Deoxygenation dynamics above the western Nile deep-sea fan during sapropel S1 at seasonal to millennial time-scales

Abstract. Ocean deoxygenation is a rising threat to marine ecosystems and food resources under present climate warming conditions. Organic-rich sapropel layers deposited in the Mediterranean Sea provide a natural laboratory to study the processes that have controlled the changes in seawater oxygen levels in the recent geological past. Our study is based on three sediment cores spanning the last 10 thousand years (10 kyr BP) and located on a bathymetric transect offshore the western distributaries of the Nile delta. These cores are partly to continuously laminated in the sections recording sapropel S1, which is indicative of bottom-water anoxia above the western Nile deep-sea fan. We used a combination of microfacies analyses and inorganic and organic geochemical measurements to reconstruct changes in oxygenation conditions at seasonal to millennial time-scales. The regular alternations of detrital, biogenic and chemogenic sublayers in the laminated sequences are interpreted in terms of seasonal changes. Our microfacies analyses reveal distinct summer floods and subsequent plankton blooms preceding the deposition of inorganic carbonates formed in the water-column during spring-early summer. The isotopic signature of these carbonates suggests year-round anoxic to euxinic bottom waters resulting in high levels of anaerobic remineralisation of organic matter and highlights their potential to reconstruct seawater chemistry at times when benthic fauna was absent. Synchronous changes in terrigenous input, primary productivity and past oxygenation dynamics on millennial time-scales obtained by our multi-proxy study show that runoff-driven eutrophication played a central role in driving rapid changes in oxygenation state of the entire Levantine Basin. Rapid fluctuations of oxygenation conditions in the upper 700 m water depth occurred above the Nile deep-sea fan between 10 and 6.5 ka BP while deeper cores recorded more stable anoxic conditions. These findings are further supported by other regional records and reveal time-transgressive changes in oxygenation state driven by rapid changes in primary productivity during a period of long-term deep-water stagnation.


carbonates formed in the water-column during spring-early summer. The isotopic signature of these carbonates 23 suggests year-round anoxic to euxinic bottom waters resulting in high levels of anaerobic remineralisation of 24 organic matter and highlights their potential to reconstruct seawater chemistry at times when benthic fauna was 25 absent. Synchronous changes in terrigenous input, primary productivity and past oxygenation dynamics on 26 millennial time-scales obtained by our multi-proxy study show that runoff-driven eutrophication played a central 27 role in driving rapid changes in oxygenation state of the entire Levantine Basin. Rapid fluctuations of oxygenation 28 conditions in the upper 700 m water depth occurred above the Nile deep-sea fan between 10 and 6.5 ka BP while 29 deeper cores recorded more stable anoxic conditions. These findings are further supported by other regional records 30 and reveal time-transgressive changes in oxygenation state driven by rapid changes in primary productivity during 31 a period of long-term deep-water stagnation. 32 Pinardi et al., 2015). Deeper water masses flow below the LIW and form by incorporation of cold surface waters 85 from the Adriatic Sea and sometimes dense waters from the Aegean Sea to the LIW (Cornuault et al., 2018; 86 Robinson et al., 1992;Roether et al., 2007). 87 The main source of freshwater to the Levantine Basin is the Nile River runoff through the active Rosetta channel 88 located in the western part of the Nile Delta (Fig. 1). Although the present Nile flow is drastically reduced compared 89 to historical times (Halim et al., 1967), runoff still leads to marked salinity gradients (halocline) in the upper 200 90 m water-depth (w-d), in particular during the summer monsoon (Fig. 1a)  For our study, we use three gravity cores that were retrieved during the oceanographic campaign P362/2 "West 110 Nile Delta" onboard the R.V. Poseidon in August 2008 (Feseker et al., 2010) (Table 1). These sediment cores are 111 located along a bathymetric section on the western Nile deep-sea fan (DSF), in the vicinity of the North Alex and 112 Giza mud volcanoes (Feseker et al., 2010) (Fig. 1b,c). Cores P362/2-73 (P73) and P362/2-99 (P99) are located 113 close to each other and ca. 75 km eastward of core P362/2-33 (P33), on the opposite side of the submarine Rosetta 114 canyon (Fig. 1b). Core P33 was recovered at 738 m water-depth, is 5.6 m-long and contains well-preserved mm-115 scale laminations in the lower 5 m of the core (see Blanchet et al., 2013Blanchet et al., , 2014, for detailed descriptions) (Fig. 1c, 116 MS27PT  Upper limit of the OMZ during S1?  2). Core P73 was collected at 569 m depth, is 5.4 m-long and contains an alternation of finely laminated (similar 117 to P33) and bioturbated intervals (Fig. 1c, 2). Core P99 is located at 396 m depth, is 5.3 m-long and contains a few 118 laminated intervals within otherwise bioturbated sediments (Fig. 1c, 2). All cores are composed of clay-rich dark-119 brown hemipelagic mud, which shows significantly lighter colours in the upper 1m of the cores. In cores P33 and 120 P73, several harder carbonate-rich layers were identified (Fig. 2). 121 A qualitative assessment of the foraminiferal assemblage was realised in the lower part of core P33 (between 75 122 and 558 cm), with a particular emphasis on benthic foraminifera.

