A detrital record of the Nile River and its catchment

This research uses analyses from Nile catchment rivers, wadis, dunes and bedrocks to constrain the geological history of NE Africa and document influences on the composition of sediment reaching the Nile delta. Our data show evolution of the North African crust, highlighting phases in the development of the Arabian–Nubian Shield and amalgamation of Gondwana in Neoproterozoic times. The Saharan Metacraton and Congo Craton in Uganda have a common history of crustal growth, with new crust formation at 3.0 – 3.5 Ga, and crustal melting at c. 2.7 Ga. The Hammamat Formation of the Arabian–Nubian Shield is locally derived and has a maximum depositional age of 635 Ma. By contrast, Phanerozoic sedimentary rocks are derived from more distant sources. The fine-grained (mud) bulk signature of the modern Nile is dominated by input from the Ethiopian Highlands, transported by the Blue Nile and Atbara rivers. Detrital zircons in the Nile trunk are predominantly derived from Phanerozoic cover rocks. Most detritus from the upstream White Nile is trapped in the Sudd marshes and contributes little to the Nile trunk. Therefore, the White Nile downstream is dominated by locally derived Phanerozoic cover. The White Nile proximal to the Gezira Fan is influenced by the fan's Blue Nile signature. Supplementary material: Sample information, analytical methods and data tables are available at https://doi.org/10.6084/m9.figshare.c.3569490

south-draining 'Qena System' dominating the Nile valley until the Messinian Salinity Crisis (Issawi & McCauley 1992). Sediment volume calculations of the delta cone have been used to infer that a connection with the Ethiopian Highlands did not occur until the Pleistocene (Macgregor 2012;Palacios 2013).
The Red Sea Hills, to the east of the present-day Nile trunk, consist of Arabian-Nubian Shield basement and Phanerozoic sedimentary rocks that consist of Palaeozoic and Mesozoic clastic rocks and Eocene carbonates, partially overlain by modern wadi sands. Although not drained by the Nile at present, the Western  Kazmin (1972), al-Miṣ rıȳah (1981, Ministry of Energy and Mines (1981) and Johnson (2014). Desert, to the west of the present-day Nile trunk, is thought by Issawi & McCauley (1992) to have once contributed detritus to the main Nile trunk, and aeolian contribution occurs today (Garzanti et al. 2015). The Western Desert comprises reworked pre-Neoproterozoic crust of the Saharan Metacraton (Abdelsalam et al. 2002), and overlying Phanerozoic sedimentary successions. Modern-day wadi sediment and aeolian dunes currently cover a large proportion of both these regions.

Geology of the Nile catchment
The geology of east Africa has largely been shaped by the events of the Pan-African Orogeny when East and West Gondwana collided to form 'Greater Gondwana' at the end of the Neoproterozoic. The Pan-African Orogeny in NE Africa involved the collision of Archaean cratons such as the Congo Craton and the Saharan Metacraton with the Arabian-Nubian Shield, a terrane comprising Neoproterozoic, juvenile, oceanic island arcs. Phanerozoic cover sedimentary rocks blanket much of NE Africa, reflecting erosion from this mountain belt and subsequent Phanerozoic tectonics. The most prominent Phanerozoic tectonic events in the region involve rifting associated with opening of Neotethys, and later inversion associated with Africa-Mediterranean convergence. Uplift of the Ethiopian Highlands and surrounding area, and the eruption of continental flood basalts, is thought to have had a strong influence on the hydrology of this area in the Cenozoic.

Palaeoproterozoic and Archaean cratons
The cratons of Central Africa are formed of various blocks of Archaean and Palaeoproterozoic crust, flanked or truncated by Palaeoproterozoic to Mesoproterozoic orogenic belts. Together, these form the Congo Craton, defined as the 'amalgamated central African landmass at the time of Gondwana assembly' (De Waele et al. 2008). The Congo Craton (Fig. 1) forms part of the West African Central Belt, which formed at 2.0-3.0 Ga and comprises orthogneiss, metasediments and granitoids of the Tanzania Craton, Gabon-Cameroon Shield, Bomu-Kibalian Shield, Kasai Shield and Angolan Shield (Cahen et al. 1984;Walraven & Rumvegeri 1993;Goodwin 1996;Tchameni et al. 2000;De Waele et al. 2008).
The term Saharan Metacraton refers to an area of pre-Neoproterozoic continental crust that has, in part, been highly remobilized during the Pan-African Orogeny (Abdelsalam et al. 2002). It extends from the Arabian-Nubian Shield in the east to the Tuareg Shield in the west and the Congo Craton in the south. More than 50% of the Saharan Metacraton is overlain by Phanerozoic sedimentary rocks and desert sands (Abdelsalam et al. 2011). The poor exposure of the region means that the Saharan Metacraton, and its relationship to adjacent blocks, is poorly understood. The southern boundary is not well defined, but is taken to be marked by the Oubanguides orogenic belt, which separates it from the Congo Craton (Abdelsalam et al. 2002). Little modern geochronology has been carried out on rocks of the Saharan Metacraton. Legacy Rb-Sr and U-Pb data quoted by Abdelsalam et al. (2002) indicate a range of late Archaean and Palaeoproterozoic protolith ages, with significant cratonic reworking and addition of new crust in the Neoproterozoic. Bea et al. (2011) reported sensitive high-resolution ion microprobe (SHRIMP) U-Pb zircon ages as old as 3.22 ± 0.04 Ga for gneisses in the Uweinat and Gebel Kamil regions of the Western Desert in southernmost Egypt.

The Arabian-Nubian Shield
The Arabian-Nubian Shield is a collage of Neoproterozoic (c. 870 -670 Ma) continental margin and juvenile island arcs, cut by voluminous granitoid intrusions as young as latest Neoproterozoic, and overlain by younger sedimentary and volcanic basins (Stern & Hedge 1985;Stern 1994;Johnson & Woldehaimanot 2003;Kusky & Matsah 2003). Only minor outcrops of pre-Neoproterozoic crust are found. Formation of the Arabian-Nubian Shield began with the initiation of subduction and development of magmatic arcs at 870 Ma. Terrane amalgamation associated with closure of the Mozambique Ocean and amalgamation of East and West Gondwana took place between c. 780 and c. 600 Ma (Johnson & Woldehaimanot 2003). To the west of the Red Sea, the oldest arc terranes (>800 Ma) of the Arabian-Nubian shield occur in the south, in Ethiopia, Eritrea and Sudan (Johnson & Woldehaimanot 2003). Ophiolitic rocks of the Eastern Desert of Egypt range in age from 810 to 720 Ma (Ali et al. 2010), and are overlain by younger volcanic and sedimentary sequences (e.g. Breitkreuz et al. 2010).
Ediacaran alluvial sedimentary rocks of the Hammamat Formation overlie the Arabian-Nubian Shield basement in the Eastern Desert of Egypt. The Hammamat Formation clastic rocks comprise terrestrial conglomerates, sandstones and mudstones in the Red Sea Hills area, and have proposed correlatives overlying the Midyan Terrane in NW Saudi Arabia (Bezenjani et al. 2014). Previous workers, using a variety of approaches (e.g. maximum depositional age determined from detrital minerals, dating of crosscutting igneous units), have proposed ages ranging from 630 to 583 Ma for these rocks in Egypt (Willis et al. 1988;Wilde & Youssef 2002;Bezenjani et al. 2014). The Hammamat Formation rocks' depositional environments, and their relationship to the Pan-African Orogeny is of some debate (e.g. Ries et al. 1983;Wilde & Youssef 2002;Abdeen & Greiling 2005;Eliwa et al. 2010;Johnson et al. 2011;Bezenjani et al. 2014); the sediments are arguably (as described in the above-cited studies and references therein) considered to be pre-collisional, syn-collisional or post-collisional, and either locally sourced and deposited in isolated basins, or of distant provenance deposited from a major fluvial system of continental proportions.

