Continuous vegetation record of the Greater Cape Floristic Region (South Africa) covering the past 300 thousand years (IODP U1479)

The flora of the Greater Cape Floristic Region (GCFR) of South Africa is a biodiversity hotspot of global significance, and its archaeological record has contributed substantially to the understanding of modern human origins. For 15 both reasons, the climate and vegetation history of south-western South Africa is of interest to numerous fields. Currently known paleo-environmental records cover the Holocene, the last glacial-interglacial transition and parts of the last glaciation but do not encompass a full glacial-interglacial cycle. To obtain a continuous vegetation record of the last Pleistocene glacialinterglacial cycles, we studied pollen, spores and micro-charcoal of deep-sea sediments from IODP Site U1479 retrieved from SW of Cape Town. We compare our palynological results of the Pleistocene with previously published results of Pliocene 20 material from the same site. We find that the vegetation of the GCFR, in particular Fynbos and Afrotemperate forest, respond to precessional forcing of climate. The micro-charcoal record confirms the importance of fires in the Fynbos vegetation. Ericaceae-rich and Asteraceae-rich types of Fynbos could extend on the western part of the Palaeo-Agulhas Plain (PAP), which emerged during periods of low sea-level of the Pleistocene.

After considerable efforts from various disciplines, detailed maps of the now submerged continental shelf area south of South 40 Africa have been constructed (Cleghorn et al., 2020) providing a view of the ecosystems on the Palaeo-Agulhas Plain (PAP), which was largely exposed during the last glacial maximum . The inferred glacial vegetation of the PAP is based on vegetation modelling  using the geological and soil maps  in connection with regional climate modelling (Engelbrecht et al., 2019). A wide landscape emerged divided by broad river systems with diverse Fynbos types in the western and southern part of the PAP. The central portion of the PAP was probably dominated by 45 open grasslands.
Despite the wealth of information, a synthetic understanding of vegetation and climate of the past glacial-interglacial cycle remains elusive. This uncertainty stems from a combination of problems including (i) uneven spatiotemporal sampling coupled with a substantial amount of poorly understood spatial variation (e.g., Chase et al.,2017;; (ii) records that reflect changes over limited spatial extend and over relative short periods of time (e.g., Chase, 2010;Marean et al., 2014); (iii) reliance on a 50 diversity of different proxies, not all of which directly inform about the climate or vegetation parameters of interest ; (iv) seemingly contradictory lines of evidence (e.g., Avery 1982;Faith, 2013) as well as conflicting interpretations of the evidence (e.g., Faith et al., 2019;Thackeray, 2020).
To provide a regional framework of climate and vegetation dynamics across multiple glacial-interglacial cycles, we report here a record of the vegetation in the GCFR using pollen and spores recovered from deep-sea sediments of IODP Site U1479 ( Figure  55 1), which were retrieved offshore almost 160 km SW of Cape Town (Hall et al., 2017). While the terrestrial paleontological and palaeobotanical record mostly reflect the environment of a relatively limited spatial area and cover relative short stretches in time (Marean et al., 2014) marine sediments collect pollen and spores from a wide terrestrial area and thus record an integrated regional signal (Dupont, 2011). Moreover, the marine sediments have a chronology that allows the assessment of vegetation dynamics with respect to orbital forcing and other climatic proxy records that are tied to the same chronology. The 60 sediments of IODP Site U1479 demonstrably provide a continuous signal extending from the late Quaternary down to the Pliocene. Our record enables us to continuously follow regional vegetation change in the GCFR over the past 300 thousand years (ka) at a temporal resolution of ca 3 ka. We investigate the vegetation response to climate changes and sea-level fluctuations of the middle to late Pleistocene climate cycles.

Modern climate and vegetation 75
The modern climate of South Africa is dominated by the atmospheric South Atlantic High-pressure system in the southwest and the easterly flow from the Indian Ocean (Tyson and Preston-Whyte, 2000). This results in a climate pattern over South Africa that can be divided according to its rainfall regime ( Figure 2) (Hijmans et al., 2005;Chase and Meadows, 2007). In most of South Africa, precipitation falls during the summer months (October to March; Summer Rainfall Zone, SRZ), while in the south-western part of the sub-continent, westerly winds bring precipitation during the winter months (April to September; 80 https://doi.org/10.5194/cp-2021-93 Preprint. Discussion started: 19 July 2021 c Author(s) 2021. CC BY 4.0 License.

