Influence of Northern Hemisphere climate and global sea level rise on the restricted Red Sea marine environment during termination I



[1] We present high-resolution paleoceanographic records of surface and deep water conditions within the northern Red Sea covering the last glacial maximum and termination I using alkenone paleothermometry, stable oxygen isotopes, and sediment compositional data. Paleoceanographic records in the restricted desert-surrounded northern Red Sea are strongly affected by the stepwise sea level rise and appear to record and amplify well-known millennial-scale climate events from the North Atlantic realm. During the last glacial maximum (LGM), sea surface temperatures were about 4°C cooler than the late Holocene. Pronounced coolings associated with Heinrich event 1 (∼2°C below the LGM level) and the Younger Dryas imply strong atmospheric teleconnections to the North Atlantic. Owing to the restricted exchange with the Indian Ocean, Red Sea salinity is particularly sensitive to changes in global sea level. Paleosalinities exceeded 50 psu during the LGM. A pronounced freshening of the surface waters is associated with the meltwater peaks MWP1a and MWP1b owing to an increased surface-near inflow of “normal” saline water from the Indian Ocean. Vertical δ18O gradients are also increased during these phases, indicating stronger surface water stratification. The combined effect of deglacial changes in sea surface temperature and salinity on water column stratification initiated the formation of two sapropel layers, which were deposited under almost anoxic condition in a stagnant water body.

1. Introduction

[2] As widely from the world ocean restricted and desert-surrounded basin, the northern Red Sea suffered extreme oceanographic changes in the past that resulted in an amplification of paleoenvironmental signals in the marine record [e.g., Hemleben et al., 1996]. This setting makes the Red Sea to an ideal location for studying the atmospheric transfer of North Atlantic millennial-scale climate events to the subtropics of the Near East. Owing to the shallow sill connection to the Indian Ocean (Strait of Bab el Mandeb), glacioeustatic sea level changes strongly affected particularly sea surface salinities in the Red Sea. Though the northern Red Sea is even today one of the areas with the maximum salinities in the world ocean, the sea level low stand during the last glacial maximum caused a further increase in salinities which exceeded the tolerance level of many planktic and benthic organisms [e.g., Winter et al., 1983; Hemleben et al., 1996]. Extensive work has been carried out in order to understand the impact of glacial to interglacial paleoenvironmental changes on planktic and benthic ecosystems in the Red Sea [e.g., Winter et al., 1983; Almogi-Labin et al., 1991, 1998; Fenton et al., 2000]. Besides planktic and benthic foraminifera, the most extensively studied organisms in the Red Sea are thecosomatous pteropods, which are extremely sensitive to changes in past water column properties and underwent major changes in their assemblages especially during glacial-interglacial transitions [Almogi-Labin et al., 1991, 1998]. Almogi-Labin et al. [1991] demonstrated on a sediment core from the central Red Sea that changes in pteropod assemblages during the last termination most likely reflect changes in water column stratification and the extension of oxygen minimum zones (OMZ). Owing to a lack of sufficient stratigraphic control and restricted resolution of the existing records, the precise timing and relationship of the Red Sea paleoenvironmental signals to glacioeustatic sea level changes, local paleoclimate and teleconnections to North Atlantic climate change are still not fully unraveled. In the present study we concentrated on an extensively 14C-AMS-dated and cm wise sampled sediment core (GeoB 5844-2) from the northernmost Red Sea off the Sinai Peninsula, a sensitive region, where important processes such as the winter deep mixing and intermediate and deep-water formation take place. Combining stable oxygen isotope measurements on multiple planktic and benthic foraminifera with alkenone paleothermometry, we reconstructed sea surface temperatures (SST) and salinities (SSS) as well as vertical gradients in δ18O (Δδ18O) for the last 22 ka. Major shifts in SST seem to be in phase with Northern Hemisphere middle- and high-latitude climatic changes, whereas SSS and vertical isotope gradients are indirectly linked to the course of the postglacial sea level rise.

2. Study Area

2.1. Climate

[3] The Red Sea area is a tectonically relative young rift basin that is surrounded by desert, separating the African from the Asian arid belt. Year-round, in the northern Red Sea winds blow from NW to NNW along the Red Sea axial through, while south of about 19°N seasonally changing winds associated with the Indian monsoon blow from SSE in winter and somewhat weaker from NNW in summer [e.g., Edwards, 1987]. In the central Red Sea during the winter convergence, precipitation is as high as 180 mm/year, but to the north, rainfall decreases dramatically (10 to 25 mm/year) and mainly relates to depressions and associated cold frontal troughs traveling eastward from the Mediterranean across the northern Red Sea. Generally evaporation exceeds precipitation in the Red Sea area and was estimated to amount on average 200 cm/year [Morcos, 1970].

2.2. Hydrography

[4] With average sea surface salinities (SSS) of about 40.6, the northern Red Sea represents the high-salinity end-member of the strongly developed north-south gradient in the Red Sea. Below the pycnocline at about 300 m lies an almost homogenous body of deep water [Woelk and Quadfasel, 1996]. Annual sea surface temperature (SST) is about 25.3°C with 28.5°C in summer and 22.5°C in winter. Deep water has an almost uniform temperature of about 21.5°C.

