The Middle to Late Miocene climatic development of Southwest Africa derived from the sedimentological record of ODP Site 1085A

B. Roters1*

(1)MARUM – Center for Marine Environmental Sciences, University of Bremen, P.O. Box 330440, 28334 Bremen, Germany*present Address: Fronterra Geosciences GmbH; Landstraßer Hauptstr. 101, 3. Altstiege, TOP A6, A-1030 Vienna, AustriaR. Henrich2

(2)Department of Geosciences, University of Bremen, P.O. Box 330440, 28334 Bremen, Germany+49-(0)421-218 9591

+49-(0)421-218 7431

bastian.roters@uni-bremen.de

Sediments from the ODP Site 1085A were studied to investigate the impacts of global cooling in the Middle and Late Miocene on the climate in Southwestern Africa. The size composition of the sediment was analysed emphasising the silt fraction. A comparison with the modern grain size distribution and suitable transport processes made it possible to assign specific transport processes to the grain size composition. Three processes are considered for transport of terrigeneous silt: While there was no evidence found for (1) transport by ocean currents, the analyses showed signals of (2) wind transport indicating dry conditions associated with a cool climate and (3) fluvial transport that points to humid and warm conditions. Three climatic phases were defined. The first phase from 13.8 to 11.8 Myr reveals a stable humid climate in Southwest Africa independent of the Antarctic glaciations. During the second phase from 11.8 to 10.4 Myr the regional climate cooled considerably but was not drier. Additionally, the climate during this phase reacted to the Antarctic glaciations. This cooling-trend continued during phase 3 from 10.4 to 9.0 Myr with a significant increase in dust input, pointing to overall drier conditions. However, fluvial transport still remained as the main source.

South West Africa; Middle and Late Miocene; Climate cooling; Aridification; Grain size analysis

Introduction

The main target of this study is to trace the marine climatic record in Southwest Africa during the Middle to Late Miocene. During this period prominent shifts towards a modern setting were achieved; ocean circulation, the onset of the Benguela Upwelling System and the desertification of the hinterland (Diester-Haass et al. 1990, 1992). For the development of the upwelling cells, seaward blowing winds were required, which prevented coeval landward intrusion of humid air masses from the South Atlantic.

1

However, during the late middle Miocene and the early late Miocene the ocean current system still differed considerably from today. An important precondition for modern thermohaline circulation was the opening of the Drake Passage during the Oligocene. The Antarctic Circumpolar Current (ACC) was not as strong as at present (Diekmann et al. 2003). But exchange of deep and surface water masses through the Panama Gateway was still possible. The deep ocean basins were ventilated by Southern Component Water (SCW) only from 15 – 12.5 Myr (Wright et al. 1992). From 12.5 to 10 Myr Northern Component Water (NCW) is evident for the first time, with changing δ13C values in the Atlantic (Wright et al. 1992). The southern hemisphere intermediate depths were affected by the Southern Component Intermediate Water (SCIW), that originated in the southern Ocean (Wright et al. 1992). North Atlantic Deep Water (NADW) did not intrude into the Cape Basin before 10 Myr (Diester-Haass et al. 2004). During the Middle Miocene the Benguela Current (BC) progressed northwards and reached the Walvis Ridge at about 10 Myr (Diester-Haass et al. 1990, 1992; Diester-Haass et al. 2002; Krammer et al. 2006; Kastanja et al. 2006).Due to the opening of the Drake Passage and the onset of the ACC Antarctica cooled progressively. In addition, during the Middle Miocene several glaciations on Antarctica occurred, which are documented by significant shifts in the benthic oxygen isotope values, the Mi-Events (Miller et al. 1991). As known from various climatic proxies, from 15 Myr on a trend of general cooling had been established that lasted finally until the Holocene (Zachos et al. 2001). Evidence for these Antarctic glaciations was also found in South Atlantic isotopic records by Westerhold et al. (2005). The δ18O-record of Westerhold et al. (2005) shows four excursions between 14 and 9 Myr, at 13.8, 13.2, 11.7 and at 10.4 Myr. These episodes indicate cooler climates and coincide with the Mi-Events Mi3, Mi4, Mi5 and Mi6.The cooling on Antarctica has a great significance for the development of the continental climate in Africa and the development of the Benguela Upwelling System. Upwelling cells are always driven by trade-winds, but the wind strength depends predominantly on the temperature gradient between polar and tropical regions. The strength of the winds has also a bearing on the Southwest African climate. Stronger winds from the South and Southwest during cool periods cause a dryer climate over the African continent (Gingele 1996). In contrast, during warmer periods, the winds are weaker and the climate tends to be more humid (Gingele 1996). In this study the results of analyses of sediment composition of the Middle to Late Miocene core section of ODP Site 1085A are presented. With the results of grain size analyses of carbonate and terrigeneous silt, shifts in the climatic settings that govern Southwest Africa and the adjoining part of the South Atlantic Ocean during the time between 14 and 9 Myr, are determined and discussed in context of global climate evolution during this critical period.This study incorporates and expands the results and findings of earlier studies by Kastanja et al. (2006) who showed, that dust input from the Namib desert started to become identificable after 11.2 Myr. This gives indications for the establishment of a major shift in the wind regime, which has been a main precondition for the development of the wind-driven Benguela Upwelling System (BUS). Based on nannofossil distribution and accumulation rates, Krammer et al. (2006) concluded, that the initiation of upwelling conditions in the Benguela region started at about 10 Myr. In a continuation of these earlier results, this study links the time before and after the development of the BUS and gives an overview

2

of its preconditions and impacts. Three phases were found that are characterised by differences in continental humidity and in influence of the Antarctic glaciations on the Southwest African climate. These can be correlated with proposed global stratigraphic events.

Materials and Methods

ODP Site 1085A

Site 1085A is situated in the Cape Basin, Southeast Atlantic Ocean (29°22.47'S, 13°59.41'E, 1713 m water depth), beneath the BC upwelling region. The mouth of the Orange River is about 260 km to the northeast of the drill-site (Fig. 1).

Fig. 1 Location of the working area, showing the position of ODP Site 1085A off the Namibian coast. Bathymetry was compiled with OceanDataView 1.0. The small map in the upper right corner shows the working area in grey in relation to the African continent. A: Arrisdrift; B: Buntfeldschuh; G: Grillental.