90
While modern rainfall in the SRZ is modulated by the sea surface temperature (SST) of the Indian Ocean (Reason and Mulenga, 1999), south-western African precipitation is affected by South Atlantic SST and the Agulhas Current. Winter rainfall in the south-western Cape comes mainly with cold fronts, especially if the westerlies have a northern position over the southern tip of Africa. Warm Atlantic SSTs provide more moisture to be advected into the continent. Moisture from the warm Agulhas 95 waters recirculates over southernmost Africa (Reason et al., 2003).
The situation on the south coast is more complicated. Both tropical-temperate troughs from the east and frontal systems from the west influence the balance between WRZ and YRZ. According to Engelbrecht et al. (2015), modern precipitation along https://doi.org/10.5194/cp-2021-93 Preprint. Discussion started: 19 July 2021 c Author(s) 2021. CC BY 4.0 License. the south coast is contributed by cut-off-lows (16% of annual rainfall), ridging high pressure systems (46%) and tropicaltemperate troughs (28%). In addition, the Agulhas Current would have influenced the coastal area propagating climate signals 100 from the tropics to the southern Cape coast (Chase and Quick, 2018).
The GCFR is situated in the WRZ and YRZ and encompasses three biomes, Fynbos (including Fynbos sensu stricto, Renosterveld and Strandveld), Succulent Karoo and Albany thicket (Figure 1). The semi-desert vegetation of the Nama Karoo grows in the SRZ northeast of the GCFR . Fynbos consists of fine-leaved (from the Afrikaans word meaning "fine bush") evergreen shrubs growing on nutrient poor soils of the south-western Cape with annual 105 precipitation between 600 and 800 mm. It is extraordinarily rich in Erica species. In addition to Ericaceae other important families are Asteraceae, Rhamnaceae, Thymelaeaceae, Rutaceae, Proteaceae and the almost endemic Bruniaceae. The most important graminoids are the Restionaceae (Cape reeds) but Cyperaceae (sedges) and Poaceae (grasses) also occur.
Renosterveld is dominated by evergreen asteraceous shrubs, notably Elytropappus rhinocerotis (renosterbos). Cliffortia (Rosaceae) and Anthospermum (Rubiaceae) are important elements and so are species of Fabaceae and Malvaceae. Along the 110 coast the littoral thicket, called Strandveld, shrubs and small trees are present that have broader leaves than the shrubs of the Fynbos. In the thicket, apart from the trees (Celastraceae, Diospyros, Morella) Restionaceae are abundant (Rebelo et al., 2006).
The Succulent Karoo vegetation contains many dwarf leaf-succulents of Aizoaceae and Crassulaceae growing in the drier parts of the GCFR, in which the sparse rain (average 170 mm/a) falls predictably during the winter months. Furthermore, the vegetation includes Asteraceae, Amaranthaceae, Euphorbiaceae and Zygophyllaceae but few grasses . 115 Albany Thicket grows in the eastern part of the GCFR and is probably less relevant for this study.
On the south and east facing slopes of the Cape Fold Belt Mountains and along the south coast Afrotemperate forest is growing, which has a multi-layered canopy and occurs in the deep gorges. Podocarpaceae species (Podocarpus latifolius and Afrocarpus falcatus) are important constituents. Afrotemperate forest needs mean annual rainfall of at least 525 mm (Mucina and Geldenhuys, 2006). In both Fynbos and Renosterveld, burning of the vegetation is important and its regeneration depends on 120 fire. Fires are less important in the Succulent Karoo and Afrotemperate forest due to insufficient fuel in the former and too much moisture in the latter (Whitlock et al., 2010).