2.3. Circulation Patterns in the Red Sea

[5] Excess evaporation and the particular Red Sea bathymetry restricting the seawater exchange with the Indian Ocean through the shallow Strait of Bab el Mandeb (about 137 m b.s.l.) gives, in a first simplified approach, rise to an antiestuarine circulation (Figure 1). Relatively fresh and nutrient rich waters enter the Red Sea through the Strait of Bab el Mandeb from the Gulf of Aden. Connected to the monsoonal wind field, inflow is strong at the surface in winter and somewhat weaker and shifted to intermediate water depths in summer, thus being part of a seasonally alternating two and three layer structure over the straits [e.g., Smeed, 1997]. Based on various observations and model studies and more recently in a comprehensive numerical model [Eshel and Naik, 1997, and references therein] it was shown that the combined effects of thermohaline and wind-forcing result in northward flowing eastern and western boundary currents (Figure 1). An integral part of this large-scale surface circulation is represented by a succession of several gyres that seem to be associated with local wind patterns and topography [Quadfasel and Baudner, 1993; Eshel and Naik, 1997]. On their way north, surface waters are getting continuously denser ending up in the northern Red Sea, where southward flowing intermediate and deep waters are initiated. Intermediate and deep water formation is largely dependent on the thermohaline preconditioning of the northward flowing upper layer water over several years and the interaction with the dense subsurface waters leaving the Gulf of Suez and to a lesser extend also the Gulf of Aqaba. Different modes of deep-water renewal are suggested. A convective mode describes the descent plumes of the Gulf of Suez outflow, which supposedly contributes up to 50% to the Red Sea deep water [Woelk and Quadfasel, 1996]. Cember [1988] postulated a second mode, i.e., to be isopycnally formed intermediate water that is injected directly beneath the pycnocline in the open northern Red Sea off the tip of the Sinai Peninsula. A third mode of shallow convection initiating intermediate waters at the collision site of the Western Boundary and Eastern Boundary Currents was suggested more recently by Eshel and Naik [1997] (Figure 1). Depending on the different authors, estimates of the average residence time of the deep water vary between 200 to 300 years [Manins, 1973] and 30 to 70 years [e.g., Cember, 1988; Woelk and Quadfasel, 1996].

Figure 1.

Map of the working area showing the surface near circulation [after Eshel and Naik, 1997], the areas of intermediate and deep water formation, and the core location. In the upper right panel the antiestuarine circulation in the Red Sea is schematically shown.

3. Material and Methods

[6] Surface sediments and gravity cores were sampled along three profiles extending from the Saudi Arabian coast to the central axis of the northern Red Sea during the R/V Meteor cruise M44/3 in spring 1999 [Pätzold et al., 1999]. For the present study we selected core GeoB 5844-2 (27°42.81′N–34°40.90′E; 963 m water depth) located off the tip of the Sinai Peninsula (Figure 1) as well as a set of surface reference samples from the area. We performed multiple stable oxygen isotope analyses (continuous 1 cm sampling), determined the alkenone paleo-SSTs, and estimated past SSSs (2 to 4 cm sampling steps) from the uppermost 2 m of the sediment core. Additionally we determined the content of total organic carbon (TOC) and carbonate and logged the sediment core using a MultiSensor-Core-Logger (MSCL) with an attached video camera, and X-ray fluorescence (XRF) scanning. The resulting data sets are available at the paleonetwork for geological and environmental data PANGAEA (see

3.1. Bulk Sediment Composition

[7] Sediment samples used for TOC and carbonate analysis, were taken in intervals of 2 and 4 cm above and below a core depth of 142 cm, respectively, and then freeze-dried and ground in an agate mortar. Total carbon (TC) and TOC were determined on untreated and decalcified samples by combustion at 1050°C using a Heraeus CHN-O-Rapid elemental analyzer. Carbonate was calculated from the difference and expressed as calcite (CaCO3 = (TC-TOC) × 8.33). The results are given in weight percent dry, salt-free sediment with a relative precision of about ±4% for TOC and ±1% for carbonate.

[8] Color/gray scale and magnetic susceptibility (MS) was determined by means of the MSCL (GEOTEK) video camera and the MS point sensor device respectively. Bulk-sediment chemistry was determined with profiling XRF measurements (0.5 to 1 cm resolution) using a XRF core-scanner [Röhl and Abrams, 2000]. By this method, element intensities of the entire suite of elements between K and Sr can be recorded. Generally, Ca and Sr intensities correlate well with the carbonate content, whereas elements like Fe, Ti, and K are related to siliciclastic components and vary directly with the terrigenous fraction of the sediment [Röhl and Abrams, 2000].

3.2. Stable Oxygen Isotopes

[9] Stable oxygen isotope measurements were performed on the planktonic foraminifera Globigerinoides ruber (white), the nonmigratory pteropod Creseis acicula, the diel-migratory mesopelagic pteropod Limacina inflata, and the benthic foraminifera Cibicides mabahethi and Bulimina marginata. Sample resolution of the records is 1 cm. To avoid size-dependent effects on the δ18O values and to reduce potential intrasample noise, 50 specimens of 350 to 400 μm diameter (measured along the longest axis) of G. ruber (white) and about 30 specimens of C. acicula and L. inflata were hand-picked and homogenized. Sample preparation was performed employing an automatic carbonate preparation system attached to a Finnigan MAT 251 mass spectrometer. Analytical internal longtime precision for δ18O was better than ±0.07‰.