The Pan-African Orogeny
During the Pan-African Orogeny, the final closure of the Mozambique Ocean led to amalgamation of the Saharan Metacraton and Congo Craton with the Arabian-Nubian Shield and the formation of the 'Trans-Gondwanan Supermountains' in the region that now forms NE Africa (Squire et al. 2006;Meinhold et al. 2013). To the south, it has been proposed (Abdelsalam et al. 2002) that collision between the Saharan Metacraton and Congo Craton resulted in the formation of the Oubanguides orogenic belt or Central African Fold Belt.

Phanerozoic sedimentary rocks
Erosion of the Trans-Gondwanan mountain belt and subsequent recycling of the eroded detritus during later inversion tectonics associated with various plate movements resulted in the deposition of a thick cover of fluvial and marine sediments overlying the amalgamated terranes, from the Cambrian onwards. Sedimentation continued until Cenozoic times, influenced by periods of tectonism. The most important of these tectonic events is the Mesozoic opening of Neotethys and later the South Atlantic, which resulted in the development of rift basins in the region, with Late Cretaceous-Early Eocene inversion of such basins as a result of the convergence of the Mediterranean with the African margin (Klitzsch & Squyres 1990;Guiraud et al. 2005;Bosworth et al. 2008).
Previous studies of the cover sedimentary rocks in regions adjacent to the Nile catchment (i.e. Libya; Meinhold et al. 2011;Altumi et al. 2013), Jordan and Israel (Kolodner et al. 2006;Morag et al. 2011)) have put forward contrasting Phanerozoic palaeodrainage models for these regions during deposition of the cover.
The textural and mineralogical maturity of the bulk of the cover, and the provenance of the sediments, has been ascribed to varying degrees of recycling, long-distance transport and/or intense chemical weathering. Particularly debated is the origin of the c. 1000 Ma zircon population, for which an obvious significant basement source in the Saharan Metacraton or Arabian-Nubian Shield has not been recognized (Kolodner et al. 2006;Meinhold et al. 2011).

Cenozoic uplift and flood basalts
Uplift of the Red Sea Hills associated with opening of the Red Sea rift, and uplift in Ethiopia associated with eruption of voluminous continental flood basalts, has had a major influence on Nile drainage.
The continental flood basalts that dominate much of the Ethiopian Highlands are associated with East African rift-related magmatism and continental break-up (Ebinger 2005;Buck 2006). Pre-rift basaltic and silicic magmatism initiated around 31 Ma (Baker et al. 1996;Rochette et al. 1998;Ukstins et al. 2002). Magmatic upwelling resulted in the uplift, extensional stresses and faulting within what is now the Ethiopian Highlands.

The provenance signal of modern Nile River sediments
The provenance of sediments in the Nile River has previously been studied using heavy mineral distributions and petrography, Sr and Nd trace isotope studies, and U-Pb analysis of detrital zircons.
Sr and Nd ratios have proved useful when recording major changes in provenance along the course of the Nile River (Padoan et al. 2011;Garzanti et al. 2013). High 87 Sr/ 86 Sr ratios and low ε Nd values in White Nile muds from Archaean cratonic sources contrast with lower 87 Sr/ 86 Sr and higher ε Nd in the Sobat River, which receives most of its sediment load from the crystalline basement and Cenozoic volcanic rocks of the Ethiopian Highlands. Isotope signatures of the Blue Nile and Atbara rivers are dominated by Ethiopian volcanic detritus and show low 87 Sr/ 86 Sr ratios and higher ε Nd values.
Heavy mineral analysis and petrography have also been used to characterize the signature of the Nile catchment area (Garzanti et al. 2006). The Blue Nile, Atbara and Tekeze rivers drain predominantly flood basalts and rhyolitic ignimbrites in their upper courses until they reach the Neoproterozoic amphibolite-facies basement of the Arabian-Nubian Shield, consisting of granitoid gneisses, staurolitebearing schists and marbles (Tadesse et al. 2003), and Phanerozoic sedimentary rocks found in the pre-rift sedimentary succession of the Blue Nile Canyon. Despite both rivers draining similar lithologies in the Ethiopian Highlands, the Atbara contains significantly less quartz than the Blue Nile.
The Victoria Nile downstream of its Lake Victoria output carries feldspatho-quartzose sand, with feldspars derived from granitoid rocks exposed locally (Garzanti et al. 2006). Downstream of the Lake Albert outlet, Albert Nile sand is quartzose with few feldspars and poor heavy-mineral suites with epidote, hornblende, kyanite, rutile and zircon. Sediment composition changes across South Sudan, and White Nile sand of all grain sizes downstream of the Sudd marshes chiefly consists of monocrystalline quartz, becoming slightly enriched in plagioclase, volcanic lithic fragments and clinopyroxene when downstream of the Gezira Fan. This fan deposit is ultimately derived from the Ethiopian flood basalts via a Pleistocene overspill from the Blue Nile (Garzanti et al. 2006).
The Red Sea Hills lie adjacent to the Nile River through much of its course through Sudan and Egypt. However, Red Sea Hills derived sediment has not been considered as an end-member in modelling of the Sr and Nd isotope composition of sediments in the Nile delta cone (Krom et al. 2002;Revel et al. 2014). Although construction of the Aswan High Dam has clearly affected the trunk Nile sediment flux downstream, comparison of petrographic data from samples collected prior to and subsequent to construction of the dam indicates that sediment composition is minimally affected (Garzanti et al. 2015). Iizuka et al. (2013) presented coupled U-Pb, Lu-Hf and O-isotope data for detrital zircons from a sample of modern Nile sand taken 'Near Cairo City'. They identified age peaks at 1.1 -0.9, 0.85 -0.7 and 0.7 -0.55 Ga, with minor groups at c. 2.6 and 2.0 Ga. They noted that grains with ages between 0.85 and 0.7 Ga had positive ε Hf values, and were likely to have been derived from juvenile crust in the Arabian-Nubian Shield. Zircons with ages of 1.1 -0.9 and 0.7 -0.55 Ga, coinciding with times of supercontinent assembly, showed wide variations in O-isotope composition and ε Hf .
Be'eri-Shlevin et al. (2014) discussed U-Pb and Hf-isotope data for detrital zircons in Quaternary to Recent Israeli coastal sands, which are believed to have been derived by a combination of longshore drift and aeolian transport from the Nile delta. They concluded that the ubiquitous presence of 0.56 -0.75 Ga detrital zircons with negative ε Hf implies that the Arabian-Nubian Shield is not the main source of Nile sands. Rather, they believed that multiple recycling through Phanerozoic sedimentary rocks that blanketed North Africa explains the age and hafnium isotope composition of the detrital zircon populations, as well as the quartz budget of the system.

