Sediment samples from ODP Site 1085 were investigated in order to obtain more information on the initiation and development of the Benguela upwelling system during the middle and upper Miocene. In particular, our intent was to establish the causes of the upwelling as well as the response of the upwelling regime to the development of the Antarctic Circumpolar Current. Based on changes in the calcareous dinoflagellate cyst association, we found an initial increase of the dinoflagellate cyst productivity, probably related to the initiation of upwelling about 11.8 Ma ago. Two distinct increases in cyst productivity in conjunction with temperature decreases of the upper water masses reflect upwelling pulses off Namibia and occur at the end of the Miocene cooling events Mi5 (about 11.5 Ma) and Mi6 (about 10.5 Ma). Both cooling events are associated with an ice volume increase in Antarctica and are thought to have led to an increase in southeasterly winds, possibly causing these two upwelling pulses. We demonstrate a decrease in dinoflagellate cyst productivity and enhanced terrigenous input via the Orange River after the Mi5 event. At about 11.1 Ma, the dinoflagellate cyst productivity increases again. The polar cyst species Caracomia arctica occurs here for the first time. This implies an influence of subantarctic mode water and therefore a change in the quality of the upwelling water which allowed the Benguela upwelling to develop into modern conditions. From about 10.4 Ma, C. arctica forms a permanent part of the association, pointing to an establishment of the upwelling regime.
 The middle-to-late Miocene epoch is characterized by major climatic shifts to colder global temperatures. These periods of cooling (Mi events) are indicated by oxygen isotopic shifts related to stepwise increases of the Antarctic ice sheet [e.g., Miller et al., 1991]. To date, many hypotheses exist about the causes of the Antarctic ice buildup [DeConto and Pollard, 2003; Holbourn et al., 2005, 2007; Pagani et al., 2005]. Two climatic features are considered to have had a significant influence on the cooling of the global climate: 1) the decrease in atmospheric CO2, and 2) the isolation from heat by oceanic currents [Ruddiman, 2008]. Along with this cooling, the intensification of the Antarctic Circumpolar Current increased the temperature gradient between the poles and the equator. This favored the development of zonal winds. Enhanced zonal winds triggered the development of coastal upwelling along eastern boundary currents, such as the Benguela upwelling associated with the Benguela current [Vincent and Berger, 1985].
 Today, the Benguela upwelling system is one of the four major eastern boundary upwelling regions of the world ocean and is driven by the southeasterly (SE) trade winds. It can be divided into a coastal and oceanic upwelling. The coastal upwelling develops when the SE trades drive an offshore surface drift and induce upwelling over most of the continental shelf. The oceanic upwelling forms at a thermal front between the warmer coastal current and the further offshore flowing Benguela current. This front is oriented along the shelf break. Both upwelling features result in zones of enhanced productivity. Studies of the Benguela oceanic upwelling have shown an initiation of high productivity in relation to upwelling during the Miocene at about 10 Ma on Walvis Ridge (DSDP 362, 532 [Diester-Haass et al., 1990, 1992; Siesser, 1980]). More recent studies along the southwest African continental margin (ODP Leg 175 [Wefer et al., 1998]) provide evidence that the high productivity and upwelling initiation began earlier, about 12–11.5 Ma [Diester-Haass et al., 2004; Paulsen, 2005]. These studies document an increase in mass accumulation rates of organic carbon and benthic foraminifera, which indicates an increase in productivity and coincide with changes in the regional climate [Diester-Haass et al., 2004]. Based on sedimentological analyses, Diester-Haass et al.  found enhanced input of terrigenous material in the marine realm originating from the Orange River as well as a high content of shelf-derived particles. They suggested that a major regression occurred and increased shelf erosion. This is supported by the study of Haq et al.  who documented a regression in the global sea level curve over this time interval. An earlier start of the upwelling also coincides with the Mi5 event, which has been determined from an increase in δ18O at Site 1085 between 11.9 and 11.5 Ma [Westerhold et al., 2005]. Westerhold  and Westerhold et al.  suggested a strong relationship between the increases in δ18O during the Mi events with enhanced downslope transport of terrigenous matter and a sea level low stand during periods of extended ice sheets on Antarctica. A relationship between the initiation of the upwelling and the ice buildup on Antarctica seems therefore reasonable. However, it remains an open question as to why the upwelling was not initiated during the Middle Miocene Climate Transition some 14 Ma ago? During this transition, there is evidence that a major growth of the East Antarctic ice sheet associated with global cooling occurred [Billups and Schrag, 2002; Flower and Kennett, 1994]. Furthermore, it is documented that the atmospheric and oceanic circumpolar circulation intensified during the transition [Shevenell et al., 2004]. This resulted in an increased meridional temperature gradient and strengthened zonal winds, which could have initiated the Benguela upwelling. However, there is yet no explanation why there is a time lag between the initiation of the Benguela upwelling and the middle Miocene climate transition. A possible explanation might be the nutrient availability of the upwelled water. Sarmiento et al.  showed that besides the intensity of the southeast trade winds, nutrient advection from the subantarctic region is also an important factor promoting productivity in the upwelling area. They demonstrated that, nowadays, subantarctic mode water, which originates in the winter mixed layers around the Southern Ocean and spreads throughout the entire Southern Hemisphere and North Atlantic Ocean, is the main source of nutrients at the thermocline. It seems that the nutrient advection by subantarctic mode waters to low-latitude upwelling areas is somehow decoupled from the middle Miocene Antarctic glaciation. An indication for that is given by Paulsen  who observed an initial influence of southern water masses from 11.1 Ma on based on the occurrence of the polar foraminifer species N. pachyderma (s) at ODP Site 1085 off Namibia.