Material and Methods
IODP Site U1479 (35°04′S, 17°24′E) is located ca 160 km southwest of Cape Town, South Africa, on a 30 km wide ridge rising ca 200 m above its surroundings on the mid-to-lower western slope of the Agulhas Bank in the Cape Basin at a water 125 depth of ca 2615 m below sea level ( Figure 1, Hall et al., 2017). The seafloor at the site is bathed in North Atlantic Deep Water, while the surface ocean is influenced by the waters of the Agulhas Current, which retroflects from the Agulhas Bank at the southern tip of Africa and then flows eastwards as the Agulhas Return Current. The Agulhas retroflection occurs east of Site U1479 (Hall et al., 2017), but the phenomenon is characterized by the frequent shedding of warm core rings that pass directly https://doi.org/10.5194/cp-2021-93 Preprint. Discussion started: 19 July 2021 c Author(s) 2021. CC BY 4.0 License. over the site. A principal motivation for drilling Site U1479 was to develop a long record of this Agulhas "leakage" of warm 130 and salty Indian Ocean water into the South Atlantic Ocean.
Nine holes were cored at Site U1479 to ensure a continuous sedimentary sequence spanning more than 230 meters (early Pliocene to recent). A spliced composite record from these nine holes was created from multiple physical properties measured shipboard, most notably sediment colour (Hall et al., 2017). Sampling for various micro-paleontological and geochemical purposes was conducted along this spliced composite. The chronostratigraphic framework for this splice derives from the 135 oxygen isotopic composition of the benthic foraminifera Planulina wuellerstorfi, measured at 5 cm intervals at the Scripps Institution of Oceanography using a Thermo MAT253 equipped with a Kiel IV carbonate preparation device. Analytical uncertainty is less than 0.08‰. The details of the full record (including the accompanying carbon isotopic analyses) are

145
The core material from the shipboard splice was largely exhausted by this initial sampling. However, in the upper part of the sequence, there was sufficient overlap among the cores recovered to construct a parallel spliced composite record. Accordingly, we selected samples from Holes B, C, G and H to obtain a record of the past 300 ka with a time step of 2 to 4 ka. Holes were correlated using the green channel values of the shipboard data ( approach involves single depth offsets for an entire core (ca 10 meter in length), it does not account for variable stretching/compression of cored intervals relative to their stratigraphic correlates in a different spliced composite. In order to make use of the stratigraphic information of oxygen isotopes measured in the shipboard splice, it is necessary to account for this relative distortion in our alternative splice. The depth of Core B1 had to be shifted 46 cm upwards and stretched by a factor 155 of 1.184 to obtain the best stratigraphic correlation with the shipboard splice. Depths of Core H1 only had to be stretched by a factor of 1.127 to achieve optimal stratigraphic correlation Supplementary Figure 1). The chronology for the spliced composite can be established in a number of ways, including transferal of common oxygen isotopic chronologies (e.g., LR04 stack, Lisiecki and Raymo, 2005). However, the chronology we employ here uses a more 165 independent approach. Throughout the ca 6 Ma long record of Site U1479, sediment colour is demonstrably coherent with climatic precession, and, therefore, the chronology can be tuned to maximize the coherence of colour channel data (e.g., RGB green) with climatic precession; the phase of the color with respect to orbital variability is fixed by radiocarbon constraints of the last 30 ka. Discrete shipboard measurements and XRF scanning indicate that the sediment colour essentially monitors variable carbonate content, but the orbital tuning approach requires no assumption about the actual mechanism through which 170 orbital variability paces the carbonate variability. The resulting age-depth model and sedimentation rates for the past 350 ka are plotted in Figure 4, illustrating that sedimentation rates implied by orbital tuning are relatively stable around a mean of ca 4.5 cm ka -1 since ca 300 ka. Furthermore, we can use the oxygen isotopic record to cross check the orbitally tuned chronology against the LR04 global isotope chronology (Lisiecki and Raymo, 2005); the differences are generally within error ( though there is isolated divergence in discrete intervals such as the penultimate deglaciation. At the top of the sequence, the 175 later Holocene is probably missing. In any case, through either approach-oxygen isotopic or orbital tuning-the sediment chronology clearly resolves the suborbital variability of the last three ice age cycles. Sample volume was measured using water displacement. Pollen samples were decalcified with cold HCl (ca 10%) and then treated with HF (ca 40%) for two days. Concentrated HCl (ca 37%) was added to keep fluor-complexes in solution. Samples were sieved over a 7-µm nylon screen while ultrasonically disaggregating organic matter. The residuals were stored in water, 180 mounted in glycerol and examined under a light microscope (magnification 400 x and 1000 x) for pollen, spores and microcharcoal. Two Lycopodium spore tablets (batch No. 177745 containing 18584 ± 829 spores per tablet) were added during the decalcification step.
Pollen grains were identified using the African pollen reference collection of the Department of Palynology and Climate Dynamics of the University of Göttingen, the photo collection of the Department of Environmental and Geographical Science 185 of the University of Cape Town, the African Pollen Database (http://apd.sedoo.fr/pollen/interface/indexPollen.html) and literature (Bonnefille and Riollet, 1980;Scott, 1982;Schüler and Hemp, 2016). Black opaque angular particles were counted as micro-charcoal.
Percentages were calculated based on the sum of pollen and spores (ranging between 141 and 360). Pollen concentration was determined based on the Lycopodium spore counts and accumulation rates by multiplying the concentration with the 190 sedimentation rate. The 95% confidence intervals of percentages were calculated following Maher (1972), those of concentration values after Maher (1981). Pollen zones were determined using stratigraphically constrained cluster analysis (unweighted pair-group average and correlation similarity index) from the PAST package (Hammer et al., 2001). The latitudinal insolation gradient (LIG) during Southern Hemisphere (SH) summer and winter was calculated using the insolation data of La04 (Laskar et al., 2004); SH summer LIG by subtracting the insolation at 30°S from that at 60°S on December 21 195 (270°) and SH winter LIG by subtracting the insolation at 30°S from that at 60°S on June 21 (90°). To compare pollen concentrations with curves of possible forcing mechanisms, such as SH summer LIG, SH winter LIG, stable oxygen isotopes and sea-level, linear correlations were calculated on the equidistantly interpolated values resampled every 3 ka. Also using PAST, spectral analysis was carried out with REDFIT (Schulz and Mudelsee, 2002), which does not need equidistantly spaced data. We employed a Hanning window, 2 times oversampling and 2 segments resulting in a bandwidth of 0.00993 ka -1 . To 200 determine phase shifts between pollen data and insolation curves, we employed Blackman-Tukey cross-spectral analysis using ARAND software (Howell et al., 2006). Data were interpolated equidistantly and resampled every 3 ka between 3 and 306 ka.
We applied 40 lags resulting in a bandwidth of 0.01111 ka -1 .
However, concentrations of wind-pollinated Poaceae pollen are relatively low (8-77 grains/ml). Charcoal particle concentration varies between 2600 and 18000 particles/ml ( Figure 5). Long micro-charcoal particles make up only a fraction of the total with ratios fluctuating between 3 and 18% but for two samples dated 3 and 7 ka, in which 21% and 25% of the micro-charcoal particles are longish. Average sedimentation rates decrease from 7 to 5 cm/ka around 300 ka and decline to ca 225 4 cm/ka after 240 ka ( Figure 4).
We performed spectral analysis on the logarithmic transformed accumulation rates [log(ARs)] to avoid interdependence between the data as is the case with percentages. We analysed the accumulation rates of total pollen and spores, micro-charcoal, Restionaceae, Asteraceae pp, Poaceae, Cyperaceae, Ericaceae, Podocarpaceae, Stoebe-Elytropappus type, Anthospermum, Pentzia-Cotula type and fern spores (Tables 2 and 3). Accumulation rates are given in Supplementary Figure 3, and significant 230 power maxima in Table 2. The spectral density (power) of the accumulation rates of these taxa exceeds the 90% Χ 2 significance level in the precession band. Additional significant power in the obliquity band is found for Restionaceae, Poaceae and Stoebe-Elytropappus type. Cross-spectral analysis revealed these taxa to be negatively correlated with the Southern Hemisphere winter latitudinal insolation gradient (SH winter LIG) lagging between 0 and 3 ka (0-42° phase shift). Highest coherency was found for the cross-spectrum of Podocarpaceae pollen accumulation rates with SH winter LIG multiplied by minus 1 (Table 3). 235

