Five sampling campaigns were carried out between May 2012 and February 2013 at water depths of 19 m (site RS-II-19, ex-fish farm location), 21 m (site RS-V-21), 306 m (site RS-IV-306), 420 m (site RS-III-420), and 694 m (site RS-I-694) (Figure 1 and Table 1). Sampling sites map was generated using GeoMapApp¹ (Ryan et al., 2009). At the deep-water sites, cores were retrieved by a multicorer from the IUI-Eilat research vessel. Near-shore sampling was performed manually by divers. The core liners were 50 cm long Perspex tubes with 9.5 cm internal diameter. Three cores were retrieved during each sampling. These cores were used for high resolution dissolved oxygen concentration measurements (Core 1), total sediment composition analysis (Core 2), pore-water extraction and chemical analyses (Core 3).

The pore-waters were extracted at 4 cm intervals by Rhizon samplers (MacroRhizon, 9 cm long, 4.5 mm diameter, Rhizosphere Research Products, Netherlands) with a filter pore size of 150 ± 30 nm, under a nitrogen atmosphere in a glove box (Seeberg-Elverfeldt et al., 2005).

The first 1 mL of every sample was discarded to avoid sample contamination. The volume of pore-water extracted at each core depth was c.a. 50–80 ml. The extraction of large pore-water volumes was crucial in order to perform numerous geochemical analyses for a variety of dissolved species. The vertical overlap between pore fluid samples retrieved with the Rhizons was calculated according to Seeberg-Elverfeldt et al. (2005) and was found to be in the range of 0–14%.

Ferrous (Fe²⁺) iron was quantified spectrophotometrically by the ferrozine method according to Stookey (1970). The minimum detection limit (MDL) of this method is 1 μmol L^-1. Manganese in pore-waters was also measured spectrophotometerically (LaMotte, SMART Spectro) using the 1-(2-Pyridylazo)-2-naphthol (PAN) method (LaMotte test kit 3660-sc) according to Goto et al. (1977). Ascorbic acid was used to reduce total manganese to dissolved Mn²⁺.

Hydrogen sulfide was measured by two methods. For samples with hydrogen sulfide concentration >1.5 μmol L^-1 samples were pre-treated with 50 g L^-1 zinc acetate at a ratio of at least 3:20 (V/V) immediately after sampling and the concentration was quantified by the methylene blue method according to Cline (1969). This method accounts for the following species: H₂S, HS^-, S^2-, S(II) in polysulfides, as well as nanoparticles of acid-soluble metal sulfides (e.g., MnS, FeS), which can pass through the pores of the Rhizon sampler. The MDL was 1.5 μmol L^-1. For samples with hydrogen sulfide concentration <1.5 μmol L^-1, quantification was performed by high-precision liquid chromatography (HPLC, Agilent Technologies 1260 Infinity) with fluorescence detection after derivatization with monobromobimane (Kosower et al., 1979; Fahey and Newton, 1987; Zopfi et al., 2004). The sulfur species measured by this method are: H₂S, HS^-, S^2- and, possibly, the S(II) of polysulfides. The MDL for the sum of these species was 2 nmol L^-1. Dissolved zero-valent sulfur (ZVS, includes dissolved and colloidal elemental sulfur and well as polysulfide zero-valent sulfur) was quanified by cyanolysis according to Kamyshny (2009a). The MDL was 60 nmol L^-1. The concentrations of polysulfides were below the detection limit of chromatographic quantification after derivatization to dimethylpolysulfanes with methyl trifluoromethanesulfonate (MDL = hundreds of nmol L^-1 for individual polysulfides, Kamyshny et al. (2006)) in all samples. The analyses of thiosulfate and sulfite were carried out simultaneously by the same HPLC analysis procedure as for low hydrogen sulfide concentrations. The MDL was 2 nmol L^-1. Free thiocyanate (SCN^-) and tetrathionate (S₄O₆^2-) concentrations were below the detection limit (60 nmol L^-1 and 500 nmol L^-1, respectively) of chromatographic analyses (Rong et al., 2005; Kamyshny, 2009a). Sulfate (SO₄^2-) was quantified by ion chromatography (DIONEX, DX500) with ASRS-300 suppressor. The eluent, which was composed of 1.8 mmol L^-1 sodium carbonate and 1.7 mmol L^-1 sodium bicarbonate, was pumped at a flow rate of 2 mL min^-1 through a guard column (AG4A-SC), an anion exchange column (AS4A-SC). The MDL of this method was 10 μmol L^-1.

