Thirty thousand years of vegetation development and climate change in Angola (Ocean Drilling Program Site 1078)

Abstract. ODP Site 1078 situated under the coast of Angola provides the first record of the vegetation history for Angola. The upper 11 m of the core covers the past 30 thousand years, which has been analysed palynologically in decadal to centennial resolution. Alkenone sea surface temperature estimates were analysed in centennial resolution. We studied sea surface temperatures and vegetation development during full glacial, deglacial, and interglacial conditions. During the glacial the vegetation in Angola was very open consisting of grass and heath lands, deserts and semi-deserts, which suggests a cool and dry climate. A change to warmer and more humid conditions is indicated by forest expansion starting in step with the earliest temperature rise in Antarctica, 22 thousand years ago. We infer that around the period of Heinrich Event 1, a northward excursion of the Angola Benguela Front and the Congo Air Boundary resulted in cool sea surface temperatures but rain forest remained present in the northern lowlands of Angola. Rain forest and dry forest area increase 15 thousand years ago. During the Holocene, dry forests and Miombo woodlands expanded. Also in Angola globally recognised climate changes at 8 thousand and 4 thousand years ago had an impact on the vegetation. During the past 2 thousand years, savannah vegetation became dominant.

change from glacial to interglacial? Does it have Northern or Southern Hemisphere timing? How much is the precipitation in Angola influenced by East Atlantic sea surface temperatures. What is the impact of changes in the Atlantic meridional overturning circulation?
The latitudinal position of the CAB is connected with the amount of rainfall and the 10 duration of the rainy season. The land-ocean temperature contrast is associated with aridity, wind strength and wind direction. Both directly affect the vegetation. Earlier studies have shown that vegetation changes in Angola are sensitively recorded by pollen and spores in marine sediments along the coast (Shi and Dupont, 1997;Dupont and Behling, 2006). The marine material allows us to compare the pollen record with 15 the sea surface temperature (SST) record from the same core and thus enabling us a direct land-sea correlation. We studied a high resolution marine core (ODP Hole 1078C) spanning the past 30 ka encompassing the last glacial maximum, deglaciation including Heinrich event 1 and the Younger Dryas period, and the Holocene. We aim to interpret changes in the pollen 20 assemblage in terms of vegetation change on the continent taking into account effects of transport of pollen by wind and rivers. The thus reconstructed vegetation changes are compared to other African environmental records. We aim at a conceptual model of changes in the latitudinal position of atmospheric and marine fronts explaining the environmental change in Angola that might be tested with numerical models. Introduction Interactive Discussion 2 Regional climate and ocean circulation and its impact on the modern vegetation of west southern Africa Present-day climate in Angola ranges from the permanently wet climate of the Congo basin in the north to the dry summer rain climates in the south. The Atlantic or west African monsoon controls the climate of the northernmost part during most of the year, 5 while the middle and southern parts receive (Atlantic) moisture only during austral summer. Angola is the southernmost region where the West African monsoon brings rain (Fig. 1). The boundary is in the form of a low pressure area, the CAB (formerly Zaire Air Boundary), that annually migrates over Angola from its southern surface position around 20-15 • S in January (austral summer) to its northern position around 9-6 • S in 10 July (austral winter) (Leroux, 1983). The CAB is part of the discontinuities that divide easterly trades and westerly monsoonal wind systems over Africa and can be regarded as a southern branch of the ITCZ. The latitudinal position and annual migration of these systems determines the timing and duration of the rainfall seasons. They are associated with the global thermal gradient and the strength and position of the Hadley Cell 15 (Nicholson, 2000). However, sea surface temperatures and ocean circulation also have an impact on the position of the ITCZ. The surface and shallow subsurface ocean circulation in the eastern Angola Basin is dominated by the South Atlantic cyclonic gyre consisting of the Benguela Ocean Current flowing northwards and westwards over the Walvis Ridge, the South Equatorial 20 Counter Current flowing eastwards, and the Angola Current flowing southwards along the coast (Fig. 1). In the marine realm, as in the terrestrial one, an important frontal system exists off the coast of Angola in the form of the Angola Benguela Front (ABF) annually migrating between 17 and 14 • S. South of the ABF, the surface Benguela Coastal Current being a tongue of the Benguela Ocean Current flows northwards along The ABF is recognizable by a sea surface temperature (SST) gradient of 4 • C per 1 • of latitude and identifiable to a depth of at least 200 m. The front is most intense within 250 km from the coast. The position and the strength of the front are maintained through a combination of factors including coastal orientation, wind stress, and opposing flows of the equatorward Benguela Coastal Current and the poleward Angola 5 Current (Shannon and Nelson, 1996). Within the limits of the coastal orientation, there is a coupling between the position of the ABF and the CAB on land. South of the front the zonal component of the wind stress over the ocean is east (SE trades) and north of the front the zonal component is west (SW monsoon). Strong SE trade winds will trigger both a cooling south of the ABF by increasing the upwelling along the coast and 10 a warming north of the ABF due to stronger intrusion of warm equatorial waters by the Angola Current (Kim et al., 2003). Intense coastal upwelling occurs only south of the ABF, whereas directly north of it upwelling is seasonal and relatively weak (Wefer et al., 1998).