Analytical methods
Samples of medium sand and mud were taken from modern rivers, sedimentary bedrock and modern wadis and dunes. Sample locations are shown in Figure 1. A sample of gneissic basement was also collected from the Saharan Metacraton.
Zircon and rutile grains were separated using standard methods, then hand-picked and mounted in epoxy discs and polished to reveal their interiors. All zircon grains were imaged using cathodoluminescence (CL) prior to analysis. U-Pb analyses for both zircon and rutile were carried out at the NERC Isotope Geosciences Laboratory (NIGL), using a single-collector Nu-Attom mass spectrometer with one of three New Wave laser systems, typically using a 35 µm laser spot. Hafnium isotope composition of zircons was measured at NIGL using a Thermo-Electron Neptune Plus mass spectrometer, coupled to a New Wave 193UC or 193FX Excimer laser. A 50 µm spot was used, targeting previously dated zircon domains.
Plagioclase and white mica were separated from the light fraction remaining after zircon and rutile separation. 40 Ar/ 39 Ar analyses were carried out at SUERC, East Kilbride, using a GVi instruments ARGUS five-collector mass spectrometer using a variable sensitivity Faraday collector array in static collection (non-peak hopping) mode (Sparks et al. 2008;Mark et al. 2009).
Mud samples for Sr, Nd and Hf analysis were leached in 10% acetic acid to remove carbonate before spiking with 149 Sm-150 Nd, 176 Lu-180 Hf, 87 Rb and 84 Sr isotope tracers. Standard dissolution methods and ion-exchange chromatography were used to separate elements of interest. Sr and Nd isotope compositions were measured at NIGL on a Thermo-Electron Triton mass spectrometer using dynamic multi-collection. Hf isotope composition was analysed in static mode on a Thermo-Electron Neptune mass spectrometer coupled to a Cetac Aridus II desolvating nebulizer. X-ray fluorescence major and trace element analysis was carried out at the Open University in Milton Keynes using standard methods. Pressed powder pellets were used for trace element measurements, and fused discs for major elements.
Petrographic analysis and heavy mineral analysis of sands and muds were carried out at the Universita di Milano-Bicocca using methods modified from Garzanti et al. (2006) and Garzanti & Andò (2007). Assemblages were described in terms of transparent Heavy Mineral Concentrations (tHMC), Zircon-Tourmaline-Rutile Index (ZTR), Mineral Maturity Index (MMI) and Hornblende Colour Index (HCI).

Petrography and heavy mineral analysis
The analyses presented below are intended to provide data for regions that are not adequately covered by published datasets, summarized by Garzanti et al. (2015).

Red Sea Hills, Egypt
Modern wadi sands (RSH03A and RSH05A) in northern Egypt are sub-litharenites ( Fig. 2), with carbonate clasts and minor chert and shale grains, and a poor to moderately poor tHMC of 1.2 ± 1.0. Epidote-amphibole-clinopyroxene transparent heavy mineral assemblages include minor garnet, zircon, rutile, staurolite and tourmaline (ZTR 10 ± 5). Phanerozoic sandstone samples (RSH08A and RSH09A) from the Red Sea Hills are quartz arenites with an extremely high ZTR of 98 -99 and tHMC between 12 and 20. The modern wadi sands contain slightly more (4 -5%) feldspar than the bedrock samples. This difference in mineralogical maturity between bedrock and wadi samples may reflect sampling bias as the modern wadi sands sample a larger area. By contrast, a sample from the Hammamat Formation (RSH14A) was found to be a sheared lithic-arkose volcaniclastic arenite dominated by epidote, with ZTR = 0 and tHMC = 73.

Western Desert, Egypt
Aeolian dune sands (WD03C, WD19C and WD20C) from the Egyptian desert west of the Nile are quartzose, with few feldspars (K-feldspar > plagioclase) and sedimentary rock fragments (limestone and subordinately siltstone-shale and chert). Very poor to moderately poor (tHMC of 0.8 ± 0.4) transparent heavy mineral assemblages include epidote, zircon, amphibole, rutile, tourmaline, clinopyroxene, staurolite, garnet, kyanite and titanite (HCI = 11 ± 1, MMI = 50, ZTR = 34 ± 13).  All samples are modern muds or mudstones, except for samples from the Nile trunk in Egypt collected prior to construction of the Aswan Dam (Shukri 1949), which are modern sands. *Data from Garzanti et al. (2013). The field for Red Sea Hills samples encompasses 11 samples. Other distributions represent single analyses, or averages of small groups of samples.

Nile trunk, Sudan
Downstream of the confluence of the Blue Nile, White Nile and Atbara, sediments (SD04A and SD05A) are quartz-dominated with plagioclase feldspar more abundant than K-feldspar, and some volcanic detritus present. Both samples were found to be rich in transparent heavy mineral assemblages (tHMC of 16.8 and 10.1) including clinopyroxene, epidote and hornblende (HCl = 6 -7). The ZTR for both Nile trunk samples was very low (0 -1).

Major and trace element chemistry
Trace element results are presented in Figure 3, normalized to the trace element concentrations of Post-Archaean Australian Shale (PAAS), compiled by McLennan & Taylor (1985). V, Cr, Co, Ni and Cu concentrations allow us to determine the relative amount and location of mafic material within the Nile catchment area.
Values of V, Cr, Co, Ni and Cu are considerably higher in the Blue Nile and Atbara rivers than in the White Nile, reflecting the significant proportion of continental flood basalts in the Blue Nile and Atbara catchments. Similarity between the Blue Nile-Atbara values and those of the Nile trunk in Sudan attest to the dominance of the Blue Nile-Atbara contribution compared with the White Nile to the trunk river downstream.
Western Desert values are similar to the White Nile values, reflecting their location in proximity to cratonic rocks. Red Sea Hills values show a greater mafic influence owing to their closer proximity to the Arabian-Nubian Shield. Of note is the difference between Red Sea Hills wadi muds collected from locations overlying Red Sea Hills Phanerozoic bedrock and Red Sea Hills wadi muds from locations directly overlying Arabian-Nubian Shield basement (samples RSH14B and RSH15B). The Ni spike in the latter most probably reflects significant derivation from the underlying ophiolitic ultramafic rocks of the Arabian-Nubian Shield.