Sediments were sliced in the field at 2 cm intervals using a core extruder and a spatula. For acid-volatile sulfur (hereafter AVS), non-S⁰ chromium reducible sulfur (CRS) and elemental sulfur measurements, c.a. 15 mL of sediment were placed in a 50 mL volume falcon tube containing 25 mL zinc acetate in the field (50 g L^-1). Additionally, for metal speciation measurements, c.a. 25 mL of sediments were placed in a 50 mL falcon tube and frozen the same day on arrival to the laboratory. For porosity measurements, 10 mL of sediment were placed in a 15 mL volume falcon tube and stored at 4°C in the laboratory.

Elemental sulfur was extracted from the sediment that had been pre-treated with zinc acetate by shaking with methanol for 16 h on a rotary shaker (Zopfi et al., 2004). The sample-to-methanol ratio was c.a. 1/20 [w/v]. After separation of the sediment by centrifugation, elemental sulfur (S⁰) in the methanol extract was quantified by HPLC. An Agilent Quaternary pump – G1311B, UV-VIS Detector (Agilent G1365D), and Prevail C18 Grace reverse phase column (250 mm × 4.6 mm × 5 μm) were used. Pure methanol (HPLC grade) was used as a mobile phase at a flow rate of 1 mL per minute. Elemental sulfur was detected at 230 nm wavelength with an MDL of 10 μmol kg^-1 (wet sediment). In two samples of surface sediments at site RS-V-21, the elemental sulfur content was high enough to allow analysis of isotopic composition. Elemental sulfur was preconcentrated by partial evaporation of methanol on a rotary evaporator, extraction with dichloromethane from the residual methanol-water mixture and evaporation of dichloromethane under gentle flow of nitrogen. Elemental sulfur was reduced to hydrogen sulfide by Cr(II) reduction in the ethanol-water medium according to Gröger et al. (2009). Evolving hydrogen sulfide was trapped in AgNO₃/HNO₃ trapping solution.

Acid-volatile sulfur and CRS distillations were performed on the sediment sample after it had been subjected to methanol extraction, so that elemental sulfur was not accounted for in the CRS distillation. Although greigite (Fe₃S₄) is an important precursor for pyrite, its content in sediments is usually low (Morse and Cornwell, 1987). Thus, the main sulfur mineral recovered by CRS was pyrite sulfur, which will be referred to as pyrite-S (or Py-S) hereafter. The sediments were subjected to a two-step distillation. AVS was extracted over two hours by boiling in 5 M HCl, which was followed by 3 h boiling with 1 mol L^-1 acidic CrCl₂ solution (extracting CRS) (Fossing and Jørgensen, 1989). The MDL was 10 μmol kg^-1 wet sediment or 50 nmol total pyrite-S.

Total iron and manganese concentrations were measured according to Aller et al. (1986). For quantification of iron and manganese SpectrAA 300/400 Series Varian AAS was used. Results are presented in moles of metal per kilogram of wet sediment after correction based on sediment porosity. The MDL was 2 μmol kg^-1 (wet sediment) for iron and 0.6 μmol kg^-1 (wet sediment) for manganese. For quantification of reactive iron species (Fe_HR) a sequential extraction scheme was used according to Poulton and Canfield (2005). Quantification was performed by AAS, and the MDL was 2 μmol kg^-1 Fe (wet sediment). Pyritic iron was measured on a subsample by the CRS protocol.

For porosity analysis, the weight and volume of wet sediment and of sediment dried at 50°C for one week were measured. The porosity was calculated as the fraction of pore-water volume in the wet sediment. Total organic carbon (TOC) was quantified by an elemental analyzer (LECO SC632). The MDL was 0.005% TOC with precision of 1%.

For preparation for sulfur isotopic analysis of AVS and pyrite-S, zinc sulfide that was formed during hydrogen sulfide distillation was converted to silver sulfide by addition of an excess of silver nitrate solution. Each Ag₂S sample was aged for at least one week and cleaned with sequential washes of 4 × 50 mL Milli-Q water, rinsed in 50 mL of 1 M NH₄OH overnight, then washed with 3 × 50 mL Milli-Q water and dried at 55°C. For the isotopic analysis of sulfate, sediments were centrifuged, the supernatant was filtered, and an excess of barium chloride solution was added to the filtrate in order to precipitate BaSO₄. The precipitate was washed with 3 × 50 mL Milli-Q water and dried at 55°C.