The core has been sampled from the early Late Miocene down to the Middle Miocene. This comprises the core sections 56X to 63X (543.36 to 621.06 meters composite depth [mcd]). From this interval 30 samples were taken with a resolution of about 100 kyr beginning at 13,891 kyr up to 10,975 kyr. Also data from sections 47X to 55X (457.36 to 534.26 meters), which were collected in higher resolution by Kastanja et al. (2006) were used for this study. This second interval comprises 118 samples from 10,967 kyr to 9,006 kyr. The age model of Westerhold et al. (2005) has been used. The sediments at Site 1085A consist mostly of terrigeneous mud and carbonate ooze (Wefer et al. 1998). In the studied interval the sedimentation rates range from 1.1 to about 8.4 cm/kyr. (Wefer et al. 1998; Westerhold et al. 2005).

Sample treatment and analyses

The samples with a volume of about 20 cm³ were divided: one half served as a reference and for combustion analysis of the bulk fraction. The other half was used for the grain size analysis. In the first instance this part was split into three

3

fractions: sand (>63 µm), silt (2-63 µm) and clay (<2 µm). The grain size classification was done in accordance with the ISO-standard 14688. To remove the sand fraction from the sediment, the sample was washed over a 63 µm-mesh sieve. The fine fraction was then divided into clay and silt using the Atterberg method based on Stoke's Law. The fractions were weighed afterwards. The bulk silt fraction was then used for the grain size analysis with a Micromeritics®

Sedigraph 5100. After the grain size analysis, carbonate was dissolved from the silt samples using 12.5% Hydrochloric Acid (HCl). The remaining material was then weighed and analysed again with the Sedigraph to determine the grain size distribution of the terrigeneous component. The grain size distribution of the carbonate silt was then calculated from the results of the bulk and terrigeneous silt analyses. Beside the grain size distribution of the carbonate silt also the carbonate content of the bulk silt fraction was determined by the weight difference between untreated and treated samples. All weights are expressed in relative terms and have been recalculated to 100 wt.%.The carbonate and organic carbon content of the bulk sediment and of the clay fraction, respectively, were measured by combustion analysis using a LECO® CS-200 analyzer. Portions of the bulk sediment and of the clay fraction, respectively, were grounded and divided in two parts. One part was directly measured to resolve the total carbon content (TC). The second part was treated with 12.5% HCl to remove the carbonate in order to measure the total organic carbon content (TOC) of the samples. The carbonate content of the samples was determined as follows (Eq. 1):

Eq. 1 ( ) [ ] [ ]%8,33%. 3 −⋅−− wtCaCO=wtTOCTC

Age model and Mass Accumulation Rates

The employed age model was generated by orbital tuning of a high-resolution composite XRF-Fe intensity record of ODP Sites 1085 and 1087 (Westerhold et al. 2005). The linear sedimentation rates (LSR) in the present study derive from this earlier age model. Mass accumulation rates (MAR) were calculated for the bulk sediment (MARbulk), the carbonate fractions (MARcc) and the terrigeneous fractions (MARterr). To calculate MAR for specific fractions, fraction contents are multiplied by MARbulk. The MAR are calculated from the LSR and the dry bulk density (DBD) (van Andel et al. 1975) (Eq. 2):

Eq. 2 [ ] [ ] [ ]3112 cmgDBDkyrcmLSR=kyrcmgMAR −−−− ⋅⋅⋅⋅⋅

The DBD (Eq. 3) is calculated from grain density (ρgc) and porosity (Φ) (Eq. 4). Φ is calculated from ρgc, the wet bulk density (ρbc) and the fluid density (ρfc, here: 1.025 g*cm-3).

Eq. 3 Φρ=DBD gc ⋅

Eq. 4fcgc

bcgc

ρρρρ

=Φ⋅⋅

4

The results of the Sedigraph analysis were plotted with Surfer for Windows 7, the surfaces for the plot were interpolated from the data points by means of a Kriging Algorithm.

Results

Granulometry

The bulk sediment is mainly composed of fine fraction grains (<63 µm). The samples contain very low amounts of sand (<3.0 wt.%). The fine fraction is composed of clay for the most part (~55 wt.%). About 40 wt.% of the bulk sediment consists of silt (Fig. 2).The sand fraction predominantly consists of planktonic and a few benthic foraminifers. A few sponge spicules are also present. Terrigeneous material is almost absent. During microscopic examination only a few quartz grains and lithoclasts were found. In the clay fraction the highest contents of terrigeneous material are found, terrigeneous clay averages to 37 wt.% over the whole section. Four excursions showing silt contents of more than 50 wt.% were found: two in the lower part at 11.3 Myr and 13.7 Myr, and two in the upper part at 10.9 and 10.5 Myr, The two latter excursions match with peaks in the MAR of nannofossil carbonate (Krammer et al. 2006), respectively (Fig. 2). The silt fraction consists of up to about 80% carbonate particles (Fig. 2). This fraction is made of mostly coccoliths (fine silt) and foraminifer fragments (coarse silt) (cf. Krammer et al. 2006).

Fig. 2 Grain size fraction contents of Site 1085A, stacked, and normalised to 100 wt.%. Light grey: terrigeneous clay; light medium grey: carbonate clay; dark medium grey: terrigeneous silt; dark grey: carbonate silt; black: bulk sand.

The carbonate plot (Fig. 3a) displays a strongly right skewed, unimodal and well sorted pattern. Only two values at 12.0 Myr and 13.3 Myr reveal a finer grain size maximum. Also in this plot it can be seen that the carbonate silt fraction is merely made up of coccoliths, displayed by a calculated mean value of 3.90 ± 0.67 µm

5

for the finer silt. Over the depth range several zones with peak concentrations over 5 wt.% were found. Two of them are located at grain sizes around 3 µm at 12.0 and 13.3 Myr. There are three zones comprising concentrations over 5 wt.% around 4 µm: from 13.8 to 12.5 Myr, from 11.5 to 10.5 Myr and between 10 and 9.6 Myr.

Fig. 3 Plots of Sedigraph results. a) top: Carbonate silt fraction; b) bottom: Terrigeneous silt fraction. x-axis: Age, y-axis: Grain size, shading shows contents of calcareous and terrigeneous silt in wt.% normalized to the bulk silt weight. The areas were compiled from the Sedigraph measurements with Surfer for Windows 7 using a Kriging algorithm.

The Sedigraph plot for the terrigeneous silt fraction (Fig. 3b) shows a right skewed, unimodal and well sorted pattern. Over the depth range the plot shows six

6

distinct frequency maxima with contributions greater than 1 wt.% to the grain size distribution. There are smaller maxima from 13.45 to 13.6, at 12.4, 11.9, 11.4 and a smaller one at 11.15 Myr. A zone of values greater than 1 wt.% spans from 10.6 to 9 Myr with frequencies of more than 2 wt.% from 9.6 Myr onwards. The frequency distribution also shows that very little coarse silt is present in the samples. From 10.2 Myr onwards, coarser grain sizes can be seen in the terrigeneous fraction (Fig. 3b), whilst the portion of coarser grain sizes in the carbonate fraction decreases after 10 Myr (Fig. 3a).

