Southern Hemisphere anticyclonic circulation drives oceanic and climatic conditions in late Holocene southernmost Africa

Due to the high sensitivity of southern Africa to climate change, a reliable understanding of its hydrological system is crucial. Recent studies of the regional climatic system have revealed a highly complex interplay of forcing factors on precipitation regimes. This includes the influence of the tropical easterlies, the strength of the southern hemispheric westerlies as well as sea surface temperatures along the coast of the subcontinent. However, very few marine records have been available in order to study the coupling of marine and atmospheric circulation systems. Here we present results from a marine sediment core, recovered in shallow waters off the Gouritz River mouth on the south coast of South Africa. Core GeoB18308-1 allows a closer view of the last ∼ 4 kyr. Climate sensitive organic proxies, like the distribution and isotopic composition of plant-wax lipids as well as indicators for sea surface temperatures and soil input, give information on oceanographic and hydrologic changes during the recorded time period. Moreover, the micropaleontology, mineralogical and elemental composition of the sediments reflect the variability of the terrigenous input to the core site. The combination of down-core sediment signatures and a catchment-wide provenance study indicate that the Little Ice Age (∼ 300–650 cal yr BP) was characterized by climatic conditions favorable to torrential flood events. The Medieval Climate Anomaly (∼ 950–650 cal yr BP) is expressed by lower sea surface temperatures in the Mossel Bay area and humid conditions in the Gouritz River catchment. These new results suggest that the coincidence of humid conditions and cooler sea surface temperatures along the south coast of South Africa resulted from a strengthened and more southerly anticyclonic circulation. Most probably, the transport of moisture from the Indian Ocean by strong subtropical easterlies was coupled with Agulhas Bank upwelling pulses, which were initiated by an increase in Agulhas Current strength.


Introduction
South Africa's regional climate has been discerned as particularly sensitive to future climate change.Over the last five decades mean annual temperatures have increased by at least 1.5 times the observed global average and both extreme rainfall events and droughts are anticipated as environmental, social and economic threats (Niang et al., 2014;Ziervogel et al., 2014).It is well known that the precipitation patterns in South Africa are influenced by shifts of the southern hemispheric westerlies (SHW), the strength of the tropical easterlies, as well as variations in ocean circulation (e.g., Chase and Meadows, 2007;Chase et al., 2011;Marzin and Braconnot, 2009;Schefuß et al., 2011;Fig. 1).The interplay and regional extent of these factors and their relation to global climate forcings at different timescales is, however, far from being understood.The main dispute currently concerns the Published by Copernicus Publications on behalf of the European Geosciences Union.
driving factors behind the easterly rainfall regimes in the southernmost part of Africa (south of ∼ 13-14 • S), which is not influenced by modern Intertropical Convergence Zone shifts.A major question in this area is whether the climate variability is synchronous with northern hemispheric variations or driven by direct insolation changes and thus antiphased to northern hemispheric signals.In the past, it was relatively uncontroversial that long term (glacial-interglacial timescale) climate variations in southernmost Africa are directly forced by local insolation (Daniau et al., 2013;Simon et al., 2015).However, recently several paleorecords have emerged showing either no evidence of direct insolation as a climatic driver (Chase et al., 2009(Chase et al., , 2010;;Dupont et al., 2011) or evidence for opposing trends (i.e., synchrony with the Northern Hemisphere) (Tierney et al., 2008;Stager et al., 2012;Truc et al., 2013).Furthermore, on shorter timescales, recent datasets reveal strong indications for a highly complex local climate system.So far, data from Wonderkrater in eastern South Africa (Truc et al., 2013) and regional pollen analyses (Scott et al., 2012) document enhanced and diminished precipitation with respect to the geographical position of the recording archive.Evidence for climate variability in phase with the Northern Hemisphere is delivered by the regional compilation of pollen records published by Chevailier and Chase (2015) as well as by the progressive Holocene aridification in northwestern Namibia (proposed by Chase et al., 2010) and the simultaneous cool and dry period in eastern South Africa (proposed by Holmgren et al., 1999 andNorström et al., 2014).Mechanisms suggested to explain the transmission of the northern hemispheric signal to southern Africa include teleconnections causing a dipole pattern of rainfall between eastern tropical and southern Africa (Norström et al., 2014) as well as sea surface temperature (SST) changes due to ocean circulation variability (Agulhas strength) (Tierney et al., 2008;Stager et al., 2012;Scott et al., 2012;Truc et al., 2013).The latter can be discerned as a major research gap in southernmost South Africa.Due to the strongly erosive Agulhas Current, very few marine records from the Agulhas Bank exist and the influence of SSTs on the regional precipitation patterns therefore remains under dispute.On the one hand, it has been suggested that decreased SSTs on the southwestern coast of South Africa lead to decreasing precipitation in southernmost South Africa (Rouault et al., 2003) and positive SST anomalies along the east coast associated with Agulhas strengthening enhance summer precipitation in the eastern South Africa (Jury et al., 1993;Dupont et al., 2011;Scott et al., 2012).On the other hand, the oxygen isotope composition of marine mollusk shells preserved in Nelson Bay archaeological cave deposits indicates that during periods of wetter conditions over the southern African interior, the Agulhas surface water temperatures were actually lower than during arid periods (Cohen and Tyson, 1995).Based on these data in combination with interannual observations, a conceptual model relating oceanic and atmospheric circulation systems of southern Africa was proposed in 1995 by Cohen and Tyson.This model predicts that during periods of stronger anticyclonic circulation, the increased alongshore winds caused coastal upwelling in the Mossel Bay area whereas the increased Agulhas strength drives upwelling over the east coast shelf edge.The database behind this model is however sparse and a large gap exists for the period between 2400 and 650 yr BP (Tyson and Lindesay, 1992;Cohen and Tyson, 1995).Our multiproxy study on core GeoB18308-1, located on the south coast of South Africa, off the mouth of the Gouritz River, 30 km west of the town Mossel Bay, aims to close this gap.In addition to providing a SST reconstruction for the past ∼ 4 kyr, this record also holds the potential for a high-resolution continental climatic reconstruction.In order to decipher the complexity of the terrigenous climatic signal, sediment provenance and transport processes are studied using catchment material in a sourceto-sink approach.