For the analysis of δ¹⁸O_SO4, barium sulfate was pyrolyzed at 1450°C in a temperature conversion element analyzer (TC/EA), producing carbon monoxide. Carbon monoxide was measured by continuous helium flow in a GS-IRMS (Thermo Finnegan Delta V Plus, at the Godwin Laboratory, University of Cambridge). Analyses of δ¹⁸O_SO4 were performed in replicates (n = 3–5) and the standard deviation of the replicate analysis is reported (∼0.4‰ 1σ). To analyze the δ³⁴S_SO4, δ³⁴S_Py-S, δ³⁴S_{AV S}, δ³⁴S_S0 values, barite or silver sulfide with an excess of vanadium pentoxide were combusted at 1030°C in a flash element analyzer (EA), and the resulting sulfur dioxide (SO₂) was measured by continuous helium flow on a GS-IRMS (Thermo Finnegan Delta V Plus Godwin Laboratory, University of Cambridge). The analytical error for the δ³⁴S analysis was determined using the standard deviation of standards run at the beginning and the end of each run (long term reproducibility of our standards is ∼0.3‰ 1σ). δ¹⁸O_SO4 is reported relative to the Vienna Standard Mean Ocean Water (VSMOW) and δ³⁴S is reported with respect to Vienna Canyon Diablo Troilite (VCDT).

In the vicinity of the north shore of the Gulf of Aqaba (sites RS-II-19 and RS-V-21), sediments contain grayish sandy particles, suggesting the source of the sediment is terrigenous material derived from adjacent rocks by erosion and/or weathering and transported by seasonal floods. The sediments from near the fish farm contained black fragments which are likely residual material from fish cages, consistent with the composition of sediment cores collected during the operation period of Ardag fish farm (Katz et al., 2002). The bulk composition of the north shore sediments is dominated by siliciclastic material with lower carbonate content (6.17%; Al-Rousan et al., 2006) compared to coral reef sediments (74.8%; Al-Rousan, 1998). The sediment from the intermediate and deep-water sites (RS-IV-306, RS-III-420, and RS-I-694) is characterized by a much finer grain size with a higher content of light brown clay and mud. These textural properties at the deeper-water sites reflect the aeolian source of particles derived from desert soils (Garrison et al., 2003).

Dissolved iron was present in the sediments of the deep-water sites (RS-IV-306, RS-III-420, and RS-I-694) at all depths below seafloor, while at the shallow water site that were not influenced by the fish farms (RS-V-21) no dissolved iron was detected below 3 cm below seafloor (Figure 2A). Dissolved manganese concentrations increase with overlying water column depth. Concentrations above 10 μmol L^-1 were detectd only at the deepest site RS-I-694 (Figure 2B).

Sulfate concentrations were constant with depth at all the sampling sites, in the range of 32–35 mmol L^-1 (data not shown). Hydrogen sulfide concentrations were below 0.34 μmol L^-1 at the deep-water sites (RS-IV-306, RS-III-420 and RS-I-694), up to 1.3 μmol L^-1 at former fish farm site RS-II-19, and up to 12 μmol L^-1 at site RS-V-21 (Figure 2C). Zero valent sulfur (ZVS) was below detection limit at sites RS-II-19 and RS-IV-306, below 200 nmol L^-1 at sites RS-III-420 and RS-I-694 and up to 301 nmol L^-1 at site RS-V-21 (data not shown). The highest thiosulfate concentrations were detected at the shallow water sites (up to 4.05 μmol L^-1 at site RS-V-21). At sites RS-IV-306 and RS-I-694, thiosulfate concentrations were <0.1 μmol L^-1 (Figure 2D). Sulfite concentrations in the pore-waters were lower than thiosulfate concentrations. At the shallow sites (RS-II-19 and RS-V-21), sulfite concentrations were 78–972 nmol L^-1. Sulfite concentrations at the deeper sites were <150 nmol L^-1 (Figure 2E).

The porosity of the sediments increased with increasing water depth as well as with depth below the sediment-water interface (Figure 3A). TOC content was in the range of 0.1-0.6 wt% at all sites. The AVS and pyrites sulfur contents of the sediment decreased with an increase in water column depth. At the shallow site not influenced by anthropogenic pollution (RS-V-21), AVS content decreased with depth below the seafloor and pyrite sulfur increased with depth below the seafloor. At site RS-II-19 (the site impacted by the fish farm) the lowest AVS concentration was detected in the upper 10 cm of sediments (Figures 3B,C). In the upper 10 cm of the sediment, elemental sulfur content was higher at the shallow water sites than at the deep-water sites (Figure 3D).