Source area and pollen transport
The floral composition of the palynological assemblage of Site U1479 indicates that it records the biomes of the GCFR of South Africa ( Figure 5, Supplementary Figure 2 (Quick et al., 2015) and/or at Vankervelsvlei (Quick et al., 2016). With exception of Podocarpaceae, the floristic composition also resembles records from the Cederberg Mountains near the border between mountain Fynbos and Succulent Karoo (Scott and Woodborne, 2007a;2007b;Meadows et al., 2010;Quick et al., 2011).

255
The sediments of marine Site U1479 are thus rich in pollen and spores from southernmost Africa. This is remarkable because at present no big rivers discharge in the vicinity of the core location and the general wind direction of the southern westerlies is landwards. Moreover, the site lies outside the influence of the SE trade winds. However, numerous smaller rivers drain the https://doi.org/10.5194/cp-2021-93 Preprint. Discussion started: 19 July 2021 c Author(s) 2021. CC BY 4.0 License. south coast of South Africa, which bring pollen and spores to the ocean. Further westward transport could derive from the 260 Agulhas Current, which also transports clays into the Cape Basin (Petschick et al., 1996).
This situation was different during sea-level low-stands, especially during the LGM, when a wide network of rivers spread over the PAP . The Breede River, now discharging into the ocean near Cape Infanta west of Blombos Cave on the south coast, watered the western part of the PAP. During the LGM, the mouth of the Breede River was probably situated almost directly south of Cape Agulhas, some 250 km east of Site U1479 . Thus, during sea-level 265 low-stands the dominant source area of pollen and spores might have shifted from the present south coast and adjacent inland to the western part of the PAP.

