Marine Micropaleontology 72 (2009) 146–156
Contents lists available at ScienceDirect
Marine Micropaleontology j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / m a r m i c r o
Factors controlling the distribution of planktonic foraminifera in the Red Sea and implications for the development of transfer functions Michael Siccha ⁎, Gabriele Trommer, Hartmut Schulz, Christoph Hemleben, Michal Kucera Institute of Geosciences, Eberhard-Karls University of Tuebingen, Sigwartstr.10, 72076 Tuebingen, Germany
a r t i c l e
i n f o
Article history: Received 20 November 2008 Received in revised form 7 April 2009 Accepted 14 April 2009 Keywords: Planktonic foraminifera Transfer functions Productivity Red Sea
a b s t r a c t The Red Sea is an extreme marine environment, with conditions limiting the application of standard geochemical proxies for the reconstruction of paleoclimate. In order to develop paleoenvironmental reconstruction methods which are not dependent on chemical signals, we investigated the distribution of planktonic foraminifera in the surface sediments and assessed the viability of constructing foraminiferal transfer functions in this basin. We find a distinct gradient in the faunal assemblage along the basin's axis, which is reflected in a high correlation between faunal composition and all considered environmental parameters (temperature, salinity, chlorophyll a concentration, stratification, and oxycline depth). As a result, transfer functions constructed by different methods (ANN, MAT, IKM, WA-PLS) appear to be able to estimate all of these parameters with a high average accuracy (15% of the parameter's range in the Red Sea). However, redundancy analysis of the distribution of foraminiferal assemblages in surface sediments alone did not yield unambiguous results in terms of which of the considered factors exerts a primary control on the foraminifera distribution and which of the observed relationships are the result of the mutual correlation among the environmental factors. To disentangle the effect of individual environmental parameters, we applied the obtained transfer functions on a newly generated Holocene record from the central Red Sea. The integration of published paleoclimate reconstructions with our data allowed us to identify productivity as the most likely primary control of the planktonic foraminifera distribution in the Red Sea. The generated transfer functions can estimate paleoproductivity with acceptable accuracy (RMSEP chlorophyll a = 0.1 mg/m3; ~ 8% of recent range), but only under such conditions in the past when circulation patterns and salinity levels in the basin were fundamentally comparable to the present day. Since productivity in the central and southern Red Sea is closely linked with the Monsoon-driven water exchange across the Strait of Bab al Mandab, the resulting reconstructions can provide indirect information on the mode and intensity of the monsoonal system in the past. © 2009 Elsevier B.V. All rights reserved.
1. Introduction Reconstructions of past environmental conditions in the Red Sea have so far been limited to two distinct aspects. The first one includes qualitative reconstructions of water-column properties, climate and circulation patterns by analyses of fossil assemblages of pteropods (Almogi-Labin et al., 1998, 1991), benthic foraminifera (Badawi et al., 2005), planktonic foraminifera (Edelman-Fürstenberg et al., 2009; Fenton et al., 2000; Halicz and Reiss, 1981; Ivanova, 1985) and coccolithophorida (Legge et al., 2006, 2008). The second aspect concerns reconstructions of salinity by means of stable oxygen isotope analysis in shells of planktonic foraminifera (Hemleben et al., 1996) combined with oceanographic modelling (Rohling, 1994; Siddall et al., 2004). The stable isotopic signals in the Red Sea are dependent on the
⁎ Corresponding author. E-mail address:
[email protected] (M. Siccha). 0377-8398/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.marmicro.2009.04.002
water exchange with the open ocean and employed in models translating isotopic records to variations in salinity and indirectly to sea level changes (Siddall et al., 2004). The special characteristics of the Red Sea basin (low productivity and extreme salinities) complicate the use and interpretation of most standard geochemical proxies. Until now, only one quantitative paleotemperature record has been obtained by analysis of alkenones from a core in the very north of the Red Sea (Arz et al., 2003). Attempts to use Mg/Ca in planktonic foraminifera have been confronted with unusually high Mg values and possible inorganic precipitation (Rohling et al., 2008). In this study we present the first planktonic foraminifera transfer function approach for paleoclimate reconstruction in the Red Sea. Auras-Schudnagies et al. (1989) has shown that the distribution of planktonic foraminifera in Red Sea sediments is characterised by a distinct gradient, which reflect the circulation pattern in the basin. This work has shown that planktonic foraminifera assemblages in the Red Sea are distinct from those in the open ocean, precluding the use of calibration data from outside the basin. The aim of this study was to develop a well
M. Siccha et al. / Marine Micropaleontology 72 (2009) 146–156
147
Fig. 1. Map and environmental gradients of the Red Sea. The map shows the positions of the surface samples (circles) and the investigated core KL9 (triangle). The environmental data are subsampled along the basins central (trench) axis. Stratification is calculated as Brunt-Väisälä frequency of the top 300 m water column. Oxycline depth is the depth of the 2 ml O2/L oxycline.
constrained calibration dataset and to use it to construct faunal transfer functions and evaluate the results for improved reconstructions of sea surface properties in the Red Sea. 2. Regional setting The Red Sea (Fig. 1) is a desert-enclosed, narrow basin of ~ 2000 km length, with a maximum width of ca. 350 km. A trench with a depth of up to 2500 m runs along the axis of the otherwise shallow sea (average depth ca. 500 m). The only connection of the basin to the open ocean is through Bab al Mandab in the south, leading to the Gulf of Aden in the Indian Ocean. Within the Bab al Mandab, while less suspicious than the just 18 km wide Perim Narrows, the Hanish Sill is with only 137 m depth the critical point for inter ocean exchange (Smeed, 2004). The Red Sea circulation system is driven by the high evaporation rates of up to 2.1 m yr− 1 in the north and the monsoon controlled seasonal winds affecting the entire basin (Sofianos and Johns, 2002). The resulting oceanography of the Red Sea is rather complex and beyond the short and simplified overview below, (for a more complete review about Red Sea oceanography, the reader is referred to Morcos, 1970; Neumann and McGill, 1962; Sofianos and Johns, 2003, 2007 and references therein; Tragou and Garrett, 1997). Two distinct circulation regimes through the Bab al Mandab alternate over the course of the year (Murray and Johns, 1997; Patzert, 1974; Siddall et al., 2004; Smeed, 1997). During October to April (Indian NE Monsoon) the circulation is two-layered; surface water from the Gulf of Aden enters the Red Sea, reaching up to 25°N, while Red Sea deep water flows out of the basin. From May to September (Indian SW Monsoon) the circulation pattern is three-layered. Above the bottom layer of outflowing Red Sea water, which is weaker than in winter, a layer of nutrient-rich intermediate water, upwelled by the Indian SW Monsoon in the Gulf of Aden, enters the Red Sea. At the surface, a thin, wind-driven layer of surface water leaves the Red Sea. The circulation pattern of deep water masses in the basin and the deep water formation in the north of the Red Sea and the adjoining Gulf of
Aqaba and Gulf of Suez are still not fully understood (Cember, 1988; Eshel et al., 1994; Manasrah et al., 2004; Woelk and Quadfasel, 1996). Fig. 2 illustrates the consequences of the circulation pattern in the Red Sea on the primary productivity of the surface water. During the summer, upwelling increases in the Gulf of Aden, but the inflowing nutrient-rich intermediate waters entering the Red Sea can fuel increased productivity only in the very southern Red Sea, south of
Fig. 2. Contour plot of surface chlorophyll a concentration over one year (averaged over 5 years) in monthly resolution. Dotted line denotes Bab al Mandab (Perim Narrows).