The position of the climatic and oceanic frontal systems is reflected in the vegetation 15 of southern Africa. The southern boundary of the Congolian rain forest corresponds to the latitudinal July (austral winter) position of the CAB. The Zambezian phytogeographical region is most extended consisting of different types of dry forest, woodland, savannah and edaphic grassland (White, 1983). In the north of Angola the transition between the Zambezian and the Congolian vegetation zones nowadays is mainly oc-20 cupied by secondary grassland and savannahs. However, surviving cloud forests of Congolian affinity exist occurring at altitudes between 350 and 1000 m. Along the west coast, climate and vegetation are strongly influenced by the SST. Especially, the low SSTs of the upwelling area south of the ABF suppress coastal rain fall (Nicholson and Entekabi, 1987 Interactive Discussion development is sensitive to climate change and shifts of the frontal systems described above. In this paper, we present a detailed pollen record of marine sediments of ODP Hole 1078C, which allows the study of the vegetation history of the past 30 ka during full glacial, deglacial, and interglacial conditions.

5
The material used is sediment from the upper 11 m ODP Site 1078 Hole C. The site is located outside the Bight of Angola at 11 • 55 ′ S 13 • 24 ′ E in 427 m deep water. The sediments are composed predominantly of a moderately bioturbated olive-gray silty clay with varying amounts of nannofossils and foraminifers (Wefer et al., 1998).
The age model is established after radiocarbon AMS measurements on foraminifer 10 tests and molluscs carried out at the Leibniz-Laboratory, University of Kiel, Germany (Kim et al., 2003;Rühlemann et al., 2004). Extension of the age model uses eight dates (Table 1) (Ybert, 1979;Bonnefille and Riollet, 1980;Sowunmi, 1973Sowunmi, , 1995, the African Pollen Database (http://medias.obs-mip.fr/pollen/), and a reference collection. Charcoal particles have been counted from the pollen slides. Samples are counted routinely up to 300 pollen grains. However, some samples turned out to be too poor 25 to reach that goal. Percentages are expressed on the total number of pollen and fern spores, which is denoted in Fig. 4.

Age model and sedimentation rates
The age depth model is shown in Fig. 2. The younger part (0-22 ka) is after Kim et al. (2003) and Rühlemann et al. (2004) and has been calibrated using Calib and the marine calibration curve (Stuiver and Reimer, 1993;Hughen et al., 2004). The lower dates (Table 2) are outside the range of the marine calibration curve and have been 15 calibrated using Fairbanks et al. (2005). Calibration ranges increase with age and the uncertainty of the calibration. Some radiocarbon ages seem to occur at different levels, which can be explained by burrowing down of shells and shell fragments. We therefore consequently chose the upper ones to include in the age model.
