A distal 140 kyr sediment record of Nile discharge and East African monsoon variability

Clay mineral assemblages in a sediment core from the distal Nile discharge plume off Israel have been used to reconstruct the late Quaternary Nile sediment discharge into the eastern Mediterranean Sea (EMS). The record spans the last ca. 140 kyr. Smectite abundances indicate the influence of the Blue Nile and the Atbara River that have their headwaters in the volcanic rocks of the Ethiopian Highlands. Kaolinite abundances indicate the influence of wadis, which contribute periodically to the suspension load of the Nile. Due to the geographical position, the climate and the sedimentary framework of the EMS is controlled by two climate systems. The long-term climate regime was governed by the African monsoon that caused major African humid periods (AHPs) with enhanced sediment discharge at 132 to < 126 (AHP 5), 116 to 99 (AHP4), and 89 to 77 ka (AHP3). They lasted much longer than the formation of the related sapropel layers S5 (> 2 kyr), S4 (3.5 kyr), and S3 (5 kyr). During the last glacial period (Marine Isotope Stages (MISs) 4–2), the long-term changes in the monsoonal system were superimposed by millennial-scale changes in an intensified midlatitude glacial system. This climate regime caused short but pronounced drought periods in the Nile catchment, which are linked to Heinrich events and alternate with more humid interstadials. The clay mineral record further implies that feedback mechanisms between vegetation cover and sediment discharge of the Nile are detectable but of minor importance for the sedimentary record in the southeastern Mediterranean Sea during the investigated African humid periods.


Introduction
The Nile in northern Africa is the longest river of the world and the dominant sediment source for the Eastern Mediterranean Sea (Milliman and Syvitski, 1992). Its drainage basin is about 3.2 × 10 6 km 2 and extends from the equator to ca. 30 • N (Fig. 1). The known about timing and intensity of former AHPs. Marine proxy data across sapropel S5 suggest that during the Eemian period (MIS 5e) the African monsoon was particularly strong and the African rain belt was shifted even further north than during the early Holocene. This resulted in flooding into the EMS through wadi systems along the wider North African margin (Rohling et al., 2002a;Osborne et al., 2008).
Different opinions exist on whether Nile sediment discharge is primarily controlled by precipitation and runoff, and thus closely linked to changes in the monsoon intensity (Wehausen and Brumsack, 2000;Emeis et al., 2000;Revel et al., 2010), or whether feedback mechanisms play an important role, with vegetation protecting soils and weathering products in the source area from being eroded (Adamson et al., 1980;15 Krom et al., 2002;Box et al., 2011).
A comprehensive effort to reconstruct Nile discharge during the last 100 ka was undertaken using the Fe content in sediments recovered from the proximal Nile delta Caley et al., 2011). Most other geochemical investigations concentrated on shorter time spans and used different parameters such as K/Al, Mg/Al, Ti/Al 20 to characterise the Nile discharge (Wehausen and Brumsack, 2000;Box et al., 2011;Weldeab et al., 2014). Several studies include the isotopic fingerprint of Nile sediments to reconstruct the Nile history (Krom et al., 1999a;Weldeab et al., 2002;Box et al., 2011;Blanchet et al., 2014).
We choose a different approach of reconstructing the Nile River sediment supply 25 throughout the last some 145 ka by mainly investigating the content of typical Nilederived clay minerals in sediment core GeoTü SL110, which was recovered from the distal Nile discharge plume off Israel (Fig. 1) Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 1999b), and is characterised by a high smectite content (e.g. Venkatarathnam and Ryan, 1971;Foucault and Mélières, 2000). By analysing a distal core from the plume, we document the Nile discharge activity as a whole rather than the activity of only a single channel in the delta.