 In this study, we aim to contribute to this discussion by providing information about both the initiation of increased productivity as well as the source of the upwelled waters for the time period just after the Middle Miocene Transition. To accomplish this, we present a detailed reconstruction of past oceanographic conditions at Site 1085 located off the coast of Namibia based on calcareous dinoflagellate cysts. Dinoflagellates represent one of the major phytoplankton groups and are abundant in all oceanic environments [Dale and Dale, 1992; Marret and Zonneveld, 2003]. Calcareous cyst-forming dinoflagellates are phototrophic/mixotrophic and live in the photic zone. The cysts are species specific and the fossilized assemblages in marine sediments reflect the environmental conditions of the surface water masses at the time of their production [Vink, 2004; Zonneveld et al., 2000]. Numerous studies have shown their suitability for paleoclimate reconstructions [Bison et al., 2009; Esper et al., 2000, 2004; Meier and Willems, 2003; Richter et al., 2007; Vink, 2004; Zonneveld et al., 2005, 2000.]
 In modern times, there is a clear differentiation in dinoflagellate associations between regions of the Antarctic Circumpolar Current, the upwelling regions and the central South Atlantic Ocean. We demonstrate in this study that the first increase in productivity in these areas occurred at 11.8 Ma. At 11.1 Ma, species typical for Antarctic Circumpolar regions appear for the first time in the Benguela upwelling region. These species constitute permanent members of the association after 10.5 Ma, indicating a pronounced shift in the nutrient availability in the Benguela upwelling area at that time.
2. Study Area
2.1. Climate and Hydrography
 Today the south African climate is controlled by two subtropical anticyclones, the South Atlantic High and the Indian Ocean High, and a continental pressure field above South Africa [Shannon, 1985]. The pressure difference between the South Atlantic High and the continental pressure field results in alongshore winds (trade winds). These trade winds are the prevailing winds offshore western South Africa today. The strength and position of the anticyclones and the resulting winds produce the dry climate of western South Africa, the Namib Desert [Gasse, 2000]. The only perennial river in South Africa is the Orange River, where the river mouth is located at 28.5°S [Nelson and Hutchings, 1983]. The Orange River, with a length of 2160 km and a catchment area of 973,000 km2, is the largest river in South Africa. [Dingle and Hendry, 1984].
 The SE trade winds control the strength of the South Atlantic Gyre. Its southern branch, the South Atlantic Current, transports temperate nutrient-poor thermocline waters eastward [Stramma and England, 1999]. The Subtropical Front extends south of the South Atlantic Current and separates the South Atlantic Current from the Antarctic Circumpolar Current [Peterson and Stramma, 1991]. The Antarctic Circumpolar Current is banded into different fronts extending from the sea surface to the seafloor, and marks the boundaries between zones with different physical, chemical and ecological properties [Rintoul, 2009]. The driving force for the Antarctic Circumpolar Current are the strong westerlies as well as air-sea exchange of buoyancy [Rintoul, 2009]. South of the African continent, the South Atlantic Current and the Antarctic Circumpolar Current converge with the Agulhas Current, which originates in the Indian Ocean (Figure 1). In this retroflection area, a part of the Agulhas Current is redirected toward the south and east, creating the Agulhas Return Current [Lutjeharms and de Ruijter, 1996; Lutjeharms, 2001]. The other part of the Agulhas Current forms eddies, which are shed at the retroflection area [Gordon, 1986] and transport warm, saline water into the South Atlantic. Occasionally, a direct transfer of Agulhas waters into the South Atlantic occur via filaments [Lutjeharms and Cooper, 1996]. The Agulhas waters and the South Atlantic Current form the main source for the Benguela Current. The Benguela Current is driven by the alongshore trade winds and can be divided into the fast equatorward flowing Benguela Coastal Current and the weaker Benguela Oceanic Current, which turns northwest at about 28°S (Figure 1). Eckman pumping leads to the development of coastal upwelling. Currently, about eight permanent upwelling cells can be observed off the southwestern African coast. In these cells, cold thermocline water wells up to the surface [Lutjeharms and Meeuwis, 1987; Shannon et al., 2001]. Surface water filaments transport these cold and nutrient-rich waters offshore. The most intense upwelling can be observed near Lüderitz (27°S) where the upwelled water is coldest and the offshore extension of the filaments is largest [Nelson and Hutchings, 1983]. The strength of the upwelling cells shows a decreasing trend toward the north and south. In the north, the Benguela Current is bounded by the poleward flowing Angola Current. This generates the Angola-Benguela Front. The Angola-Benguela Front is a permanent feature, marked by a sharp horizontal temperature gradient, and is situated throughout the year between 14°S and 17°S [Meeuwis and Lutjeharms, 1990; Shannon, 1985; Shannon et al., 2001]. The southern boundary of the Benguela Current varies considerably, but is mainly located between 36°S and 38°S [Nelson and Hutchings, 1983; Shannon, 1985].