Vegetation development
For the past 300 ka, three Terminations (i.e., rapid onset of interglacial conditions) have been recognized (Cheng et al., 2009 During the last glacial-interglacial transition, a shift from ericaceous Fynbos to a vegetation with more thicket species (Olea, Rhus, Euclea, Dodonaea, Myrsine) was found in records from the Cederberg Mountains (Scott and Woodborne, 2007a;2007b;Quick et al., 2011;Valsecchi et al., 2013). Around the Pakhuis Pass Site in the Cederberg Mountains the vegetation changed from abundant Proteaceae, Ericaceae, Stoebe-Elytropappus type, Passerina and Scrophulariaceae to increased Aizoaceae, 280 Anthospermum, Dodonaea, Rhus, Euclea (Ebenaceae), and Olea (Scott and Woodborne, 2007b). Although the details differ, the marine sediments show a similar pollen sequence and might have recorded shifts in the regional abundance of Fynbos, succulent vegetation, thicket and forest. However, it should be kept in mind that the marine sediments collect pollen and spores from a large area and that the subsequent maxima in the marine record would not represent a local vegetation succession. Asteraceae, Ericaceae, Passerina and Anthospermum tend to be more abundant during glacial stages. Cooler temperatures during glacial stages might have favoured Ericaceae (Quick et al., 2016). Podocarpaceae also seem to increase during cooler 295 stages with possibly greater moisture availability, although not during glacial maxima. Low atmospheric CO2 during MIS 8, 6 and 2 might have hampered the growth of Afrotemperate forests (Bereiter et al., 2015;Dupont et al., 2019). Probably more important is the exposure of the PAP south of the present coast during glacial stages. Especially when sea level sank below -100 m (Figure 1), a large area for colonization and growth of thicket, woodland on the floodplain, Renosterveld and Fynbos was exposed . This is suggested by pollen percentages of the Asteraceae pp (r=0.34, p<0.01) and 300 Ericaceae (r=0.58, p<0.01) correlating significantly with sea level (Bintanja et al., 2005) Table 2

summer LIG), of Asteraceae pp and Ericaceae with sea level (Bintanja et al., 2005), and of Podocarpaceae with the Southern Hemisphere winter latitudinal insolation gradient (SH winter LIG). Insolation gradients were calculated using the astronomical solution of La04 (Lascar et al., 2004). Note the reversed scales. Shadings indicate 95% confidence intervals.
The modelled LGM vegetation on the PAP does not include forest , perhaps because the climate and 320 relatively flat geomorphology of the plain would favor rapid and expansive fires (Kraaij et al., 2020), precluding the growth of extensive forest on the plain. The low Podocarpaceae pollen percentages for the LGM corroborate this. However, we find a substantial increase in the representation of Podocarpaceae during MIS 7d and MIS 5 after the last interglacial maximum (MIS 5e) when sea-level was around 40 m lower than at present. Also, Esper et al. (2004) found substantial amounts of LGM reconstruction. It is also possible that during phases of increased summer rain (see next section), Podocarpaceae extended in the Cape Fold Belt Mountains and the pollen were transported by rivers, in particular the Breede River. Wood charcoal assemblages from Elands Bay Cave prior to 20 ka gave evidence for several species characteristic for Afrotemperate forest, including Podocarpus elongates, which were probably growing along rivers and streams (Cowling et al., 1999). 330