Western Desert Archaean gneiss
The oldest bedrock sample analysed (WD16A) is a gneiss, mapped as Archaean crust of the Saharan Metacraton (Daumain et al. 1958), collected east of Uweinat in southernmost Egypt. Zircon crystals have oscillatory-zoned cores under CL, with homogeneous CLdark, possibly metamorphic overgrowths (Fig. 4 inset). Cores are commonly broken, with abrupt terminations, suggesting that the sample may have had a sedimentary protolith that has undergone high-grade metamorphism.
Fifteen U-Pb analyses of zircon cores have 207 Pb/ 206 Pb ages ranging from 2924 to 3235 Ma, with a single analysis giving a statistically younger age of 2794 Ma. It is not possible to identify statistically valid populations within the zircon core dataset owing to the small number of cores and rims found in the samples. This may reflect a detrital source for the zircons. The zircon cores have an average ε Hf of −2.3, and a weighted average depleted mantle model age (TDM Hf ) of 3479 ± 47 Ma (MSWD = 4.1, n = 15/16).
Five analyses of zircon rims form a cluster with a weighted average 207 Pb/ 206 Pb age of 2701 ± 32 Ma. Three further rim analyses yield younger 207 Pb/ 206 Pb ages, between 2524 and 2587 Ma. The zircon rims yield an average ε Hf of −10.6 and a weighted average TDM Hf of 3467 ± 45 Ma (MSWD = 2.2, n = 8), within error of the age derived from zircon cores, indicating that new zircon growth took place in a closed system during high-grade metamorphism (Flowerdew et al. 2006). The weighted average model age of all zircon analyses in this sample is 3474 ± 31 Ma (MSWD = 3.3, n = 23/24). It is notable that the Th/U ratio of cores (0.37 -1.3) is considerably higher than that in the rims (0.03 -0.14), supporting a metamorphic origin for the rims.

White Nile (Uganda)
Modern river sands were collected from the White Nile upstream of the Sudd marshes at Wadelai (N7-16S) and Murchison Falls (N7-17S) in northern Uganda. Both samples show a similar distribution of Archaean ages, with maxima at c. 2620 Ma (Fig. 5), and only a few older grains. Analyses with 207 Pb/ 206 Pb ages of 2500 -2570 Ma commonly occur as  average ε Hf value of −24, a composition that is consistent with formation by melting of Archaean crust similar to that represented by the c. 2620 Ma zircon population. The 900 -1000 Ma grains show oscillatory zoning with no evidence for core-rim relationships. Zircons with ages between 670 and 870 Ma have ε Hf between +3.5 and −15, requiring input of juvenile material in addition to Archaean crust. These grains are typically CL-light, and weakly zoned or unzoned, with Th/U of 0.22 -0.89.
Four analyses of rims on Archaean zircon cores give ages of c. 630 Ma, with an average ε Hf of −30.5 (Fig. 4). The overgrowths show no zoning under CL, and have low Th/U (≤0.03), suggesting a possible metamorphic origin. A single zircon rim, which formed on a c. 800 Ma grain, also gives an age of c. 630 Ma, with ε Hf = −14, again consistent with growth in a closed system under metamorphic conditions (Flowerdew et al. 2006). Figure 4 includes a field for NE African Archaean Basement, which is derived using the composition of zircons in the Saharan Metacraton gneiss (WD16A) and the detrital zircon grains from the White Nile at Murchison Falls (N7-17S). The slope of the observed trend is consistent with an average 176 Lu/ 177 Hf ratio in the sampled North African crust of c. 0.012. This field represents the likely composition of craton-derived grains in other sedimentary rocks of the Nile catchment, discussed below. Only four grains give apparent ages younger than 620 Ma, the youngest having a 206 Pb/ 238 U age of 596 ± 18 Ma, within error of the age of the structurally much later Um Had granite. The timing of deposition of the Hammamat Formation is unclear, with suggested ages ranging between 630 and 583 Ma (Willis et al. 1988;Wilde & Youssef 2002;Bezenjani et al. 2014). It is therefore critical to establish whether these young grains could be used to determine a maximum depositional age for the Hammamat Formation, or whether these grains may have been affected by Pb loss. Several grains with ages younger than 645 Ma were therefore revisited, and five new analyses were carried out on each grain. All grains yielded ages of c. 635 Ma or older, confirming that the original analyses had been affected by Pb loss (Fig. 7). The original analyses used a large laser spot size (35 µm, compared with 20 µm for the new analyses), which may have been more likely to intersect cracks and defects in the grains.

Phanerozoic cover rocks
Western Desert (Egypt) Samples WD04A, WD12A and WD18A (Figs 5 and 6) are Phanerozoic sandstones. These samples have a dominant zircon population at c. 600 Ma, with >60% of grains of this age showing negative ɛ Hf . Zircons between 700 and 900 Ma are dominated by more juvenile grains, with positive ɛ Hf . A population of c. 1000 Ma grains appears to fall into two distinct groups, with juvenile and more crust-dominated hafnium compositions. In addition, subordinate populations of Palaeoproterozoic and Archaean grains (totalling c. 10%) and sparse Phanerozoic grains are present.

Red Sea Hills (Egypt)
Sample RSH09A is a Phanerozoic sandstone from Wadi Kharit in the Red Sea Hills. Zircons are dominated by peaks at c. 620, 730 -

Modern wadi sands
Sample RSH07A is a modern wadi sand collected in Wadi Hammamat in the Red Sea Hills, deposited on Phanerozoic sandstone bedrock. The zircon age and ɛ Hf isotope distribution are similar to those of the Red Sea Hills Phanerozoic bedrock sample, with peaks at c. 620, 770 and 1000 Ma. Fifty per cent of 620 Ma grains and 42% of c. 1000 Ma grains have cratonic ɛ Hf values. Eight per cent of grains give Archaean or Palaeoproterozoic ages.

Blue Nile
Sample ETH02A was collected from the Blue Nile Gorge in the Ethiopian Highlands. It shows two main zircon age populations at 660 -720 and 800 -850 Ma. ɛ Hf values are dominantly juvenile, although sparse grains at 500 -600 Ma yield ɛ Hf values as low as −10. Sparse c. 1000 Ma ages are also present, but Archaean and Palaeoproterozoic ages are absent. Juvenile Cenozoic grains are present (Fig. 8)

Atbara
Sample SD06A was collected from the Atbara close to its confluence with the Nile trunk in Sudan. The most dominant population forms a peak at c. 600 Ma, with a subordinate peak at c. 800 Ma, both dominated by juvenile ɛ Hf values. Palaeoproterozoic and Archaean grains are sparse (3%). Cenozoic grains make up 6% of the population.

White Nile
Sample SD02A was collected from the White Nile downstream (north) of the Sudd marshes in Sudan, and south of the Gezira Fan. Zircons are dominated by c. 600 Ma grains, with subpopulations showing both juvenile and more cratonic ɛ Hf values. Older age peaks at c. 850 and 1000 Ma are dominated by juvenile ɛ Hf values. Only two Archaean grains were identified (c. 2%), in marked contrast to sample N7-17S (see above) from upstream (south) of the Sudd marshes.
Sample SD01A was collected downstream of the Blue Nilesourced Gezira Fan. It shares similar peak characteristics to sample SD02A, but with more dominant c. 650 and 1000 Ma populations.