Sedimentation patterns

The bulk sediment at Site 1085 is composed of calcareous shells (mainly coccolithophorids and occasional foraminifers) and terrigeneous material. The organic carbon contents are rather low (0.29 ± 0.18 %). The LSR and MAR record for the bulk sediment reveals a significant rise in sedimentation and accumulation after 11.8 Myr. Before 11.8 Myr, LSR fluctuates at low levels between 1 and 3 cm/kyr (Fig. 4a).There is an exception between 13.2 and 13.6 Myr where LSR and MAR rates are quite low. The rise in LSR and MAR and the period of low LSR and MAR are also found in the carbonate and the terrigeneous records (Fig. 4b, 4c).

Fig. 4 Linear Sedimentation Rates (bars) and Mass Accumulation Rates (black lines) of a) bulk sediment (dark grey, left), b) calcareous material (medium grey, center) and c) terrigeneous material (light grey, right) versus age.

Carbonate sediments

Carbonate is delivered to ODP Site 1085A by biogenic production, particularly from coccolithophorids. This can be seen in the grain-size composition of the carbonate fraction. The fraction contents record (Fig. 5a) shows, that variations of the bulk carbonate content are driven by variations in carbonate clay and fine silt and in particular by particles with sizes between 2 and 8 µm. Coarse carbonate silt (8 to 63 µm) is considered to consist of fragments of foraminifers (Frenz et al. 2005; McCave et al. 1995). For the sand fraction, higher contents before 11.8 Myr are observed, followed by lower contents. The sand contents are however

7

generally low in the studied samples. The signal of the coccoliths is found in the <8 µm fraction according to Frenz et al. (2005). The correlation between MAR (Fig. 6a) and the mass frequency of the coccolith fraction is quite good, especially after 12.4 Myr. The correlation between the carbonate clay MAR and its mass frequency is lower to that of carbonate silt.

Fig. 5 a) Grain size class contents of the carbonate fraction. Light grey line: silt fraction >8 µm; grey line with crosses: silt fraction <8 µm; black line: clay fraction. b) Contents of terrigeneous silt fractions. Light grey line: silt >10 µm; dark grey line: silt <10 µm; c) terrigeneous clay contents. All values are plotted as wt.% versus age and based on total sample mass.

In the section from 12.4 to 11.2 Myr lower contents, in particular for the fine silt and clay fraction, are observed. The frequency variabilities between the successive samples are more distinct. This is due to higher and more variable contents of the terrigeneous clay component (see below). After 11.2 Myr both values rise again. The fine silt reaches its maximum value at 10.8 Myr, but then contents begin to decrease down to less than 10 wt.% at 9.3 Myr. The contents of coarse silt remain stable at about 10 wt.% until 11.2 Myr. During the following 400 kyr they decrease to about 5 wt.% and stay at this level in the youngest investigated section. The carbonate clay contents are rising after 11.2 Myr and show a minimum at 10.3 Myr. Between 10.3 to 9.6 Myr carbonate clay has got very high amount with a peak value of more than 50 wt.%, while terrigeneous clay shows an opposite pattern.

Terrigeneous sediments

Terrigeneous material is almost exclusively found in the fine fraction < 63 µm. Values for the coarse silt are mostly below 2 wt.%. The contents for the fine silt fraction vary between 4 and 12 wt.% (Fig. 5b) with the highest contents at top of the section. The silt fraction MAR plot shows low contents until 11.8 Myr, after which time the MARterr-silt is slightly increasing (Fig. 6b) and reaches its maximum at 10.8 Myr. Towards the top of the section MARterr-silt decreases again. The MAR values for the terrigeneous clay fraction are higher and much more variable. After 11.8 Myr they rise steeply. Within 300 kyr the MARterr-clay increases four-fold. Before 11.8 Myr values stay below 2 g/cm2/kyr and after this time they are always higher than 2 g/cm2/kyr except of the period between 10.3 and 9.5 Myr. During this period in particular the terrigeneous clay contents are low (Fig. 5c). In addition, a sedimentation minimum is clearly displayed between 13.6 and 13.2

8

Myr. The record shows clay values of less than 14 wt.%, and very low silt values of 4.7 wt.% for the fine silt and 0.7 wt.% for the coarse silt at 12.1 Myr (Fig. 5b, 5c).

Fig. 6 a) MAR of calcareous fractions versus age. Grey line: silt fraction, black line clay fraction. b) MAR of terrigeneous fractions versus age. Grey line: silt fraction, black line clay fraction. c) MAR of organic matter in the bulk fraction versus age.

At 13.4 Myr the silt values were too low to be measured with the Sedigraph and the variation in terrigeneous sedimentation results mainly from the clay fraction (Fig. 5c). Looking at the values for fine and coarse silt, one can see that in the younger part after 11.0 Myr more fine silt is found in the samples, while coarse silt values stay below 2 wt.% (Fig. 5b). There is almost no correlation between accumulation of clay and silt throughout the whole core segment. Also for shorter intervals the correlation coefficients are low. A good positive correlation exists between fine and coarse silt from 13.8 to 11.0 Myr (R²=0.805); after 11.0 Myr the correlation coefficients again return to low values.

9

Fig. 7 Mean values for the terrigeneous silt fraction versus time. Sedigraph results were divided into a) a fine (2–10 µm) fraction and b) a coarse (10–63 µm) fraction. For each fraction the mean grain size has been calculated. The mean size of the fine silt remains within a narrow range. The plot for the coarse silt shows substantially higher mean values in the lower than in the upper part with a transition in the center. Dotted grey lines were added to clarify the trends.

The mean values of terrigeneous fine silt (Fig. 7a) vary over a narrow bandwidth. Before 11.8 Myr this band is less than 0.4 µm wide, after this point it expands to 0.6 µm. The coarse silt mean values (Fig. 7b) comprise a broader bandwidth with a clear trend to coarser means in the younger part after 11.6 Myr. Before 11.8 Myr the means vary between 14 and 17 µm, after which the values spread between 12.5 and 19 µm. This coarsening points to a stronger wind activity.