148
M. Siccha et al. / Marine Micropaleontology 72 (2009) 146–156
14°N, and the resulting chlorophyll concentrations there are markedly higher than those of the winter season. The central and northern regions of the Red Sea show their productivity maximum, still with very low chlorophyll concentrations, during the winter, apparently powered by the inflow of Gulf of Aden surface waters (rather than intermediate waters as in summer). This is consistent with the observation of high plankton biomass in the Gulf of Aden in winter by van Couwelaar (1997). In contrast to the monsoon-related summer productivity in the Red Sea, which is high but patchy, the surface inflow during the winter is more homogenous and also more voluminous than the summer inflow (to compensate evaporative loss in the north, (e. g. Siddall et al., 2002). Therefore, the inflowing surface water in winter stimulates a higher primary productivity in the Red Sea much further to the north than the Gulf of Aden water entering the basin during the summer. The origin of the weak but distinct productivity maximum in the very north of the Red Sea (north of 25°N, Fig. 2) during the winter remains unclear, but is most probably connected to mixing processes in the water column during deep water formation. Directly connected to the pattern of primary productivity is the development of an oxygen minimum zone (OMZ), more prominent in the southern part of the Red Sea, with oxygen concentrations as low as 0.5 ml/l in its core. The 2.0 ml/l oxycline lies at 75 m depth in the south (b16°N) and deepens continuously till it reaches depths of around 300 m north of 25°N (World Ocean Atlas 2001, Conkright et al., 2001; Neumann and McGill, 1962). The north–south orientation of the Red Sea in combination with restricted water exchange only at its southern end, result in strong and mutually correlated gradients of most oceanographic parameters along the basins axis (see Table 1). This strong correlation among virtually all oceanographic parameters poses a challenge to the identification of individual factors which control the distribution of planktonic organisms in the Red Sea (Auras-Schudnagies et al., 1989). 3. Materials and methods 3.1. Surface samples The distribution pattern of planktonic foraminifera species was determined by analysing 60 surface sediment samples from the Red Sea and the Gulf of Aden, collected during three cruises (see online supporting material). None of these samples showed any signs of modification of the assemblages due to carbonate dissolution. This is important to note because samples from the Gulf of Aden could potentially be affected by dissolution (Almogi-Labin et al., 2000), whereas carbonate dissolution is know to be minimal throughout the Red Sea (Almogi-Labin et al., 1998). Indeed, we have initially considered seven surface sediment samples from the Gulf of Aden, but rejected two of them because of signs of dissolution. The samples were washed over a 63 μm sieve, dry-sieved over 150 μm mesh and split with an ASC Scientific microsplitter. For each sample an aliquot containing at least 300 individual planktonic foraminifera was counted under the binocular and the relative abundances of planktonic foraminifera species were determined, following the taxonomy of Hemleben et al. (1989). The resulting database contains
more than twice as many samples as the previous study by AurasSchudnagies et al. (1989). Apart from eight counts of samples that were also included in that study, all the additional samples were taken from multicorer tubes from the top 1 cm or the top 0.5 cm. In order to achieve the highest possible level of taxonomic consistency in the counts, all of the 60 samples in this study have been counted by the authors, and the counts were checked repeatedly for taxonomic consistency. Multicorer hauls are ideal to recover undisturbed surface sediment and the recent age of multicorer coretops from the central Red Sea is supported by radiocarbon dates in Edelman-Furstenberg et al. (2009). The count data are available in the online supporting material. The resulting database of species abundances was compared with a series of environmental parameters describing the most pertinent properties of the surface waters. The analysis of open-ocean planktonic foraminifera faunas by Morey et al. (2005) suggests temperature as the dominant factor controlling the distribution of planktonic foraminifera. However, in the restricted environment of the Red Sea, many oceanographic parameters (e.g., salinity) show much greater gradients than in the open ocean and we have therefore considered the effect of not only temperature but also salinity, chlorophyll a concentration, oxycline depth and stratification (Fig. 1). The latter two variables are rarely considered when analysing the distribution of planktonic foraminifera, but they are very closely coupled to the circulation pattern in the Red Sea, which on the basis of earlier studies (Almogi-Labin et al., 1991; Auras-Schudnagies et al., 1989) seems to represent the most important factor affecting plankton distribution in the basin. Temperature, salinity and oxygen concentration data used in all following analyses represent annual average values from the 10 m depth level of the World Ocean Atlas 2001 (WOA; Conkright et al., 2001) interpolated to the position of the samples by area weighting. Investigation of vertical temperature and density gradients in the Red Sea leads to the estimation of a very shallow mixed layer depth, ranging from 30 m in the south to around 70 m in the north of the basin (Monterey and Levitus, 1997). These values reflect the layering of the circulation system but most probably not the vertical extent of the habitat of planktonic foraminifera. Therefore the Brunt-Väisälä frequency (BVF) was calculated from WOA data for each depth interval and the average of the top 300 m of the water column was then used as a measure of stratification. The BVF is the frequency with which a packet of water will oscillate when it is vertically displaced from its position in the water column. High frequencies correspond to large density differences found typically in a stratified water column, whereas low frequencies will be found in a more homogenous water column. Though the employed average of BVF is highly correlated with mixed layer depth, it does not hold any depth-related information per se but is a measure of water column stability and/or the intensity of the pycnocline. Surface water chlorophyll a concentrations were obtained from the Ocean Color Web (Feldman and McClain, 2006). We note that the surface chlorophyll a data might not necessarily represent the primary productivity or the biomass of all phytoplankton groups throughout the water column, but in the absence of vertically resolved data at appropriate resolution, the satellite data represent the best proxy for productivity available. The annual average
Table 1 Correlation of environmental variables in the surface sample dataset.
Temperature [annual mean at 10 m; WOA 2001] Salinity [annual mean at 10 m ; WOA 2001] Chlorophyll a [annual mean of 5 year period; Merged AquaMODIS-SEAWIFS] Distance to Bab al Mandab [Perim Narrows at 12.36′ °N, 43.22′ °E] Stratification [mean BVF of the top 300 m annual mean water column; WOA 2001] Oxycline depth [depth of annual mean 2 ml O2/l oxycline; WOA 2001] All values are significant at p b 0.05.