Open grasslands in the south-western Cape?
Pollen percentages for grasses are rather low (maximum of 10%, mean 5%), which is in line with glacial terrestrial pollen records of Cederberg Mountains and the south coast (Meadows and Sudgen, 1991;Scott and Woodborne, 2007a;2007b;Quick et al., 2011Quick et al., , 2015Valsecchi et al., 2013). Grass pollen is also relatively scarce at Vankervelsvlei and at Rietvlei grass pollen is only abundant during the early Holocene (Quick et al., 2015;. Modern values from the Namaqualand mudbelt 335 along the west coast of South Africa run from ca 10% Poaceae pollen near Elands Bay adjacent to Fynbos vegetation to ca 40% in front of the Orange River, which drains the grass-rich landscapes of the South African interior (Zhao et al., 2016).
Thus, the low Poaceae pollen percentages at Site U1479 are comparable to modern values derived from Fynbos and Renosterveld, and do not indicate extensive grasslands in the western part of the GCFR at any time during the past 300 ka.
In contrast to the pollen record, the Pleistocene fossil record of large-bodied herbivores from the GCFR has long been 340 interpreted as indicating an expansion of grasslands, especially during glacial phases (Klein, 1983). South coast sites dating from MIS 6 onwards (e.g., Blombos Cave, Die Kelders Cave, Klasies River, Nelson Bay Cave, Pinnacle Point) frequently include extinct or extralimital grazers (e.g., Klein, 1972Klein, , 1976Klein, , 1983Klein and Cruz-Uribe, 2000;Henshilwood et al. 2001;Rector and Reed, 2010), with very high abundances of grazers often observed during the LGM and late glacial (Klein, 1972(Klein, , 1978(Klein, , 1983Faith, 2013). Stable isotope evidence (δ13C) from ungulate grazers at Boomplaas Cave indicate consumption of 345 primarily C3 grasses during the LGM (Sealy et al., 2016), as is also the case at Nelson Bay Cave (Sealy et al. 2020). Williams et al. (2020 measured stable isotopes on teeth of micromammals and macromammals from a hyena den found at Pinnacle Point and dated to MIS 6 (penultimate glaciation). The micromammal δ 13 C indicate a C3 vegetation in the immediate surrounding of the den, while the macromammal teeth indicate more mixed C3/C4 grasslands for the surrounding plain.
To reconcile the fossil pollen records to the fossil bone records from the south coast, one could hypothesize that the climate 350 displayed a strong east-west gradient along the Palaeo-Agulhas plain from winter rainfall in the west to increased summer rainfall in the east. This pattern is indicated by the downscaling of the LGM results of coupled global climate model simulations to the region (Engelbrecht et al., 2019). Mapping of the glacially exposed PAP places the grasslands mainly in the central and eastern part of the plain with Fynbos and Renosterveld vegetation dominating the western part during the LGM Marean et al., 2020). It is envisaged that the PAP would have harboured grazers migrating from the west in winter 355 to the east in summer and back to make optimal use of grazing opportunities between winter rainfall and summer rainfall regions (Marean, 2010;Faith and Thompson, 2013;Copeland et al., 2016). However, a test of this proposed migration using https://doi.org/10.5194/cp-2021-93 Preprint. Discussion started: 19 July 2021 c Author(s) 2021. CC BY 4.0 License.
isotopic measurements of serially-drilled herbivore teeth from Pinnacle Point could not definitively corroborate a seasonal migration pattern (Hodgkins et al., 2020).
In this scenario, the pollen assemblages at Site U1479 would mainly register the vegetation of Renosterveld, Fynbos and 360 riparian woodlands on the PAP. However, this concept is inconsistent with fossil evidence of micro-and macromammals indicating that the latest Pleistocene witnessed an expansion of open grassland relative to modern times in the south-western Cape as well (Avery, 1982;Klein and Cruz-Uribe, 1987;. Glacial faunal assemblages document numerous grazers and open-country species (e.g., springbok, wildebeest, white rhino, quagga). For instance, Elands Bay Cave (Klein and Cruz-Uribe, 1987;, Swartklip 1 (Klein, 1975), Hoedjiespunt 1 (Klein, 1983), and Die Kelders Cave (Klein and Cruz-Uribe, 2000) all 365 indicate abundant grazing fauna feeding mainly on C3 grasses (Sealy et al. 2016(Sealy et al. , 2020. Also, the LGM pollen spectra from Elands Bay Cave show strongly increased grass pollen percentages (Parkington et al., 2000). Perhaps a grassy under-story, such as those often associated with the Renosterveld, were sufficient to support these grazers. Alternatively, the vegetation of the western portion of the PAP consisted more of a mixture between grassland, Fynbos and Renosterveld. However, the modern vegetation of the south-western Cape does not support grazers, such as mountain zebra (Equus zebra), very well (Faith, 2012). 370 The grass species composition in pre-Holocene Renosterveld may have differed from the modern situation (and was possibly more palatable). Unfortunately, grass pollen cannot be determined to species level so it is not possible to assess the specific grasses available to these grazers. We could also speculate that grazing pressure might have subdued flowering and pollen production of the grasses, or that reduced atmospheric CO2 enhanced grass quality (lower C:N ratio) and supported greater grazer biomass (e.g., Owensby et al., 1996). On the other hand, the grasslands of the western portion of the PAP might be 375 underrepresented in the pollen records, because other plants in the vegetation, such as Restionaceae, produced more pollen.
This may result in a pollen record that underestimates the extent of grasses in these ecosystems. A final possible explanation for the discrepancy between the pollen and bone records is that Pleistocene grazers may have browsed more than their Holocene counterparts (e.g., Stynder, 2009).