Age determination and transformation depth-to-age 133
The chronology of core P33 has been constructed based on a set of 14 radiocarbon samples (Blanchet et al., 2013). 134 We prepared another 8 samples for radiocarbon dating using planktonic foraminifera and pteropod shells, which 135 provide tie-points for cores P73 and P99 (Table 2). Only one sample in core P73 (Poz-113951) did not generate 136 enough carbon for an accurate dating. The new radiocarbon measurements were performed at the Poznan 137 radiocarbon laboratory (Poland). 138 More detailed stratigraphic constrains were obtained from correlation of the titanium over calcium (Ti/Ca) records 139 of cores P73 and P99 with core P33 (Fig. 2). It was shown that changes in sedimentation rates are coherent on the 140 western Nile DSF (Hennekam et al., 2015) and Ti/Ca records show similar patterns (Fig. S1). Six tie-points (marked 141 T1-T6 on Fig. 2) mark changes in Ti/Ca records identified in the upper parts of the cores and were used to further 142 synchronise the records of P73 and P99 with core P33 (Table 2). 143 The age-depth modelling was performed using Bacon version 2.3 (Blaauw and Christen, 2011), which enables to 144 discriminate between sections of the core with contrasting accumulation rates and to provide these as priors. 145 Convergence and mixing of the Markov chain Monte Carlo iterations used to build the age model by Bacon was 146 tested and the number of iterations was adjusted to obtain a Gelman and Rubin Reduction factor below 1.05 147 (Blaauw and Christen, 2011;Brooks and Gelman, 1998). The new age-depth model of core P33 based on 148 radiocarbon ages is similar to that presented in Blanchet et al. (2013), but age uncertainties are now available for 149 the whole core (Table 2 and Fig. 2a). Radiocarbon ages and the six tie-points were used for age modelling for cores 150 P73 and P99 using the Bacon program following a similar procedure to that used for core P33 (Blaauw and Christen, 151 2011). For the lowest part of core P73, for which there is no tie-point, we evaluated the ages and uncertainties using 152 the range of sedimentation rates observed in the other two cores for ages older than 7500 years and derived the 153 median sedimentation rate applying the maximum and minimum sedimentation rates observed as uncertainty range 154 ( Fig. 2). 155 156 https://doi.org/10.5194/cp-2020-114 Preprint. Discussion started: 2 October 2020 c Author(s) 2020. CC BY 4.0 License.

Microfacies description and scanning electron microscopy 165
A detailed examination of the microfacies was performed on the laminated part of core P33 (i.e., between 100 and 166 559 cm) and on a few samples of cores P73 and P99. Sediment blocks of 10 cm-long, 2 cm-wide and 1 cm-thick 167 were cut out of the fresh sediment with 2-cm overlaps to enable continuous microfacies analysis. Preparation of 168 thin-sections from soft and wet sediment blocks followed a standard procedure minimizing process-induced 169 disturbances of sediment micro-structures and included shock-freezing with liquid nitrogen, freeze-drying for 48 170 h, and epoxy resin impregnation under vacuum (Brauer and Casanova, 2001). 171   T1  T2  T3  T4   T5   T1  T1  T2  T2   T3   T3   T4   T5   T5

Mineralogy 178
The mineralogical composition was determined by X-ray diffractometry (XRD) on random powder samples of core 179 P33. Therefore, the rock chips were powdered to a grain size of < 63 µm, and loaded from the back side of the 180 sample holders. XRD analyses were performed with a PANalytical Empyrean x-ray diffractometer operating with 181 Bragg-Brentano geometry at 40 mA and 40 kV with CuKa radiation and PIXel3D detector at a step size of 0.013 182 °2q from 4.6 to 85 °2q and 60 sec per step. The Mineralogy was determined qualitatively with the EVA software 183 (version 11.0.0.3) by Bruker. 184

X-Ray Fluorescence scanning 185
The bulk sediment compositions for cores P33 and P99 were measured using an Aavatech TM X-Ray fluorescence 186 (XRF) core-scanner at the Institute of Geosciences of the University of Kiel (Germany) (Table S1 and S2). Non-187 destructive XRF core-scanning measurements were performed every 1 cm using a Rh X-Ray source at 10 kV and 188 0.65 mA to acquire the elements S, Cl, K, Ca and Ti. Core P73 was measured using an ITRAX XRF core-scanner 189 at GFZ Potsdam (Germany) ( Table S3). These measurements were conducted every 1 cm with a Cr-X-Ray source 190 operated at 30 kV and 60 mA to cover the same elements as acquired for core P33 and P99. Element at 50 kV, 600 µA and 50 ms using a Bruker M4 Tornado, which is equipped with a Rh X-ray source in combination 198 with poly-capillary X-ray optics generating an irradiation spot of 20 µm. Mapping results of the element K, Ca,199 and Ti representing solid-state chemical components are presented as normalized element intensities after initial 200 spectrum deconvolution (Fig. 3). However, elements that occur predominantly in pore fluids (e.g. Cl and S) are not 201 well preserved in epoxy embedded samples. 202

Stable oxygen and carbon isotopes 203
Measurements of stable oxygen and carbon isotopes were realized both at GEOMAR and GFZ (Table S4).