Gouritz River catchment area
The Gouritz River catchment is located within the yearround rainfall zone (YRZ; Fig. 1) and has a mean annual runoff of ca.488 × 10 6 m 3 (Le Maitre et al., 2009).Desmet and Cowling (1999) and Le Maitre et al. (2009), however, note that even though observed rainfall patterns in the catchment area show year-round precipitation, most of the area mainly receives rainfall during summer and autumn.Major floods, e.g., in January 1981, March 2003and in March 2004, caused by extreme rainfall events, are another characteristic of this area (Desmet and Cowling, 1999;Cowling et al., 2004).Despite the relatively small size of the catchment area (45 715 km 2 ), the altitudinal gradient is steep as the Swartberg Mountains rise abruptly above 2000 m a.s.l.within 100 km of the coast (Le Maitre et al., 2009).

Vegetation in the study area
Within the study area three dominant vegetation types are described: Fynbos, Succulent Karoo and the Nama Karoo biomes (Mucina and Rutherford, 2006; see map in Fig. 2).The Fynbos biome is characterized as a Mediterranean-type vegetation (Goldblatt and Manning, 2002).It is found especially along the southern and southwestern coast of South Africa and therefore receives most of its rainfall during austral winter.The most dominant photosynthetic pathway reported is C 3 (Vogel et al., 1978).The Succulent Karoo biome is described as being better adapted to arid conditions and higher summer temperatures (Carr et al., 2014).Due to the geographic distribution of vegetation, the Succulent Karoo biome gets most of its rainfall during winter but also receives summer precipitation in the eastern part of the catchment (Rundel et al., 1999).The last vegetation type, Nama Karoo, is described to consist mostly of C 4 grasses and, due to its occurrence in a northeastern geographical position, receives dominantly summer rain (Le Maitre et al., 2009).Studies within this region as well as in other areas of Africa revealed that these vegetation types showed differences in their nalkane composition (Rommerskirchen et al., 2006;Vogts et al., 2009;Carr, 2012Carr, , 2014;;Boom et al., 2014;Herrmann et al., 2016; map in Fig. 2).

Geology of the Gouritz River catchment
The distinctive topography of southern Africa is characterized by a high-relief inland plateau flanked by a low average elevation coastal plain.The Gouritz River catchment drains almost all sequences of the Cape: the Malmesbury Group and the Cape Granite Suite as the oldest "basement" rocks (Rozendaal et al., 1999;Johnson et al., 2006;Milani and de Wit, 2008); the thick siliciclastic sequence of the Cape Supergroup (Thamm and Johnson, 2006;Newton et al., 2006); and the Karoo Supergroup, which was terminated by the extrusion and intrusion of the extensive Drakensberg basalts and dolerites (Duncan et al., 1997).The southern cape continental shelf and the broad, shallow plain, low-relief coastal plain mantled with Pleistocene/Holocene deposits comprise the submerged and emergent portions of a continuous feature (Cawthra et al., 2015).The south coast is characterized by a series of eastward-opening log-spiral bays that extend for approximately 20-40 km between adjacent west-east-trending rocky headlands.The Gouritz River is associated with a welldeveloped, stratified, sediment wedge (∼ 10 km wide and 85 km long) (Birch, 1978) that extends predominantly westwards of the river mouth.Offshore, the Gouritz River is associated with a subdued incised valley, running across the continental shelf (Cawthra, 2014).

Oceanic circulation on the eastern Agulhas Bank
The oceanography of the Agulhas Bank is strongly influenced by the warm, fast-flowing Agulhas Current originating in the Mozambique Channel (Lutjeharms et al., 2001).The inner shelf targeted in this study is influenced by wind-driven coastal upwelling, particularly during summer (see Hutchings et al., 1995).Along-coast easterly winds drive these periodic events of short-term surface water cooling of up to 8 • C (Schumann et al., 1982;Beckley, 1983).During winter, plumes of warm Agulhas Current water have been observed to advect onto the shelf by southwesterly winds (Lutjeharms and van Ballegooyen, 1988).

Sediment coring and catchment samples
Sediment core GeoB18308-1 was taken with a vibrocorer during RV METEOR Cruise M102 in December 2013 from a protected valley fill on the continental shelf near Mossel Bay (34 • 22.39 S, 21 • 55.75 E) 4 km offshore of the Gouritz River mouth.Using multibeam bathymetry and side-scan sonar Cawthra et al. (2015) describe the morphology of the area as a generally smooth, wide continental shelf of low gradient (see location in Fig. S1 in the Supplement).The core has a total length of 4.94 m and mainly consists of fine sand and mud.
In the interval between ∼ 30 and 70 cm an event deposit revealing slumped turbidite facies with an erosive contact was identified.South African sea level reconstructions (Compton, 2006;Ramsay and Cooper, 2002) reveal that the sampling site (located in about 40 m water depth) was not significantly influenced by fluctuations during the last 5 kyr.Riverbank, flood deposit and suspension load (obtained by filtering ca. 100 L of water pumped from the rivers' main flow) samples were collected from eight locations along the Gouritz River in March 2015 in order to determine the provenance of the deposited material (see locations in Fig. 2 map).