For the calibration of the radiocarbon dates, we have applied a reservoir effect of 400 20 years as suggested by the present day marine reservoir correction database (Butzin et al., 2005). Although there are no direct measurements in the vicinity of ODP Site 1078, we think this is justifiable because the site is just outside the coastal upwelling range and the influence of the open ocean is more pronounced than at sites north or south of it (Wefer et al., 1998 years in the glacial dates would not significantly change our age model. The sedimentation rates for ODP Site 1078 that result from linear interpolation between the calibrated radiocarbon ages are also given in Fig. 2. Until 27 ka sedimentation rates range between 22 and 35 cm/ka, but are rather imprecise. Between 10-9.8 mbsf (27-22 ka) sedimentation rates drop to low values of about 4 cm/ka. This 5 estimate is supported by three ages. Moreover, a concentration of shells and molluscs have been found at those levels both in Hole C, used in this study, and in the neighbouring Hole B (Wefer et al., 1998). The very low sedimentation might have occurred during the last glacial lowest sea-level period. This seems to be much too early, but a recent publication provided evidence that the period of lowest sea-level must have 10 occurred between 26 and 21 ka ago (Peltier and Fairbanks, 2006).
After 22 ka sedimentation rates increase again to levels between 20 and 40 cm/ka and jump to levels around 1 m/ka shortly before 14 ka. We associate the strong increase in sedimentation rates with the rapid sea-level rise during Meltwater Pulse 1a dated between 14.6-14.3 ka (Fairbanks, 1989;Hanebuth et al., 2000). The levels of 15 high sedimentation rates between 8 and 3 mbsf (14-9 ka) shows raised Rhizophora pollen percentages. This increased relative abundance of mangrove pollen comes together with increased amounts of other mangrove materials and is interpreted as sedimentation of eroded mangrove peat (Kim et al., 2005). Sedimentation levels remain high until 9 ka, when they drop again to levels between 25 and 40 cm/ka. 20 Thus the history of sea-level change at ODP Site 1078 as reconstructed with our age model fits well into the global picture, which makes us more confident about the age model itself.

Results
Alkenone-derived SSTs showed a typical glacial-interglacial warming trend, ranging 25 from 21 • to 24 • C over the last 30 ka (Fig. 3). During the glacial period, alkenonederived SSTs varied between 21 • and 23 • C and decreased to the lowest value of 118 20.9 • C occurred at 16.3 ka during Heinrich Event 1. Afterwards, alkenone-derived SSTs increased towards the Holocene, showing a warm period during the Younger Dryas. During the Holocene, alkenone-derived SSTs continued to increase to a coretop value of 24.9 • C with a slight cooling around 5 ka. Based on the relative abundance of the most conclusive pollen taxa, seven pollen 5 zones (PZ) have been distinguished. In this section, each zone is briefly characterised. Average percentage values per pollen zone are given for selected pollen taxa in Table 2. Percentages of groups and selected pollen taxa are given in Figs. 4-6. Appendix A lists the pollen taxa within each group.
PZ 1: 25 samples between 11.16 (lowermost sample analysed) and 9.70 m, dated 10 between 31.7-21.9 ka. The record is poor between 9.99 and 9.81 m (26.3-22.4 ka), but not systematically different from the rest of the zone (see Table 2). Radiocarbon dates indicate very low sedimentation rates (∼4 cm/ka) for this part of the zone. Maximum percentages of the mountain elements including pollen of Ericaceae, Myrica and Ilex cf. mitis, as well as spores of the fern Anemia and liverworts (Phaeoceros, Anthoceros) 15 are found. High pollen percentages of Asteroideae, Stoebe-type, Poaceae, and other desert elements, e.g. Tribulus occur as well as raised percentages of Brachystegia. PZ 2: 19 samples between 9.70-8.80 m, dated between 21.9-18.8 ka. Gradual increase in the pollen concentration (  Aeolian transport of pollen to ODP Site 1078 is mainly expected during austral fall and winter when the lower tropospheric flow is from the SE (Dupont and Wyputta, 2003). Source areas for fluvial and aeolian pollen largely overlap.

