Between 1 and 12 February 2019, the FUNAMOX (Biogeochemical functioning in Namibian oxygen-depleted waters) research expedition with RV Pelagia (cruise code 64PE449) was undertaken on the shelf and slope in the northern part of the BUS off the coast of Namibia between 22° and 26° S (Fig. 1). In the research area, shelf waters (up to ∼ 150 mbss) are characterized by high productivity as a result of upwelling of high-nutrient and low-oxygen waters from the deep sea (Chapman and Shannon, 1987; Shannon et al., 1987; Summerhayes et al., 1995; Emeis et al., 2017). Upwelling and current patterns result in (i) a perennial OMZ with dissolved O₂ concentrations down to ∼ 50 µmolL-1 on the slope between ∼ 200–400 mbss, and (ii) seasonal bottom-water anoxia on the shelf with recurring water-column anoxia and episodic sedimentary release of hydrogen sulfide (HS⁻) (Mohrholz et al., 2014; Nagel et al., 2016; Ohde and Dadou, 2018; van Kemenade et al., 2022). The northern BUS (NBUS) is a region with its own distinct seasonality of wind forcing and upwelling (Emeis et al., 2017) with most intense oxygen depletion on the shelf in late austral summer, i.e. January–February (Brüchert et al., 2006; Mohrholz et al., 2008).

During FUNAMOX, water and sediment were sampled at shelf and slope stations between 100 and 1500 mbss, from the anoxic shelf to the oxygenated slope below the perennial OMZ. Physicochemical properties were measured throughout the water column at each station using a CTD-rosette equipped with conductivity-temperature-density (CTD, Seabird SBE 911plus), oxygen (Seabird SBE 43), fluorescence and turbidity (Seabird ECU FLNTU) sensors mounted on a rosette with 24 polypropylene water-sampling bottles (volume 12 L). The oxygen sensor was calibrated using O₂ measurements (Winkler titration) in discrete water samples capturing the full range of O₂ concentrations encountered during the cruise (∼ 0–200 µmolL-1). The sensor has an uncertainty of approximately ± 5 µmol O₂ L⁻¹ as indicated by the supplier; concentrations around 0 µmolL-1 (anoxia) should be treated with some caution. Discrete water samples were taken by remotely closing sampling bottles at selected depths for subsequent sub-sampling and (on-board) chemical analyses (see below). Surface sediments were collected with a multi-corer (Oktopus GmbH) equipped with 10 polycarbonate tubes (10 cm diameter).

Immediately after recovery of the CTD-rosette, bottles were sub-sampled for a range of dissolved species as detailed in Table 1. First, a 30 mL sample was taken with a polypropylene syringe and filtered using a 0.8/0.2 µm Pall Acropak syringe filter that was rinsed three times with sample water. Once all bottles had been sampled, water samples were taken into a laboratory container and sub-sampled for: total alkalinity (TAlk); dissolved inorganic carbon (DIC); soluble reactive P (SRP; here referred to as HPO42-, the dominant species at seawater pH of 8.2) and Si (H₄SiO₄); NH4+, NO2- and NO3- (together termed dissolved inorganic nitrogen, DIN). All these analytes were measured by segmented-flow analysis (QuAAtro auto-analyser, AA). Quality control was performed by including a 10-point calibration for each analyte (r = 0.9999–1.0000) and replicate measurement during each analytical run of an internal nutrient standard, which has been validated in various inter-calibration exercises (ICES, Quasimeme, Inter Comparison). Furthermore, two KANSO CRMs (BY, CA) were included in analytical runs for N, P and Si. Dickson CRMs were measured in all DIC and TAlk analytical runs and a correction to the results was applied based on the offset between the measured and certified value (relative standard deviation generally < 5 %). Method detection limits were determined by analysis of 10 replicate samples containing 2 % of the highest standard, limits were (in µmolL-1): 3.5 (DIC, TAlk), 0.005 (HPO42-), 0.019 (NH4+), 0.015 (NO3-), 0.003 (NO2-), 0.008 (Si). Relative standard deviation was < 1 % for replicate measurement of all species. For HS⁻, we used the methylene-blue colorimetric method (Cline, 1969). A stock solution was prepared from Na₂S, validated by iodometric titration and used for a calibration curve between 0 – 100 µmolL-1. The detection limit for HS⁻ was determined as twice the standard deviation of five milli-Q blanks: ∼ 3 µmolL-1. At stations 2 (750 mbss) and 6 (100 mbss), contrasting slope and shelf environments, in addition to SRP and DIN measurements, another set of sub-samples was taken at selected depths for analysis of total (dissolved + particulate) and total dissolved pore-water P and N, see the Supplement. The difference between total dissolved N and DIN was assumed to represent dissolved organic N (DON). The difference between total dissolved P and HPO42- was assumed to represent dissolved organic P (DOP). The difference between filtered and unfiltered totals represents particulate N and P. For dissolved Fe and Mn analysis, filtered pore-water samples were diluted 30-fold with 1 M, twice-distilled HNO₃ and analysed by induced coupled plasma mass spectrometry (ICP-MS, Thermo Scientific Element II). The HNO₃ matrix contained internal standards (In, Rh) to correct for drift. Iron and manganese calibrations, using an element cocktail prepared from certified standards, were included in each analytical run. Acid blanks were included to correct the baseline. Detection limits were < 0.1 µmolL-1 for Fe and Mn.