 Below the surface depth Benguela Current, equatorward flowing Antarctic Intermediate Water is present between 500 and 1200 m [Shannon and Nelson, 1996; Shannon et al., 2001]. At depths between 1000 and 3500 m, well oxygenated North Atlantic Deep Water flows poleward [Shannon et al., 2001]. The deepest water mass in the Cape Basin is composed of the clockwise circulating Antarctic Bottom Water at depths below 4000 m. The Antarctic Bottom Water is restricted in its northward extension by the Walvis Ridge.
2.2. Miocene Oceanography
 During the Middle Miocene the Benguela Current system was located more to the south than it is today, due to weaker trade winds as well as a weaker Benguela Current. At that time the Benguela Current flowed into the Cape Basin from off the southwest African coast, turned west within the Cape Basin, but did not reach Walvis Ridge [Diester-Haass et al., 1992]. During Antarctic glacial periods in the late Miocene, the subtropical anticyclone strengthened and the Benguela system shifted northward. The Benguela Current angled west and extended to Walvis Ridge [Diester-Haass et al., 1992; Van Zinderen Bakker, 1984]. Strengthening of the subtropical anticyclone and the Benguela Current coupled to the onset of upwelling was responsible for the beginning aridification of the Namibian desert [Van Zinderen Bakker, 1984; Van Zinderen Bakker and Mercer, 1986].
 The initiation of the Antarctic Circumpolar Current occurred during the early Miocene due to the opening of the Drake Passage to deep water flow at about 20.3–19.5 Ma [Pagani et al., 2000]. At this point the formation of cold and nutrient-rich Antarctic Intermediate Water occurred [Pagani et al., 2000], which was important for the development of an upper water temperature gradient (e.g., thermocline). This, in turn, is necessary for the formation of high-productivity areas, such as the Benguela upwelling. However, a strengthening of the Antarctic Circumpolar Current is not documented prior to 14 Ma [Shevenell et al., 2004].
3. Material and Methods
3.1. Site Description and Stratigraphic Framework
 ODP Hole 1085 A (29°22′47″S; 13°59′41″E) was drilled during Leg 175 in a water depth of 1713 m on the continental slope of southwest Africa in the southern Cape Basin along the western margin of the present-day Benguela upwelling region [Wefer et al., 1998]. It is located 250 km east of the Orange River mouth, and the sediments mainly consist of greenish-gray nannofossil-foraminifer ooze diluted by variable amounts of terrigenous silt and clay [Wefer et al., 1998]. Fifty-three samples were used for dinoflagellate studies from cores 48 X to 61 X at depths of 440–574 mbsf. Sample spacing ranged between 10 and 100 cm with a higher resolution (5–20 cm) at core depths of 488–536 mbsf (Figure 2).
 The stratigraphic framework is based on the data of Westerhold et al. , generated by orbital tuning of an XRF-Fe intensity record to an eccentricity-tilt-precession (ETP) target curve. The sedimentation rates vary between 1.0 and 7.3 cm ky−1 [Westerhold et al., 2005]. The 53 samples span the time interval of the middle-to-upper Miocene (12.4–9.2 Ma), and sample spacing corresponds to time intervals of 10–200 kyrs. During this time period, three Miocene cooling events were observed (Mi 5, Mi 6, Mi 7 [Miller et al., 1991; Westerhold et al., 2005]).