Organic sedimentation

Organic carbon contents are generally low in the older part of the record. For the bulk sediment TOC values do not exceed 0.3 wt.% until 11.2 Myr. Up to 12.2 Myr MARTOC is below 0.01 g/cm²/kyr. After that a slight rise in MARTOC is apparent, followed by a steep rise after 11.8 Myr (Fig. 6c). This rise is consistent with the findings of Kastanja et al. (2006), who also found an enhanced MARTOC

from 11.8 Myr on. It also coincides in time with the rise of MARbulk. After 11.8 Myr MARTOC values are larger than 0.02 g/cm²/kyr except during a short decline at about 9.5 Myr. The rise of MARTOC at 11.8 Myr is also a hint for the commencement of the upwelling because it coincides with the rise in MARcc. Later, no correlation of MARcc and MARTOC is apparent because values for MARcc

decline again (Fig. 6a) while MARTOC remains at a high level (Fig. 6c).

Discussion

During the Cenozoic Southern Africa underwent intervals of both, arid and humid, conditions. Arid conditions in the Namib region prevailed from the beginning of the Cenozoic to about 50 Myr (Partridge 1993). During these times

10

cross-bedded aeolian sands were deposited in the Buntfeldschuh Formation (Fig. 1) in the southern Namib indicating hyper-arid conditions (Ward et al. 1983). Humid conditions followed this phase. Palaeontological evidence was found in the Grillental Beds and at Arrisdrift (Fig. 1), showing that from 18-12 Myr a humid climate predominated in Southwest Africa (Partridge 1993). At both sites units of the Middle Miocene Tsondab Sandstone are exposed. In the Grillental Beds fossils were dated to times between 18.5 and 16 Myr. At Arrisdrift fossil ages range from 18 to 12 Myr (Partridge 1993). The records of the climatic development of SW-Africa and the SE Atlantic show that the upwelling in the Benguela region starts about 10 Myr ago (Kastanja et al. 2006).

Fig. 8: Benthic foraminiferal δ18O data from Site 1085A versus age modified from Westerhold et al. (2005). In the grey curve the original values are plotted (Westerhold 2005). Black curve represents a 19 value running average. Oxygen isotope data were normalised against PDB standard and were not adjusted for isotopic disequilibrium. Arrows mark increased δ18O-values coinciding with Mi-Events 3 to 6 as found at ODP Site 1085A (Westerhold et al. 2005).

The temperature development in the Cape Basin is expressed by the benthic foraminiferal δ18O-record of ODP Site 1085 (Fig. 8). It shows similarities to the global δ18O-record (Zachos et al. 2001). From 13.9 to 11.8 Myr smoothed values range between 1.65 and 2.1 ‰, from 11.75 to 10.2 Myr a rising trend between 2 and 2.65 ‰ is found. From 10.2 Myr on values stabilise again between 2.2 and 2.6 ‰ near the Holocene δ18O level (Westerhold et al. subm.). δ18O-increases of 0.4 to 0.7 ‰ were assigned to the Miller-Events Mi-3 to Mi-6 by Westerhold et al. (2005), while identification of Mi-7 is uncertain. These point to a connection to the climate development over and around Antarctica. The main question is, how the oceanographic changes influenced the continental climate over Africa.At about 10 Myr opal signals from the BC are detected at DSDP Sites 362 and 532 (Diester-Haass et al. 1992; 1990). This is also the time at which climatic changes initiated a desertification of SW Africa, especially in the Namib region, where a Late Miocene sand sea was re-established (Partridge 1997). Partridge (1993) noted that these desert sediments, which are preserved in the Tsondab Sandstone Formation, appeared “no earlier than about 10 Myr ago”. “Towards the end of the Mid-Miocene fluvial activity began to decline in the extensive Koa River basin of the western interior of South Africa” (Partridge 1997).

11

General patterns

According to Kastanja et al. (2006) the high carbonate concentration indicates a generally good carbonate preservation during the studied interval. In conjunction with the relatively shallow water depth (of 1713 m today) this leads to the suggestion, that the site was situated well above the Miocene Carbonate Compensation Depth (CCD). It has been shown that at ODP Site 1265 (Walvis Ridge) the carbonate lysocline varied between 2,000 and 3,000 meters during the Miocene (Kastanja and Henrich 2007). During the Pliocene and Pleistocene the CCD has risen during cold stages (Henrich et al. 2003; Howard and Prell 1994). However, those CCD elevations reached about 2000 m and depended on northern hemisphere ice caps. The similar appearance of MARcc and MARterr leads also to the suggestion that carbonate dissolution plays no significant role at Site 1085A during the investigated time period. Intervals of decreased MARcc occur due to weakened nannofossil productivity during those times (Krammer et al. 2006). The carbonate is mainly made up of nannofossils (Krammer et al. 2006; Kastanja et al. 2006.) and this is supported in this study by the presented grain size analysis. The conditions for phytoplankton productivity can be characterised as mesotrophic (Krammer et al. 2006). According to the definition of Burckle et al. (1996), the productivity of calcareous nannofossils is coupled to the sea-surface temperature, thus the temperatures were temperate to warm. These features were also found for records of the more southern ODP Sites 1088 and 1092 (Diekmann et al. 2003).

Transport mechanisms for terrigeneous silt

The terrigeneous silt fraction is dominated by fine silt. The silt fraction can be classified as accumulated sediment transported by suspension (Höppner and Henrich 1999). The small mean grain size (4.87 µm) and the low standard deviation about the mean (±0.42 µm) are the characteristic features for this fraction. Three possible mechanisms are taken into account to transport the silt.

Transport by ocean currents

Sedimented silt can be resuspended by ocean currents and then be resedimented. Published statements about transported silt are generally limited to grain sizes coarser than 10 µm, the so-called sortable silt, because below this size limit grains show a cohesive behaviour (e.g. McCave et al. 1995). The Sedigraph plot (Fig. 3b) shows, that the contribution of terrigeneous sortable silt grain sizes to the whole silt fraction is small. Therefore statements about ocean currents as a transport mechanism for terrigeneous silt are not possible, because there is nearly no terrigeneous sortable silt (McCave et al. 1995). The Sedigraph plots show continuous spectra for both (Fig. 3a;3b) silt fractions. If the sorting would have been altered by ocean currents, a part of the spectra should be missing. It seems that ocean currents do not play an identificable role as a transport mechanism for silt at ODP Site 1085A in the Middle and Late Miocene.