Nile trunk
Sample SD04A captures the signature of the Nile trunk in northern Sudan, downstream of the confluences of the Blue Nile, White Nile and Atbara, but upstream of the Aswan Dam. The detrital zircon age and ɛ Hf distribution are similar to those of the upstream Blue Nile Ar/Ar mica and plagioclase feldspar and U/Pb rutile Samples analysed include White Nile, Blue Nile and Atbara, Nile trunk, Red Sea Hills modern wadi and bedrock, and Western Desert aeolian samples.
The 40 Ar/ 39 Ar mica and plagioclase data and the U-Pb rutile data all show an overwhelming Pan-African signature (c. 600 Ma). This includes the White Nile samples upstream of the Sudd, despite having a strongly Archaean signature in U/Pb zircon analyses, and indicates strong overprinting by the Pan-African Orogeny in this region. The Blue Nile also contains Cenozoic feldspars, derived from the Ethiopian Highlands and in agreement with the presence of Cenozoic zircons also recorded in this river (see above).
Although the number of white mica analyses is small, we observe a change in signal between White Nile sands south of the Sudd marshes and the Nile trunk and Red Sea Hills to the north. The White Nile sample (N7-17S) is dominated by a single population with a mean age of c. 615 Ma. By contrast, both the Nile trunk in Sudan (Sample SD04A) and a Red Sea Hills wadi sand (RSH07A) show age peaks at c. 600 and 650 Ma.
Modern river mud samples collected from the Atbara and Tekeze rivers have Sr, Nd and Hf isotope compositions similar to average Ethiopian basalts (Figs 9 and 10; Pik et al. 1999;Meshesha & Shinjo 2010). Samples from the Blue Nile and the Nile trunk in northern Sudan plot on trends towards higher 87 Sr/ 86 Sr and lower ε Nd and ε Hf , consistent with incorporation of older sediment derived from sources other than Ethiopian basalt. Average values for Holocene Nile trunk sediments in northern Sudan, and sand samples collected in Egypt before the construction of the Aswan High Dam, also plot on this trend (Shukri 1949;Woodward et al. 2015). A sample collected from the White Nile just south of its confluence with the Blue Nile at Khartoum plots close to the Nile trunk trend.
Three modern river mud samples have what, at first glance, appear to be anomalous isotope compositions, given their locations. A sample (ETH07B) from the Uri River, Ethiopia, a tributary of the Tekeze in Ethiopia, has values unlike those one might expect from the Ethiopian plateau continental flood basalt region. However, the Uri River's local catchment drainage basin geology is in fact Phanerozoic cover (Fig. 1), consistent with its isotopic signature. Samples SD01B and SD02B are White Nile samples, yet their Sr-Nd signatures are similar to those of the Blue Nile and Phanerozoic cover respectively. This is consistent with, in the case of SD01B, its location downstream of the Gezira Fan, which is an overspill of Blue Nile detritus, and in the case of SD02B, its location downstream of the Sudd Marshes, which act as a sediment trap to the upstream White Nile, and thus its provenance is locally derived, as discussed in more detail in following sections.
Excluding the three anomalous samples described above, for which local geological anomalies provide explanation, and the White Nile samples taken from Uganda, the range in isotope composition shown by the Nile and its tributaries is rather limited: 87 Sr/ 86 Sr varies from 0.7043 in the Atbara to 0.7067 in the White Nile sample from Khartoum, ε Nd varies from 5.5 to −1.8, and ε Hf varies from 10.8 to −1.8.

Discussion
Our data document early crust-forming events, Precambrian orogenies culminating in the Neoproterozoic assembly of Gondwana, subsequent erosion and deposition of voluminous sediments, and finally Cenozoic uplift of the Red Sea Hills and Ethiopian Highlands and eruption of continental flood basalts in Ethiopia. Additionally, our data track the influence of geomorphology and local geology on the progressive evolution of the Nile sedimentary signal downstream.

Palaeoproterozoic and Archaean cratons
The Nile traverses Precambrian rocks of the Congo Craton and Saharan Metacraton. The geology of the Saharan Metacraton ( Fig. 1) is poorly understood owing to the overlying sedimentary cover and desert sands, and as a result very few modern isotope data are available to help constrain its age and tectonic evolution; the U-Pb ages reported by Abdelsalam et al. (2002) are discordant prechemical abrasion isotope dilution thermal ionization mass spectrometry ages, or model ages derived by zircon evaporation techniques, which yield poorly constrained 207 Pb/ 206 Pb ages only. However, Bea et al. (2011) reported Archaean U-Pb SHRIMP ages as old as 3 Ga for gneisses at Uweinat, with evidence for crust as old as 3.22 Ga at Gebel Kamil in southern Egypt. Similar ages have been obtained in this study for gneisses from c. 80 km east of Uweinat, within the region mapped as the meta-igneous Gebel Fig. 9. 87 Sr/ 86 Sr v. ɛ Nd isotope data for Nile river muds; Egyptian Nile trunk sands taken prior to the construction of the Aswan Dam (samples from Shukri 1949); Nile hinterland mudstones, wadi muds and aeolian sands; and Sudanese Holocene Nile trunk samples (Woodward et al. 2015). Also shown for comparison are average data from the Ethiopian Continental Flood Basalts (Pik et al. 1999). Fig. 10. ɛ Hf plotted against ɛ Nd for the same samples as shown in Figure 8. *Ethiopian basalt data from Meshesha & Shinjo (2010). Symbols as in Figure 9, except as noted.
Kamil Series by Bea et al. (2011). The sample studied here (WD16A) has evidence for a metasedimentary protolith (broken pre-metamorphic zircon cores), with evidence for zircon growth between c. 3.2 and 2.9 Ga, and subsequent metamorphism at c. 2.7 Ga. Depleted mantle Hf model ages lie between 3.4 and 3.5 Ga. We see no evidence in our data for the c. 2 Ga thermal event (the Eburnean orogeny) recorded by Bea et al. (2011).
Sample N7-17S, a modern river sand sample collected at Murchison Falls, Uganda, was sampled from within the NE Congo Block of the Congo Craton, close to the Aswa Shear Zone, which forms the boundary with remobilized crust of the Saharan Metacraton to the north (Appel et al. 2005;Katumwehe et al. 2016), adjacent to the Pan-African Central African Fold Belt. Zircons from this sample have similar Hf model ages (Fig. 4) to the Archaean gneiss sample from the Saharan Metacraton (WD16A). Zircons in the Murchison Falls sample record crystallization at c. 2.7 Ga, c. 960 Ma and c. 600 Ma; the younger age population forms rims with Hf isotope compositions that are consistent with remelting of similar Archaean crust. The fields defined by the Saharan Metacraton and modern river White Nile provide the best estimate of the composition of Precambrian cratonic basement in the Nile catchment. The 600 Ma zircon rims, and similar aged rutile, plagioclase and mica, indicate penetration of the effects of the Pan-African Orogeny deep into the craton interior, in agreement with the previous work and conclusions of Appel et al. (2005) and Schenk et al. (2007).