15
The lowermost part of the record, PZ1, covers the glacial between 31.7 and 21.9 ka. As discussed above, it includes the last glacial maximum (defined as the period of lowest global sea-level). From the high pollen percentages of mountainous elements, including Ericaceae, Myrica and Ilex accompanied with high relative abundances of pollen of Poaceae and Asteraceae, we infer that open mountainous vegetation with 20 heather species (Ericaceae), composites (Asteraceae), and grasses (Poaceae) was wide-spread ( Fig. 5). Ericaceae and Ilex mitis are typical elements from the Afromontane vegetation (White, 1983). Palgrave (2002) mentions several Myrica species from mountain areas. Apart from the Afromontane vegetation, the existence of dry woodlands, probably at lower altitudes is suggested by the occurrence of pollen of Brachys- 25 tegia and spores of Anemia. The Flora Zambesiaca (Schelpe, 1970)  between 600-1525 m. Pollen from composites and grasses probably had their sources not only from Afromontane vegetation but from desert and semi-desert vegetation along the coast, as well. Extension of desert and semi-desert is also indicated by the regular occurrence of Tribulus and Stoebe-type pollen. Is this extension a result of a northward 5 shift of the Angola Benguela Front (ABF) and increased upwelling along the coast? Alkenone measurements indicate SSTs (Fig. 3) were around 22 • C (Kim et al., 2003) 2 to 3 • lower than today. Increased upwelling, however, should have lowered the SST below the general cooling of the glacial ocean, which is estimated at 4-5 • C for the east equatorial Atlantic (Niebler et al., 2003). Comparison of several SST profiles with a coarser timely resolution between 5 • and 20 • S did not indicate a shift of the ABF north of 12 • S during the glacial (Schneider et al., 1995). Thus, there is little indication of upwelling of cooler intermediate waters.
The open mountainous vegetation and the expansion of desert and semi-desert suggest a cooler and drier climate in Angola during the glacial. Our data corroborate 15 terrestrial mountain records, especially in east Africa, of an open glacial landscape covered with a grass-and ericaceous-rich vegetation (e.g. Coetzee, 1967;Bonnefille and Riollet, 1988;Bonnefille and Chalié, 2000). However, in Congo, at lower altitudes and nearer to the equator, no mountain elements were found (Elenga et al., 1994). Forest growth in the mountains might have been impeded by low temperatures at higher 20 altitudes, dry conditions, and/or due to low glacial atmospheric carbon dioxide levels (Jolly and Haxeltine, 1997).
Between 21.9 and 18.8 ka, PZ2, the vegetation gradually changes. The relative abundance of pollen from open vegetation types (Afromontane and desert) declines, and tree pollen percentages of Podocarpus (Fig. 4) and Mallotus-type (Fig. 6) rise. 25 Probably, Podocarpus trees were spreading in the mountains, while rainforest started to expand in the northern lowlands with pioneer trees such as Mallotus (Maley and Brenac, 1998). Other Afromontane trees such as Ilex mitis and Myrica are still present (Fig. 5). The gradual increase in pollen concentration suggests higher terrestrial -

Printer-friendly Version
Interactive Discussion probably fluvial -input. Around 22 ka, the lake level of Lake Malawi rises after the glacial lowstand indicating increased precipitation in the East African mountains (Johnson et al., 2004). Furthermore, semi-deciduous forest elements became established around Lake Masoko just north of Lake Malawi suggesting the dry season became as short as 3 to 4 months (Vincens et al., 2007). Glacial woodland and bushland with 5 Afromontane elements around Lake Rukwa in Tanzania (almost 2 • of latitudes further north), however, indicates that the climate was still rather cool and dry (Vincens et al., 2005).
Expansion of rain forest in the northern lowlands of Angola also suggests increased precipitation, which is possible if the winter migration of the CAB migrated less north-10 wards than in the period before. Then the CAB position would remain sufficiently to the south to allow Atlantic rain penetrating during most of the year in the northern lowlands of Angola. The spread of forest between 22 and 19 ka is too early to be an effect of rising carbon dioxide levels, because the CO 2 rise after the glacial is dated by Monnin et al. (2001) between 17 and 15.4 ka. Moreover, the pollen diagram from Lake Barombi in 15 Cameroon indicates an opening of the forest and increase of grasslands from ca. 24 ka on (20 14 C ka;Maley and Brenac, 1998). The development north and south of the equator proceeding in opposite directions would fit with a southern average position of the ITCZ coupled to a southern position of the CAB (Fig. 7). Our interpretation fits the study of Johnson et al. (2004) inferring a southern average position of the ITCZ 20 between 23 and 19 ka based on the increased productivity in Lake Malawi coupled to northerly winds over the lake.