Temperature
Salinity
Chlorophyll a
Distance
Stratification
Oxycline depth
1.00 − 0.69 0.34 − 0.73 0.95 − 0.82
− 0.69 1.00 − 0.80 1.00 − 0.82 0.95
0.34 − 0.80 1.00 − 0.76 0.57 − 0.73
− 0.73 1.00 − 0.76 1.00 − 0.84 0.96
− 0.95 − 0.82 0.57 − 0.84 1.00 − 0.91
− 0.82 0.95 − 0.73 0.96 − 0.91 1.00
M. Siccha et al. / Marine Micropaleontology 72 (2009) 146–156 Table 2 14 C ages obtained by accelerator mass spectrometry dating of monospecific samples (Globigerinoides sacculifer) in the 250–315 μm fraction in core KL9, performed at the Leibniz-Labor AMS facility in Kiel, Germany. Lab.-ID
depth [cm]
Conventional age [yr BP]
± Error [yr]
Calibrated age [cal. yr BP]
KIA33786 KIA33785 KIA33130 KIA33131
2 24 60 84
1315 3260 6595 8955
25 30 35 40
732 2922 6976 9495
values of all variables were used in all analyses. Summer (JJA) and winter (DJF) averages were considered in the distribution analyses as well. 3.2. Downcore samples In order to test the environmental relationships recovered from the surface sediment dataset by means of transfer functions, new faunal data were produced for the Holocene of core M5/2-143 GeoTÜ KL (referred to as KL9, 19°57.6′N, 38°8.3′E, 814 m; Nellen et al., 1996). A total of 45 samples were taken at 2-cm intervals from the top 90 cm of the core. These samples were processed and counted in exactly the same way as described for the surface samples earlier. These data are available in the online supporting material. The age model of the Holocene section of core KL9 is based on linear interpolation between four 14C AMS dates (Table 2). The resulting sedimentation rates and the implied correlation of faunal patterns are in good agreement with published data from the 14C AMS dated core M5/2-174 GeoTÜ KL (referred to as KL11, 18°44.5′N, 39°20.6′E, 825 m) (Schmelzer, 1998) and multi-cores from the central Red Sea (Edelman-Fürstenberg et al., 2009). The transfer functions have been tested on Holocene material, since on glacial–interglacial timescales the planktonic faunas of Red Sea are affected by extreme variations in salinity (Fenton et al., 2000). Such variations are not captured by any of the surface samples and it is a-priori obvious that applications of any transfer function beyond the interglacial sea-level high stands will be confronted with nonanalogue faunas. 3.3. Analysis of faunal data and transfer function design In order to determine the impact of an environmental factor on the planktonic foraminifera assemblage we first employed gradient analysis using the Canoco 4.5 software package. An initial detrended canonical correspondence analysis showed a gradient length of 1.89, suggesting a direct gradient analysis with a linear response model as the most appropriate method to analyse the foraminifera data (Leps and Smilauer, 2003). The redundancy analysis was conducted with untransformed percentage data and scaling focussed on speciescorrelations. For counts of 300 individuals, the 95% confidence level for a detection limit of a species corresponds to observed abundance of 1% (Dryden, 1931; Revets, 2004). Therefore the faunal data of each sample was purged of species not reaching 1% relative abundance in order to avoid the influence of rare species which might or might not have been recorded by chance. The purged data were then recalculated to 100%. The results of the redundancy analysis were used to identify environmental factors strongly affecting the foraminifera distribution. The relationship between the faunal distribution and these environmental factors was then characterised quantitatively with the aim to develop transfer functions that will allow us to reconstruct past environmental conditions in the Red Sea from fossil foraminifera assemblages. For the identification of non-analogue samples, downcore and core–top faunal counts were analysed in a joint principal component analysis. This analysis was based on logratio-transformed data following the recommendations of Pollard and Blockley (2006).
149
Our calibration dataset of 60 samples is relatively small for the development of transfer functions and could lead to method-specific bias (Kucera et al., 2005). In addition, the foraminifera distribution data in the Red Sea can be expected to show a high autocorrelation (Auras-Schudnagies et al., 1989), which is known to lead to underestimation of error rates in certain methods (Telford et al., 2004). Therefore, we employed four different approaches with different sensitivity to the potential sources of error mentioned above: the modern analogue technique (MAT; Hutson, 1980), the method after Imbrie and Kipp (IKM; Imbrie and Kipp, 1971), the weighted averaging partial least square regression (WA-PLS; ter Braak and Juggins, 1993) and the artificial neural networks approach (ANN) introduced by Malmgren and Nordlund (1997). For the first three approaches we used the C2 software package (Juggins, 2003). For the ANN approach we used the NeuroGeneticOptimizer© 2.6.142 to develop back propagation networks consisting of a maximum of four neurons in one hidden layer. A genetic algorithm was applied to optimise the network structure (number of neurons and type of transfer function) by training a population of 75 networks over 25 generations, with a maximum of 1000 learning epochs. In each case, the best networks included four neurons in the hidden layer. Network fitness was based solely on the test set, consisting of 50% of the surface dataset. The test set samples were randomly chosen by the software and each partition was manually checked for consistency in the coverage of the environmental gradient. The ANN results are reported as the mean of the five best networks obtained from five different training set partitions for any single factor. In case of the MAT approach, we calculated the Bray-Curtis-dissimilarity and used the weighted average of the three most similar samples. In the WA-PLS we used the number (usually just one) of components, which yielded the lowest RMSEP (root mean squared error of prediction). For the IKM we used the simple Kaiser-Guttman criterion (eigenvalue N 1.0) to limit the number of factors to three and included the quadratic terms in the regression. For all methods except ANN, the validation of the transfer functions was performed by bootstrapping the calibration dataset (1000 cycles). The limited number of recent surface samples and their uneven distribution made the random selection of a validation subset unsuitable; instead we relied on the bootstrap approach for crossvalidation. The prediction error was expressed as RMSEP and estimated as the average RMSE (root mean square error) of the bootstrapping cycles. Due to the high time-demand for the training of artificial networks, we did not attempt to validate the results by a comparable bootstrapping, but presented only the average RMSE and R2 values for the test sets of the five partitions. 4. Results 4.1. Faunal distribution of planktonic foraminifera A total of 30 different species (see Appendix A) of planktonic foraminifera in the size fraction N150 μm were identified in the surface samples, of which the three most abundant species, G. sacculifer, G. ruber and G. glutinata contributed 73.3% of all individuals. Since
Table 3 Taxonomic units used for faunal data analysis. Globigerina bulloides Globigerinita glutinata Neogloboquadrina pachyderma merged with Neogloboquadrina incompta Globigerinoides ruber Globigerinoides sacculifer Globigerinella siphonifera merged with Globigerinella calida Globoturborotalita tenella Gulf of Aden species incl. Globorotalia menardii Neogloboquadrina dutertrei
150
M. Siccha et al. / Marine Micropaleontology 72 (2009) 146–156
most of the species occurred only at very low relative abundances, the mathematical analyses of the data were based on a reduced dataset of the species, which were considered most important. A distinct pair of species occurring only in the Gulf of Aden samples and only found in the southernmost of the Red Sea samples, make up less than 1% of the assemblage and were lumped together into one category. In order to achieve maximum consistency in the data, the species G. siphonifera
and G. calida, which are difficult to differentiate, were treated as one category, as was N. pachyderma and N. incompta. The latter two differ only in coiling direction (Darling et al., 2006) and the taxonomic significance of this character outside the polar and subpolar waters could not yet be established. The data reduction leaves us with a matrix of eight taxonomic units encompassing eleven species (Table 3), serving as mathematically and ecologically reasonable
Fig. 3. Contour plots of relative abundances of taxonomic units in the surface sample data. The contouring was done by kriging using the Surfer 8.05 software. Note that the values for the Gulf of Aqaba and the Gulf of Suez are extrapolated, as no actual counts were included in the analyses from those regions (confer Fig. 1).