The Hammamat Formation
The depositional and tectonic environment of the Hammamat Formation, which overlies the Precambrian Arabian-Nubian Shield basement, is disputed (e.g. Ries et al. 1983;Wilde & Youssef 2002;Abdeen & Greiling 2005;Eliwa et al. 2010;Johnson et al. 2011;Bezenjani et al. 2014). Proposed depositional ages lie close to the time of the Pan-African Orogeny, and both pre-collisional and postcollisional settings have been proposed. Local versus long-distance provenance is debated. Previous work has recorded detrital zircon grains with ages as young as 585 Ma in the Hammamat Formation (Wilde & Youssef 2002), which has implications for assigning a maximum depositional age to these rocks. The volcanic arenite sample from the Hammamat Formation studied here contains four zircon grains that give 206 Pb/ 238 U ages younger than 620 Ma. In detail, these grains are slightly discordant, and repeat analysis (five analyses for each grain) demonstrated that these grains had indeed undergone Pb loss, and had crystallization ages of 635 -640 Ma (Fig. 7). This is consistent with the age of the youngest concordant zircon population with consistent hafnium isotope composition in the dataset, of c. 640 Ma. In the light of this, we suggest that any maximum depositional ages for the Hammamat Formation based on sparse young grain ages should take into account the possibility that such grains have undergone Pb loss.
Hammamat Formation zircon data analysed as part of this study (Figs 5-7) can be explained by derivation from underlying local Arabian-Nubian Shield bedrock, which is dominated by zircons with juvenile hafnium isotope compositions derived from 870 -630 Ma oceanic arc rocks (e.g. Stern et al. 2010). A small number of pre-Neoproterozoic detrital zircons are found in the Hammamat Formation. Wilde & Youssef (2002) suggested that possible source areas for these pre-Neoproterozoic grains could be the Central and South Eastern Desert, or further afield in the Arabian-Nubian Shield in parts of Saudi Arabia where sparse zircons of this age have previously been recorded. However, we believe that such distal sources are not required to explain our data because it is possible that these zircons have been derived from the more proximal Egyptian Arabian-Nubian Shield (Ali et al. 2009). The Arabian-Nubian Shield bedrock in Egypt incorporates some inherited pre-880 Ma igneous zircons with juvenile ɛ Hf values indicating either incorporation of sediments during subduction along a passive margin or inheritance from the mantle source region, or both (Stern et al. 2010).
The volcanic lithic arenite composition and epidote-rich nature of the Hammamat bedrock sample (RSH14A) supports our interpretation that it is derived directly from Arabian-Nubian Shield arc rocks. Furthermore, the composition and zircon U-Pb ages of the sample are in stark contrast to the overlying post-collisional Phanerozoic cover sequences in the region, which are quartz arenites with ZTR values as high as 99 and have diverse zircon U-Pb age spectra, including a typical 600 Ma peak as described below. We therefore see no requirement to invoke long-distance transport of detritus to form the Hammamat Formation.

Phanerozoic sedimentary cover rocks and overlying modern sediment
Our analyses provide the first isotopic data from the Phanerozoic sedimentary cover bedrocks in Egypt. Samples from Phanerozoic sandstone cover sequences in the Western Desert and Red Sea Hills of Egypt are mineralogically mature. They compare well with Phanerozoic cover successions in surrounding regions across North Africa and the Middle East in terms of both zircon age spectra (e.g. compare with data from North Africa; Avigad et al. 2003Avigad et al. , 2012, and zircon ε Hf characteristics, namely the typical 'double plunge' to negative values at c. 600 and c. 1000 Ma (e.g. compare with data from Jordan; Morag et al. 2011), first recognized as characteristic in Israeli coastal sands by Be'eri-Shlevin et al. (2014).
Red Sea Hills Phanerozoic cover rocks and modern wadi samples do not consist entirely of material recycled from the underlying Arabian-Nubian Shield basement. They are derived, at least in part, from areas other than the underlying Arabian-Nubian Shield basement. This is indicated by the following: (1) the occurrence of >2000 Ma zircons, which are extremely rare in the Arabian-Nubian Shield (Stern et al. 2010); (2) the increased mineralogical maturity and more complex zircon populations compared with the underlying locally sourced Hammamat Formation (see above); (3) the presence of a significant 1000 Ma population with cratonic ε Hf values, which is incompatible with derivation from the juvenile arc of the Arabian-Nubian Shield.
A similar conclusion was reached by Morag et al. (2011), who looked at ε Hf values of the 1000 Ma zircon population in cover sequences of Israel and Jordan. Although their data showed a preponderance of grains with negative ε Hf values, our data from Egypt show two populations at c. 1000 Ma, one with positive ε Hf values and one with negative values. Whereas the population with positive values would be compatible with derivation from an as yet unrecorded arc from within the Arabian-Nubian Shield, the provenance of the population with negative ε Hf values remains enigmatic. The cratonic c. 1000 Ma zircons found in the Mesozoic cover of both Western Desert and Red Sea Hills samples can be most simply explained by recycling from underlying Palaeozoic cover sequences, but the original basement source is debated. As pointed out by Kolodner et al. (2009), long-distance transport from suitable cratonic source areas far to the south are ruled out by the paucity of >2000 Ma grains, which would also be present if derived from such a region. Avigad et al. (2003) proposed that the source could be material transported from East Africa towards the margins of Gondwana by Neoproterozoic glaciers. We suggest that the original source may have been from a region since rifted off Gondwana, which may have also supplied zircons to, for example, the Neoproterozoic Arkenu Formation cover sediments of Libya, which contain grains of an appropriate age (Meinhold et al. 2011).
Phanerozoic bedrock samples and Red Sea Hills modern wadi sand that partially covers the bedrock have similar zircon U-Pb age spectra and show a similar 'double plunge' of ε Hf characteristics to negative values at c. 600 and c. 1000 Ma (Figs 5 and 6). This similarity suggests that the origin of the modern sand is broadly as recycled material from Phanerozoic sedimentary rocks that the sand mantles. Some component of aeolian transport undoubtably contributes in the deposition of such modern sands. However, the extent of such transport distance is difficult to determine given the similarity of Phanerozoic sedimentary cover signature across vast distances of North Africa. Western Desert dune sands were not analysed for zircon U-Pb, and thus an assessment of far-field aeolian input distinct from that recycled from Phanerozoic sandstones cannot be assessed. The similarity in bulk-rock data between the Western Desert mudstone and the aeolian dune sand leads us to tentatively suggest that, as in the Red Sea Hills, composition of the modern sand is broadly similar to Phanerozoic sedimentary bedrock.