3.3. Alkenone Paleothermometry

[10] To determine past SST variations in the northern Red Sea, we measured the alkenone unsaturation index UK′37 as defined by Prahl et al. [1988]. Alkenones were extracted from 5–10 g portions of the freeze-dried and homogenized sediment using ultrasonication (UP200H sonic disruptor probe, S3 Micropoint, 200 W, amplitude 60%, pulse 0.5 s) and successively less polar mixtures of methanol and methylene chloride (CH3OH, CH3OH/CH2Cl2 1:1, CH2Cl2, each for 3 min). Prior to extraction, the samples were spiked with squalane as internal standard. After each extraction, the suspensions were centrifuged and the supernatants combined. The combined extracts were washed with demineralized water to remove sea salt and methanol, dried over Na2SO4, and concentrated to near-dryness under N2. The residue was redissolved in 200 μl of CH2CL2, purified by elution (3 × 500 μl CH2CL2) through a commercial silica gel cartridge (Varian Bond Elut), and saponified to eliminate possibly interfering fatty acid methyl esters. Saponification was performed with 300 μl of 0.1 M KOH in 90/10 CH3OH/H2O at 80°C for 2 hours. The neutral fraction containing the alkenones was obtained by partitioning into hexane (3 × 500 μl), concentrated under N2, and finally taken up in 25 μl of a 1:1 CH3OH/CH2Cl2 mixture (or pure CH2CL2 for on-column injection).

[11] The extracts were analyzed by capillary gas chromatography using two gas chromatographs (HP 5890A and HP 5890 series II) equipped with 60 m columns (J&W DB1, 0.32 mm × 0.25 μm and J&W DB5MS, 0.25 mm × 0.1 μm), split/splitless and on-column injection, and flame ionization detection. Helium was used as carrier gas. The oven temperature of the split-GC was programmed from 50 to 250°C at 25°C min−1, then to 310°C at 1.5°C min−1 (holding time 26 min), and finally to 320°C at 30°C min−1 (holding time 15 min). For the analyses with on-column injection, we additionally employed a deactivated fused silica precolumn (Chrompack, 1.8–2.5 m, 0.53 mm ID) and the following temperature program: initial time 5 min at 40°C, 40 to 250°C at 25°C min−1, 250 to 310°C at 1°C min−1 (holding time 25 min), then to 320°C at 30°C min−1 (holding time 5 min).

[12] Quantification of the alkenones was achieved using squalane as internal standard and Turbochrome 4 as analytical software. The alkenone unsaturation index UK′37 was calculated from UK′37 = (C37:2)/(C37:3 + C37:2), where C37:2 and C37:3 are the di- and tri-unsaturated C37 methyl alkenones. Alkenone concentrations in the sediments vary between 25 and 1300 ng/g. The measurements obtained by split (1:20) and on-column injection did not show any systematic difference. They agreed within ±0.02 UK′37 units (i.e., ±0.3°C) and were averaged. For conversion into temperature values, we used the culture calibration of [Prahl et al., 1988] (UK′37 = 0.034T + 0.039), which has been validated by core-top compilations [e.g., 1998] and comparison with other paleotemperature proxies [Müller et al., 1997; Bard, 2001]. Surface sediments (0–1 cm) collected by multiple corers in the study area had UK′37 values between 0.938 and 0.945, or 26.5 ± 0.1°C in terms of temperature. These values are in agreement with atlas SST values for this region [Levitus and Boyer, 1994] ranging midway between the mean annual (25.3°C) and summer (27.7°C) temperatures (Figure 2).

Figure 2.

Alkenone temperature and stable isotope temperatures based on surface sediment samples from the northern Red Sea compared to atlas winter (December, January, and February), summer (June, July, and August), and annual mean temperatures [Levitus and Boyer, 1994].

3.4. Paleosalinity Estimates

[13] In paleotemperature equations, SST is expressed as a function of the difference in the stable oxygen isotope composition of equilibrium precipitated carbonate (δ18Oc) and ambient seawater (δ18Ow). For our purpose we used the HL (high light) linear paleotemperature equation of Bemis et al. [1998],

equation image

where T is paleotemperature, δ18Ow is δ18O of ambient seawater, and δ18Oc is δ18O of carbonate measured versus VPDB. This has been experimentally established for the symbiont-bearing planktonic foraminifera Orbulina universa and is in good agreement with calibration data also from other shallow dwelling symbiotic foraminifera [Bemis et al., 1998].

[14] δ18Ow is known to relate linearly to SSS but with slopes differing from region to region [Fairbanks et al., 1992]. In the Red Sea a well corresponding δ18Ow-salinity relation was independently determined by Craig [1966] and Andrié and Merlivat [1989]. Valid for both surface and intermediate/deep Red Seawaters the authors established the equation:

equation image

[15] If δ18Oc and SST (measured by an independent method, i.e., alkenone method) are known, SSS can be subsequently calculated solving the equation (1) for δ18Ow and in a second step equation (2) for salinity. To reconstruct salinity values for the past, corrections of the δ18Ow have to be performed accounting for the global effect owing to water storage in the continental ice sheets. We removed the so-called “ice effect” using the glacioeustatic sea level record from [Fairbanks et al., 1992]. Finally, we augmented salinity values, accounting also for the sea level-dependent salinity changes of the global ocean with ΔS = sea level 35/3800 [Rostek et al., 1993] (LGM ∼ 1.1 psu).