M. Siccha et al. / Marine Micropaleontology 72 (2009) 146–156
151
parameter, indicating that the shapes of environmental gradients remain essentially the same throughout the year. Since seasonality as we can assess it does not seem to add any detectable variation to the dataset, all further analyses use annual average values of the environmental parameters. 4.2. Transfer functions
Fig. 4. Plot of the redundancy analysis of processed assemblage data and environmental parameters. The arrows present the annual mean of the respective parameter, connected circles indicate the arrow tips of winter DJF (open circles) and summer JJA (crossed circles) season.
input for the transfer functions. These eight taxonomic units comprise 95.4% of the initial foraminifera assemblage data. The planktonic foraminifera distribution patterns in our samples are consistent with the results of the study of Auras-Schudnagies et al. (1989). The species G. sacculifer, G. tenella and to a lesser extent N. pachyderma and N. incompta are typical for the northern Red Sea, while G. glutinata and G. bulloides are typical for the southern Red Sea. G. siphonifera and G. calida are well present throughout the basin, as is G. ruber, which is also abundant in the Gulf of Aden. Within our study area the Gulf of Aden species, including G. menardii and N. dutertrei, are restricted to their name-giving basin and the southernmost part of the Red Sea. Noteworthy is the high degree of spatial complementarity between the occurrence of G. sacculifer and G. glutinata (Fig. 3). The high correlation among environmental parameters in the surface sample dataset (Table 1) is clearly visible in the redundancy analysis (Fig. 4, Table 4). It showed the dominance of one single parameter (71.1% of the variance), represented by the horizontal axis, responsible for the distribution of foraminifera species. This parameter is highly correlated with salinity, oxycline depth, stratification and chlorophyll a concentration and can roughly be equated to the geographical position along the basin's axis. The second axis explains only 5.4% of the variance and correlates highest with temperature and stratification. Further axes add only negligible amounts to the explained variance (1.9% and 0.8%, respectively) of the species distribution. In order to evaluate the effect of seasonal variation in the investigated parameters on the foraminifera distribution, the redundancy analysis was carried out including seasonal values of each parameter (Fig. 4). In all cases, there is no major difference in the position of the annual and the seasonal values of the same
Since the redundancy analysis failed to identify unambiguously one of the parameters as the most significant factor controlling the foraminifera distribution, we constructed transfer functions with the four described methods for all five environmental parameters (Table 5). All the different types of transfer functions reproduced the environmental parameters of the surface sample dataset accurately (average R2 = 0.88). The accuracy of the estimation follows the respective parameter's correlation with the first axis of the redundancy analysis. Chlorophyll a concentration is an exception to this pattern as it appears to be estimated better than its correlation with the first axis suggests. A comparison of the different methods showed that after ANN, the validation of which is, however, not fully comparable due to the lack of bootstrap estimates of prediction errors, MAT performed second best, followed by WA-PLS and IKM. Considering transfer functions' general frailty to spatial autocorrelation (Telford et al., 2004; Telford and Birks, 2005), which can be expected to be especially pronounced in the Red Sea, it can be assumed that our validation still significantly underestimates the true RMSEP. 4.3. Results for the Holocene of core KL9 The Holocene data from core KL9 contained 20 species with similar proportions as represented in the surface sediment dataset (Fig. 5). This is clearly seen in Fig. 6, which shows the results of a PCA of the logratio transformed data of combined surface and Holocene sample sets. The first component of the PCA contrasts the Gulf of Aden and southern Red Sea assemblages from the northern surface sediment samples and all Holocene samples. The second axis broadly captures the faunal gradient from the northern (component 2 negative) to the central Red Sea (component 2 positive). The broad overlap between the two datasets indicates that at no time during the Holocene did the planktonic foraminifera fauna deviate from assemblages found in the Red Sea at the present day. In fact, all Holocene assemblages find their most similar coretop counterparts in the region north of 16°N, indicating that the fauna of the southern Red Sea, has never expanded beyond the position of core KL9 during the last 10,000 years. The absolute abundance of planktonic foraminifera remained after a clear increase in the earliest Holocene relatively stable during the last 7000 years (Fig. 5). The three most abundant species however, show distinct trends for the observed time interval. Relative abundances of G. ruber and G. glutinata increased monotonously until a core depth of 25 cm (~3000 years). This increase is at the expense of G. sacculifer, whose relative abundance decreased accordingly. After
Table 4 Results of the redundancy analysis of the surface sample dataset. Axis 1 Eigenvalue Species-environment correlation Cumulative percentage variance of species data of species-environment relation Correlation Temperature Salinity Chlorophyll a Stratification Oxycline depth
0.711 0.949 71.1 89.7 0.567 − 0.900 0.754 0.865 − 0.892
Axis 2 0.054 0.828 76.5 96.5 − 0.649 0.111 − 0.497 0.157 0.197
Axis 3 0.019 0.567 78.4 98.9 0.028 0.123 0.043 0.166 0.001
Axis 4
Captured variance
0.008 0.447 79.2 99.9 0.062 − 0.054 0.028 0.074 − 0.020
28.7% 64.1% 59.4% 46.8% 63.1%
152
M. Siccha et al. / Marine Micropaleontology 72 (2009) 146–156
Table 5 Error estimates for constructed transfer functions.
Salinity [psu]
Temperature [°C]
Chlorophyll a [mg/m3]
Stratification [s2]
Oxycline depth [m]
MAT IKM WA-PLS ANN⁎ MAT IKM WA-PLS ANN⁎ MAT IKM WA-PLS ANN⁎ MAT IKM WA-PLS ANN⁎ MAT IKM WA-PLS ANN⁎
RMSE
R2
0.270 0.347 0.278 0.190 0.527 0.761 0.600 0.402 0.090 0.062 0.072 0.059 0.000E-05 0.000E-05 0.000E-05 0.000E-05 29.9 37.4 31.3 21.8
0.941 0.899 0.945 0.957 0.867 0.722 0.827 0.879 0.875 0.940 0.921 0.722 0.882 0.810 0.841 0.891 0.890 0.824 0.887 0.913
this turning point the relative abundance of G. sacculifer increased again, in this case at the expense of G. glutinata, while the relative abundance of G. ruber remains approximately at the same level. G. siphonifera, G. calida, N. pachyderma, N. incompta and G. tenella show no expressed changes in their abundance, except for a peak in the abundance of G. tenella at 32 cm core depth. Virtually no specimens of G. bulloides and the Gulf of Aden species were found in the Holocene of core KL9; this situation is typical for the central Red Sea today (Fig. 3).
Fig. 5. Plots of relative abundance of six taxonomic units and number of planktonic foraminifera per gram sediment (3p running mean) for the Holocene section of core KL9. G. bulloides and the Gulf of Aden species group were not encountered in this core section. Triangles indicate samples dated by 14C AMS.
RMSEP (boot) 0.333 0.541 0.366 0.634 1.841 0.724 0.096 0.129 0.090 1.498E-05 2.882E-05 1.598E-05 34.0 51.3 36.7
RMSEP (% of average)
RMSEP (% of range)
0.9 1.4 0.9 0.5 2.3 6.7 2.6 1.5 34.5 46.3 32.2 21.3 11.0 21.1 11.7 7.5 13.5 20.4 14.6 8.7
7.9 12.8 8.7 4.5 15.1 43.8 17.2 9.6 8.3 11.2 7.8 5.1 11.3 21.7 12.0 7.7 11.7 17.7 12.6 7.5
Transfer function reconstructions of the five environmental parameters from the species counts are shown in Fig. 7. For each reconstructed parameter, the Holocene patterns show a broad similarity between the individual methods. The agreement among the methods is highest for chlorophyll a, and lowest for temperature, where for much of the late Holocene, ANN and WA-PLS yielded temperature estimates up to 2 °C colder than MAT and IKM. The estimates of the stratification and oxycline depth also deviate among the transfer function approaches, in particular for the sample with a peak abundance of G. tenella (cp. Figs. 5, 7). In this sample, the salinity parameter also shows the largest discrepancy among the methods, whereas the chlorophyll a values are virtually identical. A comparison of the reconstructed values with averages measured today at the core position shows a general tendency towards an over-respectively too low estimates of salinity and temperature for WA-PLS and IKM approaches, whereas the reconstructed chlorophyll a concentrations clustered more closely around the recent value for all methods. In terms of internal variability of the reconstructed records, MAT values
Fig. 6. PCA of logratio transformed data of all samples from the surface and KL9 Holocene dataset.