Wind-borne transport

At present time, the air is an important medium to transport silt from Southwest Africa into the Southeast Atlantic. The grain size of dust derived from ocean cores and dust samplers depends on the distance to the sediment source and 4 to 5 µm-sized dust, as detected in the core samples here, is normally found several

12

hundreds of kilometers away from its source (Boven and Rea 1998). The distance of ODP Site 1085A to the coast is only about 260 km, so coarser grain sizes are to be expected. Accordingly studies of present-day terrigeneous silt from this region display coarser grain sizes for dust originating from the Namib desert (Stuut et al. 2002; Stuut and Lamy 2004). The mode of their samples is at 15 µm (Stuut et al. 2002). The mean sizes for the terrigeneous silt >10 µm found in this study vary between 13 and 19 µm (Fig. 7), which matches quite well with the data from the mentioned studies. However, the amount of this fraction is quite low, so wind transport should not be taken into consideration as the main process to incorporate terrigeneous silt to Site 1085A.

Fluvial transport

Fine grained silt (2-4 µm) in hemipelagic sediments off Southwest Africa originates from rivers (Rea and Hovan 1995). With respect to the position of ODP Site 1085A off the Orange River mouth, and the measured grain sizes, river discharge is interpreted as the main source for the terrigeneous material. The high fine fraction content in the terrigeneous fraction obscures the view on other sediment transport possibilities. In this study the mean grain sizes for terrigeneous silt <10 µm vary between 3.6 and 4.2 µm (Fig. 7). Therefore wind transport and silt resuspension play a minor role for the composition of the terrigeneous silt at this site. Thus, the supply of terrigeneous material is governed by fluvial input.

Transport mechanisms for terrigeneous clay

Terrigeneous clays contribute importantly to the sediments found at ODP Site 1085A (Fig. 2). In the younger part between 10.5 and 10.2 and after 9.6 Myr, respectively, more than 45 wt.% of the bulk sediment consist of clay. Over the whole interval values vary between 20 and 50 wt.%. The sources, which deliver clay minerals to the site are more diverse compared to the silt fraction and this is reflected in the transport pathways.Smectite and illite are the two dominating clay minerals at ODP Site 1085A (Robert et al. 2005). Smectite derives from the Karoo lavas and the Cape Province in South Africa. It is transported by the Orange River especially during high precipitation (Bremner and Willis 1993; Robert et al. 2005). Illite is also contained in the Orange River's sediment load, but much more illite is transported from the Namib desert, by wind or flash floods (Bremner and Willis 1993). Another source of illite and chlorite are Antarctica and South America, these particles are transported by ocean currents (e.g. ACC and BC to the Southeast Atlantic (Diekmann et al. 1996; Diekmann et al. 2004). Small kaolinite amounts were also found which should have been transported by ocean currents like the NADW or Angola Current (Robert et al. 2005), because kaolinite is a product of tropical weathering. It is difficult to trace sources and transport pathways of illite, chlorite and kaolinite here, because they are manifold. But ocean current transport is more important for the clay fraction than for the silt fraction.

13

Climatic and palaeoceanographic implications derived from sedimentation patterns during the Middle Miocene (13.8 to 11.8 Myr) – Phase 1

The LSRbulk (Fig. 4) in the first phase ranges between 1 and 3 cm/kyr. These rates are very low compared to Late Miocene and Quaternary LSR (Kastanja et al. 2006; Diester-Haass et al. 2004). In most sediment sections during this period the MARterr and the MARcc are positively correlated. There seem to be no factors in these intervals, that favour terrigeneous influx against marine productivity or vice versa. Two periods of negative correlation are identified. From 13.5 to 13.3 Myr and 12.7 to 12.5 Myr, carbonate production is higher than terrigeneous influx. The first period coincides with very low LSR and MAR, particularly for the terrigeneous fraction. In contrast, the second period coincides with a longer period of high LSR and MAR, particularly for the carbonate fraction. When comparing the MAR curve with the timing of the Mi-Events 3 and 4 at ODP Site 1085 (Westerhold et al. 2005), no correlated excursions are found. During the period of lower benthic δ18O-values from 13.6 to 13.2 Myr (Fig. 8) (Westerhold et al. 2005) very low MAR and LSR values are found. Miller et al. (1991) showed, that the Miocene δ18O-record is controlled mainly by the effects of waxing and waning of the Antarctic ice sheets. Low δ18O values indicate therefore a small ice extent over Antarctica and, globally, warm temperatures. Therefore a warm climate coincides with low oceanic productivity in the Southeast Atlantic and low sediment influx from the Orange River. The long period of high LSR and MAR between 12.8 and 12.4 Myr is situated in a period of rather stable δ18O values of about 1,9 ‰ (Fig. 8) (Westerhold et al. 2005). MAR values for this period indicate that biogenic carbonate production is enhanced during these times, but terrigeneous influx stays low. The pelagic carbonate production benefits from this stable Antarctic temperature conditions which are reflected in the δ18O values, while continental climate is unaffected.Phase 1 ends near the stratigraphic border between the Serravallian and the Tortonian at 11.6 Myr (Gradstein et al. 2004) and near the Mi5-event which has been dated at this site at 11.7 Myr (Westerhold et al. 2005). During Phase 1 there are no indications for an upwelling zone in the ocean off SW-Africa. In addition, no evidence for strengthened trade winds and no significant rise in biogenic carbonate production or TOC is found (Fig. 4c and Fig. 6c). The MARterr is not coupled to the δ18O values, which reflect Antarctic ice coverage, so the climate in SW-Africa is not governed by waxing or waning of Antarctic ice sheets. Therefore the differences in LSR and MAR can only relate to local climatic variations. The variations in LSR and MAR, in particular for terrigeneous sediments, are quite small, showing a stable climate during this phase. Denudation is low, there is no verifiable information concerning wind-borne silt in the core section, what means that there are no indications for a desertification. Also sediment influx from the Orange River is low, as demonstrated by the low MARterr, and governed by clays. A rather stable, humid climate is hypothesised, that facilitates a close vegetation cover, which minimises denudation.

14

Climatic and palaeoceanographic implications derived from sedimentation patterns from the latest Middle Miocene to the early Late Miocene (11.8 to 10.4 Myr) – Phase 2