Sediments from the deep-water sites were enriched in total manganese in the upper two centimeters of the core. At RS-I-694 site total manganese content reached 26.0 mmol kg^-1 of wet sediment. In cores taken from shallow-water sites (RS-II-19 and RS-V-21), the total manganese content was below 7 mmol kg^-1 of wet sediment (Figure 3E). Sediments in the deep-water sites were enriched in iron compared to shallow-water sites, especially in the upper 14 cm of the sediment. The highest total iron content was detected at site RS-II-420 (Figure 3F). Traditionally, highly reactive iron is defined as the sum of iron species which are reactive toward hydrogen sulfide and iron which has already reacted with sulfide (e.g., pyrite iron). The content of highly reactive iron, which in our study includes pyrite iron, acetate-, hydroxylamine-, dithionite-, and oxalate-extractable iron, increases with water depth (Figure 3G). As concentrations of pyrite are higher at the shallow-water sites (Figure 3C), concentrations of iron species which are still reactive toward hydrogen sulfide increase further with water depth. At the shallow-water sites (RS-II-19 and RS-V-21) concentrations of highly reactive iron slightly increased with depth (Figure 3G). Sodium acetate-extractable iron, which accounts for Fe(II) carbonate minerals such as siderite and ankerite represented less than 10% of the highly reactive iron at all water depths. Dithionite-extractable iron (goethite + hematite + akaganéite) represented the most abundant fraction of highly reactive iron in the sediments of the Gulf of Aqaba (data not shown).

The isotopic composition of the Gulf of Aqaba seawater sulfate was measured in five samples. Three samples were taken at the surface at various locations and two samples were retrieved from 20 and 700 m depths. The isotope composition of sulfate at various locations and water depths within the Gulf of Aqaba is similar (δ³⁴S = 20.3 ± 0.2‰, δ¹⁸O = 8.6 ± 0.2‰), and does not differ more than the error of the analytical method. At all sites, except at the former fish farm (site RS-II-19), the δ³⁴S of sulfate near the top of the sediment was similar to the overlying water (20.3–20.5‰). At these sites, the δ³⁴S increased with depth below the seafloor to 21.2–23.1‰. At the former fish farm site, sulfate sulfur was isotopically lower than seawater sulfate, possibly due to re-oxidation of isotopically light reduced sulfur minerals formed during fish farm operation (Figure 4A). At all depths bsf, the δ¹⁸O of sulfate was higher than seawater sulfate (9.4–10.4‰ in the upper 2 cm of sediment vs. 8.6% in the water column). For the deep-water sites, a slight increase in δ¹⁸O of sulfate was observed with depth (Figure 4B).

The sulfur isotopic composition of AVS is only reported for the shallow-water sites, as at the deep-water sites the AVS content was too low to perform sulfur isotopic analysis. The most ³⁴S-enriched AVS was detected at both shallow-water sites in the uppermost samples: δ³⁴S = –18.4‰ to –19.3‰. At 17–29 cm bsf, the δ³⁴S of AVS was in the range of –21.6‰ to –25.4‰ (Figure 4C). Enough pyrite-S for sufur isotopic analysis was retrieved from all sediments at the two shallow-water sites, three samples from site RS-IV-306 from 14 to 20 cm, and one sample from site RS-I-694 at 18–20 cm (this point cannot be seen on the figure as it coincides with the point for RS-IV-306 profile). Pyrite sulfur was isotopically lower in δ³⁴S than AVS sulfur. At deepwater sites RS-IV-306 and RS-I-694, pyrite was isotopically lower in δ³⁴S than at the shallow-water sites (Figure 4D). The isotopic composition of elemental sulfur was measured in the 0–2 cm and 2–4 cm intervals in the sediments from site RS-V-21and the δ³⁴S was –20.9‰ and –22.4‰, respectively.

The concentration of total manganese in the solid sediment increases with the depth of the overlying water column (Figure 3E). This increase is likely caused by greater contribution of aeolian dust to the sediment further from the shore. The atmospherically transported material from the Sahara and adjacent deserts has a high manganese content, as much as 880 ppm (16 mmol kg^-1) in Saharan dust (Mendez et al., 2010). In contrast, the shallow sediment has a dominance of fluvial input with lower manganese content. An enrichment in solid manganese in the upper 2 cm of the sediment was observed in the deep-water sites (Figure 3E). This solid manganese enrichment results from the oxidation of reduced aqueous manganese (Mn²⁺), which is produced by microbial manganese reduction coupled to organic carbon oxidation. This reduced Mn²⁺ produced deeper in the sediment diffuses toward the water-sediment interface where it is oxidized to manganese oxides. The only site at which concentrations of dissolved manganese were higher than concentrations of dissolved iron was deepwater site RS-I-694 suggesting a dominance of the manganese cycle at this location.