Vegetation associated with fire 380
Elevated micro-charcoal concentrations seem associated with Fynbos vegetation (Figure 7, Table 2). Significant positive linear correlations (p<0.01) were found between micro-charcoal and pollen concentrations of Asteraceae pp (r=0.42), Restionaceae (r=0.44) and Ericaceae (r=0.40), but not with Stoebe-Elytropappus type. Micro-charcoal and Pentzia-Cotula type concentrations show a weaker correlation, however this is still significant (r=0.27, p<0.05). There is a slight correlation with Podocarpaceae (r=0.21), but not with Amaranthaceae nor Aizoaceae pollen concentrations. The micro-charcoal record 385 indicates that, as expected, fires were more frequent in asteraceous and ericaceous Fynbos vegetation but maybe less frequent in the Renosterveld .

Orbital forcing
The spectral analysis results indicate strong influence of precession and some influence of obliquity (Tables 2 and 3; Figure   8). We use here the SH winter LIG as an expression of the precessional variability. The precession component in the pollen accumulation rates is not an artefact of the age model based on tuning of the colour data as the age-model derived sedimentation rates are rather constant. Moreover, spectral analysis of pollen percentages which are independent of changes in sedimentation, 400 also reveals precessional variability (not shown, but see the correlation between Podocarpaceae pollen percentages and the SH winter LIG in Figure 6). Cross-spectral analysis indicates a negative correlation between the SH winter LIG and the accumulation rates of fern spores and pollen of Pentzia-Cotula type, Podocarpaceae, Stoebe-Elytropappus type, Ericaceae, Cyperaceae, Poaceae pp, Asteraceae pp, and Restionaceae (Table 3).
The precessional forcing might be associated with the SH winter LIG, which forces the latitudinal temperature gradient 405 Zhao et al., 2020). Partridge et al. (1997) previously proposed precessional forcing of rainfall in eastern South Africa. Only more recently, studies of independently dated marine sediments from the western Indian Ocean confirmed the importance of precessional forcing on the discharge of southeast African rivers, such as Limpopo and Tugela Rivers (Simon et al., 2015;Caley et al., 2018). It should be noted that SST of the Agulhas waters in the western Indian Ocean do not show a precessional rhythm (Caley et al., 2011;. Sensitivity modelling using maximum and minimum precession showed that 410 austral summer precipitation in southeast Africa is higher when precession is at maximum resulting in maximum austral summer insolation Simon et al., 2015) increasing the extension of the SRZ and probably also of the YRZ. This in turn corresponds to minima in the SH winter LIG (i.e., the difference in insolation between high-and mid-https://doi.org/10.5194/cp-2021-93 Preprint. Discussion started: 19 July 2021 c Author(s) 2021. CC BY 4.0 License. southern latitudes during austral winter). The model experiment also indicates that maximum precession, i.e., minimal SH winter LIG, induces weakening of the Southern Hemisphere westerlies (Simon et al., 2015). 415  Table 2). Increased river discharge of rivers in the YRZ during maximum precession could explain the precessional variability in the accumulation rates of other taxa (Table 2). For variability in ocean conditions, Esper et al. (2004) found strong precessional variability in the dinoflagellate cyst record of core GeoB3603-2 (located ca 15 km east of Site U1479) indicating that stratified oligotrophic waters prevailed when precession 430 was weak (SH winter LIG strong). Table 3. Results of the Blackman-Tukey cross-spectrum analysis (3-306 ka, step 3 ka, 40 lags, bandwidth 0.01111 ka-1) executed with ARAND software (Howell et al. 2006). Phase is given for SH winter LIG multiplied by minus 1 against the log(AR) in cm 2 ka -1 for fern spores, selected pollen taxa, total pollen and spores, micro-charcoal. 95%-level of non-zero coherency is 0.84 (denoted in 435 bold); 80%-level of non-zero coherency is 0.70. These taxa exhibit a negative correlation with SH winter LIG lagging by up to 2.6 ka.
(SH winter LIG)x(-1) Accumulation rates of Podocarpaceae pollen show highest coherency with SH winter LIG and the maximum in accumulation rates occurs 1.5 ± 0.2 ka after the minimum of SH winter LIG (phase lag of 23° ± 3 between precessional forcing and vegetation 440 response, Table 3). Our results are in line with a transient simulation of monsoon climate over the past 280 ka focusing on precessional variability (Kutzbach et al., 2008). This simulation indicated that the South African monsoon (in the SRZ) responded with a phase lag of slightly less than one month (30°) to maximum December insolation, which is very close to the response of Podocarpaceae pollen accumulation rates.
Obliquity forcing could be explained by the latitudinal temperature gradient during summer (Davis and Brewer, 2009) Table 4.
Not only did variability and relative abundance of pollen taxa change, but also the rhythm in the variability differs between the Piacenzian and Pleistocene. Spectral analysis of percentages, concentration and accumulation values of the most important pollen taxa showed precessional (24-18 ka) variability but no power in the obliquity band (40-42 ka) during the Piacenzian 490 (Zhao et al., 2020), whereas the Pleistocene record shows significant power at the obliquity period in the pollen accumulation rates of Restionaceae, Cyperaceae and Poaceae (Table 2). The appearance of the obliquity rhythm in the Pleistocene pollen record might be an effect of the much larger climate change between glacial and interglacial periods, but could also be the result of vegetation on the PAP during periods of low sea level during the Pleistocene. During these low sea-level periods, especially when global sea-level dropped below 100 m, exposed regions became available for Asteraceae-rich vegetation such as Renosterveld, thicket and woodland on the floodplains. Also, drought-adapted pioneers might have found suitable habitats on the PAP during Pleistocene glaciations. The dramatic 500 environmental changes that must have accompanied the flooding of the PAP during Terminations would have induced both erosion and changes in the vegetation reflected in the pollen record.