Sedimentary patterns on the western Nile deep-sea fan 252
The deeper core P33 exhibits a near-continuous 500-cm laminated interval between ca. 100 cm core depth and the 253 bottom of the core, with high Ti/Ca ratios and high sedimentation rates (between 100 and 350 cm/ka) ( (ca. 20-30 µm) as K-rich sublayers (Fig. 3a, 5b). 272 Light sublayers (LL) contain predominantly carbonate and can be subdivided into two main types (Fig. 3a). The 273 light-layer type 1 (LL1) are diffuse sublayers of silt-sized (ca. 10-20 µm) and unsorted particles containing biogenic 274 carbonates (coccoliths and foraminifera) and quartz grains (Fig. 3a). Foraminifera shells generally contain small 275 grains of iron sulphides, which may have formed on organic residues (Fig. S4). Some of these sublayers present a 276 slight cementation. The light sublayers type 2 (LL2) contain well-sorted fine (ca. 1-10 µm) and needle-shaped 277 calcite minerals, mixed with detrital silicate grains including clay minerals (Fig. 3). The detrital assemblage in LL2 278 sublayers is similar to that of dark layers but contains a larger amount of calcite (Fig.3c). LL2 sublayers are abiotic, 279 have a wavy internal structure and contain concentrated lenses of carbonate grains (Fig. 3a). Some of these 280 sublayers show a sharp lower boundary and either an upward decrease or increase in grain concentration (Fig. 3a). 281 In the lower part of core P33, three prominent LL2 sublayers are cemented and consolidated (thereafter denoted 282 "hard layers" HL) and are generally thicker than the softer LL2 sublayers (Fig. 2a). Four similar HL are also found 283 in core P73 (Fig. 2b). The presence of organic matter is observed either within LL or at the base of DL.  Stable d 18 O and d 13 C isotope measurements were performed on a set of 6 LL2 and HL sublayers ranging in 294 thickness from 0.1 to 0.8 mm (Fig. 4). Cemented HL have a strongly depleted d 13 C (-10 to -15 ‰) and an enriched 295 which do not correspond to either biogenic carbonates or regional methanogenic carbonates (Fig. 4). 299 The sublayers are deposited in a remarkably constant succession over the length of the laminated interval (Fig. 5). 300 This succession is particularly clear in the thicker laminations and follows the pattern from bottom to top: LL1, 301 DL1, LL2 and DL2. Sublayers LL1 and DL2 are always present in the succession, whereas DL1 is sometimes very 302 thin or not detectable and LL2 can be missing (Fig. 5b). In the lower part of the core (between 260 and 570 cm) 303 sub-layers are on average thicker and thinner in the upper part of the laminated interval. At our core sites, we did 304 not identify any remains of diatoms. 305 Anomalously thick layers (mm-to cm-scale) were identified throughout the record and labelled event layers (EL). with high sedimentation rates (> 100 cm/ka, Fig. 2) in all cores with suboxic to anoxic conditions at all water depths 316 between 9.5 and 7.5 ka BP and stable oxic conditions that establish between 7.0 and 6.0 ka BP (Fig. 6). The 317 occurrence of lycopane is an additional indicator of depositional anoxic conditions given that this biomarker 318 degrades under oxic conditions (Sinninghe Damsté et al., 2003) (Fig. 6b,g). Changes in the assemblages of benthic 319 foraminifera (Fig. 6h)  1999; Schmiedl et al., 2003). Finally, the S/Cl records (Fig. 6a,d,f)  hypoxic intervals that might have be partly reoxidized (Larrasoaña et al., 2006). All proxy-records document 330 suboxic to anoxic conditions at all water depths between 9.5 and 7.5 ka BP and stable oxic conditions that establish 331 between 7.0 and 6.0 ka BP (Fig. 6). 332 In the shallowest core P99, several short-term faintly-laminated intervals are observed between 9 and 8.3 ka BP 333 and between 7.6 and 7.4 ±0.3 ka BP (Fig. 6c), which contain well-preserved laminations and a peak value of 334 lycopane mass accumulation rates (MAR) (up to 0.013 µg/cm 2 .a) at 7.5 ±0.3 ka BP (Fig. 6b). High S/Cl values 335 occur between 9 and 8.5 ka BP, but are interrupted by a strong drop around 8.7 ka BP. Values decrease after 8.5 ka 336 BP, but remain slightly elevated between 8.5 and 7.2 ka BP (Fig. 6a). 337 In intermediate depth core P73, laminations occur as a nearly continuous interval between 9.2 ±0.6 and 7.4 ka BP 338 with an interruption between 8.3 and at 7.6 ka BP (Fig. 6e). High S/Cl values occur between 9.6 and 7.5 ka BP, 339 with a strong drop in the non-laminated interval around 7.9 ka BP. The S/Cl record shows a gradual decrease 340 between 7.3 and 6 ka BP, after which the record remained low and only varied slightly (Fig. 6d). 341 In the deepest core P33, laminations occur continuously between 9.6 and 8.1 ka BP, followed by an interval of faint 342 laminations until 7.3 ka BP (Fig. 6i). Lycopane MAR follow the lamination pattern with high values in the 343 laminated interval between 9.6 and 8.1 ka BP (between 0.1 and 0.35 µg/cm 2 .a) and a gradual decrease in the faintly-344 laminated interval until 7.3 ka BP (from ca. 0.01 µg/cm 2 .a to less than 0.001 µg/cm 2 .a) (Fig. 6g). The S/Cl record 345 shows generally higher values between 10 and 7.5 ka BP with lower values between 9.5 and 9.0 and around 8.8, 346 8.3, and 8 ka BP (Fig. 6f). The S/Cl record rapidly decreases between 7.5 and 6.5 ka BP and then remains at low 347 values in the younger part of the core. The laminated interval between 9.6 and 8.1 ka BP is devoid of benthic 348 foraminifera (except at 8.55 ka BP), after which infaunal species reappear (with a peak around 7.8 ka BP) but their 349 presence remains very volatile until ca. 7.4 ka BP (Fig. 6h). Epifaunal species indicative of well-oxygenated and 350 oligotrophic bottom waters are absent throughout the early Holocene record but a few specimens are observed at 351 ca. 7.5 and 6.7 ka BP (Cibbides spp. and Planulina spp., respectively). 352