Age model
Ages were estimated from nine total organic carbon (TOC) samples, two bivalves, one piece of wood and one crab claw (Table 1).The age model used in this study is based on the radiocarbon ages of non-redeposited material only (Fig. 3).
The cleaning procedures as well as the accelerator mass spectrometry (AMS) measurements were carried out in the Poznań Radiocarbon Laboratory, Poland, and Beta Analytic Rawww.clim-past.net/13/649/2017/Clim.Past, 13, 649-665, 2017 Table 1.AMS radiocarbon analyses of bulk TOC and carbonate material from core GeoB18308-1.Depending on sediment type (terrestrial or marine identified using all available data) the Southern Hemisphere calibration curve (SHCal13) (Hogg et al., 2013) or the modeled ocean average curve (Marine13) (Reimer et al., 2013) was used.To the latter a R of 146 ± 85 14 C (Dewar et al., 2012)   diocarbon Dating Laboratory Florida, USA.Depending on sediment type (terrestrial or marine identified using all available data) either the Southern Hemisphere calibration curve (SHCal13) (Hogg et al., 2013) or the modeled ocean average curve (Marine13) (Reimer et al., 2013) were used to calibrate the radiocarbon ages.The marine R is assumed to be close to the south west coast R (146 ± 85 14 C years) published in 2012 by Dewar et al. (Mike Meadows, personal communication, 2017).The software Bacon (Blaauw and Christen, 2011) was used to calculate an age model.We refer to median age estimations in this paper; please note the associated uncertainty indicated in Fig. 3 and Table 1.

Inorganic geochemistry
Inorganic geochemical analyses were performed down-core (2 to 10 cm resolution) as well as on soil, riverbank, flood deposit and suspension load samples collected from eight Gouritz catchment locations (ECT-1 to ECT-8) in March 2015 (Fig. 2, Table 2).Elemental profiles were collected using MARUM XRF Core Scanner II (AVAATECH serial no. 2) equipped with an Oxford Instruments 50W XTF5011 rhodium X-ray tube, a Canberra X-PIPS silicon drift detector (SDD; model SXD 15C-150-500) run at a 150 eV resolution and a Canberra digital spectrum analyzer (DAS 1000) at the MARUM -University of Bremen.The vertical resolution was 1 cm with two generator settings (30 kV, 1 mA, 20 s; 10 kV, 0.2 kV, 20 s) for detection of different elemental groups.For calibration purposes the XRF-spectrometer measurements were completed on 28 selected dried and ground sediment samples using a PANalytical Epsilon3-XL XRF spectrometer (methodology in Govin et al., 2014).Using these discrete ED-XRF measurements and the procedure proposed by Lyle et al. (2012) we were able to normalize the XRF scanning data (counts).Good correlations (R 2 above 0.7) of XRF scanner and XRF analyses of discrete samples results were found for a variety of elements including Ca and Fe.

Organic geochemistry
The continuous high-resolution, XRF scanning dataset (Fig. S2) indicates little variability in the lower half of core GeoB18308-1.Based on this finding subsamples for biomarker analysis were taken at higher resolution in the upper half of the core.

Plant-wax analyses
Dried and ground samples were extracted using a DIONEX accelerated solvent extractor (ASE 200) at 100 • C, 1000 psi with a 9 : 1 mixture of dichloromethane (DCM) to methanol for 5 min and three extraction cycles.Prior to extraction, squalane and C 46 -GDGT (glycerol dialkyl glycerol tetraether) were added as internal standards.Asphaltenes were removed from the total lipid extracts using Na 2 SO 4 columns and elution with hexane.Extracts were saponified for 2 h at 85 • C in 500 µL of a 0.1 M KOH solution in methanol (MeOH).Neutral lipids were recovered by liquidliquid extraction using hexane.Lipid fractions were separated over a silica gel column (10 % deactivated) using hexane (hydrocarbons), DCM (ketones), and DCM : MeOH 1 : 1 (polar fractions).Unsaturated compounds were removed from the hydrocarbon fractions by column chromatography over AgNO 3 -coated silica using hexane.The saturated hydrocarbon fractions containing n-alkanes were injected in splitless mode at 260 • C into a Thermo Scientific Focus gas chromatograph equipped with a DB-5ms column (30 m × 0.25 mm, 0.25 µm film thickness, Agilent Technologies, Palo Alto, USA) coupled to a flame ionization detector (GC-FID).
The oven was held at 70  to 320 • C, and remained at this temperature for 16.5 min.An external calibration standard containing n-alkanes of known concentrations was analyzed every six samples.Based on the repeated standard analyses, the precision of quantification is calculated to 5 %.The ratio between C 29 and C 31 n-alkanes (Norm31) was calculated using Eq. ( 1): The carbon preference index (CPI) (Bray and Evans, 1961) was calculated according to Eq. ( 1 (2)

Compound-specific carbon and hydrogen isotope analyses
The fractions containing n-alkanes were used for compoundspecific carbon and hydrogen isotope analyses.Compoundspecific stable hydrogen isotope measurements were performed on a Trace GC (Thermo Fisher Scientific, Bremen, Germany) coupled to a MAT 253 IRMS (Finnigan MAT, Bremen, Germany) via a pyrolysis reactor operated at 1420 • C. The PTV injector was maintained at 45 • C at injection and then heated to 340 • C to transfer the sample onto the GC column.Compounds were separated on a Rxi-5ms silica column (30 m × 0.25 mm, 0.25 µm film thickness, Restek, Bellefonte, USA).The GC was maintained at 120 • C for 3 min then heated to 200 • C with 30 • C min −1 , then at 4 • C min −1 to 320 • C and held for 24 min.H 2 reference gas was used for isotope calibration.The δD values are expressed in per mill relative to VSMOW.The H3+ factor varied around 6.12 ± 0.02 ppm nA −1 .Long-term repeated analysis of the external standard mixture with 16 n-alkanes rendered a precision (1σ ) of ±3 ‰ and an average accuracy of 0 ‰.The internal standard squalene had an accuracy and precision of 0 and 2 ‰, respectively.Carbon isotope compositions of the n-alkanes were analyzed on the same type of GC coupled to a MAT 252 IRMS via a modified GC/C III combustion interface operated at 1000 • C. Injector and GC setting were similar to that for δD analysis.Calibrated CO 2 reference gas was used for isotope calibration.Values are expressed in δ 13 C relative to VPDB.Long-term repeated analysis of the external standard mixture with 16 n-alkanes rendered a precision (1σ ) of ±0.3 ‰ and an average accuracy of 0.4 ‰.The internal standard squalene had an accuracy and precision of 0.4 and 0.2 ‰, respectively.