3.5. Surface Water Density Calculation

[16] For the determination of the sea surface water density we applied the UNESCO International Equation of State of Seawater as described in the work of Fofonoff and Millard [1983]. Densities (in kg/cm3) are expressed as sigma 0 (= density-1000) units.

4. Chronostratigraphy

[17] Carbonate samples of mainly monospecific composition (Globigerinoides sacculifer, Orbulina universa, Creseis acicula, Creseis virgula, and Limacina inflata) were dated with a 14C Accelerated Mass Spectrometer (AMS) (Table 1, Figure 3). Carbonate hydrolysis and CO2 reduction of the samples were performed at the University of Bremen, Germany, and the 14C AMS measurements were done at the Leibniz Labor in Kiel, Germany [Nadeau et al., 1997]. 14C ages were reservoir corrected using the CALIB 4.3 calibration software [Stuiver and Reimer, 1993] with the integrated INTCAL98 calibration curve. According to the Marine Reservoir Correction Database compiled by P. Reimer (, a regional deviation from the global reservoir effect (ΔR) of ∼100 years was considered. Ages are reported in calibrated calendar years. With average sedimentation rates of 8 cm/ka, the dating results indicate an undisturbed, continuous sedimentation for the last 22 ka (Figure 3) achieving an average time resolution of ∼125 years for the stable isotope measurements and ∼250 years for the alkenone paleotemperature record and the paleosalinity estimates.

Figure 3.

Sedimentation rates and age depth relation of the sediment core GeoB 5844-2 showing the original radiocarbon dates and the calibrated ages. Radiocarbon ages were calibrated using the CALIB 4.3 software [Stuiver and Reimer, 1993] considering a regional deviation from the global reservoir effect of about ΔR ∼ 100 a.

Table 1. 14C Ages Obtained by Accelerator Mass Spectrometry Dating of Predominantly Monospecific Samplesa
Laboratory IDCore Depth, cmForam/Pteropod Species14C AMS Age, Years BP±Error, YearsCalibrated Age, Calendar Years BP
  • a

    All samples from core GeoB 5844-2 were dated at the Leibniz-Labor AMS facility in Kiel, Germany [Nadeau et al., 1997].

KIA110880G. sacculifer101535516
KIA1109010O. universa2490802002
KIA1109120G. sacculifer3920453764
KIA1109230G. sacculifer5325505584
KIA1109440G. sacculifer63201006658
KIA1127555G. sacculifer8525558922
KIA1127670G. sacculifer99905010632
KIA1127785Creseis/Limacina spp.120305513430
KIA11279100Creseis acicula131206014497
KIA11280125Creseis acicula1496012017205
KIA11281150Creseis acicula170309019587
KIA11283175Creseis acicula1858010021371
KIA11284200Creseis/Limacina spp.2054012023626
KIA11285225Creseis virgula2209016025433

5. Results

5.1. Bulk Sediment Composition

[18] The investigated sediments mainly consist of nannofossil clayey muds to nannofossil oozes. Increased Fe and Ti intensities as well as magnetic susceptibility indicate higher contents of terrigenous material during the LGM and the late glacial (Figure 4). At around 9 ka a sharp decrease is observed. Carbonate content is 40 to 50 wt % until ∼10.6 ka. After ∼9 ka values a Holocene level up to 75 wt % is reached (Figure 4). From about 15.6 to 14.4 ka (a core interval of ∼12 cm) the sediment, consisting mainly of terrigenous components and partly encrusted shells of the pteropod Creseis acicula, is intermixed with lithified carbonates (aragonite and Mg-calcite) forming a hard layer [e.g., Milliman et al., 1969]. TOC content during the LGM is ∼0.8 wt % and decreases within the hard layer to ∼0.4 wt %. On top of the hard layer the sediment turns almost black for about 1600 years (∼16 cm) and TOC shows maximum values of up to 2.9 wt %. A second dark layer with maximum TOC values of ∼1.1 wt % is centered around 10.6 ka (2–3 cm). In the following the two intervals are referred to as Red Sea sapropel 1a and 1b (RS1a, RS1b). Thereafter TOC content decreases to typical Holocene values of ∼0.3 wt %.

Figure 4.

Results on the sediment core GeoB 5844-2 showing: (a) the stable oxygen isotopes on the shallow dwelling foraminifera G. ruber (white) and C. acicula (combined record), the deep dwelling pteropod L. inflata, and the benthic foraminifera C. mabahethi and B. marginata (combined record), (b) the carbonate content, (c) the sediment color reported as gray scale values, (d) the XRF measured element intensities of Fe and Ti, and (e) the magnetic susceptibility, radiocarbon dating levels are indicated by black arrows.

5.2. Stable Oxygen Isotopes

[19] Faunal assemblages in the Red Sea were subject to major fluctuations during the last glacial and the transition to the Holocene. The most extreme restriction in faunal diversity occurred during the late glacial so-called aplanktic zone, where only some pteropod species could survive the extremely high salinities. In order to establish continuous isotope records for the last 22,000 years, we therefore combined the planktic records of the surface dwelling foraminifera Globigerinoides ruber (white) and nonmigratory pteropod Creseis acicula, and the benthic records of Cibicides mabahethi and Bulimina marginata. A third record of the diel-migratory mesopelagic pteropod Limacina inflata was measured down to 90 cm core depth (Figures 4a and 5). An average offset between C. acicula and G. ruber (white) from paired isotope measurements was determined to be about 1.9 ± 0.2‰, that is 0.5‰ more than reported by Hemleben et al. [1996], but in good agreement with measurements from the western tropical Atlantic (1.7‰; H. W. Arz, University of Bremen, unpublished data). Accordingly, we adjusted δ18O values of C. acicula to G. ruber by subtracting 1.9‰. L. inflata values were solely corrected for the fractionation offset between aragonite and calcium carbonate of about 0.6‰ [Grossman and Ku, 1986]. Paired measurements of the two benthic foraminifera showed that there is virtually no offset in δ18O between the species.