Deglaciation
The pollen diagram for the period between 18.8 to 15.4 ka, PZ3, covering Heinrich Event 1, is dominated by Podocarpus (Fig. 4). Podocarpus trees produce much aero-25 dynamically well-equipped pollen and are, therefore, overrepresented in the pollen record, especially when wind transport is important. Still, there are few other elements represented save for Mallotus and Tetrorchidium from the rain forest (Fig. 6) Thus, although rain forest remained present in the northern lowlands, Podocarpus forest might have been locally abundant in the mountains. Unfortunately, the marine record of Podocarpus for this period is not consistent, because a comparable Podocarpus pollen maximum is missing in the nearby core of GeoB1016, but present in core GIK16867 at 3 • S (Dupont et al., 1996). The terrestrial pollen records in Cameroon 5 and Congo indicate a mixture of semi-deciduous forests and grasslands but neither an open savannah nor a Podocarpus forest (Elenga et al., 1994;Maley and Brenac, 1998;Elenga et al., 2004). Precipitation estimates based on deuterium of terrestrial plant waxes from a core off the Congo River mouth indicate minimal humidity in the Congo Basin (Schefuß et al., 2005). 10 During this period SSTs reach a minimum of 21 • C and the temperature difference with data from a site located 5 • of latitude further south (GeoB1023) is minimal (Kim et al., 2003). Kim et al. (2003) concluded that because of the small SST difference between 12 and 17 • S, the SE trade winds must have been weak. During the same period, biomarker based annual temperature estimates for the Congo Basin indicate  Fig. 4). Another implication is that upwelling over the site intensified. This upwelling would have increased the reservoir effect in the radiocarbon dates and the material consequently might be several centuries younger than dated by our age model. The pollen zone, however, would still include the period of Heinrich Event accumulation of all pollen types, but Rhizophora pollen from the mangroves is overrepresented (Fig. 4). A raised representation of mangroves both in relative pollen abundance and by biomarkers during this time and is explained by erosion of coastal mangrove peat during transgression and sea-level rise (Kim et al., 2005).
After Heinrich Event 1, between 15.4 and 10.0 ka (PZ4), the representation of the lowland rain forest becomes stronger. Not only pioneer trees such as Mallotus, Celtis, and Alchornea occurred, but also Tetrorchidium and Lophira, indicating increase and maturing of the forest. During this period -just before the Holocene -the lowland rain forest probably reached its southernmost extension (Fig. 6). Apart from the expansion of the lowland rain forest, elements from dry forest and woodlands, such as 10 Burkea-type and Combretaceae, and from the Miombo woodlands, such as Uapaca and Brachystegia, are increasingly represented. The Podocarpus record goes slightly down and that of Olea rises indicating a change in the mountain forest composition. The Younger Dryas period is not distinguished as such in our pollen record, but part of a longer period of favourable conditions for forest growth. 15 Most African records show a first return to moister conditions after the glacial around 15 ka, or slightly later (e.g. Maley and Brenac, 1998;Gasse, 2000;Gasse and Van Campo, 2001;Barker et al., 2004). Deuterium based precipitation estimates are high between 15 and 12 and low between 12 and 11 ka (Schefuß et al., 2005). In the Congo Basin increased precipitation resulted subsequently (at 13.5 ka) in a large palaeodis- suggesting the climate changed from a mild short (3-4 months) dry season to a distinctive dry period of no less than 5 to 6 months (Vincens et al., 2007). Other records from southern Africa either indicate wetter conditions (Shi et al., 2000) or no specific changes that might be associated with the Younger Dryas. The strong development of the rain forest and the initial development of the woodlands in Angola suggest an early southward shifting of the CAB during winter and summer (Fig. 7). The summer latitude might have been comparable to those of today and the winter latitude might have been somewhat further to the south causing a southward shift of the area with more or less humid conditions all year round. SSTs 5 rose strongly indicating that the ABF, and with it the area of coastal upwelling, probably shifted poleward, south of the site. The large difference in SSTs between 12 and 17 • S suggest strong trade winds, in particular during the Younger Dryas period (Kim et al., 2003).