Analysis of GDGTs
The concentrations of glycerol dibiphytanyl glycerol tetraethers (GDGTs) were determined at the Department of Geoscience, University of Bremen.Polar fractions containing GDGTs were filtered through a 0.45 µm PTFE filter and weighed before analysis.The instrument used to determine GDGT concentrations was an Agilent 1200 series highperformance liquid chromatograph (HPLC) coupled with an atmospheric pressure chemical ionization interface (APCI) to an Agilent 6120 quadrupole mass spectrometer (MS).A detailed description of the method this is based on can be found in, for example, Hopmans et al. (2000).GDGTs were eluted over a Prevail Cyano column (Grace, 3 µm, 150 mm × 2.1 mm) maintained at 30 • C using a gradient of solvent A (nhexane) and B (5 % isopropanol in n-hexane) from 80 % solvent A (5 min) followed by a linear increase to 36 % solvent B in 40 min.The flow rate was 0.2 mL min −1 .GDGTs were detected in positive-ion mode of the APCI-MS and selective ion monitoring (SIM).The APCI spray-chamber specifications were as follows: nebulizer pressure 50 psi, vaporizer temperature 350 • C, N 2 drying gas flow 5 L min −1 and 350 • C, capillary voltage −4 kV and corona current +5 µA.Peak areas of the target compounds were used to compute the following proxies.The TetraEther indeX TEX H 86 was used as a proxy to estimate SSTs (Kim et al., 2010).The TEX H 86 is based on the ratio between isoprenoidal GDGTs of 86 carbon atoms, containing one, two and three cyclopentane moieties (I, II and III) as well as the regioisomer of crenarchaeol, a GDGT with 4 cyclopentane moiety (V ).These compounds represent membrane constituents of planktonic archaea, which were found to shift in abundance with changing SSTs (Schouten et al., 2002).TEX H 86 was calculated with Eq. ( 2), whereas for SST reconstruction Eq. (4) was used (Kim et al., 2010).
The branched and isoprenoid tetraether (BIT) (Eq.5) was derived from the relative abundance of branched glycerol dialkyl glycerol tetraethers (brGDGTs) and the isoprenoid GDGT crenarchaeol (Hopmans et al., 2004).While brGDGTs are thought to be produced by soil bacteria, crenarchaeol is known to be a biomarker for planktonic archaea.
Hence, the index is widely used as a proxy for soil input (Weijers et al., 2014, and references therein).
A relationship between BIT and the tetraether index TEX H 86 (Fig. S3) is not evident, ruling out a possible alteration of the TEX H 86 signal near big river flows as indicated by Weijers et al. (2014).Comparison of core-top TEX H 86 -based SST estimates from GeoB18308-1 with satellite-derived SSTs (MODIS-A ftp://podaac-ftp.jpl.nasa.gov)reveals similar temperatures (Locarini et al., 2013).Reconstructed SST values are therefore assumed to reflect mean annual values.

Grain size
Particle size distribution was obtained by laser diffraction, using a Malvern Mastersizer 2000 fitted with a Hydro 2000G dispersion unit.Scattered light data were recorded from 2000 to 5000 snapshots of 10 µs.A polydisperse mode of analysis and a refractive index of 1.533 with an adsorption of 0.1 were chosen.Size data collection was performed at constant obscuration in the range 10-20 %.Visible shell fragments were removed prior to measurement.

Micropaleontology
Microfossils from 16 samples from core GeoB18308-1 were analyzed.The sample volumes vary between 1 and 9 mL around an average of 5 mL and represent a 1 cm thick horizon each.After wet sieving, the > 200 µm size fraction of the dried residues was picked under a low-power stereomicroscope for all ostracods and foraminifers.Other microfossils or fragments of larger forms were counted at group level and only selected specimens of those were picked as a reference.If possible, all taxa were identified to the species level, relying on Benson and Maddocks (1964), Dingle (1992Dingle ( , 1993Dingle ( , 1994)), Dingle and Honigstein (1994) Martens et al. (1996) for Ostracoda, and mainly on Lowry (1987) and Schmidt-Sinns (2008) for Foraminifera.Relative abundances (percentages) are calculated for samples with at least 30 specimens to keep a large number for statistical analysis; samples contain about 100 specimens on average.Statistical analysis was carried out using the program package PAST (Hammer et al., 2001).A principal component analysis (PCA) was used for identifying main factors structuring the changing association composition within the core (Fig. S4).Association data are composed of relative abundance data grouping selected species on the genus level if ecologically reliable.Only taxa occurring in at least two samples and with a proportion of more than 5 % in at least one sample are considered.Before running the PCA, a Spearman rank correlation was applied for identifying highly correlating taxa, which was not detected.