The most remarkable feature of the second phase are the rising LSR and MAR values for the carbonate, terrigeneous and organic matter fractions. This phase begins a little bit earlier than the Tortonian and its end coincides with the Mi6 event at 10.4 Myr (Westerhold et al. 2005). At 10.5 Myr a prominent sea-level fall occurred (Miller et al. 2005; Haq et al. 1987), that marks the sequence boundary between the supercycles TB2 and TB3 (Haq et al. 1987). The LSRbulk increases to 4 cm/kyr on average, with peak values of more than 8 cm/kyr. MARterr and MARcc

are both rising, but the coupling between the records is not as strong as in phase 1. We find a direct correlation of the MARterr to the southern hemisphere climate. At 11.7 Myr Antarctica was glaciated, this event is termed Mi5 (Miller et al. 1991) and induced a fall in sea level (Haq et al. 1987). The shelves were exposed and terrigeneous material was transported directly to the deep ocean. Westerhold et al. (2005) quote estimates of sea-level fall ranging between 43 and 85 m by various authors (Fairbanks and Matthews 1978; Bemis et al. 1998; Billups and Schrag 2002; John et al. 2004; Miller et al. 1998). Calcareous and terrigeneous material was also eroded from the exposed shelves and accumulated on the slope due to the glaciation of Antarctica (Kastanja et al. 2006). One would assume a rising delivery of terrigeneous silt and a coarsening of grain sizes due to shorter transport distances during the sea-level lowstand. The contents of terrigeneous silt are stable (Fig. 2,3b). The mean grain sizes reveal a positive peak for fine and especially for coarse silt during the Mi5 event (Fig. 7). The coarse silt peak points to enhanced dust transport during the cold phase. It is likely, that shelf sediments were remobilised by seaward winds. The sea-level lowstand did not alter the fluvial transport very much, because the coarsening of the fine silt means at 11.7 Myr is not prominent. This glaciation on Antarctica can also be identified in the benthic δ18O values (Fig. 8) (Westerhold et al. 2005) by steadily rising values between 11.8 and 10.5 Myr. From 11.7 Myr onwards MARTOC is rising slightly. Strong cooling of the Southern Atlantic may have favoured oceanic productivity off Southwest Africa. At 11.2 Myr the first indications for the commencement of upwelling in the Benguela Region are found, e.g. MARTOC is rising rapidly at this time (Fig. 6) as has been described by Kastanja et al. (2006). However, initiation of upwelling seems not to be accompanied by a concomitant desertification of Southwest Africa. There is no significant rise in wind transported silt visible (Fig. 5b). In contrast, the levels of terrigeneous fine silt are rising, they double between 11.0 and 10.4 Myr (Fig. 5b). The MAR for terrigeneous clay falls rapidly at 11.5 Myr, after it had reached a peak value of 3.5 g/cm2/kyr (Fig. 5b). The MARterr-clay

plot shows a second maximum of 4.0 g/cm2/kyr at 10.5 Myr. The fraction contents for the carbonate silt fraction <8 µm show a rising trend after 11.5 Myr, in the same manner as the values for carbonate clay. At 11.3 and 10.9 Myr contents of the fine carbonate silt fraction reach peak values for this phase. This is an effect of increased marine productivity and points to the beginning upwelling. On the other hand the δ18O-record of Westerhold et al. (2005) points to a glaciation on Antarctica at 11.2 Myr and this may have induced a drier climate in southwest Africa. Therefore the river runoff was reduced as documented by the lower MARterr-clay. The MARterr-clay minimum for phase 2 occurs at 10.9 Myr. The means of coarse terrigeneous silt rise at 11.0 Myr pointing to increased wind strength

15

probably as a result of the glaciation on Antarctica (Fig. 7). However, the contribution of coarse terrigeneous silt is rather low.In this phase the sedimentation responds to the trend of general cooling. Input of fine carbonate silt, carbonate clay and terrigeneous clay rises as the Southern Ocean cools. The beginning of the upwelling around offshore Southwest Africa is documented by a significant rise of MARTOC, especially from 11.2 Myr onwards. The regional climate in Southwest Africa's hinterland is affected by oceanic cooling and tends to be more humid, as shown by the rising MAR in terrigeneous clay and river-borne silt contents. Therefore desertification cannot be expected in this region during these times. Obviously seaward winds do not prevail and humidity can spread from the ocean to the land. To the end of the second phase a trend of general decline in LSR and MAR of all fractions is visible (Fig. 4, 6). At the same time a change in sediment composition is recognisable (Fig. 5). Firstly, the fraction content of coarse calcareous silt decreases by about 50 %.

Climatic and palaeoceanographic implications derived from sedimentation patterns during the early Late Miocene (10.4 to 9.0 Myr) – Phases 3a, 3b

At the beginning of phase 3a at 10.4 the MARcc and MARterr behave contrarily (Fig. 4; 6). Starting at 10.3 Myr contents of fine calcareous silt are lower than those of calcareous clay. This is caused by a change in coccolith production, as coccolithophorids are the main CaCO3 producers in this location (Krammer et al. 2006). At 9.6 Myr phase 3b begins, that is characterised particularly by a decrease in carbonate clay contents and an increase in terrigeneous clay and terrigeneous fine silt (Fig. 5). This means, that fluvial sediment input gains importance in the upper part of the studied section and probably also more clay is transported by the BC. The decrease in terrigeneous clay content in phase 3a is not only due to simultaneously elevated values of calcareous clay (Fig. 5a), since the same pattern is also displayed by low accumulation of terrigeneous clay during this interval (Fig. 6). At 9.6 Myr a short phase of very low mass accumulation rates for bulk sediment (Fig. 4a) is evident. After 9.6 Myr the MARterr-clay rises again together with MARterr-silt due to a sea level regression (Diester-Haass et al. 2004; Haq et al. 1987). During this regression the shelves were exposed and consequently the Orange River mouth moved seawards. Hence, the fluvial sediment supply at Site 1085A was more proximal and thus increased considerably. The regression is also responsible for the decline of MARcc (Diester-Haass et al. 2004) together with a further decline in coccolith production (Krammer et al. 2006). In the earlier part of Phase 3 the MARTOC values also decline (Fig. 8), in line with the sedimentation minimum. However, after 9.6 Myr the values rise again together with a rise in terrigeneous clay MAR. This points to a connection between terrigeneous clay and organic matter during this phase, probably because of the better preservation in a clay dominated sedimentation regime.At the beginning of phase 3 the low coarse terrigeneous silt values (Fig. 6) indicate, that there was not an increase in aridity, because there is no significant increase in coarse silt that would provide evidence for a greater source area. On the other hand, the mean size of coarse, terrigeneous silt rises at the beginning of phase 3 (Fig. 7), providing evidence for increasing wind strength. Later the contents of coarse terrigeneous silt rise (Fig. 5b), which is linked to the increase of wind strength and to a greater source area. This points to the initiation of

16

aridification in Southwest Africa. However, even at this time fluvial supply remains the dominant process at this site.