3.2. Sample Procedure
 From each sample, 1 cm−3 sediment was dried at 60°C. About 0.25 g of the dried material was dispersed in tap water and sieved through 63 μm and 20 μm sieves. The fractions coarser 63 μm and smaller 20 μm contained no dinoflagellate cysts and were not analyzed. The fractions between 20 and 63 μm were transferred to a test tube and centrifuged at 3200 rpm for 6 min. The overlying water was removed using a pipette, and the material was transferred to an Eppendorf cup where the volume was standardized to 1 ml. A fixed amount of material was placed upon a microscopic slide, mounted in glycerine jelly and sealed air tight using paraffin wax. Calcareous-walled dinoflagellate cysts were counted using a light microscope with polarized light [Janofske, 1996]. At least one slide per sample was counted. If a slide contained less than 150 specimens, additional slides were analyzed. We calculated relative cyst abundances as well as cyst accumulation rates (cysts cm−2 ky−1) for all samples.
 Based on the ecological characteristics of species in modern environments three different environmental indicators were distinguished.
Calciodinellum levantinum and Caracomia arctica are restricted to colder water conditions [Vink, 2004]. Calciodinellum levantinum prefers a temperature range of 13°C–20°C [Vink, 2004]. Caracomia arctica is currently limited to polar regions south of the subpolar front at temperatures ≤ 10°C [Streng et al., 2002; Vink, 2004].
3.2.3. Nutrient Indicators
Leonella granifera, Thoracosphaera heimii, Caracomia arctica and Calciodinellum levantinum are species with high abundances in areas characterized by enhanced nutrient availability in the surface waters, such as upwelling areas, river plumes, etc. Of these species, Leonella granifera is exclusively found in areas characterized by terrestrial input of fluvial origin [Richter et al., 2007; Vink, 2004; Wendler et al., 2002a].
Scrippsiella regalis, Pernambugia tuberosa, Pirumella parva and one unidentified species either just occurred in low numbers or their ecological significance is not known. They are therefore grouped as “others.”
3.3. Statistical Analysis
 The Modern Analogue Technique (C2 data analysis [Juggins, 2007]) was chosen in order to highlight and quantify the relationship between dinoflagellate cyst species distribution and environmental parameters. The Modern Analogue Technique allows for a prediction of environmental conditions in the past based on the dinocyst assemblage through the correlation of current environmental parameters with a recent dinocyst assemblage. We compared our Miocene cyst distribution with a data set of recent dinoflagellate cysts and environmental parameters, such as temperature and phosphate in the surface waters of the South Atlantic (Figure 3) [Vink, 2004].
 At the beginning of the dinoflagellate cyst record from 12.4 Ma to about 11.9 Ma, the sedimentation rates as well as the accumulation rates of all cyst species have low values. Nutrient indicators are present throughout this time in low absolute abundances. Leonella granifera, characteristic for terrestrial input in the marine realm, is present in low concentrations. The reconstructed phosphate concentrations of the surface waters show low values as well. The warm water indicator, C. albatrosianum, is the most dominant member of the dinoflagellate cyst association at that time with high relative abundances. Sea surface temperatures increase at the beginning of the record and remain at relatively high temperatures. With the exception of a few recordings of C. levantinum between 12.3 and 12.2 Ma, the cold water indicators as well as the polar species are absent (Figure 3).
 From about 11.9 Ma on, the sedimentation rate increases until a maximum is reached at about 11.4 Ma. The dinoflagellate cyst accumulation rates also demonstrate an increasing trend and reach a maximum at about 11.5 Ma.
 The cyst association changes contemporaneously with a distinct maximum of the nutrient indicators in both relative abundances and accumulation rates at 11.5 Ma. An increase at that time is also reflected in the phosphate concentration. The accumulation rates of the warm water indicator, C. albatrosianum, increase from about 11.8 Ma onward until a maximum at 11.4 Ma. However, its relative abundance remains consistent with two short intervals of lower values at about 11.7 and 11.5 Ma. The cold water indicator, C. levantinum, presents a distinct maximum at 11.5 Ma, when the reconstructed temperatures decrease to a minimum. C. arctica is still absent at that time.
 The sedimentation rates as well as the dinoflagellate cyst accumulation rates decrease between 11.4 and 11.1 Ma. The accumulation rates of the nutrient indicators overall remain low in this time interval but show two distinct short-lasting maxima at 11.3 and 11.2 Ma. The accumulation rates of the terrestrial input indicator L. granifera show a similar, but more pronounced pattern with two short spikes of increased absolute and relative abundances at 11.3 and 11.2 Ma. At 11.2 Ma, the phosphate concentrations also exhibit a short increase. The warm water indicator C. albatrosianum follows the decreasing trend of the total dinocyst accumulation rates until about 11.1 Ma. The sea surface temperature remains at generally higher values with a short decrease at about 11.1 Ma. The cold water indicator C. levantinum and the polar species C. arctica are absent.