Dynamics of primary productivity 360
Changes in primary productivity off the Nile mouth was traced using MAR of C37-alkenones and crenarchaeols in 361 cores P99 and P33 (Fig. 7). Alkenones are produced in surface waters by haptophyte algae, in particular the  (Fig. 7a,b), although the alkenone MARs do not show large peaks between 369 9 and 8.5 ka BP (Fig. 7a). There is a clear correspondence between the occurrence of peaks in MAR of organic 370 compounds and the presence of faint and well-preserved laminations (Fig. 7c). 371 In deeper core P33, the sampling resolution is higher and the signal is spikier (Fig. 7d,e). The MARs of both 372 crenarchaeols and alkenones are higher in the laminated interval between 9.6 and 8.2 ka BP (±0.1 ka BP) and in 373 the faintly laminated interval between 8.2 and 6.5 ka BP (±0.1 ka BP) (Fig. 7f). 374 In nearby and deeper cores MS27PT and GeoB7702-3 (Fig. 7 g,h), crenarchaeol MARs are also higher between 10 375 and 8 ka BP, but MARs are highest between 10 and 8.5 ka BP in core MS27PT while a marked peak occurred in 376 core GeoB7702-3 at ca. 8.1 ka BP (Castañeda et al., 2010; Ménot et al., 2020). 377

Dynamics of river runoff and sediment input 378
The changes in river runoff and sediment input to the Nile DSF were reconstructed using major elemental contents, 379 MAR of branched GDGTs (brGDGTs), odd long-chain n-alkanes and triterpenoids, the oxygen isotopic 380 composition of foraminifera and sedimentation rates (Fig. 8) of terrigenous input and grain size (Fig. 8a,e,h). We also used the recalculated oxygen isotopic composition of the 383 seawater (d 18 Osw) as a tracer of freshwater input, with lower values reflecting a larger influence of Nile-derived 384 waters on the shelf (Blanchet et al., 2014) (Fig. 8i). Finally, runoff is estimated using MAR of biomarkers specific 385 of terrestrial organic matter, such as brGDGTs, odd long-chain n-alkanes and triterpenoids (Fig. 8c,k).  The Ti/Ca and Ti/K records of all three cores ratios show very similar patterns with relatively high amounts of Ti 399 in the Early Holocene (ca. 10-7.5 ka BP) and the Late Holocene (1-0 ka BP) (Fig. 8a,e,h). Especially in well 400 laminated cores P73 and P33, high-frequency fluctuations in the Ti/Ca records during the Early Holocene are 401 related to the presence of Ca-rich laminations (Fig. 5). Although the magnitude is not similar in each core, rapid 402 decrease in Ti/Ca records around 7.2 ka BP marks a reduction of the lithogenic input ( fig. 8a,h) and is followed by 403 a more gradual decrease until ca. 3-4 ka BP. This gradual decrease is also recorded in the d 18 Osw (Fig. 8i). The 404 decrease in lithogenic input at 7.2 is accompanied by a simultaneous increase of K-rich material ( fig. 8a,e,h). A 405 stepwise increase in Ti/Ca and Ti/K (and to a lesser extend in d 18 Osw) is observed in all cores between 3-3.5 and 1 406 ka BP. 407 Sedimentation rates (SR) were calculated linearly between dated points (Table 2) (Fig. 8b,f,j). They are of the same 408 order of magnitude and vary synchronously in the three cores with highest SR during the sapropel (i.e., >100 cm/ka 409 between ca. 10 and 7 ka BP). An abrupt decrease is recorded in cores P99 and P73 around 7 ka BP whereas SR 410 decreased more gradually from 9 to 7 in core P33. After 7 ka, the SR remained below 20 cm/ka in all three cores. 411 The MAR of n-alkanes and triterpenoids are a factor 10 higher in core P33 than in core P99 (resp. up to 7 and 0.6 412 µg.cm -2 .a -1 ) but brGDGTs MAR are of a similar order of magnitude in both cores (resp. up to 0.3 µg.cm -2 .a -1 and 4 413 µg.cm -2 .a -1 ) (Fig. 8c,k). Several pulses of terrestrial organic matter input are observed in shallowest core P99 at 9.1, 414 8.8, 8.5, 8.2 and 7.5 ka BP (all dates ±0.25 ka BP) (Fig. 8c). There is a good agreement between the occurrence of 415 peaks in MAR of organic compounds and the presence of faint versus well-preserved laminations (Fig. 8d). In 416 deeper core P33, the MAR are higher in the laminated interval between 9.6 and 8.2 ka BP (±0.1 ka BP) but both 417 brGDGT and high plant n-alkanes and triterpenoids show low MARs in the faintly laminated interval between 8.2 418 and 6.5 ka BP (±0.1 ka BP) (Fig. 8k,l). 419 5 Discussion 420