Figure 5.

Vertical profile of equilibrium δ18Ow values calculated using atlas values [Levitus and Boyer, 1994] and the paleotemperature equation of Bemis et al. [1998] and the Red Sea δ18Ow-salinity relation of Andrié and Merlivat [1989]. Average calcification depths of the foraminifera and pteropods used in this work for stable oxygen isotope measurements are indicated.

[20] Large changes are recorded for both the planktic and benthic δ18O series during the last 22 kyrs with LGM-Holocene amplitudes of ∼6.1 and ∼5.4‰, respectively (Figure 4a). The recorded amplitudes are exceeding the mean-ocean glacial-interglacial change (∼1.2‰) by a factor of 4 to 5 and are evidently the expression of the glacial salt buildup in the arid semi-enclosed Red Sea basin [e.g., Hemleben et al., 1996]. In order to gain complementary information about the past water column stratification, we calculated vertical gradients in δ18O as the difference between the planktic and the benthic δ18O records on the one hand (Δδ18Op-b), and the shallow planktic and deep planktic δ18O (reaching only back to 14 ka) on the other hand (Δδ18Op-p). Core top measurements show that, in agreement with expected values, the modern gradient for both the Δδ18Op-b and Δδ18Op-p is about 1‰ (Figure 6). During the past 22 kyrs the gradients show strong deviations of 0.5 to 1‰ from the modern mean indicating significant changes in past water column stratification.

Figure 6.

Modern distribution of the vertical δ18O gradient in the northern Red Sea as reflected by measurements on 14 surface samples on the planktic foraminifera G. ruber (white), the benthic foraminifera C. mabahethi, and the pteropod L. inflata. Horizontal lines represent average values for each species.

5.3. Paleotemperature Record

[21] SST variations between 16.7°C and 25.9°C were reconstructed in the northern Red Sea for the period from 22 to 8 ka (Figure 7). The average temperature for the LGM (22 to ∼19 ka) was 22.5°C, that is about 4.2°C colder than during the late Holocene (last 5000 years) and in the modern Red Sea [Arz et al., 2003]. At about 18 ka, the SST began to drop even below the LGM level reaching a minimum value of 16.4°C at around 16.5 ka. Subsequently, the temperature gradually recovered to 21°C at about 14.6 ka. As the maximum cooling event was represented only by one data point, we confirmed this temperature value by an additional extraction of material from the same core level. After a sharp temperature rise to almost Holocene values (25.4°C) at around 14.6 ka, the temperature decreased again since about 13.5 ka forming a second minimum (23.7°C) at around 11.8 ka and recovered then gradually until about 10 ka.

Figure 7.

Results for sediment core GeoB 5844-2 showing: (a) UK′37 temperatures based on split/splitless and on-column alkenone measurements and the mean values, (b) qualitative information on the state of preservation and abundances of coccoliths in the sediment, (c) concentration of C37 alkenones, and (d) content of total organic carbon (TOC). Radiocarbon dating levels are indicated by black arrows.

[22] A major problem in assessing alkenone paleotemperatures in the Red Sea is whether calibrations established in the modern ocean also apply to the highly saline glacial surface waters in this region. Although the alkenone producing coccolithophorids Emiliania huxleyi and Gephyrocapsa oceanica are relatively salinity-tolerant species, they may not have tolerated salinities in the order of 50 psu during the LGM [Winter et al., 1983]. However, qualitative smear slide examinations (Figure 7b) indicate that both taxa were present throughout the core interval of interest here, although in variable abundances and preservation stages (H. Legge, University of Bochum, personal communication, 2003).

[23] Hence these species appear to tolerate higher salinity levels as previously thought. We are confident therefore that the alkenone record of our core derived from these species and that the E. huxleyi calibration of Prahl et al. [1988] may also be used to estimate paleotemperatures for the LGM in the Red Sea.

5.4. Paleosalinity Record

[24] During the LGM (22 to 19 ka) paleosalinities were significantly increased to ∼50.6 psu, i.e., about 10 psu higher than during the late Holocene [Arz et al., 2003]. After about 17 ka, surface waters stepwise freshened to a pronounced minimum of ∼37 psu around 9 ka. During this interval, small salinity minima are also evident at around 16.4, 13.7, and 11.4 ka.