10
During the early Holocene, between 10.0 and 7.80 ka (PZ5), dry forest and Miombo woodlands became increasingly important while rain forest elements retreated, but for the pioneer tree Alchornea (Fig. 6). Miombo woodland is nowadays the prevalent vegetation on well drained soils of the main plateau in Angola and other parts of the Zambezian region (White, 1983). The woodlands are dominated by species of Brachys- 15 tegia, often accompanied by Julbernardia or Isoberlinia (included in the Berlinia-type). Associates are Afzelia, Uapaca, and Pericopsis. Decline in the representation of most rain forest elements suggest a northward shift of the winter CAB more or less to its modern position. The occurrence of Artemisia afra throughout the period and that of Protea around 8 ka might indicate disturbance of mountainous vegetation. Mangrove 20 peat erosion increased during Meltwater Pulse 1b around 10 ka (Flemming et al., 1998) and the high total pollen input afterwards is associated with more river discharge.
Forest composition changed at 7.8 ka. Until 3.7 ka (PZ6) Podocarpus dominates in the mountains and other mountainous elements, such as Olea, are not represented (Figs. 4, 6). The representation of grasses is minimal indicating that open vegeta-25 tion had become rare (Fig. 5). Dry forests included Adansonia digitata and Lannea (Fig. 6). The high representation of Brachystegia, Afzelia, Berlinia-type, and Uapaca suggests a large expansion of the Miombo woodland. However, pollen percentages of 126 Rhamnaceae, probably from Ziziphus, and spore percentages of Pteris are also raised, indicating increased swamp forest vegetation. Among the genus Brachystegia there is one species B. laurentii occurring in wetter types of the rain forest. Also among Afzelia, the rain forest species A. africana might be present, but this species occurs nowadays in the Congolian rain forest only north of the equator (White, 1983). The genus Uapaca 5 includes rain forest species, as well. Thus, an alternative explanation to the spreading of Miombo woodland would be the expansion of wetter types of rain forest. In both interpretations, savannahs were replaced by closed forest. At 3.7 ka, the vegetation drastically changed again. The occurrence of Rhizophora pollen grains (Fig. 4) during this period indicates pollen transport from the north as 10 the modern southernmost distribution of mangroves is along the River Longa (10 • S, Spalding et al., 1997). Supply of Rhizophora pollen through erosion of older mangrove peat strata is less likely, because sea-level did not rise much during this period.
Typical for the past 3.7 ka (PZ7) is the dramatic increase in the representation of Cyperaceae and grasses (Fig. 4). After 2 ka, the relative abundance of Poaceae pollen 15 is even higher than during glacial times. The increase in Cyperaceae pollen percentages might be a localised phenomenon, the increase in grass pollen percentages is more regional as it is found as well in GeoB1016, located further offshore on the lower continental slope at 3411 m water depth (Shi and Dupont, 1997;Dupont et al., 2007). We find the high representation of grasses associated with more forest elements but 20 not with increased relative abundance of desert or mountain elements. We, therefore, infer a strong increase in savannah. The modern savannah forest mosaic of northern Angola might have come into existence around 2000 years ago. The patchy character of the forests is underlined by the increased representation of light-loving trees such as Alchornea and the oil palm, Elaeis guineensis (Fig. 6). 25 Other changes are found in the composition of the dry forest, in which Burkea, Hymenocardia, Protea, and Artemisia afra must have been prominent (Fig. 6). The latter three taxa are associated with fire. Artemisia afra often colonizes burnt areas in mountains (Beentje, 1994), Protea grows in mountainous grasslands that are burnt, and Hymenocardia acida is fire hardy (White, 1983). However, there is no increase in the amount of charcoal particles relative to the amount of pollen and spores. The charcoal record is especially high during the glacial (Fig. 6), but indicates that fire always played a role in the vegetation development. We can only explain the occurrence of more fire hardy elements by assuming a change but not necessarily an increase in the fire 5 regime that favoured