation of alkenones: implications for the measurement of palaeo sea surface temperatures. Paleoceanography 13, 42±49.
Ikehara, K., Kikkawa, K., Katayama, H., Seto, K., 1994. Late Quaternary paleoceanography of the Japan Sea: a tephro-chronological and sedimentological study. In: Proceedings of 29th International Geological Congress, Part B, Kyoto, 24 August- 3 September 1992, pp. 229±235.
Ishiwatari, R., Hirakawa, Y., Uzaki, M., Yamada, K., Yada, T., 1994. Organic geochemistry of the Japan Sea sedimentsÐ 1: Bulk organic matter and hydrocarbon analyses of core KH-79-3, C-3 from the Oki ridge for paleoenvironmental assessments. Journal of Oceanography 50, 179±195. Ishiwatari, R., Yamada, K., Matsumoto, K., Houtatsu, M.,
Naraoka, H., 1999. Organic molecular and carbon isotopic records of the Japan Sea over the past 30 kyr. Paleoceano-graphy 14, 260±270.
Keigwin, L.D., Gorbarenko, S.A., 1992. Sea level, surface sali-nity of the Japan Sea, and the Younger Dryas event in the Northwestern Paci®c Ocean. Quaternary Research 37, 346± 360.
Kitagawa, H., van der Plicht, J., 1998. Atmospheric radio-carbon calibration to 45,000 yr B.P.: Late Glacial ¯uctua-tions and cosmogenic isotope production. Science 279, 1187± 1190.
Koizumi, I., 1984. Diatom: Analysis of core KH-79-3, C-3. Gekkan-Kaiyou 6, 547±553 (in Japanese).
Marlowe, I.T., Brassell, S.C., Eglinton, G., Green, J.C., 1990. Long-chain alkenones and alkyl alkenoates and the fossil coccolith record of marine sediments. Chemical Geology 88, 349±375.
Matsui, H., Tada, R., Oba, T., 1998. Low-salinity isolation event in the Japan Sea in response to eustatic sea-level drop during LGM: Reconstruction based on salinity-balance model. Quaternary Research 37, 221±233 (in Japanese). Oba, T., Horibe, S., Kitazato, H., 1980. Analysis of the
paleoenvironment since the last glacial age based on two cores from the Japan Sea. Koukogaku-to Shizenkagaku 13, 31±49 (in Japanese).
Oba, T., Kato, M., Kitazato, H., Koizumi, I., Omura, A., Sakai, T., Takayama, T., 1991. Paleoenvironmental changes in the Japan Sea during the last 85,000 years. Paleoceano-graphy 6, 499±518.
Oba, T., Murayama, M., Matsumoto, E., Nakamura, T., 1995. AMS-14C ages of Japan Sea cores from the Oki Ridge. The Quaternary Research 34, 289±296 (in Japanese).
City in the Japan Sea based on oxygen, carbon and nitrogen isotopes. The Quaternary Research 35, 349±358 (in Japa-nese).
Prahl, F.G., Muehlhausen, L.A., Zahnle, D.L., 1988. Further evaluation of long-chain alkenones as indicators of paleo-ceanographic conditions. Geochimica et Cosmochimica Acta 52, 2303±2310.
Rosell-MeleÂ, A., 1998. Interhemispheric appraisal of the value of alkenone indices as temperature and salinity proxies in high-latitude locations. Paleoceanography 13, 694±703. Sawada, K., Handa, N., Shiraiwa, Y., Danbara, A., Montani,
S., 1996. Long-chain alkenones and alkyl alkenoates in a coastal and pelagic sediments of the northwest North Paci®c, with special reference to the reconstruction of Emiliania huxleyi and Gephyrocapsa oceanica ratios. Organic Geo-chemistry 24, 751±764.
Sikes, E.L., Farrington, J.W., Keigwin, L.D., 1991. Use of the alkenone unsaturation ratio Uk
37to determine past sea surface temperatures: core-top SST calibrations and methodology considerations. Earth and Planetary Science Letters 104, 36±47. Sikes, E.L., Volkman, J.K., Robertson, L., Pichon, J.-J., 1997. Alkenones and alkenes in surface waters and sediments of the Southern Ocean: Implications for paleotemperature esti-mation in polar regions. Geochimica et Cosmochimica Acta 61, 1495±1505.
Sun, M.Y., Wakeham, S.G., 1994. Molecular evidence for degradation and preservation of organic matter in the anoxic
Black Sea Basin. Geochimica et Cosmochimica Acta 58, 3395±3406.
Tada, R., Irino, T., Koizumi, I., 1999. Land-ocean linkages over orbital and millennial timescales recorded in late Qua-ternary sediments of the Japan Sea. Paleoceanography 14, 236±247.
Teece, M.A., Getli, J.M., Leftley, J.W., Parkes, R.J., Max-well, J.R., 1998. Microbial degradation of the marine prym-nesiophyte Emiliania huxleyi under oxic and anoxic conditions as a model for early diagenesis: long chain alka-dienes, alkenones and alkyl alkenoates. Organic Geochem-istry 29, 863±880.
Volkman, J.K., Barrett, S.M., Blackburn, S.I., Sikes, E.L., 1995. Alkenones inGephyrocapsa oceanica: Implications for studies of paleoclimate. Geochimica et Cosmochimica Acta 59, 513±520.
Weaver, P.P.E., Chapman, M.R., Eglinton, G., Zhao, M., Rutledge, D., Read, G., 1999. Combined coccolith, for-aminiferal, and biomarker reconstruction of paleoceano-graphic conditions over the past 120 kyr in the northern North Atlantic (59N, 23W). Paleoceanography 14, 336± 349.
Yamada, K., Ishiwatari, R., Matsumoto, K., Naraoka, H., 1997.d13C records of diploptene in the Japan Sea sediments over the past 25 kyr. Geochemical Journal 31, 315±321. Yoshimura, S., 1976. Limnology. Seisan Gujyutsu Sentah,