Seasonal dynamics over the Nile deep-sea fan as derived from microfacies analyses 421
As shown by the microfacies analyses, the total layer thickness is mainly controlled by terrigenous inputs (DL) 422 (Fig. 5). The calcium carbonate-rich layers (LL) only represent a minor contribution in terms of sediment budget. 423 As shown in Fig. 5b, the sequence of dark and light sublayers (DL2, LL1, DL1, LL2) is consistent throughout the 424 laminated interval, even though sub-layers DL1 and LL2 can be very thin and sometimes absent. We propose that 425 the different sublayers recorded distinct depositional regimes throughout the year and represent an annual cycle.   The detrital DL2 sublayers are overlain by coarse-grained light sublayers (LL1) that are enriched in biogenic 445 carbonates, mostly foraminifera and coccoliths (Fig. S4). This biogenic carbonates assemblage resemble that of the 446 so-called "Nile blooms", which occurred in September-October prior to the construction of the Aswan Dam in 447 1965 (Nixon, 2003). Fertilisation of the surface waters off the Rosetta mouth by large nutrients input during the 448 summer Nile floods triggered an algal bloom, upon which zooplankton grazed. This was marked by the production 449 of carbonate organisms such as coccoliths and foraminifera, like in LL1 sublayers (Halim et al., 1967). Occasional, 450 these layers contain quartz grains that might be related to autumn-early winter dust storms (Goudie and Middleton, 451 2001). 452 The following sublayer DL 1 is generally a fine-grained homogenous dark layer, which is enriched in potassium 453 and reflects winter and spring deposition. Analysis of the present suspension load of the Nile River shows that the 454 suspended load of the main Nile is enriched in clay particles during low Nile discharge periods (i.e., from late 455 autumn to early summer) (Billi and el Badri Ali, 2010; Garzanti et al., 2006). 456 Sublayers DL1 are often overlain by a final sublayer of homogenous fine calcite grains (LL2) that is sometimes 457 cemented (HL). The internal draping structure of these layers suggests that these calcite grains formed close to the 458 seafloor, either in seawater or in highly porous sediments at the water-sediment interface. The mineralogical 459 assemblage in these layers is similar to the detrital-rich layers but shows an enrichment in calcite (Fig. 3c). SEM 460 pictures suggest that microcrystalline calcite might have formed as a coating on the detrital grains, which could 461 have served as nucleus (Fig. 3b). Signs of sedimentary disturbance associated with diagenetic carbonate formation 462 during sediment compaction are only observed in the cemented layers. Furthermore, these LL2 sublayers are also 463 identified both in cores P73 and P99 with sublayer structures similar to those of core P33 (Fig. S3). This points to 464 a regionally ubiquitous formation of authigenic carbonates in the water column, whereas post-depositional 465 diagenetic processes are more pervasive and local. These layers are linked to the late spring-early summer season 466 and the potential formation pathways and implications for seawater chemistry will be discussed in section 5.2. 467