5
The investigated sediment core GeoTü SL110 was recovered from the SE Levantine Sea (Fig. 1, Table 1) during cruise M51/3 of the German research vessel " Meteor" in 2001(Hemleben, 2002. It was retrieved at the Israeli continental slope off Haifa from a water depth of 1437 m. The sediments of the 652 cm long core mainly consist of muds with traces of pteropods and foraminifers. The core top contains basically undisturbed 10 surface sediments as indicated by the presence of an oxidised layer. The muds show frequent diffuse colour changes, mainly between grey and olive grey. Darker sapropel layers (greenish black, dark greenish grey, dark grey) occur at 35-61 cm (S1, with a possible interruption at 44-45 cm), 500-514 cm (S3), 558-570 cm (S4) and 598-622 cm (S5, including pre-sapropelic layer). 15

Methods
The lightness (L * ) of core SL110 was determined in 1 cm steps with a Minolta colour spectrophotometer immediately after opening of the core. The content of total organic carbon (TOC) was measured on 174 ground bulk sediment samples with an Eltra METALYST-CS-1000-S after removal of carbonate with 20 HCl.
The core was sampled in 1 cm intervals for investigations of the grain size composition of the terrigenous sediment fraction and of the clay mineral assemblages. The samples were oxidized and disaggregated by means of 5 % hydrogen peroxide. Carbonate was dissolved by 10 % acetic acid. We isolated the terrigenous sand fraction by Introduction sieving the samples through a 63 µm mesh. The silt fraction (2-63 µm) was separated from the clay fraction (< 2 µm) in settling tubes. The analyses of the clay mineral composition followed standard methods (e.g. Ehrmann et al., 2007). We added MoS 2 as an internal standard to the clay suspension. Texturally oriented clay mounts were solvated with ethylene-glycol vapour at a 5 temperature of 60 • C and then X rayed with a Rigaku MiniFlex system (CoKα radiation; 30 kV; 15 mA). We analysed the samples in the range 3-40 • 2Θ with a step size of 0.02 • 2Θ and a measuring time of 2 s step −1 . In addition, we analysed the range 27.5-30.6 • 2Θ with a step size of 0.01 • 2Θ and a measuring time of 4 s step −1 in order to better resolve the (002) peak of kaolinite and the (004) peak of chlorite. We evaluated 10 the diffractograms by using the MacDiff software (Petschick 2001). After adjusting the diffractograms to the MoS 2 peak at 6.15 Å, we deconvoluted the peak doublets smectite/chlorite (17/14 Å), palygorskite/illite (10.5/10 Å) and kaolinite/chlorite (3.58/3.54 Å). We based the semi-quantitative estimations of the clay mineral abundances on the integrated peak areas of the individual peaks and weighting factors (Biscaye, 1964;15 1965). For palygorskite, we used the same factor as for chlorite and kaolinite.

Age Model
For establishing a stable oxygen isotope stratigraphy we analysed tests from the 250-355 µm size fraction of the surface-dwelling foraminifer species Globigerinoides ruber (white) of about 130 samples using a Kiel IV online carbonate preparation line con-20 nected to a MAT 253 mass spectrometer. The values are reported in ‰ relative to VPDB using the delta notation. The reproducibility was better than ±0.06 ‰ (1σ shells of planktonic foraminifera (G. ruber, G. bulloides, G. sacculifer, O, universa) that represent the age of the surface waters. We applied an eastern Levantine Sea Delta-R of 3 ±66 years (Marine Reservoir Database). The radiocarbon ages were converted to calendar years using the Marine13 data base (Reimer et al., 2013; Table 2). We identified a hiatus at 598 cm, at the top of the S5 interval, based on several ar-5 guments. The lower part of the dark interval between 622 and 604 cm is characterised by only about 0.8 % TOC (Fig. 3a), by the occurrence of a relatively diverse benthic foraminiferal assemblage containing G. orbicularis, G. translucens and C. laevigata, and by relatively high δ 18 O values (Fig. 3a). Despite the dark colour, this indicates a pre-sapropelic interval at the transition between Marine Isotope Stages (MIS) 6 and 10 5 rather than a full sapropel . The TOC concentrations start to rise to ca. 1.4 % at 603 cm. We therefore assume the base of sapropel S5 at 603 cm, with an age of 124.0 ka (Bar-Matthews et al., 2000;Rohling et al., 2002b;Morigi, 2009;Osborne et al., 2010). The TOC values drop abruptly to 0.5 % at 596 cm. Thus, the TOC concentrations remain below the typical values of 2-3 % in S1, S3, S4 (Fig. 3a) 15 and in S5 of nearby core KL83 . Also the analysis of the benthic foraminifera showed an extremely thin anoxic phase in SL110, compared to a 27 cm thick S5 in KL83 . Thus, most of S5 is missing in SL110 due to a hiatus that was probably caused by a slump as indicated by the sharp and curved lithological boundary at 598 cm. By comparing our δ 18 O record with standard curves 20 and the record of MD84-641 (