[25] Paleo-δ18Ow (and paleosalinity) reconstructions always combine a number of assumptions and measurements which altogether may produce substantial errors in the records and results must therefore be handled with caution. In the particular case of the Red Sea data, the amplitude of the estimated variations in δ18Ow (gacial-interglacial change of ∼4.7‰) is one order of magnitude higher than errors associated with the used methods and equations. Other potential errors are introduced in the δ18Ow calculation owing to the fact that paleotemperatures and stable oxygen isotopes are determined on different planktic groups, which potential differ in their depth habitat and seasonal distribution. The planktonic foraminifera G. ruber (white), the pteropod C. acicula, and the coccolithophorid Emiliania huxleyi are generally assumed to be surface-near dwellers. Our core top measurements from the northern Red Sea show that both alkenones and oxygen isotope temperatures represent about the same season, lying midway between the annual average and maximum summer temperature (Figure 2). Finally, reconstructed salinities are largely based on the assumption of a constant relation between δ18Ow and salinity. We suggest that this assumption is valid in the particular case of the highly evaporative Red Sea basin because we assume that only one major water source, the inflowing Indian Ocean water, contributes to the isotopic signature of the Red Seawater and freshwater input is largely negligible. However, Rohling [1994] advised some caution with this assumption, because other factors, such as mean wind speed, may have altered the relation in the past.

6. Discussion

6.1. Red Sea Last Glacial Maximum

[26] Generally, the LGM is documented in the northern Red Sea as a so-called “aplanktic zone,” a period were planktic foraminifera become extinct and an almost monospecific C. acicula fauna established [e.g., Winter et al., 1983; Almogi-Labin et al., 1998]. The lowered glacial sea level accompanied by a drastic reduction in the Red Sea-Indian Ocean water exchange, and to some extend also an increased aridity in the area [Arz et al., 2001], was responsible for anomalous high salinities in the Red Sea [e.g., Winter et al., 1983; Almogi-Labin et al., 1996; Hemleben et al., 1996; Geiselhart, 1998].

[27] According to our alkenone paleotemperature record, LGM SSTs were about 4.2°C colder than present suggesting that CLIMAP Project Members [1981] underestimated the glacial cooling by ∼2°C. Salinity estimates in the northern Red Sea were so far only based on stable isotope measurements and ecological tolerance information on foraminiferal species [e.g., Hemleben et al., 1996]. Our paleosalinity calculation for the LGM (∼50.6 psu, i.e., 10 psu more than modern and late Holocene values, Figure 8) however generally agrees with most recent results from a complex model of the glacial Red Sea (∼50 psu) [Gyldenfeldt, 1999] and SSS as determined by Hemleben et al. [1996] (53–55 psu). Our glacial paleosalinity estimate sufficiently explains the disappearance of planktic foraminifera and other organisms at that time (e.g., upper salinity limit for G. ruber and G. sacculifer is ∼49 psu; [Hemleben et al., 1989]. Nevertheless, Fenton et al. [2000] pointed out that the aplanktic zone, which actually began much earlier, in our core at around 30 ka (last MIS 3 appearance of G. ruber white), could partly be attributed to an northward expansion of the southern Red Sea OMZ. This expansion has been related to the glacial intensification of the NE monsoon regime that started as early as 60 to 70 ka [Fenton et al., 2000, and references therein]. Because pteropod assemblages, like other planktic groups, during the LGM were strongly reduced to only a few epipelagic species (Creseis acicula, Creseis virgula), their use to reconstruct water column properties in order to confirm a possible expansion of the OMZ is strongly biased [Almogi-Labin et al., 1998]. Calculated Δδ18Op-b values during the LGM are slightly lower than during the late Holocene (∼0.8‰, Figure 8), indicating that the ventilation of the water mass might have been comparable or even stronger than today and the OMZ was not significantly different that the present one. Model results [Gyldenfeldt, 1999] also suggest a very well aerated Red Sea during the LGM. Additionally glacial TOC values of ∼0.4 to 0.8 wt % may indicate a slightly increased productivity but are still relatively low for typical OMZ (Figure 7). Another striking feature worth to be mentioned is the abundant co-occurrence of the large size diatom Cosinodiscus oculus iridis in the glacial sediment (described also by Geiselhart [1998]), a species that is generally associated with coastal environments but more recently is described also as a deep chlorophyll maximum species contributing to the “fall dump” diatom flux related to the breakdown of summer stratification [Kemp et al., 2000]. High glacial abundances of this species may indicate increased productivity and a more recurrently and pronounced seasonal contrast between summer stratification and winter deep mixing at that time. The idea of a well-aerated northern Red Sea within an arid environment is supported by generally increased XRF element intensities of Fe and Ti, which point to an increased input of eolian-transported terrigenous material (Figure 4).

Figure 8.

Comparison of the (a) stable oxygen isotope record of the GISP2 ice core [Grootes and Stuiver, 1997] with (b) the reconstructed sea surface δ18Ow (salinities), (c) the alkenone paleotemperatures, (d) the vertical gradients in δ18O (calculated as the difference between planktic and benthic records, Δδ18Op-b, and between shallow planktic and deep planktonic records, Δδ18Op-p) from core GeoB 5844-2, and (e) the global sea level curve for the last 22 kyrs [Fairbanks et al., 1992]. Bold lines (Figures 8b and 8d) represent 5 point moving averages of the respective records. Dotted vertical bars show the meltwater pulses 1a and 1b [Fairbanks et al., 1992], and gray shaded vertical bars show the duration of the YD and the Heinrich event 1 [Bard et al., 2000].