Heavy mineralogy
We have studied the mineralogy of three point-counted samples (from down-core depths of 11 and 285 cm and from catchment sample ECT2-1) at the Laboratory for Provenance Studies, University of Milano-Bicocca.From a quartered aliquot of each bulk sample (5-10 g), sediments were wet sieved with a standard 500 µm sieve in steel and with handmade special tissue net sieves of 15 µm.The fraction > 500 µm and < 15 µm were dried after sieving and weighed for a quantitative estimation of each granulometric class.Heavy minerals were after separated by centrifuging in sodium polytungstate (density 2.90 g cm −3 ) in the 15-500 µm size window and recovered by partial freezing with liquid nitrogen.Heavy minerals after separation were also weighed.An appropriate amount of heavy minerals was split and mounted with Canada balsam (n = 1.54), and 200 to 250 transparent heavy-mineral grains were point-counted at suitable regular spacing (100 µm) under a polarizing microscope to obtain real volume percentages (Galehouse, 1971).During point counting we have also studied surface textures on detrital grains by polarizing microscope to estimate chemical dissolution (Andò et al., 2012).Dubious grains were checked and properly identified by an inVia Renishaw Raman spectrometer equipped with a green laser (532 nm) and a 50×LWD objective, in the spectral range (144-4000 cm −1 ) referring to Andò and Garzanti (2013).Heavy-mineral concentration was calculated as the weight percentage of total heavy minerals (HMC) and transparent heavy minerals (tHMC).

Isotope geochemistry
The CPI of all samples ranged from 7.5 to 14.4, indicating generally fresh, hardly degraded material.Down-core as well as in the catchment samples (Fig. 4e, g; Fig. 5; Table 2; Table 3) the average relative contributions of the long-chain nalkanes are n-C 29 ∼ 15-23 %, n-C 31 ∼ 40-50 %, and n-C 33 ∼ 20-24 %.Together, these compounds accounted for 80-90 % of the total n-alkanes.The most abundant n-alkane is the n-C 31 alkane.Due to the consistent and strong predominance of the n-C 31 alkane further discussion is focused on this compound.δ 13 C analyses of the n-C 31 alkane revealed only minor differences (−28.5 to −26.7 ‰ VPDB).The precision is 0.1 ‰ on average with a maximum of 0.3 ‰.Similarly, the Norm31 showed values varying only minimally from 0.75 to 0.82.Larger differences (ranging from −143 to −127 ‰) could be detected in the δD composition of the n-C 31 alkane.Average precision is 1 ‰.

Heavy mineralogy
In the studied samples (Fig. S5), heavy-mineral concentrations in silt and sand fraction (15-500 m) range from very poor (< 0.5 wt % HMC) or poor (0.5 wt % ≤ HMC < 1 wt %) to moderately rich (2 HMC).In Gouritz River catchment sample ECT2-1, clinopyroxenes dominate with subordinate ultrastable zircon, tourmaline, rutile (ZTR), apatite, garnet and titanite; common amphibole; and rare orthopyroxene.The sample, from 11.5 cm depth in core GeoB18308-1, has a very similar heavy-mineral assemblage with slightly lower content of clinopyroxene and more epidote, very rare chloritoid and orthopyroxene.Red flowers of hematite are detected and corrosion features of clinopyroxenes are similar to the Gourtiz River assemblage.The sample, from 285 cm depth in Mossel Bay core GeoB18308-1, is very similar in composition to the Gouritz River but hematite is extremely rare.

Microfossil distribution and PCA results
A diverse foraminifer (at least 46 species) and ostracod (60 species) fauna characteristic for a sublittoral environment was found in the studied core.Beside skeletons of marine invertebrates, continental taxa like plant remains, charophyte oospores, insects, fruits and seeds occur as well, but in low numbers.A continental input is also reflected by six freshwater and four brackish water ostracod taxa occurring in several depths but in rather low numbers.Results of the loading plot analysis of the first principal component show that freshwater ostracods, charophytes, and fruits and seeds coming from continental waters or even of terrestrial origin are associated with high PC1 values.Microfossil results can thus be summarized in an index for estuarian inflow (based on PC1, which is best explained by fluvial input) (Fig. S4).

Down-core variations
Results of AMS-14 C determination are presented in Table 1.The basal age of core GeoB18308-1 is ∼ 4100 cal yr BP (Fig. 3).The continuous high-resolution, XRF scanning dataset (Fig. S2) indicates little variability in the lower half of core GeoB18308-1.However, a time interval of abrupt increase in Fe, clay and silt content (up to 55 %) with more estuarine-inflow-related microfossils (increased PC1), higher BIT (from ∼ 0.06 to 0.81) and enriched δD values of the n-C 31 alkane (up to −127 ‰) can be observed during the period between 950 and 650 cal yr BP (Fig. 4a-c, e-f).These trends are accompanied by a slight depletion in δ 13 C of the n-C 31 alkane (up to −28.5 ‰) and to generally lower but highly variable TEX H 86 -based (from ∼ 0.49 to 0.41) SST values (12.2 to 17.1 • C) (Fig. 4d, g).

Linking catchment depositional processes to rainfall regimes
The catchment samples analyzed for plant wax isotopic composition were all taken at lowland locations.However, samples from horizons of darker, finer material identified as soils show distinct δD C 3 1 differences (∼ 10 ‰ VSMOW) relative to samples identified as flood deposits by their lighter, coarser facies (Fig. 5).As hydrogen used for biosynthesis of plant waxes originates directly from the water taken up by the plants, isotope changes measured in these compounds are related to isotope composition of precipitation (Sessions et al., 1999).This indicates that the plant waxes contained in soil horizons were synthesized under conditions distinct from those under which plants waxes contained in flood deposits were synthesized.δD signatures can also give an indication of the respective conditions; in the Gouritz River catchment the extreme elevation difference is most likely the main influence on rainfall δD signatures which becomes deuterium-depleted with altitude (ca.10-15 ‰ per 1000 m; Gonfiantini et al., 2001).The relatively deuterium-depleted flood deposits (δD C 31 values of ∼ −138 ‰ vs. ∼ −127 ‰ in soil horizons) indicate that all analyzed flood deposits contain a considerable amount of upper-catchment material.
Despite the dominant summer rains in most of the catchment area (Le Maitre et al., 2009), SHW-related precipitation events in the otherwise arid winters have been described as the main precipitation signal influencing the Seweweekspoort record in the upper most Gouritz River catchment by Chase et al. (2013).Since plant material synthesized in the upper part of the catchment characterizes the δD signature of paleoflood deposits sampled in the lowlands, we suggest that these winterly SHW related precipitation events in the headlands are a main cause for large Gouritz River flood events.