Results
The raw data of our investigations on sediment core GeoTü SL110 are stored in the PANGAEA data base at the Alfred Wegener Institute for Polar and Marine Research, Bremerhaven, Germany (http://doi.pangaea.de/10.1594/PANGAEA.848291) and are presented in Fig. 3, which also shows the positions of the sapropel layers S1, S3, The lightness data (L * , Fig. 3a) reflect subtle colour changes within the core. The 10 darkest intervals correlate with the sapropels (L * ca. 37). However, also large parts of the other muds are relatively dark (L * ca. 40-45). The lightest interval occurs at 625-640 cm, corresponding to MIS 6. TOC contents generally fluctuate between 0.4 and 0.7 %, but increase to 2.5-3 % in the sapropel layers S1, S3 and S4, and to 1.5 % in pre-sapropelic layer of S5. 15 The terrigenous sand content is generally < 4 %, the silt/clay ratio < 1.0. The data sets show the same general pattern ( Fig. 3a) with finer grained intervals around 0-60, 490-515, 550-570, 600-620 cm and enclose the sapropel layers. A distinct coarse interval occurs at 625-640 cm. Between 480 and 110 cm we observe an upward coarsening trend. The high silt/clay ratio at 103-102 cm reflects a thin silt layer.

20
The clay mineral assemblage of SL110 (Fig. 3b)  roughly resembling that of kaolinite. Palygorskite occurs in trace amounts throughout the core, but jumps to 4 % at 625-640 cm.

Discussion
According to our age model (Fig. 2, Table 2), sediment core SL110 has an extrapolated basal age of ca. 145 ka. The linear sedimentation rates are 2.2 to 8.8 cm ka −1 . They are 5 low during late MIS 6 and MIS 5, and distinctly higher in MIS 3 to MIS 1. Sedimentation rates for nearby core MD84-641, calculated from data by Fontugne and Calvert (1992), show a similar pattern (Fig. 2), but are distinctly lower than in SL110 due to its further offshore position. The non-carbonate sediment fraction of seafloor surface sediments in the region 10 of SL110 is dominantly derived by long-distance sediment transport from the Nile (Venkatarathnam and Ryan, 1971;Stanley et al., 1997;1998;Krom et al., 1999b;Weldeab et al., 2002;Hamann et al., 2009). Storm waves and coastal currents erode coarse sediments from the delta. Geostrophic and wave-induced near-shore and shelf currents displace this material in a large counter-clockwise current along the coast to- 15 wards the Levantine Sea ( Fig. 1) and deposit it in the shallow waters off Israel. The fine fraction, in contrast, is derived from direct input from the Nile into the sea and may be transported also further offshore and further to the north (e.g. Stanley et al., 1997;1998;Venkatarathnam and Ryan, 1971;Foucault and Mélières, 2000). Aeolian influx from the Sahara and riverine supply of small Near East rivers and wadis are only of mi-20 nor importance (Stanley et al., 1997;Sandler and Herut, 2000). The enhanced glacial sedimentation rates observed in core SL110 might be a result of stronger currents and/or a shift of the currents and therewith of the high-accumulation areas to a more seaward position due to a lower sea level. These processes may also be responsible for the coarsening of the sediments as indicated by the terrigenous sand content and 25 the silt/clay ratios (Fig. 3a). The increasing glacial sedimentation rates theoretically also could be explained by increasing aridity resulting in an enhanced influx of aeolian dust. Endmember modelling of the terrigenous silt fraction of the nearby core SL112 (892 m water depth), which is located closer to the coast, indicates increased dust fluxes during the LGM and the late glacial (Hamann et al., 2008). However, such an influx is not documented in the clay 5 mineral record of SL110, e.g. by the concentration of wind-blown palygorskite (Fig. 3b). In addition, glacial sedimentation rates at site SL112 were lower when compared to the Holocene, thus supporting the interpretation of a glacial-interglacial shift of the highaccumulation zone.