6.2. Impact of Heinrich Event 1 on the Northern Red Sea

[28] Accompanied by the massive iceberg release and the consequent accumulation of ice-rafted detritus in the glacial North Atlantic [e.g., Bond et al., 1992; Broecker et al., 1992], the so-called Heinrich Events, sea temperatures dropped not only in the North Atlantic, but also in remote areas like the Mediterranean Sea and the Indian Ocean [Bard et al., 2000; Cacho et al., 2001; Kudrass et al., 2001]. At about 18 ka LGM conditions were terminated by the last of these glacial cooling events, the Heinrich Event 1, lasting from about 18 to 15.5 ka [Bard et al., 2000].

[29] With the onset of H1, in the northern Red Sea SSTs gradually cooled with a maximum cooling centered around 16.5 ka (a cooling of >10°C with respect to late Holocene values; Figure 7). SSSs were slightly reduced (49‰) and the Δδ18Op-b gradient was not significantly different than during the LGM. The timing of the maximum cooling observed at our core site is closely corresponding to the Soreq Cave stable isotope record from Israel [Bar-Matthews et al., 1999], which shows a post LGM cooling peak at 16.5 ka. The same maximum cooling related to H1 and centered around 16.5 ka is documented also in the SW Aegean Sea [Geraga et al., 2000] (foraminiferal assemblage estimates) and the Alboran Sea and Tyrrhenian Sea [Cacho et al., 2001] (alkenone SSTs). According to Cacho et al. [2001] one possible explanation for the temperature drop below LGM temperatures in the Mediterranean Sea is a direct influence of advected cold polar water masses through the Gulf of Cadiz. However, Cacho et al. [2001] suggested also that to a certain extent the cooling signal could have been transferred atmospherically from the northern high-latitude Atlantic through the westerly wind belt to these areas. Such a process could possibly have extended also to the Near East region. There is, however, strong evidence that the H1 related climate deterioration affected also the Asian Monsoon system. Alkenone-based SST reconstructions on sediment cores from the northeastern Arabian Sea [Schulz et al., 1998; Schulte and Müller, 2001] and the Bay of Bengal [Kudrass et al., 2001] indicate a cooling of the surface waters synchronous to the H events and also to some weaker stadials. Together with the reconstructions from the Chinese Loess Plateau [e.g., An Zhisheng, 2000], it therefore appears that concomitant to the North Atlantic Heinrich events, the monsoon repeatedly switched from a SW monsoon dominated to a more NE monsoon dominated system. As Red Sea/Indian Ocean water exchange at that time was strongly restricted by the low glacial sea level, it is most unlikely that the observed SST drop in the northern Red Sea is attributed to an import of colder Arabian Sea surface waters. More likely is an atmospherically induced cooling of surface waters related to either the westerlies or an enhanced NE monsoon. Lake level reconstructions from the glacial Lake Lisan, the present Dead Sea, indicate that more humid conditions prevailed during the Heinrich Events [Stein et al., 1999]. Assuming that an enhanced NE monsoon would cause rather arid conditions in that area but SSSs during the maximum H1 cooling are slightly reduced, a stronger influence of the westerly wind system on the northern Red Sea region is more plausible.

[30] The end of the H1 event is associated in the Red Sea sediments with a significant change in the sediment composition related to the formation of a hard layer. The hard layer is thought to be the result of inorganic carbonate precipitation in a high-salinity/high-temperature benthic environment [e.g., Milliman et al., 1969] but latest results also suggest a microbial mediation of carbonate precipitation [Brachert, 1999]. During formation time of the hard layer, in our record both SSTs and SSSs increased again (Figure 8). Assuming comparable changes in the benthic environment, hard layer formation in the northern Red Sea may represent a direct response to the post-H1 temperature and salinity rise.

6.3. Bølling-Allerød-Younger Dryas Climate Transitions and Deglacial Sea Level Rise

[31] Concomitant with the transition into the RS1a sapropel at around 14.5 ka, the alkenone temperatures record an abrupt temperature rise of about 4.5°C (to 25.4°C) within a period of less than 250 years. This seems to be almost synchronous to the abrupt transition leading into the Bølling-Allerød warm period like it is documented in the GISP2 ice core (transition GS-2 to GI-1) [Björck et al., 1998]. A clear SST minimum (∼11.8 ka) falls within the Younger Dryas cold period (GS-1).Unlike the ice core records, the YD related cooling in our record is not terminated by an abrupt warming. Instead, temperature is gradually increasing from 11.6 to 10 ka by about 2°C, paralleling the gradual warming trend after the abrupt warming at the end of the YD in the GISP2 record. Similar geometry is reported also from alkenone records in the Mediterranean Sea [Cacho et al., 2001], which are interpreted to represent a warming lead with respect to the high-latitude signal. Nevertheless, the overall correspondence of the Red Sea surface temperatures with the high-latitude signal and especially the Mediterranean records [Cacho et al., 2001] during the termination I suggest a prominent link to changes of the high- and middle-latitude atmospheric circulation, i.e., the northern westerlies.

[32] With a delay to the abrupt temperature rise around 14.5 ka, SSSs decrease by around 3 psu and are accompanied by a significant increase in the Δδ18O gradient (from 1‰ to 2.5‰) (Figure 8) both pointing to a surface freshening and a strongly stratified water column at that time. During the YD, SSSs are slightly increased. The reduced Δδ18Op-b and especially Δδ18Op-p values during the YD (Figure 8) point to a cooling-induced deep convection during this probably also more arid period with a mixed layer deepened well below the modern level. A second, but minor freshening peak around 11.4 ka is accompanied again by a small increase in Δδ18Op-b (1.4‰), suggesting, although less pronounced, another phase of increased stratification of the uppermost water column.