Sediment provenance indicators
Leaf wax n-alkane distributions and δ 13 C C 31 Different vegetation types show variations in the n-alkane distribution of their leaf waxes.In general, it is thought that plants adapted to higher aridity produce longer-chain wax components than those in habitats of temperate regions (Gagosian and Peltzer, 1986).Therefore, the distribution of n-alkane chain length is widely used as an environmental proxy and the n-alkane distribution ratio "Norm31" can indicate changes of the source area (Carr et al., 2014).The stable carbon isotopic composition of organic matter reflects the isotopic composition of the carbon source as well as the discrimination between 12 C and 13 C during biosynthesis (Collister et al., 1994).In particular, compound-specific δ 13 C of long-chained n-alkanes shows variations with changes in vegetation type (Collister et al., 1994).In this study, the n-C 31 alkane is the most abundant n-alkane of the plant waxes and thus used as an indicator of vegetation type in addition to Norm31 ratios.Carr et al. (2014) show vegetation-specific distributions of n-alkane homologues within the arid-zone South African flora (Fig. 2).In the case of Gouritz River catchment samples and GeoB18308-1 down-core samples, plant waxes indicate a dominant Karoo vegetation signature according to Carr et al. (2014) and Hermann et al. ( 2016) (Fig. 2).This vegetation type is dominant in the northern parts of the catchment area (see map in Fig. 2).The flood deposits, soil and suspension load samples analyzed in this study, however, indicate that even in the lower catchment area the Karoo vegetation signature is dominant in the flood as well as in the soil deposits (Table 2, Fig. 2).This may be attributed to (a) difficulties in attributing vegetation types to n-alkane distributions caused by crassulacean acid metabolism (CAM) plants existing in South Africa's southernmost vegetation (Boom et al., 2014)  Low-resolution records are marked by colored dots.SSTs for samples with BIT above 0.3 were not calculated.The indicator of estuarian inflow is deduced from factor loadings along axes 1 of a PCA on microfossil distribution of core samples.For comparative purposes regional paleoenvironmental records are plotted: (h) aridity index from Seweweekspoort (Chase et al., 2013) and (i) indirect humidity index from Groenvlei (Wündsch et al., 2016).
beyond the scope of this study.Instead, n-alkane distributions and their isotopic values are used as provenance indicators.In the Gouritz River catchment as well as in downcore samples both indicators of vegetation type have similar signatures (identical average Norm31 values of 0.79  2).We therefore infer that the sediments deposited at the GeoB18308 site originate directly from the Gouritz River catchment area.

Heavy mineralogy
Heavy-mineral analysis represents an independent powerful tool in provenance studies.In the Gouritz River catchment as well as in samples from GeoB18308-1 clinopyroxenes dominate with subordinate ultrastable zircon, tourmaline, rutile (ZTR), apatite, garnet and titanite, common amphibole and rare orthopyroxene.The significant contribution of ultrastable ZTR to the heavy-mineral assemblages of GeoB18308-1 samples reflects the Gouritz River Cape Supergroup sandstone-dominated geology.The abundance of corroded clinopyroxene indicates weathering of doleritic dikes located only in the uppermost parts of the Gouritz River catchment.Although mineral and organic loads do not necessarily derive from the same source area, when used as provenance indicators both suggest a (upper) Gouritz River catchment provenance for the sediments deposited at the GeoB18308-1 site.
5.2 GeoB18308-1 paleoclimate record 5.2.1 ∼ 4000-950 cal yr BP -stable arid conditions The oldest part of the 18308-1 paleorecord (∼ 4058-950 cal yr BP) is characterized by a relatively low Fe/Ca ratio and a domination of the sand size fraction (mud and silt content below 20-30 %).We associate this with winnowing of fine-grained material by the strong current field inherent to the coastal Agulhas Bank.Heavy-mineral analyses from this part of the core (285 cm) show that hematite is extremely rare, suggesting a dry period without formation of iron oxides in the catchment area.Further afield in South Africa, pollen data from Lake Eteza (South African east coast) reported relative dryness for this period.This was deduced from evidence for decreasing trees and shrubs vegetation accompanied by increasing herbaceous plants (Neumann et al., 2010;Scott et al., 2012).The high-resolution XRF records indicate little variability in the composition of the supplied sediments which is in accordance with the relatively stable conditions recorded in the millennia prior to ∼ 1000 cal yr BP in the speleothem layer width from the Cold Air Cave (Holmgren et al., 1999) as well as pollen data from Wonderkrater and Rietvlei Dam (Scott et al., 2012) in southeastern Africa.