Seasonal changes in physical and chemical properties of seawater 468
The seasonal sequence of sublayers depicted above for core P33 provides new insights into the variability of 469 chemical and physical parameters of seawater offshore the Nile mouth. The DL2 sublayers were deposited during 470 the summer floods of the Nile River. According to present-day and historical observations (Halim et al., 1967), 471 summer floods lead to the formation of a freshwater plume above an underlying highly saline seawater mass (Fig.  472 1a). During sapropel S1, this low-salinity subsurface water mass extended much further offshore in the whole 473 eastern Mediterranean basin, as shown by species-specific oxygen isotopes of planktonic foraminifera deposited 474 south of Crete (Tang and Stott, 1993). The large input of freshwater is thought to have led to a strong density-475 driven stratification of water masses in the eastern Mediterranean (Kemp et al., 1999). However, the important 476 sediment loading of Nile-derived freshwater that entered the relatively saline eastern Mediterranean probably 477 played an important role in driving the mixing of fresh-and seawater, through double-diffusive mixing and the 478 formation of salt fingers (Parsons et al., 2001). In addition, these processes are also thought to drive nutrient 479 exchange across stratified water-masses (Oschlies et al., 2003). The release of nutrients and their exchange across 480 waters masses during summer floods fuelled primary productivity in the whole eastern Mediterranean (Mojtahid et 481 al., 2015) and triggered "Nile blooms" in autumn, clearly evidenced in our record by zooplankton shells in LL1 482 sublayers. High levels of primary productivity probably triggered a gradual eutrophication of the underlying water 483 masses due to bacterial oxidation of sinking organic matter. The deposition of fine-grained clay-rich particles 484 following these "Nile blooms" occurred during the late autumn to early summer low discharge regime of the Nile 485 River (DL1 layers). The ability to deposit clay-sized particles as laminated layers suggests a hydrodynamically 486 stagnant water mass, which might have reinforced the gradual eutrophication of deeper water masses. 487 The occurrence of bottom water-derived authigenic calcite sublayers (LL2) in core P33 provides a unique 488 opportunity to investigate seawater chemistry using their oxygen and carbon isotope signatures (Fig. 4). The d 13 C 489 signatures of soft LL2 calcite sublayers suggest that they formed in euxinic (sulfidic) bottom waters dominated by 490 respiratory processes (i.e., degradation of organic matter) similar to the present-day Black Sea (Fry et al., 1991;491 Nägler et al., 2011). LL2 d 13 C values range from -11 to -5 ‰, which is comprised between the d 13 C of bulk organic 492 matter in core P33 (-17 to -20 ‰) and bottom water DIC in the euxinic Black Sea (-6.3 ‰, Fry et al., 1991) (Fig.  493  4). The bathymetric d 13 C gradients from surface to bottom waters in core P33 is also similar to that of the present-494 day Black Sea. The d 13 C ranges from 0 ‰ at the surface (as indicated by G. ruber d 13 C, Fig. 4)  likely prevented the oxidation of methane into CO2 and led to the precipitation of carbonate crusts with a very 504 depleted d 13 C signature (-25 to -45 ‰), typical of hydrocarbon sources and clearly distinct from that of authigenic 505 carbonate layers found in core P33 (Fig. 4). However, the incorporation of methane generated at the sulphate-506 methane transition zone from fermentation of sedimentary organic matter would provide a likely source of depleted 507 DIC to account for the low d 13 C of HL. It is therefore possible that methane-rich conditions developed temporarily 508 in bottom waters and led to the formation of the HL sublayers found in cores P33 and P73 (Fig. 2)  Blue Nile-dominated/low-salinity water, which suggests that the formation of calcite in LL2 layers is related to the 515 flood season. This also supports the hypothesis of an efficient diapycnal mixing during summer floods discussed 516 earlier (Parsons et al., 2001). A potential mechanism could be the rising inflow of freshwater at the end of the 517 spring/beginning of the summer, providing cations to the sulphate-depleted and bicarbonate-rich waters that 518 developed as a result of anaerobic oxidation of organic matter off the Nile mouth (as derived from d 13 C signature 519 of LL2). At the end of the dry (winter) period, the Nile River water is indeed enriched in cations (Ca 2+ , Mg 2+ ) 520 (Dekov et al., 1997). In absence of additional evidence, we therefore propose that the onset of the summer flood 521 constituted a potential trigger to precipitate calcite by supplying cations to stagnating, anoxic to euxinic alkaline 522 waters. The existence of euxinic water masses was suggested for sapropel layer S5 (