M. Siccha et al. / Marine Micropaleontology 72 (2009) 146–156
Fig. 7. Annual mean salinity, temperature, chlorophyll a concentration, stratification (Brunt-Väisälä frequency) and oxycline depth (2 ml O2/l) of the Holocene record of KL9 as reconstructed by the different types of transfer functions. Dashed horizontal lines denote recent values of the parameter at the core position. Triangles indicate samples dated by 14C AMS.
were least variable followed by ANN, WA-PLS and IKM, which produced the largest scatters in the parameter estimates. In general, the reconstructed Holocene records show two notable features. The first is a short excursion around 3 ka BP, characterised by lower salinity, slightly higher temperature and a pronounced maximum in chlorophyll a concentration. The second corresponds to the early Holocene period between 7 ka BP and 9.6 ka BP where the transfer function estimates indicate increased salinity and oxycline depth and decreased temperatures, stratification and productivity. The very beginning of the Holocene record is characterised by estimates close to recent values. 5. Discussion 5.1. Factors controlling the distribution of planktonic foraminifera in surface sediments The results of multivariate analyses (Fig. 4) suggest that the contribution of individual environmental parameters to the prominent gradient in planktonic foraminifera abundances in Red Sea sediments cannot be strictly mathematically deconvolved, due to their high mutual correlation (Table 1). Of the typically considered oceanographic factors, sea surface salinity shows the highest explanatory
153
power for the recent distribution of planktonic foraminifera in the Red Sea (64.1% of the species variance, Table 4). However, neither analyses of global distributions of planktonic foraminifera (Morey et al., 2005) nor laboratory experiments (Bijma et al., 1990) indicate a strong dependence of planktonic foraminifera on salinity. The strongest evidence against this factor comes from the analysis of species disappearances during sea level lowstands (Fenton et al., 2000) in the Red Sea itself during the late Quaternary. Throughout MIS 3, when sea level was approximately 60 m lower than today (Shackleton, 1987; Siddall et al., 2003) and salinity in the Red Sea significantly higher (Hemleben et al., 1996; Rohling, 1994; Thunell et al., 1988), G. ruber remained the dominant species, whereas the species which nowadays seem to be associated with high salinities, G. sacculifer, had almost disappeared from the Red Sea at that time. Therefore, we conclude that the high correlation of salinity with the first axis in the redundancy analysis (Fig. 4) is coincidental and thus is not based on ecological controls of the species assemblage by this parameter. Similarly, sea surface temperature, which is the dominant controlling factor for the distribution of planktonic foraminifera in global oceans (Morey et al., 2005), seems to have little effect on the Red Sea fauna. The comparatively small explanatory power (28.7% of the species variance) of temperature is supplemented by the fact that the associations of species abundances with temperature trends in the Red Sea contradict open-ocean observations in which G. sacculifer is associated with higher SST than G. ruber (e.g. Prell, 2003). Close interrelations can be expected for the factors chlorophyll a (59.4% of the species variance), depth of the 2 ml O2/l oxycline (63.1%) and the stratification of the water column (46.8%), whose gradients are dependent on the nutrient transport of Red Sea circulation system. Fenton et al. (2000) argued that the extent and intensity of the OMZ in the Red Sea is critical for the ratio of G. sacculifer to G. ruber, and, due to the large abundances of these two species, may affect the foraminifera assemblage as a whole. There are suggestions in the literature that G. sacculifer reproduces at slightly greater depths that G. ruber (Bijma and Hemleben, 1994; Reiss et al., 1999), and therefore could be more negatively affected by the upper reaches of the OMZ. To this date no studies investigating the effects of lower oxygen concentrations on planktonic foraminifera exist, yet extensive work on benthic foraminifera assemblages indicate that oxygen concentrations above 1 ml/l do not affect the composition of benthic foraminifera species (Bernhard and Gupta, 1999). Furthermore the reconstructed depths of the 2 ml/l oxycline remain lower than 150 m throughout the Holocene record, which is well below the inferred reproduction depth of G. sacculifer of 80 m (Bijma and Hemleben, 1994). Edelman (1996) and Edelman-Fürstenberg et al. (2009) regard stratification, next to the properties of the OMZ, as controlling factor for the G. sacculifer to G. ruber ratio. However, ecological (Ravelo et al., 1990; Schiebel et al., 2004) as well as geochemical studies (Farmer et al., 2007) from the tropical Indian and Atlantic Ocean indicate that the depth habitats of these species are strongly overlapping. Thus, it appears very unlikely that the depth of the oxycline or stratification exert a primary control on the assemblages. The conspicuous complementarity of the salinity and temperature reconstructions reflects the correlation between these two parameters in the calibration dataset. In an analogous manner the reconstructions of water column stratification resemble the temperature curve and those of the oxycline depth that of salinity. Thus we conclude that the transfer function reconstructions for salinity, SST and also oxycline depth and stratification, which are all showing similar patterns, merely reflect the current combination of these variables that is found in the northern Red Sea, where the modern fauna is most similar to the early Holocene assemblages in core KL9. The chlorophyll a reconstruction differs in shape from the other parameters. Not only is it the only parameter which ignores the G. tenella peak at 3.8 ka BP, it is also the only parameter which emphasizes the faunal shift, at around 3 ka BP, where the dominance in the foraminifera assemblage changes. This event leads to distinct
154
M. Siccha et al. / Marine Micropaleontology 72 (2009) 146–156
peak in the chlorophyll a reconstruction which clearly stands out from the rest of the record. The assumption that productivity (approximated by chlorophyll a concentration) controls the planktonic foraminifera assemblage (Watkins et al., 1996) in the Red Sea introduces no contradictions with open-ocean observations. Species like G. glutinata and G. bulloides, which are typical for regions of increased coastal productivity including in the Arabian Sea and the eastern equatorial Atlantic (Schiebel et al., 2001; Thiede, 1975), are associated with this factor in the redundancy analysis of the Red Sea assemblages (Fig. 4). Under normal oceanic conditions, stratification would be expected to be negatively correlated with productivity, as a more stratified water column is less conducive to vertical mixing and transport of nutrients from deeper waters to the photic zone. However, in the Red Sea basin highest productivity is found in the southern area where stratification is also strongest (Fig. 1), because of the layered water exchange across the Strait of Bab al Mandab. We hypothesize that productivity is affecting the planktonic foraminifera assemblages in the Red Sea independent from local stratification (cf. Table 1, Fig. 4). Since the observed productivity gradient results from the same processes that also control the remaining variables (circulation pattern), all variables are highly correlated and cannot be deconvolved by analysis of recent data. A way to corroborate this hypothesis would be the analysis of fossil assemblages showing the response of the Red Sea plankton to oceanographic conditions different from the present day. 5.2. Interpretation of Holocene paleoenvironmental reconstruction results in the Red Sea The two consistent features in our Holocene transfer function reconstructions in core KL9 (Fig. 7) are linked to changes in the G. sacculifer/G. ruber ratio, which shifted towards G. ruber at 3 ka BP and towards G. sacculifer between 7 and 9.6 ka BP. If the increased abundances of G. sacculifer in the early Holocene reflected increased salinity, as suggested by the transfer function reconstructions, then this pattern would be the result of the post-glacial sea level rise that lasted until 7 ka BP (Fleming et al., 1998). In this case, G. sacculifer abundances (and salinity reconstructions) should have been the highest in the earliest part of our record. Yet, the abundance of G. sacculifer is at approximately recent level in the oldest two samples. The lower abundance of planktonic foraminifera suggests unfavourable living