Young zircons derived from the Ethiopian Highlands
Cenozoic zircons were found in the Blue Nile (sample ETH2A), the Tekeze and Atbara rivers (ETH6A, ETH8A, SD06) and downstream in the Nile trunk in Sudan (SD04). Grains of such age were not recorded in the White Nile or modern wadi sediment from the Red Sea Hills (Fig. 8). The obvious source for these grains is the Cenozoic Ethiopian volcanic province in the Ethiopian Highlands ( Fig. 1), which includes a variety of volcanic units from flood basalts to shield volcanoes, and bimodal compositions with significant ignimbrites and rhyolitic airfall tuffs intercalated with the basalt lava flows (Ukstins et al. 2002;Kieffer et al. 2004;Prave et al. 2016). Our Cenozoic grain ages extend from c. 23 to 33 Ma, with population peaks at 25 and 30 Ma. These compare well with the timing of a major period of bimodal and silicic volcanism in Ethiopia, dated at c. 25 -30 Ma (Ukstins et al. 2002). Furthermore, Hf analyses that we undertook on both our samples and grains from Lake Tana Rhyolites of the Ethiopian Highlands, previously analysed for U-Pb by Prave et al. (2016), show an excellent match (Fig. 8). Ukstins et al. (2002) documented pre-rift bimodal magmatism in Ethiopia from 31 to 24 Ma. They recorded a major decrease in volcanism between 25 and 20 Ma, which they associated with the transition from pre-to syn-rift volcanism triggered by the separation of Africa and Arabia. The zircons we have analysed would, by their definition, be of pre-rift origin.
Effects of geomorphology, local bedrock geology and aeolian input on the isotope signal of the Nile River North of Uganda, the White Nile passes through an extensive area of marshland (the Sudd), which has trapped most of the detritus from the river since 2.7 Ma (Williams & Talbot 2009). This is clearly reflected in the detrital zircon data. Zircons in Ugandan White Nile modern river sediments, south of the Sudd, are dominated by Archaean grains with evidence for crustal reworking and new zircon growth at c. 960 and 600 Ma. A craton-dominated signature is also displayed by the highly negative ɛ Nd values and high 87 Sr/ 86 Sr values of the White Nile muds.
By contrast, north of the Sudd at Kosti, the detrital signature of White Nile zircons is much less cratonic. In sample SD02A, located north of the Sudd but south of the Gezira Fan, Archaean grains are almost completely absent, and instead the zircon populations show strong similarities to Blue Nile and/or Phanerozoic cover sediment signatures. A mud sample from the same location has Sr-Nd-Hf values similar to those of Phanerozoic cover (Figs 9 and 10). Potential contributing sources to this White Nile sample north of the Sudd could be the Sobat River and/or alluvial fans draining into the river from the Nuba Mountains.
The River Sobat drains similar lithologies to the Blue Nile. Published data show it to have 87 Sr/ 86 Sr values of 0.708 -0.712 and ɛ Nd values of −1.6 to −9.1 (Padoan et al. 2011), spanning the range of our data for modern muds from Phanerozoic cover, as sampled in the Red Sea Hills. However, the River Sobat flows through an extensive region of marshland north of the Sudd (the Machar Marshes) and thus was not considered to be a significant source of sediment to the Nile trunk downstream by Padoan et al. (2011).
Sample SD02 was collected north of Kosti, close to the southern limit of the Gezira Fan, so significant input of Blue Nile material from the fan is not thought to be likely, and, furthermore, such input would not be compatible with the bulk sediment Sr and Nd data. However, from the Last Glacial Maximum (c. 20 ka) onwards, a major alluvial fan (the Khor Abu Habl alluvial fan) has drained into the river from the Nuba Mountains to the west . Ephemeral rivers feeding the fan drain Proterozoic basement rocks and overlying Phanerozoic siliciclastic cover sequences, thus compatible with the isotopic signatures observed in this sample.
Further downstream, a mud sample at SD01B, collected just to the south of Khartoum, has an ɛ Nd of −1.8 and 87 Sr/ 86 Sr of 0.7067, a signature with affinity to the Ethiopian Continental Flood Basalts. This is due to the sample location being adjacent to the Gezira Fan (Williams & Talbot 2009), which formed during the Late Pleistocene by overspill from the Blue Nile and contributes today to White Nile sediment load. Significant contribution from the palaeo-Blue Nile sourced Gezira Fan is also reflected in the zircon age spectrum of sample SD01A (Fig. 5) where the dominant peak at 800 Ma strongly resembles the modern Blue Nile signature of sample SD03A in Sudan.
Further downstream to the north, the Blue Nile, and then the Atbara, join the Nile trunk. Both the Blue Nile and the Atbara (with its tributary the Tekeze) drain, from south to north downstream, Cenozoic volcanic rocks, Phanerozoic cover, Arabian-Nubian Shield and Proterozoic basement with Pan-African mineral cooling ages (Mock et al. 1999). Sr-Nd-Hf isotope data indicate that the Atbara has a higher proportion of volcanic detrital contribution to the river sediment compared with the Blue Nile. This is also reflected in petrographic data showing a higher proportion of clinopyroxene and olivine (Garzanti et al. 2006), and a higher proportion of V and Cr in the Atbara compared with the Blue Nile (Fig. 3). By contrast, a sample collected from a tributary of the Tekeze (the Uri River) in an area of Phanerozoic cover rocks has an isotope composition typical of Phanerozoic cover successions (Figs 9 and 10). Variation in the contributing bedrock lithology is also illustrated by the significant change in proportions of major zircon populations in the Blue Nile downstream. Notably, the region of Arabian-Nubian Shield cut by the Blue Nile in NW Ethiopia contains granitoids with U-Pb zircon ages in the ranges 650 -700 and 800 -880 Ma, and juvenile ɛ Nd values (data summarized by Johnson 2014). These populations are clearly represented in our Blue Nile dataset (Fig. 5), with prominent peaks at 800 -830 Ma in samples ETH02A and SD03A, persisting into the Nile trunk at Dongola (sample SD04A). Zircons of 680 Ma are also abundant in sample ETH02A from the Blue Nile, and in the southern Nile trunk (SD04A).
The Nile trunk sample downstream of the White Nile confluence with the Blue Nile and Atbara, at Dongola in Sudan (SD04A), shows a U/Pb zircon signature much the same as that of the Atbara and Blue Nile combined. Garzanti et al. (2006) suggested a greater input to the trunk Nile from the Blue Nile compared with the Atbara, based on petrographic data. The greater similarity in Sr-Nd and Cr and V values between the Blue Nile and Nile trunk, compared with the Atbara and Nile trunk (Figs 3 and 9), is consistent with this proposal.
By the time the Nile River reaches northern Egypt (sample BS01A; Figs 5 and 6), its signature has changed to more closely resemble that of typical Phanerozoic sedimentary cover in terms of zircon age spectra, with more Palaeoproterozoic to Archaean grains, and the double plunging ε Hf to negative values at c. 600 and c. 1000 Ma, also clearly observed in a Nile sample collected near Cairo by Iizuka et al. (2013). This downstream change can be explained by the Nile's route in this stretch of the river, which flows across the Phanerozoic sedimentary bedrock cover and overlying modern sediment of the Red Sea Hills to the east and the Western Desert to the west. Any sediment input to the Nile from the west would more likely be transported to the river by aeolian processes, and from the east would include flash flood transportation, given the higher topography of the Red Sea Hills compared with the Western Desert.
Overall, when considering the entire length of the Nile, undoubtably aeolian transportation plays a part in delivering detritus to the river. However, such input does not mask the dominant bedrock and geomorphological controls that we have demonstrated to effect the changes in the river's signature downstream, for both sand and mud grade material.
Finally, we consider that the contribution of the White Nile to the Nile delta is small, the result of both its significantly lower mean annual discharge compared with the Blue Nile and sediment trapping in the Sudd Marshes. Both Sr-Nd and trace element values of the trunk Nile are more similar to those of the Blue Nile-Atbara than to those the White Nile (Figs 3, 9 and 10). Additionally, the white micas of the trunk Nile show similarity in Ar-Ar age distribution to grains collected from the Red Sea Hills, and dissimilarity to the age spectra from the White Nile south of the Sudd. Although pre-Neoproterozoic zircon grains make up to 97% of sediment in the White Nile south of the Sudd, such aged grains in the Nile trunk are not required to be delivered by the White Nile. Pre-Neoproterozoic grains are in fact extremely sparse in the White Nile north of the Sudd (average 2%). Pre-Neoproterozoic grains are relatively common in the Phanerozoic cover sedimentary rocks and modern wadi sediments sampled in the Red Sea Hills (average 9%) and the Western Desert (average 12%), and are also found in the Blue Nile (average 1%) and in sediments from the Atbara and Tekeze (average 6%) (Fig. 5). Any of these regions could have supplied the pre-Neoproterozoic zircons recorded in the Nile trunk. Figure 11 shows all of the U/Pb and Hf isotope data derived from Nile hinterland samples as part of this study. The largest peak at c. 600 Ma relates to the assembly of East and West Gondwana, when Neoproterozoic juvenile intra-oceanic island arcs of the Arabian-Nubian Shield accreted and finally collided with the Saharan Metacraton and Congo Craton (Johnson et al. 2011). Fig. 11. Probability density plot illustrating the Hf isotope-time evolution of the Nile source regions supplying detrital zircons discussed in this study. The density of the data distribution is calculated using a modified version of the MATLAB implementation of the Kernel Density Estimation procedure supplied by Botev et al. (2010) using bandwidths equivalent to the typical analytical uncertainty of the U/Pb ages (±20 Ma) and ɛ Hf values (±1 epsilon units). Contours are generated by the MATLAB contour3 function and plotted using the plot3d function. Also shown are the compilation of African detrital data from Condie & Aster (2010), and the modern African river data of Iizuka et al. (2013) filtered using the discordance criteria applied in this paper. This can be seen in samples derived from as far afield as the White Nile in Uganda.