Spatial and temporal variability of deoxygenation in the eastern Mediterranean during sapropel S1 533
Based on the observations of seasonal sublayer deposits and in particular the occurrence of authigenic calcite, we 534 derived that bottom waters offshore the Nile mouth were anoxic almost all-year round during sapropel S1 535 deposition and potentially reached euxinic and methane-rich states in spring. As similar sublayers of authigenic 536 calcite were found in laminated intervals of cores located at shallower depths (P73 and P99, Fig. S3), we propose 537 that similar chemical conditions occurred intermittently at depths ranging between 400 and 700 m on the western 538 Nile deep-sea fan. 539 By comparing available markers of deoxygenation in the three sediment cores (in particular the S/Cl records and 540 occurrence of laminations), we can draw a picture of changes in seawater chemistry at centennial-to millennial-541 scale along a bathymetric transect (Fig. 9d). We have added two deeper cores MD04-2726 and MS27PT (retrieved 542 at 1060 and 1390 m w-d, resp.) for which S/Cl ratios were published previously (Ménot et al., 2020;Revel et al., 543 2015). This allows exploring changes in oxygenation state along a bathymetric transect within a newly-defined 544 water mass located between 500 and 1800 m w-d in the Eastern Mediterranean during sapropel S1 deposition, 545 labelled Sapropel Intermediate Water (SIW) (Zirks et al., 2019). In order to determine potential forcing factors, 546 this bathymetric and temporal record of paleo-oxygenation will be compared with regional reconstructions of 547 terrigenous input and primary productivity (Fig. 9a,b). As seen by microfacies analyses (section 5.1, Fig. 3), 548 calcium mainly occurs in the marine biogenic and authigenic carbonate component (coccolithophores, foraminifera 549 and authigenic calcite layers), whereas K and Ti occur in respectively finer-and coarser-grained detrital layers 550 (Fig. 3). 551
Titanium reflects the erosion of volcanic rocks in the Blue Nile watershed (Blanchet et al., 2013;Garzanti et al., 561 2015), while potassium is enriched in clay minerals such as smectite and is related to low-energy Nile runoff (Billi 562 and El Badri 2010). Therefore, we confirm the interpretation the Ti/Ca and Ti/K records as tracers of relative 563 variations of terrigenous input and grain size (Blanchet et al., 2013) (Fig. 8a,e,i). The dynamics of terrigenous input 564 and primary productivity are coherent over the western Nile DSF as shown by the good alignment of Ti/Ca ratios 565 measured in all cores (Fig. 9a) (Fig. 7, 9b). Biomarker records can be affected by post-567 depositional oxidation, especially by the downward diffusion of oxygen into the reduced sapropelic sediments (i.e., 568 a so-called burn-down process, Rutten and de Lange, 2003). However, comparable trends between biomarker 569 records and other markers of primary productivity from regional archives that are not affected by burn-down 570 processes (i.e., flux of planktonic foraminifera in core PS009PC, Mojtahid et al., 2015) suggest that productivity 571 played a larger role than post-depositional oxidation in modulating the biomarker fluxes in our records (Fig. 9b). All cores from the western NDSF show a pronounced decrease in oxygenation between 10 and 6.5 ka BP, marked 583 by higher S/Cl records and the absence of benthic foraminifera, followed by relatively stable and oxic conditions 584 in the younger sections of the cores (Fig. 9d). Anoxic (but not euxinic) conditions are also found at more distal 585 locations along the path of the Levantine Jet as indicated by high vanadium to aluminium (V/Al) ratios between 10 586 and 7 ka BP in cores PS009PC and 9509 (located at 552 and 884 m water depth, resp.; Fig. 9e) (Hennekam et al., 587 2014;Matthews et al., 2017). At the scale of the Levantine basin, suboxic to anoxic conditions developed during 588 deposition of sapropel S1 at depths ranging from 500 to >3000 m and euxinic conditions in water-depths >2000 m 589 (ODP site 967D, Azrieli-Tal et al., 2014) and potentially in cores P33 and P73 offshore the Nile mouth (Fig. 10). 590 There are inconsistencies as to whether the onset of anoxia occurred synchronously at all depths (Schmiedl et al., 591 2010) or spread from shallower to deeper depths between 10 and 9 ka BP (Zirks et al., 2019), which may at least 592 partly be explainable by dating uncertainties. 593 A complete reoxygenation is achieved at all depths by 6.5 ka BP (and by 6 ka at 1060 m water depth, Revel et al., 594 2015). At the scale of the eastern Mediterranean, the recovery of benthic ecosystems appears to have followed a 595 time-transgressive pattern, with shallower sites depicting oxic conditions as early as 8 ka BP (Schmiedl et al., 596 2010). Such a pattern is also evidenced on the western NDSF with deoxygenation conditions being maintained 597 until ca. 6.5-6 ka BP at depths >1000 m while shallower sites already started recovering around 7.5 ka BP (Fig.  598   9d). a pronounced shoaling to 400 m w-d is observed around 8.5 ka (Fig. 9d). Intermediate depth cores also show 606 variations in the oxygenation state with lower S/Cl records around 9.2 and 8.3 ka BP recorded in both cores P33 607 and P73 (although more prominent in P33) and at 8.7 ka BP recorded in core P33 ( Fig. 6 and 9c). Millennial-scale 608 variations in bottom-water oxygenation were also recorded at similar depths by nearby cores PS009PC and 9509 609 as evidenced by fluctuations in the RSER (Fig. 9e)  likely been more stable during S1a, as suggested by continuously laminated sapropels and relatively constant S/Cl 612 records in deeper cores MD04-2726 and MS27PT (Fig. 6, 9d). Due to complex diagenetic processes involving 613 sulphur cycling, the S/Cl may not have recorded rapid changes in the oxygenation state as shown by RSER 614 (Matthews et al., 2017;Tachikawa et al., 2015). However, the presence of continuous laminations throughout the 615 sapropel sequence strongly suggests stable anoxic conditions at depths >1000 m. An anoxic water-mass was present 616 between 500 and 3000 m w.-d. in the eastern Mediterranean during S1a, although two sites around Eratosthenes 617 Seamounts showed larger variations in oxygenation state between suboxic conditions at ca. 900 m water depth and 618 euxinic conditions at ca. 2550 m water depth (Fig. 10). 619 620 https://doi.org/10.5194/cp-2020-114 Preprint. Discussion started: 2 October 2020 c Author(s) 2020. CC BY 4.0 License. Figure 10. Maps of deoxygenation conditions in the Eastern Mediterranean during different intervals of sapropel S1: S1a, S1 622 interruption and S1b. Oxygenation conditions: oxic (orange), suboxic (light grey), anoxic (grey) and euxinic (sulfidic, dark grey) in 623 archives located at various water depths between 400 and 3000 m (bathymetric scale at the right of the map for S1b). The dark grey 624 contour indicates the water-depth 1800m, which is considered the lowest bathymetric extend of the SIW (Zirks et al., 2019)  Mediterranean generally referred to as sapropel S1 interruption, which has been related to a drastic decrease in Nile 631 runoff around the 8.2 ka event (Blanchet et al., 2013;Rohling et al., 2015). In nearby cores located above 900 m 632 water depth at the cost of Israel, the timing of this reoxygenation event was similar ( Hypoxic conditions developed again on the western NDSF around 7.8-7.5 ka BP (sapropel S1b) (Fig. 9d). If bottom 642 waters were suboxic at 740 m as indicated by the presence of a few benthic foraminifera and faint laminations in 643 core P33, laminated intervals and high lycopene contents in cores P73 and P99 suggest that anoxic conditions 644 existed intermittently between 600 and 400 m water depth during S1b (Fig. 6a,b, 9d). At deeper sites, conditions 645 remained anoxic until ca. 6.7 ka BP when laminations stopped in core MS27PT (Fig. 9d). This suggests the 646 existence of split anoxia, during which both shallower and deeper water masses were deprived of oxygen due to 647 the respiration of rapidly-sinking organic detritus, while intermediate water masses remained suboxic (Bianchi et 648 al., 2006;Rush et al., 2019). In most cores from the eastern Mediterranean, suboxic to anoxic conditions were re-649 established by 7.8 ka BP but generally lasted only a few hundred years and were rather unstable ( Recent modelling experiments indicated that the formation of a basin-wide oxygen depletion in the Mediterranean 654 Sea during sapropel S1 required a multi-millennial deep-water stagnation (Grimm et al., 2015). It was recently 655 determined that the increasing sea-level during the deglaciation allowed the inflow of fresher Atlantic-derived for modelling experiments to include nutrient loading as a forcing factor, since it is presently rising due to fertilizer 688 use in the Nile valley (Nixon, 2003;Powley et al., 2016). 689