conditions in general and salinity was probably closer to the tolerance level of G. sacculifer. Therefore, the analysis of the Holocene record further support the hypothesis, that salinity itself does not control the assemblage composition of planktonic foraminifera in the Red Sea. The transfer-function results of decreased productivity in the earliest Holocene are supported by two issues affecting the circulation regime. Firstly the lower sea level had decreased the water exchange through Bab al Mandab and thus the import of nutrients. Secondly there is good evidence that during the early Holocene (Holocene insolation/thermal maximum) the Indian SW monsoon was amplified and the ITCZ extended further north than at present (Fleitmann et al., 2007; Haug et al., 2001). This fuelled the upwelling in the Arabian Sea (Gupta et al., 2003) and the Gulf of Aden. Under the conditions of a pronounced and/ or prolonged Indian SW monsoon, the present day Red Sea circulation would be expected to import more nutrient-rich waters from the Gulf of Aden leading to more fertile conditions in the very south of the basin while nutrient levels in the central and northern Red Sea would be diminished due to decreased winter circulation (cf. Fig. 2). The estimation of a reasonable weighting for these two influences is difficult. Sea level reached recent values at about 7 ka BP, while the translation of the ITCZ is a continuous process. An examination of the absolute abundance data of planktonic foraminifera (Fig. 5) yields further support to the transfer function results of a lower productivity. Compared to the period after 7 ka BP, early Holocene planktonic foraminifera abundances per gram sediment are on average about two-
thirds, indicating lower foraminifera productivity in the central Red Sea. The apparent discrepancy between low absolute abundance of planktonic foraminifera and higher productivity reconstructed for the oldest two samples could be explained by unsteady conditions during the establishment of planktonic foraminifera populations after the aplanktonic period of the last glacial. A shift in size distribution towards smaller species like G. glutinata and the general frailty of samples with low total abundances (b1000 ind. g/sed.) towards bias in sample processing could have affected these samples. Although all of the Holocene samples contain assemblages which find good analogues in the present-day Red Sea, the two oldest Holocene samples are apparently the product of non-analogous conditions of a recolonization period. This leads to incorrect interpretation of the assemblage by transfer functions. The work of Fenton et al. (2000) showed that the prerequisite of an analogous planktonic foraminifera fauna is not given for large parts of the analysed record, especially during glacials. In summary, we can expect our developed transfer functions to yield realistic reconstructions of productivity only in interglacials, when sea-level and thus salinity was not significantly different from the present day state. 6. Conclusions In this study we analysed the planktonic foraminifera distribution in recent surface sediments of the Red Sea. The strong environmental gradient along the basin's axis is mirrored in the faunal assemblages, leading to a significant correlation of assemblage composition with all investigated surface water parameters. Redundancy analysis shows that salinity could explain the largest amount of variance in the faunal dataset followed by productivity in form of chlorophyll a concentration and the related parameter oxycline depth. Stratification and temperature could explain the least amount of variance. The transfer functions reproduce the recent distribution with good to excellent accuracy. The application of these transfer functions on a newly generated Holocene record from the central Red Sea also showed a strong interdependency among the reconstructed variables, except of productivity, which showed a distinct pattern and the highest degree of consistency between the different transfer function methods tested. A comparison with published ecological and paleoclimate data allowed us to dismiss salinity as the forcing factor and gave support to the concept of a productivity-controlled foraminifera assemblages. Quantitative reconstructions of this parameter are only viable under the assumption of similarity of foraminifera fauna and climatic conditions between the recent calibration set and the time period to be investigated. It has been shown that this prerequisite is often (i.e. during glacials) not given, due the Red Sea's sensibility to climatic changes, notably sealevel variations. During the interglacials, when sea-level was similar to the present, planktonic foraminifera faunas in the Red Sea can be used to reconstruct past productivity regimes. Due to the connection between the Red Sea's circulation and the Indian monsoon, reconstructions of productivity in the central parts of the basin can thus yield indirect information on the intensity and mode of the Indian Monsoon in the past. Acknowledgements This work was funded by the Deutsche Forschungsgemeinschaft (DFG grants KU 2259/3-1 “RedSTAR” and HE 697/7, 17, 27). Alexander Floria and Sofie Jehle helped in preparation of the samples. We thank A. Almogi-Labin and an anonymous referee for helpful comments on the manuscript. Appendix A List of planktonic foraminifera species encountered in modern surface sediments in the Red Sea.
M. Siccha et al. / Marine Micropaleontology 72 (2009) 146–156
Species included in transfer functions Globigerina bulloides (d'Orbigny), 1826 abundant in the southern Red Sea and the Gulf of Aden, associated with productivity in the Red Sea, which is consistent with the behaviour of this species in the open ocean (Thiede, 1975) Globigerinella calida (Parker), 1962 common in small numbers throughout the area of investigation. The counts of this species were merged with those of G. siphonifera due the difficulty and subjectivity in their differentiation Globigerinita glutinata (Egger), 1893 abundant, strong increasing gradient of occurrence from north to south in the Red Sea, linked to productivity, consistent with the results of Watkins et al. (1996) Globorotalia menardii (Parker, Jones and Brady), 1865 rare, occurring only in the southern Red Sea and Gulf of Aden; Auras-Schudnagies et al. (1989) showed that this and other species reflect the advection of water masses from the Gulf of Aden. Its abundance is higher when larger size fractions are investigated, but in the 150 μm size fraction it is extremely rare in the Red Sea. Neogloboquadrina incompta (Cifelli), 1961 common in small numbers throughout the area of investigation with higher numbers in the northern Red Sea Neogloboquadrina pachyderma (Ehrenberg), 1861 common in small numbers throughout the area of investigation with higher numbers in the northern Red Sea. It could possibly reflect advection and small population of the unusual N. pachyderma form that thrives in the upwelling cells of the western Arabian Sea (Schiebel et al., 2004), as the abundance of sinistral Neogloboquadrina individuals is much higher than what would be expected of the known anomalously sinistral specimens of N. incompta (Darling et al., 2006). Globigerinoides ruber (d'Orbigny), 1839 second most abundant species in this study, occurring abundantly throughout the area of investigation, no differentiation between morphotypes was made, no pink specimens were found Globigerinoides sacculifer (Brady), 1877 most abundant species in this study, strong declining gradient of occurrence from north to south in the Red Sea, also occurring in small numbers in the Gulf of Aden, no differentiation between morphotypes (sacculifer vs. trilobus) was made Globogerinella siphonifera (d'Orbigny), 1839 abundant throughout the area of investigation Globoturborotalita tenella (Parker), 1958 common in small numbers throughout the area of investigation with higher numbers in the northern Red Sea Neogloboquadrina dutertrei (d'Orbigny), 1839 rare, occurring only in the southern Red Sea and Gulf of Aden, see G. menardii Species excluded from transfer functions Globorotalia anfracta (Parker), 1967 rare, mostly in the southern Red Sea and Gulf of Aden Globigerinoides conglobatus (Brady), 1879 very rare, seven individuals (out of 23,818 counted individuals in the surface dataset; see online supporting material) Globoquadrina conglomerata (Schwager), 1866 very rare, one individual Beella digitata (Brady), 1879 very rare, eight individuals, of which seven come from the Gulf of Aden Globigerina falconensis Blow, 1959 very rare, twelve individuals, of which eight come from the Gulf of Aden
155