Smectite is the typical clay mineral in Nile-derived sediments (Venkatarathnam and Ryan, 1971;Maldonado and Stanley, 1981;Stanley and Wingerath, 1996a;Foucault and Mélières, 2000). According to the compilation by Hamann et al. (2009), the Nile assemblage contains up to 85 % smectite. Kaolinite is the second important clay mineral with concentrations in the order of 20 %. Illite contents are < 10 %, and chlorite occurs in trace amounts. The Nile incorporates almost its entire suspension load close to its 15 source (Adamson et al., 1980). The smectite comes from the catchment of the Blue Nile and Atbara rivers and originates from weathering of Cenozoic volcanic rocks on the Ethiopian Plateau (Stanley and Wingerath, 1996a). Sediments of the Egyptian wadis discharging into the Nile have clay mineral assemblages with typically 40-55 % kaolinite, 35-45 % smectite and 5-15 % illite (Stanley and 20 Wingerath, 1996b;Hamann et al., 2009; chlorite not determined, thus, actual kaolinite concentrations may be somewhat lower). The kaolinite is mainly derived from the erosion of kaolinite-rich Eocene, Palaeocene and Mesozoic sediments and lateritic soils in the wadis (Stanley and Wingerath, 1996b;Bolle et al., 2000).
The composition of the aeolian clay mineral assemblage reaching the SE Levantine The clay mineral assemblage in the seafloor surface sediments of the SE Levantine Sea off southern Israel consists of up to > 70 % smectite, 10-25 % kaolinite, 10 % illite and < 10 % chlorite (Hamann et al., 2009) and thus reflects the Nile source. The uppermost samples of core SL110 with ca. 60 % smectite, 22 % kaolinite, 8 % illite and 9 % chlorite fit well into this distribution pattern.
According to the standard method, percentages of the clay minerals smectite, illite, chlorite, kaolinite and palygorskite add up to 100 %. A change in the abundance of one mineral therefore causes also changes in the concentration of the others. Hence we prefer to discuss clay mineral ratios. We use Sm r = smectite/(illite+chlorite) as the main proxy for Nile discharge provided by the Blue Nile and Atbara, and 10 Ka r = kaolinite/(illite+chlorite) as an indicator for the contribution of the wadis.