∼ 950-650 cal yr BP -Medieval Climate Anomaly
At ∼ 950 cal yr BP a shift occurred to slightly lower SSTs (Fig. 4d) in the study area.At the same time, fluvial deposition became more dominant in the record: there is a strong increase in terrigenous input described by a higher index of fluvial input based on microfossil composition, a higher Fe content, an increased BIT (from ∼ 0.1 to 0.3) as well as a 2fold increase in clay and silt content (to 40-60 %) (Fig. 4a-c,  f).This increase in terrigenous input may either indicate a decrease in Agulhas strength, an increase in sediment delivery via the Gouritz River or a combination of both.Simultaneously a slight shift (0.5 ‰) towards lower average δ 13 C C 31 values in the sediment core can be observed.Heavier isotope 13 C is more depleted in C 3 plants, which are less adapted to aridity as opposed to C 4 (Collister et al., 1994).However, the observed shift is very minor and a change in vegetation type is unlikely to occur within the timescale in question.It is more probable that the δ 13 C values became more depleted as the water-use efficiency of plants decreased in moister climatic conditions (Pate, 2001;Ehleringer and Cooper, 1988) gel, 1992;Holmgren et al., 1999;Thomas and Shaw, 2002) as well as Wonderwerk Cave (Brook et al., 2015), Braamhoek Wetland, Free State (Norström et al., 2009(Norström et al., , 2014)), Lake Eteza, coastal KwaZulu-Natal (Neumann et al., 2010), Lake Sibaya, KwaZulu-Natal (Stager et al., 2013), Blydefontein Basin, Kikvorsberge (Scott et al., 2005), Katbakkies Pass, Swartruggens Mountains, southwestern Cape (Chase et al., 2015) and northeastern South African baobab trees (Woodborne et al., 2015).The few existing records of the YRZ record similar trends; a continuous rise in precipitation was found in δ 15 N of hyrax middens at Seweweekspoort (Chase et al., 2013) and more evidently in the indirect (TOC-based) humidity record at Groenvlei -a Wilderness lake (Wündsch et al., 2016) (Fig. 4h-i).In contrast to the large array of available continental datasets, marine records are rare.The only SST record published for the area (Cohen and Tyson, 1995) does not include data for the time period in question.However, it does provide a conceptual model of oceanatmosphere interplay to be tested (see Fig. 6).The authors postulate a periodically strengthened and more southerly South Indian Ocean anticyclone and South Atlantic anticyclone having an onshore as well as an offshore effect: (1) reinforcing the tropical easterly influence causing extended warm wet spells in the South African summer rainfall zone and (2) increasing the frequency of eastward ridging highs and, thus, alongshore winds driving the coastal upwelling on the eastern Agulhas Bank.terval of increased humidity in the Gouritz River catchment inferred in this study for the time interval of the Medieval Climate Anomaly serves as data to validate the conceptual model by Cohen and Tyson (1995) for which thus far no Medieval Climate Anomaly evidence by SST data had been available.

Conditions after ∼ 650 cal yr BP -Little Ice Age and beyond
Continuous sedimentation at the core site was interrupted at ∼ 650 cal yr BP.One or more erosive event deposits are inferred from the sedimentology (erosive contact at ∼ 60-70 cm; fine sand with intercalated organic layers and lumps from ∼ 30 to 60 cm) and an age reversal indicated by the two radiocarbon dates in this interval (∼ 1466 cal yr BP at a depth of 31 cm ∼ 640 cal yr BP at a depth of 60 cm; Fig. 3; Table 1).Due to the discontinuous nature of the deposition we are only able to curtail the timeframe of deposition to having taken place between ∼ 650 cal yr BP (the youngest age in the event deposit) and a post-bomb date ∼ 13.5 cm above the event deposit.From the redeposited sediment package three samples have been analyzed for organic geochemistry (see Table 3).The average values over the possible timeframe of deposition is plotted in Fig. 5.The high BIT (∼ 0.7) indicates that the redeposited package can be characterized as reworked soil material.The averaged δD C 31 signature of ∼ −135 ‰ is comparable to that of Gouritz River paleoflood deposits described in Sect.5.1.1.We therefore suggest that the origin of the event-deposited material in core GeoB18308-1 is similar to the origin of these terrestrial paleoflood deposits, i.e., the upper parts of the Gouritz catchment (see Sect. 5.1.1).The shift in δ 13 C C 31 towards slightly more depleted values in the event-deposited material in core GeoB18308-1 (average in the redeposited unit: ∼ −28 ‰ VPDB) furthermore indicates that the n-C 31 alkanes contained in the event deposit were produced by plants under less water stress (see Collister et al., 1994;Ehleringer and Cooper, 1988) than those deposited before ∼ 650 cal yr BP.We therefore infer an increase in upper catchment rainfall inducing floods for the time period of the event deposit(s) (∼ 300-650 cal yr BP).This roughly falls into the timeframe of the Little Ice Age (LIA) recorded as humid phase throughout the South African winter rainfall zone (Meadows et al., 1996;Benito et al., 2011;Stager et al., 2012;Weldeab et al., 2013).Most regional studies associate this increase in rainfall with a northward shift of the SHW (Tyson and Preston-Whyte, 2000;Chase and Meadows, 2007).A global study using high-resolution records (Lechleitner et al., 2017) also documents a southward shift of the ITCZ during the LIA interval; however, no data from the southern hemispheric midlatitudes are included and we suggest that the effects of this shifts are limited to locations north of our study site.In the uppermost Gouritz catchment (Seweweekspoort site) a major SHW-sourced rainfall regime has been documented (Chase et al., 2015).Desmet and Cowling (1999) indicate that despite the general summer rainfall zone regime in the Gouritz catchment, the SHW supplies additional rainfall in extreme events.We suggest that an increase in these extreme SHW-sourced rainfall events produced large floods during the LIA (∼ 300-650 cal yr BP).After ∼ 650 cal yr BP continuous sedimentation is re-established with sediment properties returning to pre-hiatus conditions.The return to "normal conditions" recorded at the GeoB18308-1 site after the LIA time interval (i.e., after ∼ 300 cal yr BP) does not indicate any recent major Gouritz River flood events (e.g., the 1983 mega-flood documented by Damm and Hagedorn, 2010), underlining the magnitude of the LIA events.