Globorotalia inflata (d'Orbigny), 1839 very rare, one individual Globorotaloides hexagonus (Natland), 1938 very rare, five individuals from the Gulf of Aden Globigerinita minuta (Natland), 1938 rare, occurring throughout the area of investigation Globoturborotalita rubescens (Hofker), 1956 common in small numbers throughout the area of investigation, the majority showed the typical pink colouration Globorotalia scitula (Brady), 1882 rare, mostly in the southern Red Sea and Gulf of Aden Globorotalia truncatulinoides (d'Orbigny), 1839 very rare, one individual Globigerinita uvula (Ehrenberg), 1861 very rare, eight individuals Gallitellia vivans (Cushman), 1934 very rare, thirteen individuals, of which eleven come from one sample in the southern Red Sea Hastigerina digitata (Rhumbler), 1911 very rare, fourteen individuals Hastigerina pelagica (d'Orbigny), 1839 common in small numbers throughout the area of investigation Orbulina universa (d'Orbigny), 1839 common in small numbers throughout the area of investigation Pulleniatina obliquiloculata (Parker and Jones), 1865 very rare, three individuals from the Gulf of Aden Turborotalita humilis (Brady), 1884 very rare, three individuals from the Gulf of Aden Turborotalita quinqueloba (Natland), 1938 Appearing sporadically in a quarter of the samples, regularly only in the very southern Red Sea and the Gulf of Aden. Disregarded for transfer functions, because most individuals found in our samples were below 150 μm in size. They appeared in the N150 μm size fraction as attachments to larger particles and other foraminifera, preventing a representative count. In samples were quantification was attempted, the abundance of this species never exceeded 12%, the average abundance was 1.3%. Appendix B. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.marmicro.2009.04.002. References Almogi-Labin, A., Hemleben, C., Meischner, D., Erlenkeuser, H., 1991. Paleoenvironmental events during the last 13,000 years in the central Red Sea as recorded by pteropoda. Paleoceanography 6 (1), 83–98. Almogi-Labin, A., Hemleben, C., Meischner, D., 1998. Carbonate preservation and climatic changes in the central Red Sea during the last 380 kyr as recorded by pteropods. Marine Micropaleontology 33, 87–107. Almogi-Labin, A., Schmiedl, G., Hemleben, C., Siman-Tov, R., Segl, M., Meischner, D., 2000. The influence of the NE winter monsoon on productivity changes in the Gulf of Aden, NW Arabian Sea, during the last 530 ka as recorded by foraminifera. Marine Micropaleontology 40, 295–319. Arz, H.W., Lamy, F., Pätzold, J., Müller, P.J., Prins, M., 2003. Mediterranean moisture source for an early-Holocene humid period in the Northern Red Sea. Science 300, 118–121. Auras-Schudnagies, A., Kroon, D., Ganssen, G., Hemleben, C., Van Hinte, J.E., 1989. Distributional pattern of planktonic foraminifers and pteropods in surface waters and top core sediments of the Red Sea, and adjacent areas controlled by the monsoonal regime and other ecological factors. Deep-Sea Research 36 (10), 1515–1533. Badawi, A., Schmiedl, G., Hemleben, C., 2005. Impact of late Quaternary environmental changes on deep-sea benthic foraminiferal faunas of the Red Sea. Marine Micropaleontology 58 (1), 13–30. Bernhard, J.M., Gupta, B.K.S., 1999. Foraminifera of oxygen-depleted environments. In: Gupta, B.K.S. (Ed.), Modern Foraminifera. Kluwer Academic Publishers, Dordrecht, pp. 201–216. Bijma, J., Hemleben, C., 1994. Population dynamics of the planktic foraminifer Globigerinoides sacculifer (Brady) from the central Red Sea. Deep-Sea Research I 41 (3), 485–510.
156
M. Siccha et al. / Marine Micropaleontology 72 (2009) 146–156
Bijma, J., Faber Jr., W.W., Hemleben, C., 1990. Temperature and salinity limits for growth and survival of some planktonic foraminifers in laboratory cultures. Journal of Foraminiferal Research 20 (2), 95–116. Cember, R.P., 1988. On the sources, formation and circulation of Red Sea deep water. Journal of Geophysical Research 93 (C7), 8175–8191. Conkright, M.E., et al., 2001. World Ocean Atlas 2001. Darling, K.F., Kucera, M., Kroon, D., Wade, C.M., 2006. A resolution for the coiling direction paradox in Neogloboquadrina pachyderma. Paleoceanography 21 (2), PA2011. Dryden, A.L.J., 1931. Accuracy in percentage representation of heavy mineral frequencies. Proceedings of the National Academy of Sciences, USA 17 (5), 233–238. Edelman-Fürstenberg, Y., Almogi-Labin, A., Hemleben, C., 2009. Palaeoceanographic evolution of the central Red Sea during the late Holocene. The Holocene 19 (1), 117–127. Edelman, Y., 1996. Reconstruction of Paleocenaographic Settings During the late Holocene in the central Red Sea. Hebrew University, Jerusalem, Israel. 172 pp. Eshel, G., Cane, M.A., Blumenthal, M.B., 1994. Modes of subsurface, intermediate, and deep water renewal in the Red Sea. Journal of Geophysical Research 99, 15,941–15,952. Farmer, E.C., Kaplan, A., de Menocal, P.B., Lynch-Stieglitz, J., 2007. Corroborating ecological depth preferences of planktonic foraminifera in the tropical Atlantic with the stable oxygen isotope ratios of core top specimens. Paleoceanography 22 (3), PA3205. Feldman, G.C. and McClain, C.R., 2006. Ocean Color Web, SeaWIFS/Chlorophyll a concentration, 07/2002–06/2006. In: N. Eds. Kuring and S.W. Bailey (Editors), http://oceancolor.gsfc.nasa.gov/. NASA Goddard Space Flight Center, Washington, USA. Fenton, M., Geiselhart, S., Rohling, E.J., Hemleben, C., 2000. Aplanktonic zones in the Red Sea. Marine Micropaleontology 40, 277–294. Fleitmann, D., et al., 2007. Holocene ITCZ and Indian monsoon dynamics recorded in stalagmites from Oman and Yemen (Socotra). Quaternary Science Reviews 26, 170–188. Fleming, K., et al., 1998. Refining the eustatic sea-level curve since the Last Glacial Maximum using far- and intermediate-field sites. Earth and Planetary Science Letters 163, 327–342. Gupta, A.K., Anderson, D.M., Overpeck, J.T., 2003. Abrupt changes in the Asian southwest monsoon during the Holocene and their links to the North Alantic Ocean. Nature 421 (6921), 354–357. Halicz, E., Reiss, Z., 1981. Paleoecological relations of foraminifera in a desert-enclosed sea — The Gulf of Aqaba (Elat), Red Sea. Marine Ecology 2 (1), 15–34. Haug, G.H., Hughen, K.A., Sigman, D.M., Peterson, L.C., Röhl, U., 2001. Southward migration of the intertropical convergence zone through the Holocene. Science 293, 1304–1308. Hemleben, C., et al., 1996. Three hundred eighty thousand year long stable isotope and faunal records from the Red Sea: influence of global sea level change on hydrography. Paleoceanography 11 (2), 147–156. Hemleben, C., Spindler, M., Anderson, O.R., 1989. Modern Planktonic Foraminifera. Springer-Verlag, New York. 363 pp. Hutson, W.H., 1980. The Agulhas current during the late Pleistocene: analysis of modern faunal analogs. Science 207, 64–66. Imbrie, J., Kipp, N.G., 1971. A new micropaleontological method for quantitative paleoclimatology: application to a late Pleistocene Caribbean core. In: Turekian, K.K. (Ed.), Late Cenozoic Glacial Ages. Yale University Press, New Haven, pp. 71–181. Ivanova, E.V., 1985. Late Quaternary biostratigraphy and paleotemperatures of the Red Sea and the Gulf of Aden based on planktonic foraminifera and pteropods. Marine Micropaleontology 9 (4), 335–364. Juggins, S., 2003. C2 Data Analysis. University of Newcastle, Newcastle, UK. Kucera, M., et al., 2005. Reconstruction of sea-surface temperatures from assemblages of planktonic foraminifera: multi-technique approach based on geographically constrained calibration data sets and its application to glacial Atlantic and Pacific Oceans. Quaternary Science Reviews 24, 951–998. Legge, H.L., Mutterlose, J., Arz, H.W., 2006. Climatic changes in the northern Red Sea during the last 22,000 years as recorded by calcareous nannofossils. Paleoceanography 21 (1), PA1003. Legge, H.L., Mutterlose, J., Arz, H.W., Pätzold, J., 2008. Nannoplankton successions in the northern Red Sea during the last glaciation (60 to 14.5 ka BP): reactions to climate change. Earth and Planetary Science Letters 270 (3–4), 271–279. Leps, J., Smilauer, P., 2003. Multivariate Analysis of Ecological Data using CANOCO. Cambridge University Press, Cambridge. 269 pp. Malmgren, B.A., Nordlund, U., 1997. Application of artificial neural networks to paleoceanographic data. Palaeogeography, Palaeoclimatology, Palaeoecology 136, 359–373. Manasrah, R., Badran, M., Lass, H.U., Fennel, W., 2004. Circulation and winter deepwater formation in the northern Red Sea. Oceanologia 46 (1), 5–23. Monterey, G.I., Levitus, S., 1997. Climatological Cycle of Mixed Layer Depth in the World ocean. U.S. Gov. Printing Office. NOAA NESDIS. Morcos, S.A., 1970. Physical and chemical oceanography of the Red Sea. Oceanography and Marine Biology: An Annual Review 8, 73–202.