 From about 11.1 Ma, we find an increase of the sedimentation rate, which remained higher until about 10.5 Ma. The accumulation rate of all dinoflagellate cysts shows the same pattern except a short minimum at 10.6 Ma. The nutrient indicators have low absolute and relative abundances except one maximum at about 10.5 Ma. L. granifera, the terrestrial input indicator, shows an increasing trend until 10.7 Ma and then a decrease until 10.4 Ma. The phosphate concentrations show maxima at 11.1 and 11.0 Ma as well as at 10.6 and 10.5 Ma. The warm water indicator increases from about 11.1 Ma until reaching a maximum at about 10.8 Ma. Afterward, it decreases to a minimum at about 10.4 Ma. The cold water indicator C. levantinum occurs in low abundance at 11.6 Ma but a distinct maximum at 10.5 Ma. A distinct change in the association occurs with the first appearance of the polar species C. arctica at 11.1 Ma. This species is present in short intervals at 11.0, 10.8, 10.6 and 10.5 Ma as well. The temperature curve shows minima coeval with the occurrence of C. arctica at 11.1, 11.0, 10.6 and 10.5 Ma.
 The sedimentation rate decreases after 10.5 Ma and remains at lower values with the exception of one increase at about 9.9 Ma. The total dinoflagellate cyst accumulation rates along with the nutrient indicators also follow this trend, but show a minimum at 9.6 Ma and a slight increase afterward until 9.2 Ma. L. granifera increases from about 10.4 Ma on until a maximum at 9.9 Ma, only to decrease subsequently to a minimum at 9.6 Ma. Afterward, there is a slight increasing trend of L. granifera visible. Phosphate values drop abruptly at about 10.4 Ma and increase thereafter until 10.3 Ma. The phosphate concentration then remains at lower values until 10.7 Ma when it increases again until the end of our record. The warm water indicator C. albatrosianum increases from 10.4 Ma, is present in low values until 9.8 Ma and then decreases again. The cold water indicator C. levantinum decreases after 10.5 Ma and vanishes after 9.9 Ma. C. arctica increases in absolute and relative abundances from 10.4 Ma on with two minima at 9.9 and 9.6 Ma. The reconstructed sea surface temperatures show a minimum at 10.3 Ma, then increase. After about 9.7 Ma, they decrease until the end of the dinoflagellate cyst record.
5.1. Possible Alteration of the Cyst Assemblages
 By establishing a paleoenvironmental reconstruction, it is assumed that the calcareous cyst association in sediments reflects upper ocean conditions. However, this is inaccurate if the initial signal has been altered because of the dislocation of cysts by ocean currents and/or species-selective dissolution. Sediment trap studies on modern cyst production and deposition revealed that sinking velocities of calcareous dinoflagellate cysts are fast enough that a large lateral displacement of cysts is unlikely [Richter, 2009]. So far, detailed surveys of modern calcareous cyst distributions have not reported any indications of pronounced relocation of calcareous cysts. Associations deposited in bottom sediments reflect in detail the upper water conditions and transitions between upper water masses such as frontal systems [Vink, 2004; Zonneveld et al., 2000]. We therefore assume that no selective dislocation has affected the dinoflagellate cyst association.
 Dissolution of calcareous dinoflagellate cysts can occur during settling through the water column or after deposition at the sediment-water interface, thereby affecting the primary cyst association. In modern sediments, calcareous cysts are found to be relatively robust against calcite dissolution and can even be observed in sediments from below the lysocline [Vink et al., 1999]. Freshly produced calcareous cysts are covered by a thin organic layer which protects the cysts against calcite dissolution [Karwath, 2000]. The only indication for depth related selective dissolution of species were given over a depth transect through the lysocline in the western equatorial Atlantic [Vink et al., 1999]. Site 1085 is at depth of 1713 m and lies well above the lysocline. We therefore assume that depth related dissolution did not alter the cyst association.
 In the upper water column up to 60% of the carbonate may dissolute by either biological processes within guts of grazers or increased degradation of organic matter within marine snow or aggregates [Milliman et al., 1999]. Jansen and Wolf-Gladrow  suggested that calcite dissolution by grazing does not constitute to the major part of carbonate dissolution in the water column. Carbonate dissolution due to organic matter oxidation, however, can have an effect on cyst distribution especially in areas of high surface water productivity. Wendler et al. [2002b] documented a relationship between productivity related organic matter decay and cyst dissolution in a sediment sample transect of NE Arabian Sea. They found the small porous spheres of Thoracosphaera heimii to be most prone to dissolution, followed by Calciodinellum levantinum, Calciodinellum albatrosianum and Leonella granifera in the order of dissolution sensitive to more resistant cysts [Wendler et al., 2002b].