Conclusions

The grain size proportions of the terrigeneous and calcareous fractions of samples covering a 4.9 Myr long interval from the Middle to Late Miocene core section were investigated. No desert such as that seen today existed at the Southeast Atlantic coast or in the hinterland during these times. The results of this study point to a steppe landscape in this region, that evolved into desert conditions from the Late Miocene on. This is demonstrated by the grain size distribution of the terrigeneous silt fraction. The silt composition points to predominantly fluvially transported silt, because the grain sizes are too small for aeolian or bottom current transport and the distribution is unimodal (Fig. 3b). The fluvial input dominates the terrigeneous silt fraction generally, which is visible in the fine silt contents (Fig. 5b). The dust is responsible for the minor, coarse part only, which is a new result compared to the work of Kastanja et al. (2006).Clay minerals, particularly smectite, are also contained in the Orange River's sediment load, which supports the idea of fluvial transport. But they are also transported by winds and ocean currents. The climate in Southwest Africa during most of this times was humid and the land was covered with vegetation. Three phases were found in the record, one from 13.8 to 11.8 Myr, the second phase from 11.8 to 10.4 Myr and a third one from 10.4 to 9.0 Myr. During the first phase no evidence was found for a link between the climatic evolution of Antarctica and Southwest Africa. For the second phase a decrease of sea surface temperatures (Hodell et al. 2002) in the Southern Ocean is associated with higher terrigeneous and calcareous sedimentation rates (Fig. 4). At 11.2 Myr MARTOC rises dramatically, which may point to the initiation of the Benguela Upwelling system. But at this early stage the upwelling is not accompanied by dust influx. During the phase 3a carbonate production decreases, as does MARbulk. This runs in parallel with a sea level fall. After 9.6 Myr in phase 3b MARterr and MARTOC are enhanced. The climate cooled and the amount of river transported sediment points to at least a semi-humid climate in the hinterland. The mean values for coarse, terrigeneous silt rise in phase 3, which points to the beginning of a drier climate and a desertification of the SW African coastal area.

Acknowledgements

We are thankful for various comments from the colleagues in the SedPal working group. David Heslop is thanked for additional comments and proofreading. Critical reviews by Dan Bosence and an anonymus reviewer improved the quality of this paper. Helga Heilmann and Brit Kockisch are thanked for laboratory assistance. Dorothee Koch is thanked for work on sample preparation. This research used samples and/or data provided by the International Ocean Drilling Program (IODP). This study was funded through the DFG Research Center / Excellence Cluster “The Ocean in the Earth System”. This is MARUM publication no. MARUMXXXX.

ReferencesBemis BE, Spero HJ, Bijma J, Lea DW (1998) Reevaluation of the oxygen isotopic composition of planktonic foraminifera: experimental results and revised paleotemperature equations. Paleoceanography 13(2):150-160

17

Boven KL, Rea DK (1998) Partioning of eolian and hemipelagic sediment in Eastern Equatorial Pacific core TR 163-31B and the Late Quaternary paleoclimate of the Northern Andes. Jour of Sed Res 68(5):850-855

Billups K, Schrag DP (2002) Paleotemperatures and ice volume of the past 27 Myr revisited with paired Mg/Ca and 18O/16O measurements on benthic foraminifera. Paleoceanography 17(1):3-1-3-11

Bremner JM, Willis JP (1993) Mineralogy and geochemistry of the fraction of sediments from the Namibian continental margin and the adjacent hinterland. Mar Geol 115:85-116

Burckle LH, Mortlock R, Rudolph S (1996) No evidence for extreme, long term warming in Early Pliocene sediments of the Southern Ocean. Mar Micropaleontology 27:215-226

Diekmann B, Fälker M, Kuhn G (2003) Environmental history of the south-eastern South Atlantic since the Middle Miocene: evidence from the sedimentological records of ODP Sites 1088 and 1092. Sedimentology 50:511-529

Diekmann B, Fütterer DK, Grobe H, Hillenbrand CD, Kuhn G, Michels K, Petschick R, Pirrung M (2004) Terrigeneous Sediment Supply in the Polar to Temperate South Atlantic: Land-Ocean Links of Environmental Changes during the Late Quaternary. In: Wefer G, Mulitza S, Ratmeyer V (eds.) The South Atlantic in the Late Quaternary – reconstruction of material budgets and current systems. Springer, Berlin, pp 375-399

Diekmann B, Petschick R, Gingele FX, Fütterer DK, Abelmann A, Brathauer U, Gersonde R, Mackensen A (1996) Clay Mineral Fluctuations in Late Quaternary Sediments of the Southeastern South Atlantic: Implications for Past Changes of Deep Water Advection. In: Wefer G, Berger WH, Siedler G, Webb DJ (eds.) The South Atlantic: Present and Past Circulation. Springer, Berlin, pp 621-644

Diester-Haass L, Meyers PA, Rothe P (1990) Miocene History of the Benguela Current and antarctic ice volumes: Evidence from rhythmic sedimentation and current growth across the Walvis Ridge (Deep Sea Drilling Project Sites 362 and 532). Paleoceanography 5(5):685-707

Diester-Haass L, Meyers PA, Rothe P (1992): The Benguela Current and associated upwelling on the southwest African margin: a synthesis of the Neogene-Quaternary sedimentary record at DSDP Sites 362 and 532. In: Prell CP, Meister KC (eds). Upwelling Systems: Evolution since the Early Miocene. Geol. Soc. Spec. Publications, London, pp 331-342

Diester-Haass L, Meyers PA, Vidal L (2002) The late Miocene onset of high productivity in the Benguela Current upwelling system as part of a global pattern. Marine Geology 180(1-4):87-103

Diester-Haass L, Meyers PA, Bickert T (2004) Carbonate crash and biologic bloom in the late Miocene: Evidence from ODP Sites 1085, 1086, and 1087 in the Cape Basin, southeast Atlantic Ocean. Paleoceanography 19(1):PA1007

Fairbanks RG, Matthews RK (1978) The marine oxygen isotope record in Pleistocene corals, Barbados, West Indies. Quat Res 10:181-196

Frenz M, Baumann KH, Böckel B, Höppner R, Henrich R (2005) Quantification of foraminifer and coccolith carbonate in South Atlantic surface sediments by means of carbonate grain-size distributions. Jour of Sed Res 75(3):464-475

Gingele FX (1996) Holocene climatic optimum in Southwest Africa - evidence from the marine clay mineral record. Paleogeography, Paleoclimatology, Paleoecology, 122:77-87

Gradstein FM, Ogg JG, Smith AG (2004) A Geologic Time Scale 2004. Cambridge University Press, Cambridge, 598 p

Haq BU, Hardenbol S, Vail PR (1987) Chronology of fluctuating sea levels since the Triassic. Science 235:1156-1167

Henrich R, Baumann KH, Gerhardt S, Gröger M, Volbers A. (2004) Carbonate preservation in deep and intermediate water masses in the South Atlantic: evaluation and geological record (a review). In: Wefer G, Mulitza S, Ratmeyer V (eds.) The South Atlantic in the Late Quaternary – reconstruction of material budgets and current systems. Springer, Berlin, pp 645-670