Supercontinent assembly and the Pan-African Orogeny
The influence of the Pan-African Orogeny is also seen in U/Pb rutile, and Ar/Ar plagioclase and mica ages, which reflect postorogenic cooling. This is in agreement with hornblende, muscovite, amphibole, biotite and sericite 40 Ar/ 39 Ar data from Arabian-Nubian Shield bedrock (Johnson et al. 2011).
In detail, the zircon U-Pb and Hf-isotope data show two maxima at c. 840 -780 and 700 -600 Ma with juvenile ε Hf values, which represent two major episodes of arc development within the Arabian-Nubian Shield. The reduction in data density between the two episodes may relate to reduced subduction-related magmatism at this time, perhaps resulting from collisions between arcs within the Arabian-Nubian Shield. Arc collision is documented by the emplacement of ophiolites at this time (Ali et al. 2010), and may be represented in our dataset by c. 750 Ma detrital zircons from the Hammamat Formation, which have the highest (most juvenile) ɛ Hf values in the entire dataset.
The final closure of the Mozambique Ocean is documented by the marked switch to more negative ε Hf values in the c. 600 Ma population, owing to the intrusion of post-collisional granites derived by melting of continental crust within the suture zone.
A notable feature of the data is the discontinuity of the field of c. 1000 Ma grains with juvenile ε Hf versus those with more cratonic influence (i.e. ε Hf < 0). Although the ultimate source of these grains remains unknown, our data suggest two distinct ultimate sources, one from an oceanic-arc setting and a second involving remelting of a cratonic source.
The most cratonic (i.e. lowest ε Hf ) grains within each age population in our dataset plot close to the 3500 Ma evolution line for North African crust derived using our basement data from the Saharan Metacraton in Egypt and Congo Craton in Uganda.
The supercontinent cycles that led to the creation of Gondwana (650 -500 Ma) are represented in all of the samples collected as part of this study (Fig. 11). The isotope signatures of the samples analysed from the Nile River and its hinterland contain Mesoproterozoic and older zircons that may be related to earlier supercontinent cycles. However, the location of the Congo Craton in reconstructions of Rodinia (De Waele et al. 2008) is a matter of debate and the lack of exposure and reliable geological data from the Saharan Metacraton means that its location in Rodinia and earlier supercontinent configurations is unknown. Although our data contain some c. 2000 Ma zircons, these cannot be directly related to the formation of the supercontinent 'Nuna' (aka Colombia) because Nuna did not involve North African cratons (Rogers & Santosh 2002).

Conclusions
Gneissic basement of the Saharan Metacraton in southern Egypt and detrital zircons derived from rocks of the Congo Craton in northern Uganda both indicate an age of crust formation between 3.0 and 3.5 Ga, with subsequent melting and/or metamorphism at c. 2.7 Ga.
Our combined U-Pb and Hf isotope zircon dataset documents the evolution of the North African crust, in particular highlighting phases in the development of the Arabian-Nubian Shield and its collision with the Saharan Metacraton. The data show two phases of arc magmatism, at c. 840 -780 and 700 -600 Ma, with a reduction in magmatism at c. 750 Ma, during a period of arc collisions within the Arabian-Nubian Shield.
The Hammamat Formation, sampled in Wadi Hammamat, has a maximum depositional age of 635 Ma, based on reanalysis of anomalously young zircon grains, which are shown to have undergone Pb loss. Provenance analysis indicates that the Hammamat Formation is locally derived. By contrast, the Phanerozoic cover sedimentary rocks overlying the Arabian-Nubian Shield are derived, at least in part, from distal sources beyond the Arabian-Nubian Shield. Phanerozoic cover sedimentary rocks that blanket much of NE Africa represent an important source of detritus to the Nile River and are characterized by the presence of a major zircon population with both juvenile and crustal ɛ Hf values reflecting the Pan-African orogenic event, and a significant (9 -12%) population of pre-Neoproterozoic zircon grains. The original source of c. 1000 Ma grains in the Phanerozoic sedimentary cover rocks remains enigmatic, but we have identified that there are two distinct populations: one indicating a juvenile ultimate source and one a cratonic source.
In the modern Nile drainage, there is considerable evolution downstream, controlled predominantly by changes in local geology and geomorphology. The provenance signature of the White Nile is dramatically different upstream and downstream of the Sudd marshes as a result of sediment trapping. South of the Sudd, White Nile sediments are craton-derived. North of the Sudd, at Kosti, the signature of the White Nile is dominated by material derived from Phanerozoic sandstones supplied via alluvial fans to the west of the river. Further north, south of Khartoum, White Nile sediment composition is affected by its proximity to the Pleistocene Blue Nile sourced Gezira Fan. The Blue Nile's and Atbara's signatures are influenced predominantly by input from the Ethiopian flood basalts in terms of their bulk-rock signature, and by proximity to the Arabian-Nubian Shield in terms of zircon characteristics. A further shift in sediment signature in terms of zircon characteristics is seen by the time the Nile reaches northern Egypt, reflecting the passage of this stretch of the river through Phanerozoic cover sedimentary rocks and overlying modern sands of the Red Sea Hills and Western Desert. By contrast, the bulk isotopic Sr-Nd-Hf data show little downstream evolution and remain dominated by mafic input from the Ethiopian Highlands as far south as northern Egypt.