 Observations on carbonate sedimentation off southwest Africa indicated that most of the decreases in carbonate concentrations during the studied time interval is not caused by dissolution [Diester-Haass et al., 2004; Kastanja et al., 2006; Krammer et al., 2006]. Rather, they are diluted as a result of major increases in clastic input from the Orange River during global sea level regression and/or changes in the productivity of the calcareous nannoplankton. In contrast to dissolution, dilution would not change the association composition but instead would influence the total cyst accumulation rates.
 Indications for a change to an enhanced carbonate dissolving environment at Site 1085 are given from about 10.2 Ma until about 9.7 Ma. At these point Diester-Haass et al.  found an increase in the benthic foraminifera accumulation rate (BFAR). The authors additionally concluded that sediment pore waters experienced suboxic conditions, based on increasing mass accumulation rates of total organic carbon as well as enhanced pyrite concentrations (Figure 4). The suboxic conditions and BFAR increases indicate a stable, high-productivity area during that time. The dinoflagellate cyst showed enhanced fragmentation during this time interval, which could be an additional indication for carbonate dissolution. We, furthermore, find a decrease in the phosphate values, implying a decrease in productivity, as well as an increase in temperature. Both is in contradiction to the findings of enhanced productivity by Diester-Haass et al.  and the decrease in temperature during that time interval shown by different authors [e.g., Westerhold et al., 2005; Rommerskirchen et al., 2011]. Dissolution of the species used as the nutrient and temperature indicators could have altered the palaeoenvironmental reconstruction. The species Thoracosphaera heimii, used as a nutrient indicator, is known to be most prone to dissolution [Wendler et al., 2002b; Vink et al., 1999]. Dissolution of this cyst could change the calculation for the reconstructed phosphate values and could have led to a decrease of the values. For the reconstructed temperatures we used Calciodinellum levantinum as the cold water indicator and Calciodinellum albatrosianum as the warm water indicator. Dissolution of the more sensitive cyst C. levantinum in comparison to C. albatrosianum [Wendler et al., 2002c] could have changed the primary signal and led to enhanced calculated temperatures. So far no observations have been made about the dissolution sensitivity of Caracomia arctica. However, the larger size of C. arctica in comparison to the other cysts, as well as the relatively thick wall of the spheres [Streng et al., 2009] implies a higher resistance against dissolution.
 We therefore suggest that species selective dissolution has altered the dinoflagellate cyst association in the time interval of about 10.2–9.7 Ma and had an influence on the reconstructed environmental parameters, such as temperature and phosphate, but did not influence the polar indicator.
5.2. Modern Ecology of Calcareous Dinoflagellates
 The modern ecology of calcareous-walled cyst producing dinoflagellates has been studied in surface sediments, sediment traps, and culture experiments [e.g., Vink et al., 2002; Vink, 2004; Meier and Willems, 2003; Richter et al., 2007; Richter, 2009; Karwath, 2000; Zonneveld et al., 2000]. Several environmental factors are known to be important for cyst production and germination, including temperature, salinity, trophic conditions and the stratification of the upper water column. It has been shown that individual cyst species have different paleoenvironmental significance and could therefore be used as species specific indicators, as described in section 3.2.
 In our sediment samples from ODP Site 1085, the dinoflagellate cyst assemblage consists mainly of the species C. albatrosianum, T. heimii, L. granifera and C. levantinum, which are known to increase in abundance during times of enhanced productivity. The dinoflagellate cyst accumulation rates, therefore, can be used with the dinoflagellate cyst concentrations as a signal of productivity.
5.3. Benguela Upwelling Intensity Related to Productivity Changes
 Based on the calcareous dinoflagellates cyst assemblages, we find the first indication for enhanced upper ocean productivity between 11.8 and 11.4 Ma, illustrated by an increase in the dinoflagellate accumulation rates as well as a remarkable change in association. In this interval, nutrient and cold water indicators show a pronounced increase in accumulation rates and proceed to dominate the cyst association. This indicates that upper waters became colder and more nutrient rich. Previous to this interval (12.4–11.8 Ma), sedimentation rates, dinocyst accumulation rates, and the nutrient indicators are low and the warm water indicator dominates the cyst association. This implies that relatively warm, oligotrophic conditions prevailed in the surface waters and suggests lower productivity. This is consistent with the study of Krammer et al.  who came to similar conclusions based on the observation of minimal values in the accumulation rate of coccoliths during this time interval.