Conclusions
The unique position of the core site directly offshore of the Gouritz River mouth at a central location on the South African south coast allows for a combined analysis of variability in marine processes as well as terrestrial input at a high temporal resolution.In order to reliably reconstruct terrestrial climatic change from an offshore record, measurements were performed down-core as well as on material from the Gouritz River catchment for ground truthing in a source to sink approach.Samples from the river and its flood deposits reveal that the sediments at our core site predominantly originate from the Gouritz River catchment.Furthermore, they indicate that paleoflood deposits are sourced from heavy rainfall events in the highlands, whereas soil formation pro-cesses are often linked to regular rainfall in the lowlands.The GeoB18308-1 sediment sequence records stable conditions on land and offshore prior to the Medieval Climate Anomaly (∼ 950-650 cal yr BP), which in turn has been interpreted as a humid interval in the (lowland) regions of the Gouritz River catchment.The subsequent Little Ice Age (∼ 300-650 cal yr BP) interval is characterized by major flood events attributed to (SHW) storm tracks influencing the upper catchment area.These paleoclimatic reconstructions correspond to available continental records for the region.This first regional SST dataset for the 2400-650 cal yr BP timeframe allows us to test the conceptual model of oceanic-atmospheric interplay (Cohen and Tyson, 1995) during short-term late Holocene climatic anomalies.In accordance with this model (see Fig. 6), our results indicate that variability in South Indian Ocean anticyclone location and strength drives both tropical easterly influence on South African summer rainfall zone climate and coastal upwelling on the eastern Agulhas Bank.During a poleward displacement and strengthening of the SIA (e.g., MCA situation) the stronger easterly (alongshore) winds cause the coastal upwelling, which is common during summer on the eastern Agulhas Bank, to become more frequent.

Figure 2 .
Figure 2. The map is an enlargement of Fig. 1 and shows vegetation types in and around the drainage basin (after Mucina and Rutherford, 2006 and Scott et al., 2012) of the Gouritz River.Squares indicate localities of Gouritz River samples; diamond indicates marine sediment core location.Plots showing regional biome/ecoregion soil and vegetation average n-alkane distributions from Carr et al. (2015) are indicated on the left, whereas average n-alkane distributions of riverine and marine samples from this study are plotted above the map.

Figure 3 .
Figure 3. Left: core log of GeoB18308-1 (internal structures with recorded colors) demonstrating the presence of an event deposit (blue box).Right: the calibrated 14 C dates (transparent blue) and the age-depth model (darker greys indicate more likely calendar ages; grey stippled lines show 95 % confidence intervals; red curve shows single "best" model based on the weighed mean age for each depth).The green box marks horizons of increased terrestrial input where a terrestrial calibration curve (SHCal13: Southern Hemisphere; Hogg et al., 2013) was used.Otherwise, Marine13-modeled ocean average (Reimer et al., 2013) with a δR following Dewar et al. (2012).Redeposited material (i.e., a crab claw at 123 cm depth, a bivalve at 285 cm and the reworked package at 30-60 cm depth) were excluded from the age-depth modeling.Medieval Climate Anomaly (MCA) and the Little Ice Age (LIA) are indicated.
, (b) the presence of succulent and Karoo plants in areas classified as "Fynbos biome", and (c) an overprint of the lower catchment signature by depositional material originating from the upper catchment.Understanding these processes in detail is www.clim-past.net/13/649/2017/Clim.Past, 13, 649-665, 2017 δD of n-alk ane C31 in [‰ VSMOW]

Figure 4 .
Figure 4. Organic and inorganic down-core geochemistry of GeoB18308-1 (a-g) as indicators for fluvial input, SST, SRZ and aridity.Low-resolution records are marked by colored dots.SSTs for samples with BIT above 0.3 were not calculated.The indicator of estuarian inflow is deduced from factor loadings along axes 1 of a PCA on microfossil distribution of core samples.For comparative purposes regional paleoenvironmental records are plotted: (h) aridity index from Seweweekspoort(Chase et al., 2013) and (i) indirect humidity index from Groenvlei(Wündsch et al., 2016).

Figure 5 .
Figure 5. Variations in δD of the n-C 31 -alkane (‰ VSMOW) in the distinct horizons of paleoflood vs. soil formation horizons.Gouritz River tributary cut bank at sample location ECT-3 is depicted as an example illustrating these alternating horizons.
was applied.
* Redeposited material -not included in age model.

Table 2 .
n-Alkane isotopic composition and distribution descriptive parameters averaged (with standard deviation).

Table 4 .
Mineralogical composition of heavy minerals in silt and sand fraction (15-500 m) of marine sediments from GeoB18308-1 and Gouritz River catchment.
Nash et al. (2016) Whyte (2000) of n-C 31 alkanes exported from the catchment suggest a shift towards more humid conditions on land after 950 cal yr BP.At the same time, δD values of the leaf wax n-C 31 alkane shift to −129 ‰ indicating deuteriumenriched precipitation during this time.This likely indicates a shift to lower-altitude source regions.Enriched δD C 31 values would be an indication for plant material mainly de-rived from more vegetated lowlands as opposed to plant material derived from the upper catchment during major flood events (see Sect. 5.1.1).Within the error margin of our age model, the shift we see at ∼ 950 ± 120 cal yr BP towards a more humid lower Gouritz River catchment is a general eastern South African expression of the northern hemispheric trend termed Medieval Climate Anomaly (MCA) summarized byTyson and Lindesay (1992),Tyson and Preston- Whyte (2000)andNash et al. (2016).A large array of continental records document this humid period throughout the South African summer rainfall zone (SRZ) (Talma and Vo-