The good correlation between the Ka r and the Sm r data ( Fig. 4; r 2 = 0.73) confirms that the kaolinite in SL110 is of fluvial origin (Venkatarathnam and Ryan, 1971;Stanley and Wingerath, 1996a;Hamann et al., 2009), and not derived by aeolian influx from the Sahara, as in other regions of the EMS (e.g. Chester et al., 1977;Foucault and Mélières, 2000;Ehrmann et al., 2013). The correlation implies that during times of enhanced Nile runoff and smectite transport from Ethiopia, the erosion of kaolinite-rich sediments and soils in the wadis was generally also active, and that during dry periods in Ethiopia the wadis were dry, too. This confirms earlier findings based on the interpretation of satellite images and 20 sedimentological investigations, which showed that extensive drainage systems were active in the eastern Sahara during humid phases, especially of Eemian and Holocene time (e.g. Pachur and Hoelzmann, 2000;Rohling et al., 2002b;Osborne et al., 2008). Furthermore, oxygen isotope data from speleothems of Soreq Cave (Bar-Matthews et al., 2000) imply that rainfall extended beyond the Sahara. 25 The clay mineral data of SL110 document a highly variable smectite and kaolinite input through time (Figs. 3b, 4) and show similar distribution patterns as the Fe record from the Nile delta Caley et al., 2011). Fe is supposed to be derived from weathering of volcanic rocks in the Blue Nile headwaters in the Ethiopian highlands, and thus changing Fe contents may reflect changing precipitation regimes. Furthermore, the smectite and kaolinite distribution patterns are inversely correlated to the record of Saharan dust in the Aegean Sea (Ehrmann et al., 2013). The oldest part of the record displays minimum Sm r and Ka r values (Fig. 4) pointing to extremely dry conditions for the penultimate glacial maximum at ca. 142-132 ka of 5 MIS 6 (core interval 646-624 cm, age uncertainties > 135 ka because of extrapolation). The extreme drought possibly correlates with Heinrich Event 11, which is known as a very cold period in the Western Mediterranean Sea (Martrat et al., 2014;. Obviously, only minor amounts of Nile suspension load reached the coring site off Israel. Instead, maximum palygorskite concentrations (Fig. 3b) and high proportions 10 of terrigenous silt and sand (Fig. 3a) document an increased aeolian influx from the Sahara. This arid phase seems to have been more severe than that of the Last Glacial Maximum. After the maximum drought of MIS 6 the record of SL110 show a sharp transition to a sapropelic event embedded in an AHP.
It is well known that sapropel formation was coupled to precession-related max- 15 ima in northern hemisphere summer insolation causing a northward shift of the ITCZ, an intensification of the African monsoon system and enhanced precipitation in the Ethiopian highlands (e.g. Rossignol-Strick, 1983;Rohling, 1994;. In SL110, high Sm r , Ka r and low silt/clay ratios indicate phases with enhanced Nile and wadi discharge of fine-grained suspension to the Levantine 20 Sea accompanying formation of sapropels S5, S4, S3 and S1. The enhanced rainfall under a warm climate caused stronger chemical weathering of the volcanic rocks and thus smectite formation. The enhanced river runoff contributed to the stratification of the water column and facilitated the transport of nutrients and weathering products to the EMS promoting enhanced surface water productivity at least in areas close to the Introduction the Aegean Sea (Ehrmann et al., 2013) reveal that each phase of sapropel formation occurred within a longer AHP ( Fig. 4; AHPs numbered according to the corresponding sapropels). The Sm r maximum linked to AHP 5 shows a sharp onset after the glacial maximum of MIS 6, simultaneously with the start of the pre-sapropelic layer of S5 (Fig. 4). It indi-5 cates a northward moving rain belt resulting in enhanced monsoonal rainfall in Ethiopia and Blue Nile runoff as early as 132 ka at a time of increasing insolation but under the influence of penultimate glacial boundary conditions. Also the Ka r values clearly increase with the beginning of the pre-sapropelic layer. However, they do not exhibit such distinct maximum values as during the early phases of the younger AHPs. Most 10 of sapropel S5 and the termination of AHP 5, however, are not preserved in SL110 due to the hiatus at ca. 122-115 ka. Furthermore, because of the curved top of the sapropel and the technique of sampling the core in parallel slices, we did not receive the typical sapropel signature but a mixed sapropel/post-sapropel signal (Fig. 4).
According to the Fe and the dust record, AHP 4 lasted from > 105 ka to ca. 95 ka 15 Ehrmann et al., 2013). The Sm r and Ka r data of SL110 indicate that AHP 4 started during insolation rise at 113 ka and ended at 94 ka, with sapropel formation occurring at 105-99.5 ka (Fig. 4). AHP 3 started just before the insolation maximum and lasted ca. 86-74 ka based on the Sm r and Ka r maxima and the minima in the silt/clay ratio and the terrigenous sand 20 content. These ages are in good accordance with those obtained from the dust record in the Aegean Sea (Ehrmann et al., 2013). Sapropel S3 has a shorter duration and spans the time interval 82-77.5 ka.