 The increase in productivity at 11.8 is in agreement with the sedimentological observations from Diester-Haass et al. , which documented an increase in carbonate mass accumulation rates (MAR CaCO3) and in the mass accumulation rates of organic carbon (MAR TOC) at this time. Enhanced organic carbon concentrations coupled with enhanced pyrite concentrations at that time suggests the development of a pronounced oxygen minimum zone (Figure 4) [Paulsen, 2005]. Further support comes from Twichell et al. , who showed an increase in Corg/Ntotal values, which is interpreted to be the result of an increase in the total organic carbon export production of marine origin. At 11.5 Ma, we find a maximum in the accumulation rates of all dinoflagellate cysts as well as in the absolute and relative abundances of the nutrient indicator species. The reconstructed phosphate values also show an increase. Enhanced productivity and faunal changes are also reflected in the foraminiferal record of Paulsen . Coincident to this is a decrease in sea surface temperatures, as reconstructed from the dinoflagellate transfer function (Figure 3), and a pulse of the cold water indicator C. levantinum. The time interval of 11.8–11.4 Ma ago has been delineated as the Antarctic glaciation period Mi5 [Miller et al., 1991; Westerhold et al., 2005]. The enhanced cooling during this time could have led to an increase in wind strength [Roters and Henrich, 2010] and, subsequently, to an upwelling pulse. After 11.4 Ma, productivity tends to decrease, as reflected by a decrease in dinoflagellate cyst accumulation rates. The abrupt drop in the cyst accumulation rates is contemporaneous to short spikes of maximum values of the nutrient indicators and that of the terrestrial input indicator L. granifera. This suggests that surface waters received higher amounts of nutrients accompanied by an intensified input of terrestrial material at these times. Our suggestion of enhanced input of terrigenous material at 11.3 and 11.2 Ma is supported by the study of Kastanja et al. , who found high values of terrigenous input at about 11.3 Ma. There is strong evidence that the Namib Desert was not fully developed and that most terrestrial material was transported into the research area via the Orange River [Diekmann et al., 2003; Partridge, 1993; Roters and Henrich, 2010; Siesser, 1980]. The increase in nutrient availability in the upper water might therefore be the result of enhanced fluvial input during that time.
 From about 11.1 until 10.4 Ma, we observe another increase in sedimentation rates and accumulation rates of the cysts, which indicates an additional increase in productivity. An increase in marine primary productivity was also documented by Rommerskirchen et al.  through enhanced TOC values between 11 and 10 Ma. Westerhold et al.  observed increasing oxygen isotope values between 10.7 and 10.4 Ma, with a marked increase at 10.4 Ma at Site 1085. They suggest this signified a decrease in temperature at that time and was related to the cooling event Mi6. Similarly to the observed Mi5 succession, at the end of the Mi6 there is another maximum of the nutrient indicators as well as another pulse of the cold water indicator C. levantinum, which suggests another maximum in productivity at that time. Phosphate values show two maxima as well, at about 10.6 and 10.5 Ma. Further support of enhanced productivity in the area about 10.5 Ma ago comes from the coccolith record of Krammer et al. , who report the occurrence of the upwelling indicators R. minuta and R. haquii in high abundances. We, therefore, suggest another pulse of upwelling occurred at the end of Mi6 at about 10.5 Ma.
 After 10.5 the dinoflagellate cyst accumulation rates decrease, whereas Diester-Haass et al.  found a change to a permanent high productivity area during that time and suggested the establishment of the Benguela upwelling (Figure 4). From about 10.2 Ma until about 9.6 Ma the cyst accumulation rate as well as the reconstructed phosphate values and temperatures were altered by species selective dissolution and therefore do not show the primary signal as described in section 5.1.
5.4. Evidence of a Shift in the Source Waters of the Benguela Upwelling
 The first indication for changing conditions in the upwelling regime occurs at 11.1 Ma when the polar species C. arctica appears in the ODP 1085 record. C. arctica is found sporadically and in low abundance until 10.5 Ma. After this time, the species becomes a permanent member of the association. The presence of C. arctica implies an influence of subpolar/polar water masses at the studied site.
 The appearance of Antarctic calcareous dinoflagellate cyst species coincides with the first occurrence of the polar foraminifer species Neogloboquadrina pachyderma (s). This species is a permanent member of the plankton and is found in increasing abundance from 11.1 Ma onward [Paulsen, 2005]. Other evidence comes from the coccolith record of Krammer et al.  who showed a distinct increase in abundance of the cold water coccolith, Coccolithus pelagicus.
 One possible cause for the presence of polar water species in the Benguela region could be enhanced advection of Antarctic intermediate waters, which are the nutrient source of the upwelled waters. It was shown that, currently, subantarctic mode water is the main source for the nutrient supply in the thermocline of the southeast Atlantic [Sarmiento et al., 2004]. In this case, C. arctica could reflect a change in the source of the upwelled water, promoting conditions more similar to today. However, C. arctica is not present in upwelling regions today despite the fact that numerous detailed surveys exist for these regions [Esper et al., 2000; Vink, 2004; Zonneveld et al., 2000]. Furthermore, there is no evidence that dinoflagellates are laterally transported over long distances via intermediate or deep water currents. An explanation for the absence of C. arctica in the Benguela region today could be a change in its ecology or in its habitat. It is possible that C. arctica nowadays is not formed in the photic zone of the area where subantarctic mode water originates as a result of changed conditions in the source region or changed ecological adaptations of the species.