18

Hodell DA, Gersonde R, Blum P (2002) Leg 177 Synthesis: Insights into Southern Ocean Paleoceanography on Tectonic to Millennial Timescales. In: Gersonde R, Hodell DA, Blum P (eds). Proc. ODP, Sci. Results, 177, pp 1-54 Available at http://www-odp.tamu.edu/publications/177_SR /VOLUME/SYNTH/SR177SYN.PDF. Cited: 2007-06-04

Holz C, Stuut JBW, Henrich R (2004) Terrigeneous sedimentation processes along the continental margin off NW Africa: implications from grain size analysis of seabed sediments. Sedimentology 51:1145-1154

Höppner R, Henrich R (1999) Kornsortierungsprozesse am argentinischen Kontinentalhang anhand von Siltkornanalysen.- Zentralbl Geol Pal 7-9:897-905

Howard WR, Prell WL (1994) Late Quaternary CaCO3 production and preservation in the Southern Ocean: implications for oceanic and atmospheric carbon cycling. Paleoceanography 9(3):453-482

John CM, Karner GD, Mutti M (2004) δ18O and Marion Plateau backstripping: combining two approaches to constrain late middle Miocene eustatic amplitude. Geology 32(9):829-832

Kastanja MM, Diekmann B, Henrich R (2006) Controls on carbonate and terrigenous deposition in the incipient Benguela upwelling system during the middle to the late Miocene (ODP Sites 1085 and 1087). Paleogeography, Paleoclimatology, Paleoecology 241:515-530

Kastanja MM, Henrich R. (2007) Grain-size variations in pelagic carbonate oozes from the Walvis Ridge–SE Atlantic Ocean (ODP Site 1265): A low resolution Miocene record of carbonate sedimentation and preservation. Marine Geology 237(1-2):97-108

Kastanja MM, Stuut JBW, Westerhold T, Henrich R (subm) End-member modelling of siliciclastic grain size distributions: the Late Miocene record of terrigeneous sedimentation processes along the continental margin off SW Africa (ODP Site 1085). Sedimentology (submitted)Koopmann B (1981) Sedimentation von Saharastaub im subtropischen Nordatlantik während der letzten 25000 Jahre. In: Diester-Haass L, Sarntheim M (eds). Factors contributing to Late Glacial and Holocene sedimentation on the continental shelf and slope off NW Africa, Banc d'Arguin. Meteor-Forschungsergeb 35:53-59

Krammer R, Baumann KH, Henrich R. (2006) Middle to late Miocene fluctuations in the incipient Benguela Upwelling System revealed by calcareous nannofossil assemblages (ODP Site 1085A). Paleogeography, Paleoclimatology, Paleoecology 230:319-334

McCave IN, Manighetti B, Robinson SG (1995) Sortable silt and fine sediment size composition slicing: Parameters for paleocurrent speed and paleoceanography.- Paleoceanography 10(3):593-610

Miller KG, Kominz MA, Browning JV, Wright JD, Mountain GS, Katz ME, Sugarman PJ, Cramer BS, Christie-Blick N, Pekar SF (2005) The phanerozoic record of global sea-level change. Science 310: 1293-1298

Miller KG, Mountain GS, Browning JV, Kominz M, Sugarman PJ, Christie-Blick N, Katz ME (1998) Cenozoic global sea level, sequences, and the New Jersey transect: results from coastal plain and continental slope drilling. Reviews of Geophysics 36(4):569-601

Miller KG, Wright SD, Fairbanks RG (1991) Unlocking the ice-house: Oligocene-Miocene oxygen isotopes, eustacy, and margin erosion.- J Geophys Res 96,B4:6829-6848

Partridge TC (1993) The Evidence for Cainozoic Aridification in Southern Africa. Quaternary International 17:105-110

Partridge TC (1997) Cainozoic environmental change in southern Africa with special emphasis on the last 200 000 years. Progress in Physical Geography 21(1):3-22

Rea DK, Hovan SA (1995) Grain size distribution and depositional processes of the mineral components of abyssal sediments: Lessons from the North Pacific. Paleoceanography 10(2):251-258

Robert C, Diester-Haass L, Paturel J (2005) Clay mineral assemblages, siliciclastic input and paleoproductivity at ODP Site 1085 off Southwest Africa: A late Miocene–early Pliocene history of Orange river discharges and Benguela current activity, and their relation to global sea level change. Mar Geol 216:221-238

19

Stuut JBW, Prins MYR, Schneider RR, Weltje GJ, Jansen JHF, Postma G (2002) A 300-kyr record of aridity and wind-strength in Southwestern Africa: inferences from grain-size distributions of sediments on Walvis Ridge, SE Atlantic. Mar Geol 180:221-233

Stuut JBW, Lamy F (2004) Climate variability at the southern boundaries of the Namib (southwestern Africa) and the Atacama (northern Chile) coastal deserts during the last 120,000 yr. Quat Res 62:301-309

van Andel TH, Heath GR, Moore TC (1975) Cenozoic history and paleoceanography of the Central Equatorial Pacific Ocean. Geol Soc Am, Boulder

Ward JD, Seely MK, Lancaster IN (1983) On the antiquity of the Namib. South African Jour of Sc 79:175-183

Wefer G, Berger WH, Richter C et al. (1998) Proc. Ocean Drilling Programme, Init. Reports, Vol. 175. Ocean Drilling Programme, College Station, pp 1-1477

Weigelt E, Uenzelmann-Neben G (2004) Sediment deposits in the Cape Basin: Indications for shifting ocean currents?. AAPG Bull 88(6):765-780

Westerhold T (2005) (Appendix A7) Stable isotope data of ODP Hole175-1085A-42X to -63X (382.40-594.12 mbsf). Available at: http://doi.pangaea.de/10.1594/PANGAEA.230283 (Cited: 2008-05-21)

Westerhold T, Bickert T, Röhl U (2005) Middle to Late Miocene oxygen isotope stratigraphy of ODP Site 1085 (SE Atlantic): new constrains on Miocene climate variability and sea level fluctuations. Paleogeography, Paleoclimatology, Paleoecology, 217(3-4):205-222

Wright JD, Miller KG, Fairbanks RG (1992) Early and Middle Miocene stable isotopes: implications for deepwater circulation and climate. Paleoceanography, 7(3):357-389

Zachos J, Pagani M, Sloan C, Thomas E, Billups K (2001) Trends, Rhythms, and Aberrations in global climate 65 Myr to Present. Science 292:686-693

20