Despite the hiatus in the SL110 record, by assuming a duration of 124-119 ka for S5 (Bar-Matthews et al., 2000;Rohling et al., 2002b;Morigi, 2009;Osborne et al., 2010), 25 our data indicate that the dry phase between AHP 5 and AHP 4 lasted only ca. 6 ka, and the one between AHP 4 and AHP 3 only ca. 8 ka. It is interesting to note that a similar duration of the dry phases can be deduced from the pollen record of Tenaghi Philippon in Greece, although this remote site is governed dominantly by northern climatic factors (Tzedakis, 2005;Pross et al., 2015). After a further short dry period at 74-68 ka, Sm r , Ka r and silt/clay ratios indicate humid phase AHP 2 and enhanced Nile runoff in early MIS 4 at 68-64 ka. It is of similar magnitude as the interglacial AHPs discussed before and also starts during increasing 5 insolation (Fig. 4). In contrast to the previous AHPs this one is not accompanied by a sapropel, dark sediment colour or enhanced TOC values. However, it coincides with the common occurrence of shallow infaunal Uvigerina peregrina. This highly opportunistic species is adapted to the deposition of high amounts of fresh phytodetritus (Koho et al., 2008). In analogy to nearby SL112, the occurrence of this species at site SL110 during 10 AHP 2 is likely fuelled by pronounced algal blooms related to seasonally enhanced Nile runoff and nutrient supply (Schmiedl et al., 2010). However, in core SL110 we have strong indications that AHP 2 did not culminate in a sapropel formation. Sediment and nutrient delivery by the Nile was cut-off by a drought linked to Heinrich Event 6 occurring just at the time of maximum insolation (Fig. 4). This drought, possibly combined with 15 stronger winds and better ventilation of the deep waters, reduced sediment discharge into the Mediterranean Sea and hampered stagnation of the deep water and sapropel formation.
Lourens et al. (1966) described a thin sapropel layer S2, corresponding to AHP 2, south of Crete that occurs after the 60 ka insolation maximum. It was argued that this 20 sapropel is not visible any more in most sedimentary sequences of the eastern Mediterranean Sea because of post-depositional oxidation of the organic matter (Emeis et al., 2000;Löwemark et al., 2006).
It is not understood why our clay mineral signal has no counterpart in the Fe signal from the Nile delta . Possibly Nile discharge to the Eastern Mediter- 25 ranean Sea at that time was not via the western Rosetta branch, but via the eastern Damietta branch, and thus site MS27PT starved of sediment. It also cannot be ruled out that a match is obscured because of different age models, insofar as the Fe 72 ka peak corresponds to our well-dated 68 ka clay mineral peak (Fig. 4) The slightly enhanced Sm r and Ka r values at 58-48 ka possibly document the late phase of AHP 2 after the termination of Heinrich Event 6, reflecting the re-occurrence of slightly enhanced Nile River runoff. Other indications for somewhat increased temperature, humidity and Nile influx during this time interval come from the Fe record in the Nile delta   The spread of deciduous and other trees during the same time period likely supported the migration of anatomically modern humans from Africa into the Levant (Müller et al., 2011). In the Sm r data of core SL110, this time interval documents only a moderate increase of Nile runoff (Fig. 4) suggesting that the observed warming and freshening of the eastern Mediterranean region was predominantly related to a high-latitude climate change. 15 Thus, the sediments of SL110 document that the long-term climate regime governed by the African monsoon and characterized by pronounced AHPs during the last interglacial period was replaced as of the occurrence of Heinrich Event 6 by a distinctly different glacial climate system governed by short-term, millennial-scale climate changes having their origin in the North Atlantic and Arctic regions. Former studies showed 20 that sudden freshwater influxes into the northern Atlantic Ocean during the Younger Dryas and the Heinrich Events not only weakened the thermohaline circulation and oceanic heat transport to the northern high latitudes (e.g. Bond et al., 1993), but also had major consequences for the Mediterranean Sea environment. Increased aridity and even North African mega-droughts