The concentration of total sedimentary solid iron, similar to total manganese, increases as the depth of the overlying water column increases, except for the deepest site (RS-I-694) (Figure 3F). As with manganese, this increase may be explained by greater contribution of aeolian dust away from the shore. The fine-grain sediment that is transported from the Sahara and adjacent deserts by wind has a high iron content (Chase et al., 2006). The lack of a significant increase of solid total iron in the surface sediment layer is likely due to low iron sulfide solubility, which causes retention of iron in the sediment even at low concentrations of hydrogen sulfide in the pore-waters (Figures 2A,C).

Highly reactive iron is traditionally defined as the fraction of iron in the sediment that can react with hydrogen sulfide to produce FeS or FeS₂ (Berner, 1970). There are different pools of iron that are defined based on their reactivity toward hydrogen sulfide. Ferrihydrite and lepidocrocite react with hydrogen sulfide within hours. The most abundant iron minerals in dust from the Saharan Desert, such as hematite and goethite, react with hydrogen sulfide on time scales of days (Poulton et al., 2004) and years (magnetite) (Canfield et al., 1992). For iron in sheet-silicate minerals it may take up to millions of years to react with hydrogen sulfide (Canfield et al., 1992; Raiswell and Canfield, 1996). Highly reactive iron concentrations are lower at the shallow water sites (RS-II-19 and RS-V-21) than at the sites overlain by deeper waters. We suggest that the difference in the reactive iron content could be due to the difference in the highly reactive iron content of dust (relatively high) and desert soil and rock material transported by flash floods (relatively low), which predominantly impact shallow-water sediments. Although detailed studies of reactive iron speciation of minerals and soils in the various locations in arid Arava valley are not yet published, total iron content in granite was recently reported for two locations near Eilat: Wadi Shelomo and Mt. Rehavam with 0.30–0.76% Fe₂O₃ and 0.12–1.60% Fe₂O₃, respectively (Eyal et al., 2010). The iron in these locations is mostly associated with biotite (Bogoch et al., 1997). The detailed study of iron speciation in the seasonal stream beds, which is out of scope of this study, should be performed in order to evaluate this hypothesis.

The concentration of hydrogen sulfide in the pore-water decreased with an increase in water column depth from 0.5 to 12 μmol L^-1 at the shallow-water sites to ≤30 nmol L^-1 at site RS-I-694. The latter concentration was detected by HPLC with a fluorescence detector and is a concentration we usually consider to be below the detection limit of standard spectrophotometric techniques (1 μmol L^-1 – Cline, 1969). Despite extremely low concentrations of hydrogen sulfide in the deeper sediments, we use three lines of evidence to confirm the presence of a sedimentary sulfur cycle in the Gulf of Aqaba.

An integrated scheme of processes affecting cycling of redox-sensitive elements in the Gulf of Aqaba is presented in Figure 7. Sediments overlain by deep water receive predominantly input of aeolian dust, which has high manganese and reactive iron content. At these water depths, microbial iron and manganese reduction is preferred over sulfate reduction. In spite of a microbial cycle dominated by iron and manganese cycling, there is microbial sulfate reduction and trace amounts of pyrite accumulating. However, the vast majority of the hydrogen sulfide produced is reoxidized in the presence of the high concentration of iron and manganese oxides to sulfide oxidation intermediates, which are further oxidized or disproportionated to form the terminal oxidation product, sulfate. These processes induce the high concentration of sulfide oxidation intermediates in pore waters and leave fingerprints in oxygen isotope composition of the sulfate. Thus, at these deep water sites the sulfur cycle is mostly cryptic.

Shallow water sites are influenced by a combination of aeolian dust input and fluvial iron and manganese inputs. At these sites, the manganese (especially close to the sediment-water interface) and reactive iron contents are lower than at the deep-water sites. Lower reactive iron content allows relatively high amounts of hyrdogen sulfide to be present in the pore-waters and to be further preserved in the form of pyrite.

In summary, this work shows that in arid environments, such as the Gulf of Aqaba, aeolian dry deposition from the surrounding deserts results in an increased flux of reactive iron and manganese to the sediments. Such increase leads to fast reoxidation of hydrogen sulfide and prevents pyrite formation in the sediments. Future work should concentrate on the detailed study of speciation of iron and manganese in the aeolian and fluvial sedimentary sources. Another direction of future research should include the study of impact of aeolian dry deposition on sedimentary sulfur cycle in other marine systems affected by aeolian dust deposition. Such research efforts are required in order to understand whether these considerations are unique to the Gulf of Aqaba or may be applied to other marine systems situated in arid regions.