these specific trees. The development in Angola fits the general trend of a humid early and middle Holocene followed by a drier late Holocene. The savannah minimum and probably most humid phase in Angola from 7.8 to 3.6 falls during the maximum northward extension of the Sahelian vegetation from 7.8 to 5.7 ka (Hoelzmann et al., 2004) but lasts 10 longer. The humid phase in Angola ended around the same time as the second of two aridification steps (around 6.7-5.5 ka and 4.0-3.6 ka) indicated by lake levels of the Sahara-Sahel region (Gasse, 2000). Only the first aridification step is recorded in the dust flux record of ODP Hole 658C as the end of the African Humid period (DeMenocal et al., 2000;Adkins et al., 2006). Also west equatorial pollen records and the deu- 15 terium record for the Congo Basin indicate a sharp drop in humidity around 3 ka, which favoured the opening of the forest and the spread of savannahs (Elenga et al., 2004;Schefuß et al., 2005). A reforestation after 1 ka as suggested by the pollen records from Congolian sites (Vincens et al., 1999) is not found in our record. In East Africa Holocene precipitation levels became increasingly variable after 4 ka and declined af-20 ter 2 until 0.5 ka (Bonnefille and Chalié, 2000). For southern Africa numerous proxies indicate a temperature optimum shortly before 7.5 ka and increase in summer rainfall. Aridification set in soon after 5 ka (Scott and Lee-Thorp, 2004).
Thus, the Holocene climatic changes proceed largely parallel in northern, southern and equatorial Africa. Important changes in the average position or latitudinal migration 25 of the tropical frontal systems (ITZC and CAB), however, should have caused changes in the opposite direction. It seems that these frontal systems were in place during the Holocene and that the aridification of the past 4 to 2 ka cannot be explained by a southward shift of the average position of the CAB and ITZC over Africa.

Conclusions
The marine pollen record of ODP Hole 1078C provides a detailed history of the Angolan vegetation development since the last glacial, which we interpret as follows.
-From 31.7 ka (the beginning of the record) until 21.9 ka open mountainous vegetation with heather species, composites, and grasses was important, but dry 5 woodlands also existed. Desert and semi-desert extended rather far to the north.
-Between 21.9 and 18.8 ka Podocarpus trees spread in the mountains, while rain forest started to expand in the northern lowlands.
-From 18.8 to 15.4 ka rain forest remained present in the northern lowlands and Podocarpus forest might have been locally abundant in the mountains.

10
-After 15.4 ka the Olea record indicates a change in the mountain forest composition. Until 10.0 ka rain forest was important, but woodland elements were also present.
-The Holocene started with dry forest and Miombo woodlands expanding while rain forest elements retreated.

15
-After a disturbance around 8 ka Podocarpus forest dominated between 7.8 and 3.7 ka. During this period, either Miombo woodland or wetter types of rain forest expanded and replaced open savannah vegetation.
-Savannahs became increasingly important after 3.7 and again after 2 ka. Forests probably became patchy growing light-loving and fire hardy trees. After the cool and dry climate of the last glacial the change to more humid conditions started very early in Angola (22 ka). The early amelioration is probably related to the early warming on Antarctica. The CAB might have shifted to the south (Fig. 7).
Between 18.8 and 15.4 (around Heinrich Event 1), low SSTs indicate a northern position of the ABF implying permanent upwelling along the coast and strong SE trade 5 winds. Conditions were cool but probably not arid. During this early stage there is a different development in Angola than in the Congo Basin or in West Africa where the climate was cold and dry.
After Heinrich Event 1, the climate in Angola became wetter and warmer and possibly even wetter than during most of the Holocene. No return to glacial conditions was found 10 for the Younger Dryas period. The average position and the latitudinal migration of the CAB and ITCZ might have been further south than during the Holocene (Fig. 7).
During the early Holocene the average position of the CAB was probably already close to its present position and the modern West African monsoon system in place.