during such events can be seen in marine 25 sediments of the North Atlantic Ocean (Mulitza et al., 2008), speleothem data and lake level drops of Lake Lisan (palaeo-Dead Sea; Bar-Matthews et al., 1999;Bartov et al., 2003) and of African lakes (Gasse et al., 2008;Shanahan et al., 2015). Cold water influx from the North Atlantic Ocean to the Mediterranean Sea intensified. Sea CPD 11,2015  surface temperatures and deep-water temperatures decreased and deep-water ventilation fortified in the northwestern Mediterranean Sea (Cacho et al., 1999; and probably also in the EMS. Enhanced northward flux of wind-transported sediment from the Sahara is documented for both the Western and the EMS (Moreno et al., 2002;Bout-Roumazeilles et al., 2007;Hamann et al., 2008). Strongly decreased Sm r and 5 Ka r values in core SL110 (Fig. 4) indicate reduced Nile sediment discharge during Heinrich Events, and therewith major droughts in the headwaters of the Blue Nile and diminished erosion and river runoff. During these events the ITCZ and the African Rain Belt moved southward. Both Lake Tana, the source of the Blue Nile, and Lake Victoria, the source of the White Nile, desiccated during the Younger Dryas and the Heinrich 10 Event 1 (Lamb et al., 2007;Marshall et al., 2011). Minimum Sm r and Ka r values as well as the occurrence of a silt layer of probably aeolian origin (Fig. 3a) imply that Heinrich Event 1 was the driest interval since the MIS 6 glacial maximum (cf. Stager et al., 2011). The 65 • N insolation curve shows a further low-amplitude maximum at ca. 38 ka. It is 15 also reflected by slightly enhanced smectite and kaolinite abundances in core SL110, ranging 38-30 ka and by the Fe record in the Nile delta . No sapropel formation was recorded following the weak insolation maximum at 38 ka. Thus, the two potential humid periods linked to the glacial insolation maxima at 60 and 38 ka are only vaguely expressed in the EMS, because their effects were much less distinct and 20 because they were suppressed by the glacial climate regime, particularly the impacts of cold and dry stadials. The clay mineral data of SL110 and of nearby core SL112 (Hamann et al., 2009) document a gradual onset of AHP 1 at about 15 ka (Fig. 4), which coincides with the overflow of Lake Tana, the source of the Blue Nile (Lamb et al., 2007;Marshall et 25 al., 2011). At the same time, Saharan dust influx to the EMS decreased (Ehrmann et al., 2013). S1 at ca. 11-6.5 ka and the peak Blue Nile runoff at 10-9 ka (Weldeab et al., 2014) are only poorly reflected in the clay mineral data. Precipitation generally declined gradually after sapropel formation throughout the Holocene, the Nile River runoff responded quasi-linearly to the changes in rainfall (Weldeab et al., 2014;Blanchet et al., 2014), and the AHP 1 ended at ca. 5 ka (deMenocal et al., 2000). However, the clay mineral data do not mirror this trend but show continued increasing Sm r and Ka r values and Nile sediment discharge until ca. 2 ka. Sediment discharge thus is decoupled from water runoff. Possibly, human activity altered the natural signal. Saharan population in-5 creased shortly after 11 ka and reached a maximum at 7.5 ka (Manning and Timpson, 2014). Major land-use changes occurred with the advent of Neolithic farming some 8 ka ago (Williams, 2009). The clay mineral and grain size data of SL110 indicate an enhanced riverine influx of suspension from the Blue Nile and Atbara to the Mediterranean Sea during the AHPs 10 and therefore document enhanced weathering of the Ethiopian basalts, erosion of soils and transport of the suspension load to the sea. This view is supported by several investigations (Wehausen and Brumsack, 1998;Emeis et al., 2000;Revel et al., 2010;Blanchet et al., 2014). However, it is in conflict with some other studies (Adamson et al., 1980;Krom et al., 1999a;2002;Box et al., 2011) that argued for a lower Blue Nile 15 sediment discharge during the warm and humid Holocene AHP and higher fluxes during dry climate phases. They reasoned that at present the limited vegetation cover on the Ethiopian highlands allows extensive erosion when the summer monsoon rains fall. During increased monsoon activity, in contrast, longer periods of rain, longer growing seasons and larger vegetation-covered areas reduced soil erosion.