 Another possible explanation for the presence of C. arctica in the late Miocene sediments could be a shift of the Polar Frontal Zone northward, which allowed C. arctica to penetrate into the Benguela area. A northward movement of the Polar Frontal Zone at different times in the past has already been suggested by several authors [McIntyre et al., 1989; Pagani et al., 2000; Paulsen, 2005]. Pagani et al.  postulated a migration of the Polar Frontal Zone during the middle and upper Miocene (from about 14 Ma). This resulted from the development of the Antarctic Circumpolar Current and is thought to have caused the appearance of a relatively strong proto-Antarctic Intermediate Water current, which pushed the Polar Frontal Zone north. Paulsen  observed a northward movement of the Polar Frontal Zone at about 10.3 Ma on the northern slope of Meteor Rise in the subantarctic sector of the Southern Ocean. Diekmann et al.  described increases in sedimentation rates at Agulhas Ridge and Meteor Rise from about 11.2–9.5 Ma, which indicated lateral sediment transport via pulses of the Antarctic Circumpolar Current. However, a northward migration of the Polar Frontal Zone up to 29°S has not yet been documented.
 A second possible explanation for the transport of C. arctica via polar surface waters into the studied region could be the interaction between the warm Agulhas Current and the Antarctic Circumpolar Current. Today, the Agulhas Current mixes with the South Atlantic Current and the Antarctic Circumpolar Current at a retroflection area south of Africa. With this mixing, large rings develop that migrate as eddies into the South Atlantic Ocean [Gordon, 1986] by the South Atlantic Current. The development and movement of these rings has been studied in detail by several authors [Ansorge and Lutjeharms, 2003; Boebel et al., 2003a, 2003b; Gordon, 1985; Hardman-Mountford et al., 2003; Lutjeharms, 2001; Lutjeharms et al., 2003; Treguier et al., 2003]. However, today these rings are drifting slightly south of the studied site. Thus, a broader Antarctic Circumpolar Current would be necessary to transport the Agulhas eddies farther north.
 Despite inconclusive evidence from the dinocyst record, the most reasonable scenario for the occurrence of Antarctic polar species at 11.1 Ma seems to be transport via the subantarctic mode water. All records discussed so far reflect the influence of polar waters from 11.1 Ma on [Paulsen, 2005]. This indicates that a change in nutrient conditions related to an enhanced advection of subantarctic mode water is a likely explanation for the time lag between the Middle Miocene Climate Transition at 14 Ma and the establishment of the Benguela upwelling conditions after 11.1 Ma. The mechanism behind the change and how subantarctic mode water reached the upwelling site at that time remains a point that requires further study.
 Based on changes in the calcareous dinoflagellates cyst association in middle to late Miocene age sediments of the continental margin off Namibia (ODP 1085), we come to the following conclusions:
 We document a period of higher productivity off southwest Africa from 11.8 to 11.4 Ma. At the end of the cooling event Mi5 (at about 11.5 Ma), we find a strong increase in the cyst productivity and a decrease in temperature. We interpret this as reflecting a pulse in upwelling, probably related to enhanced glaciations on Antarctica. The same pattern is found at the end of the cooling event Mi6 at about 10.5 Ma.
 Between the two cooling events (Mi5 and Mi6), the cyst productivity decreases and the terrestrial indicator, L. granifera increases. This was most likely caused by enhanced terrigenous input from the Orange River.
 At 11.1 Ma, the first occurrence off Namibia of the polar species C. arctica indicates the influence of subantarctic water masses in the upwelling region.
 After 10.5 Ma, C. arctica occurs continuously, implying that subantarctic mode water is the stable source for the upwelled water, as it is today. The development of the Benguela upwelling much later than the Middle Miocene Climatic Transition might be explained by a late intensification of the Circumpolar Current. This was associated with a northward movement of the subantartic ocean frontal system and the enhanced mixture of nutrient-rich water into the Agulhas rings.
 The simultaneous decrease in dinoflagellate cyst productivity at this same point as well as the increase of the phosphate values and the decrease of the reconstructed temperatures are caused by species selective dissolution.
 Thanks are given to Petra Witte for her help with the SEM analyses. We thank Lieselotte Diester-Haass for providing data on the carbon and coarse fraction of the sediments from ODP Site 1085, as well as foraminiferal accumulation rates. Thanks to Thomas Westerhold for making available the data for the age model. We thank Kara Bogus for improving the language. We gratefully acknowledge thoughtful comments from Christopher Charles, one anonymous editor, and two anonymous reviewers. This work was funded through the Deutsche Forschungs Gemeinschaft International Graduate College “Proxies in Earth History” program.