[33] As demonstrated by Hemleben et al. [1996] on the basis of the empirical sea strait model of Assaf and Hecht [1974], a strong dependence on sea level changes exists for the Red Sea SSS. If compared to the eustatic sea level curve of Fairbanks et al. [1992], the two freshening peaks in our SSSs record around 14.5 ka and 11.4 ka closely correspond to the postglacial meltwater pulses MWP1a and MWP1b (Figure 8). The model-predicted salinity changes for each of the pulses, both representing a sea level rise of more than 24 m (MWP1a ∼2.9 psu, MWP1b ∼1.5 psu), are close to the reconstructed salinities (MWP1a ∼3.5 psu, MWP1b ∼1.6 psu). The associated Δδ18Op-b/Δδ18Op-p anomalies, suggesting an increased stratification, most probably relate to the increased surface-near entrance of less saline waters into the Red Sea that subsequently are removed to intermediate depths, reducing the gradient again to values around 1‰ (Figure 8).

6.4. Formation of the Red Sea Sapropels

[34] Both sea surface temperatures and salinities play a crucial role in seasonal deep mixing and accordingly, intermediate and deep water formation processes in the northern Red Sea. The additive effect of a strong warming and freshening of the surface waters that relates to the stepwise sea level rise, i.e., the fast changes of the surface water density (Figure 9), could have suppressed regular deep water formation. Resulting stagnant bottom water conditions probably led to the formation of the RS1a and RS1b sapropels. The period of RS1a formation corresponds to the I-4e interval as determined by Almogi-Labin et al. [1991] for the central Red Sea, which is described as an interval with well-stratified surface water conditions. RS1a seems to be terminated with the onset of the surface water cooling associated with the YD leading to a increased rate of deep water formation and ventilation. Else than expected, the top of RS1b falls in a period of continuously decreasing surface water densities (Figure 9). Presently, the shallow Gulf of Suez (with an average water depth of 35 to 50 m) contributes up to 50% to the Red Sea deep water. According to the eustatic sea level curve of Fairbanks et al. [1992] the Gulf of Suez was flooded around 11 to 10 ka. This may have led to the initiation of a additional and more important mode of deep-water formation, terminating the stagnant bottom water conditions responsible for the formation of the RS1b. The absence of sapropel sediments during the anomalous low SSS reconstructed for the early Holocene [Arz et al., 2003] probably relates to the same process.

Figure 9.

Calculated surface water density changes in the northern Red Sea over the past 22 kyrs compared to the TOC content and the brightness of the sediments. Note the reversed TOC scale. Intervals of dark sediment color and enhanced TOC values are labeled as Red Sea sapropel 1a and 1b (RS1a and RS1b).

7. Conclusions

[35] 14C AMS dated core GeoB 5844-1 provides a continuous, stratigraphically well-constrained, high-resolution record of environmental changes in the northern Red Sea during the past 22 ka. Multiple stable oxygen isotope measurements (foraminifera, pteropods), paleotem perature determinations, and paleosalinity calculations, as well as information about the sediment composition (e.g., carbonate, TOC) add to a complex picture of late glacial and deglaciation events in the northern Red Sea, which generally relate to the global changes in the climate system during this period.

[36] 1. Sea surface temperatures in the northern Red Sea are strongly connected to the general temperature development of the Northern Hemisphere since the LGM. Most likely, the westerly wind belt is responsible for transferring the middle- to high-latitude signals to this area. LGM temperatures were >4°C colder than the late Holocene. During Heinrich I event temperatures even exceeded LGM values by another 3–6°C. When compared to the Greenland GISP2 ice core record, an almost synchronous transition to the Bølling-Allerød period is recorded. The Younger Dryas signal seems to be marked by more gradual transitions in the Red Sea.

[37] 2. Northern Red Sea salinities are clearly dominated by the postglacial sea level rise. LGM salinities were as high as 50.6 psu and the stepwise increase in sea level is documented in two “low salinity” anomalies related to an increased near-surface inflow into the Red Sea of “relatively fresh” seawater during the MWP1a and MWP1b [Fairbanks et al., 1992].

[38] 3. Vertical gradients in stable oxygen isotopes closely resemble the repeated intrusion of lower saline seawater indicating a stronger stratification during these periods. The YD cold period is documented by strongly reduced gradients indicating a deepening of the mixed layer.

[39] 4. Two sapropel layers indicate a repeated propagation of anoxic condition into the basin (14.6–13.2 ka and 11–10.4 ka), both occurring during periods of fast decreasing surface water densities that probably prevented a regular deep mixing and intermediate and deep water formation. Since the flooding of the Gulf of Suez, an additional important source of deep water was initiated, resulting in well-oxygenated bottom waters.


[40] We thank Monika Segl and her team for stable isotope measurements and Hella Buschhoff and Dietmar Grotheer for the geochemical analyses. We are also grateful to Frank Lamy, Thomas Felis and two anonymous reviewers for comments and suggestions that greatly improved the manuscript. We also acknowledge the generous grant of permission for conducting research in the territorial waters of the Kingdom of Saudi Arabia with the German research vessel METEOR in March/April 1999. This work was funded by the Deutsche Forschungsgemeinschaft (grants PA 492/4-1 and PA 492/4-2) and as part of the DFG Research Center “Ocean Margins” of the University of Bremen, RCOM0076.