20
Our data indicate that an increase in precipitation and Nile suspension delivery happened long before sapropel formation, during phases of increasing insolation. However, in the case of AHP 4 Sm r and Ka r maxima occur just before and after sapropel S4 formation, each accompanied by low silt/clay ratios also indicating enhanced riverine suspension (Fig. 3a). Thus, maximum riverine sediment influx does not coincide 25 with maximum water discharge during sapropel formation. This is possibly due to a feedback mechanism as postulated, for example, by Adamson et al. (1980) andKrom et al. (2002). Vegetation feedbacks were obviously not active in glacial time. climate changes linked to Dansgaard-Oeschger cycles and Heinrich Events find immediate response in the sediment delivery to the EMS. The different timing between the Saharan dust record and the Nile discharge record of the humid phases (Fig. 4) may reflect regional differences in the climatic history. The northward migrating tropical rain belt may first have reached the Ethiopian high-5 lands, where the smectite signals has its origin, and with a time lag the central Sahara, there inhibiting effective dust production. The differences may also be due to different processes being at work. The Nile discharge is mainly driven by annual floods, fluvial erosion and weathering. It therefore may have reacted immediately on a northward shift of the ITCZ and the rain belt. The dust record, in contrast, is mainly a function of aridity, wind strength and, probably most important, the availability of deflatable material. The spreading and retreat of a vegetation cover plays an important role and may react non-linearly and with a different time lag on a shift of the ITCZ than river discharge (Claussen et al., 2013;Ehrmann et al., 2013). Thus, recently Blanchet et al. (2014) could show for the end of AHP 1 that the vegetation retreat from the Sahara was much 15 faster than the decrease in Nile runoff during the southward migration of the rain belt.

Conclusions
-Clay minerals are suitable tools to reconstruct the late Quaternary Nile sediment discharge to the EMS. The abundance of smectite reflects the runoff of the Blue Nile and Atbara, whereas kaolinite abundance reflects the discharge from wadis.

20
The correlation of smectite and kaolinite abundances implies that during times of enhanced Nile suspension discharge the wadis also were active sediment sources.
-High smectite and kaolinite abundances and low terrigenous silt/clay ratios indicate phases of enhanced precipitation and suspension influx from the Blue Nile, (AHPs), which are accompanied by the formation of sapropel layers. This largely contradicts the view that sediment delivery during humid periods was reduced because an extensive vegetation cover hampered soil erosion.
-Precipitation and suspension discharge started much earlier and ended later than sapropel formation. Maximum sediment discharge does not coincide with maxi-5 mum water discharge during times of sapropel formation because of minor vegetation feedback.
-The time lag between the increase/decrease of Nile suspension discharge and the decrease/increase in the dust record (SL143) can be explained by a successive migration of the rain belt, possibly combined with a vegetation feedback.

10
-The dry periods between the interglacial AHPs were relatively short, each lasting only ca. 6-8 ka.
-The sedimentary record of core SL110 demonstrates the interplay of two competing climate systems. The long-term climate regime was governed by the African monsoon and characterized by pronounced AHPs during the last interglacial period. This system was replaced in the last glacial by short-term, millennial-scale climate changes having their origin in the North Atlantic region. They are characterised by short but pronounced drought periods linked to Heinrich Events and more humid interstadials. Precipitation in the Nile catchment, river runoff and sapropel formation can be hampered or even inhibited under the influence of the 20 North Atlantic climate system, particularly under glacial boundary conditions.

Data Availability
The raw data to this article is available at http://dx.