Morey, A.E., Mix, A.C., Pisias, N.G., 2005. Planktonic foraminiferal assemblages preserved in surface sediments correspond to multiple environment variables. Quaternary Science Reviews 24, 925–950. Murray, S.P., Johns, W., 1997. Direct observations of seasonal exchange through the Bab el Mandab Strait. Geophysical Research Letters 24 (21), 2557–2560. Nellen, W., et al., 1996. MINDIK (Band II), Reise Nr. 5, 2.Januar–24.September 1987. Universität Hamburg. Neumann, A.C., McGill, D.A., 1962. Circulation of the Red Sea in early summer. Deep-Sea Research 8, 223–235. Patzert, W.C., 1974. Wind-induced reversal in Red Sea circulation. Deep-Sea Research 21, 109–121. Pollard, A.M., Blockley, S.P.E., 2006. Some numerical considerations on the geochemical analysis of distal microtephra. Applied Geochemistry 21, 1692–1714. Prell, W.L., 2003. The Brown University Foraminiferal Database (BFD). PANGAEA. Ravelo, A.C., Fairbanks, R.G., Philander, S.G.H., 1990. Reconstructing tropical atlantic hydrography using planktonic foraminifera and an ocean model. Paleoceanography 5 (3), 409–431. Reiss, Z., Halicz, E., Luz, B., 1999. Late-Holocene foraminifera from the SE Levantine Basin. Israel Journal of Earth Sciences 48, 1–27. Revets, S.A., 2004. On confidence intervals from micropalaeontological counts. Journal of Micropalaeontology 23 (1), 61–65. Rohling, E.J., 1994. Glacial conditions in the Red Sea. Paleoceanography 9 (5), 653–660. Rohling, E.J., et al., 2008. High rates of sea-level rise during the last interglacial period. Nature Geoscience 1 (1), 38–42. Schiebel, R., Waniek, J.J., Bork, M., Hemleben, C., 2001. Planktic foraminiferal production stimulated by chlorophyll redistribution and entrainment of nutrients. Deep-Sea Research I 48, 721–740. Schiebel, R., et al., 2004. Distribution of diatoms, coccolithophores and planktic foraminifers along a trophic gradient during SW monsoon in the Arabian Sea. Marine Micropaleontology 51, 345–371. Schmelzer, I., 1998. High-frequency event-stratigraphy and paleoceanography of the Red Sea. Ph.D. Thesis, University of Tuebingen, Tuebingen, Germany, 124 pp. Shackleton, N.J., 1987. Oxygen isotopes, ice volume and sea level. Quaternary Science Reviews 6, 183–190. Siddall, M., et al., 2003. Sea-level fluctuations during the last glacial cycle. Nature 423, 853–858. Siddall, M., et al., 2004. Understanding the Red Sea response to sea level. Earth and Planetary Science Letters 225, 421–434. Siddall, M., Smeed, D.A., Matthiesen, S., Rohling, E.J., 2002. Modelling the seasonal cycle of the exchange flow in Bab El Mandab (Red Sea). Deep-Sea Research I 49, 1551–1569. Smeed, D.A., 1997. Seasonal variation of the flow in the strait of Bab al Mandab. Oceanologica Acta 20 (6), 773–781. Smeed, D.A., 2004. Exchange through the Bab el Mandab. Deep-Sea Research II 51, 455–474. Sofianos, S.S., Johns, W.E., 2002. An Oceanic General Circulation Model (OGCM) investigation of the Red Sea circulation: 1. Exchange between the Red Sea and the Indian Ocean. Journal of Geophysical Research 3196. Sofianos, S.S., Johns, W.E., 2003. An Oceanic General Circulation Model (OGCM) investigation of the Red Sea circulation: 2. Three-dimensional circulation in the Red Sea. Journal of Geophysical Research 3066. Sofianos, S.S., Johns, W.E., 2007. Observations of the summer Red Sea circulation. Journal of Geophysical Research 112 (C6), C06025. Telford, R.J., Birks, H.J.B., 2005. The secret assumption of transfer functions: problems with spatial autocorrelation in evaluating model performance. Quaternary Science Reviews 24 (20–21), 2173–2179. Telford, R.J., Andersson, C., Birks, H.J.B., Juggins, S., 2004. Biases in the estimation of transfer function prediction errors. Paleoceanography 19 (4), PA4014. ter Braak, C.J.F., Juggins, S., 1993. Weighted averaging partial least squares regression (WA-PLS): an improved method for reconstructing environmental variables from species assemblages. Hydrobiologia 269/270, 485–502. Thiede, J., 1975. Distribution of foraminifera in coastal waters of an upwelling area. Nature 253, 712–714. Thunell, R.C., Locke, S.M., Williams, D.F., 1988. Glacio-eustatic sea-level control on Red Sea salinity. Nature 334, 601–604. Tragou, E., Garrett, C., 1997. The shallow thermohaline circulation of the Red Sea. DeepSea Research I 44, 1355–1376. van Couwelaar, M., 1997. Zooplankton and micronekton biomass off Somalia and in the southern Red Sea during the SW monsoon of 1992 and the NE monsoon of 1993. Deep-Sea Research II 44 (6–7), 1213–1234. Watkins, J.M., Mix, A.C., Wilson, J., 1996. Living planktic foraminifera: tracers of circulation and productivity regimes in the central equatorial Pacific. Deep-Sea Research II 43 (4–6), 1257–1282. Woelk, S., Quadfasel, D., 1996. Renewal of deep water in the Red Sea during 1982–1987. Journal of Geophysical Research 101 (C8), 18155–18165.