Conclusions
Deep-sea sediments from IODP Site U1479 retrieved from SW of Cape Town provide a 300-ka long record of pollen, spores and micro-charcoal from the Greater Cape Floristic Region indicating developments in Fynbos, succulent vegetation, thicket 505 and forest over the course of three glacial-interglacial cycles. The micro-charcoal record suggests that fires occurred more in Fynbos than in other vegetation types of the GCFR. Our data indicate a substantial change in vegetation during the Terminations of glacial periods. During sea-level low-stands, the vegetation on the western part of the PAP would have consisted of asteraceous and ericaceous Fynbos including Renosterveld as well as thicket and woodland. Our results corroborate the vegetation modelling for the PAP during the last glacial maximum. However, we also infer extension of 510 Podocarpaceae forest during parts of MIS 7 and MIS 5-4. The record exhibits relatively little grass pollen, which is difficult to reconcile with fossils of grazers found at Pleistocene paleontological and archaeological glacial sites across the GCFR.
Precession variability of the Podocarpaceae record, in particular, indicates increased summer rain and extension of the yearround rainfall zone in southernmost Africa at the cost of the winter rainfall zone during phases of maximum precession, when local summer insolation was high and the latitudinal insolation gradient in austral winter was weak. While the pollen 515 accumulation rates of most taxa only exhibit a precessional rhythm, those of Poaceae, Restionaceae and Stoebe-Elytropappus type show both precession and obliquity related variability. This differentiation suggests a heterogeneous response of different vegetation types in the GCFR to global climatic changes.
Comparison with the Pliocene (Piacenzian) record of the same site indicates that the Pleistocene-Holocene winter rainfall zone decreased in size compared to the Pliocene. Reduced extension of the winter rainfall zone along the south coast could have 520 been the result of weak westerlies in combination with cool SST, which decreased precipitation during winter. However, insolation forcing and precession variability remained drivers of rainfall distribution and seasonality. The large sea-level fluctuations of the late Pleistocene intermittently exposed the Palaeo-Agulhas Plain, but seemingly did not alter the rainfall patterns.

Acknowledgements 525
The study was financially supported by the Deutsche Forschungsgemeinschaft (DFG grant DU211/7) and the National Science Foundation (Award #1826666). LMD thanks Nikolaos Ioannidis for his help with the palynological preparations.

Author contributions
LMD designed the study, carried out the palynological analysis and wrote the draft, CC carried out the isotope analysis and constructed the age model. All authors contributed to the discussion and the final manuscript.