Conclusions 690
By combining microfacies analyses with downcore geochemical measurements, our study provides a first 691 estimation of changes in oxygenation conditions of the bottom waters off the western mouth of the Nile River in 692 the eastern Mediterranean. The regular seasonal alternation of detrital, biogenic and chemical sublayers in the 693 laminated sequence deposited during sapropel S1 in a set of cores from different water depths on the western Nile 694 deep-sea fan are interpreted in terms of seasonal changes (Fig. 5c). Strong summer Nile floods during S1 led to the 695 deposition of thick (up to a few mm) silt-sized detrital sublayers that dominate total layer thickness. The deposits 696 of our study thus confirm the large influence of Nile runoff on the eastern Mediterranean realm during sapropel 697 deposition. The large floods triggered surface blooms of planktonic foraminifera and coccoliths in autumn, which 698 are similar to historical "Nile blooms" described during the last natural flood of the Nile River (Halim et al., 1967). 699 The subsequent deposition of clay-rich detrital sublayers was associated with the low discharge regime of the Nile 700 during winter. The occurrence of inorganic carbonate sublayers in several laminated cores from the western Nile 701 deep-sea fan suggests that bottom waters reached a supersaturation state for calcite. The depleted d 13 C signature of 702 these sublayers points to the existence of anoxic to euxinic (and sometimes methane-rich) bottom waters above the 703 fan accompanied by a high level of anaerobic remineralisation of organic matter leading to high alkalinity. The 704 most likely process initiating the deposition of these layers was the onset of the strong Nile floods as a consequence 705 of increased precipitation, which supplied sufficient amounts of cations to the seawater. Our pilot measurements 706 on the sub-millimetre layers underpin their potential to reconstruct seawater chemistry at times when no benthic 707 fauna existed. 708 On millennial time-scales, we show that variations in oxygenation dynamics followed changes in primary 709 productivity driven by nutrient fertilisation during high Nile runoff. Deoxygenation above the fan shoaled to water 710 depths as shallow as 400 m offshore the Nile mouth and varied on centennial-to millennial-scales in the upper 750 711 m water depth. In contrast, the records of the cores located below 1000 m water-depth reflect more stable anoxic 712 conditions between 10 and 6.5 ka BP. The development and fluctuations of anoxic conditions during sapropel S1 713 are coherent regionally and therefore suggest a common control. Even though multi-millennial development of 714 deep-water stagnation was posited as a necessary prerequisite to basin-scale anoxia in the eastern Mediterranean 715 (Cornuault et al., 2018;Grimm et al., 2015), our records and data compilation show that changes in primary 716 productivity in the surface drove the rapid changes in oxygenation state through eutrophication processes. Indeed, 717 tight temporal links between regional productivity records, oxygenation and runoff dynamics, as well as the 718 evidence of stronger deoxygenation close to the Nile mouth point to a pivotal role of runoff-driven fertilisation. 719 Furthermore, the rapid switch towards anoxic conditions around 10 ka BP suggests the existence of thresholds or 720 tipping points, which still remain elusive. Such processes should be further explored using modelling experiments 721 incorporating new boundary conditions, such as nutrient loading from Nile runoff, either for forecasting future 722 deoxygenation dynamics or for understanding feedback processes in the past. 723

Supplement Link 732
For the present submission, supplementary figures are provided in a separate file. 733 Author contribution 734 CLB designed the study, measured and analysed the data. RT measured and analysed the XRF elemental contents. 735 AES assisted with XRD measurements. SS, MF and AB provided guidance, lab space and logistic support for 736 biomarker measurements, sedimentology and micro-facies analyses, respectively. CLB wrote the manuscript, to 737 which all co-authors contributed. 738