At selected stations covering the redox transect (2, 3, 4, 6, 7, 8, 10), three multi-core tubes (10 cm diameter) were equipped with Presens (GmbH) oxygen spots and, after core recovery from the seafloor, capped with custom-made gas-tight lids with two ports and a stir motor to which a small stir bar was magnetically attached. The core tubes contained ∼ 25 cm sediment and 35 cm overlying water. The overlying water was continuously stirred (60 rpm) and O₂ was measured every 5 min using a Presens OXY-4 setup with fibre optic cables and sensor spots, operated with the Presens Measurement Studio 2 software. One port of the lid was connected to a 10 L carboy with water from the deepest CTD/rosette bottle collected at the station, usually 5–10 m above the seafloor. The second port was used as sampling port, through which 20 mL samples of the overlying water were taken with an acid-cleaned, 30 mL polypropylene syringe at different time intervals (six sampling points equally spaced over an incubation period of ∼ 10 h; see time series in the Supplement). During sampling, the extracted volume was replenished with bottom water from the connected carboy. The water samples were immediately filtered using 0.2 µm acid-cleaned PES syringe filters and sub-sampled for: DIC, TAlk, major and minor elements (metals), NH4+, NO2-, NO3-, HPO42-, Si, as described in Table 1. The first few mL of sample were used to rinse the filter. At each sampling point, only 2 % of the water was replaced (< 15 % over the whole incubation), minimizing the impact of an evolving difference in chemistry between the water in the container and in the core incubation. Cores were checked throughout the incubation for signs of sediment suspension which would indicate over-stirring. Whole-core benthic fluxes were calculated as described in the Supplement.

Bulk solid-phase elemental profiles showed contrasts in sediment chemistry between the shelf and slope sites (Fig. 2). The CaCO₃ content was relatively low (12 ± 8 wt %, range 0–40 wt %) in sediments on the shelf, while TOC was strongly enriched (mean 12 ± 5 wt %) with a maximum of 26 wt %. The highest CaCO₃ and TOC contents were measured at station 10, which also stood out because of the high and variable P content of up to 1500 µmolg-1 (4.6 wt %), correlating with Ca content. Total contents of Fe, P, S, Mo and V (determined by ICP-MS analysis of sediment digests) were (considerably) higher in shelf sediments compared to slope sediments, where the CaCO₃ content was high (69 ± 7 wt %) and TOC was relatively low: 5 ± 2 wt % (range 3–11 wt %). Solid-phase manganese (Mn) concentrations were very low at all stations, with a maximum of 6 µmolg-1 (data not shown). The shelf stations showed downward increasing TOC / N, most notably at stations 6 and 10 where values increased from 4–7 to 10–11. The molar TOC / N ratios revealed no down-core trend but considerable scatter in slope and OMZ sediment with values of 8–10 and 10–12, respectively (Fig. 3). The molar TOC / P ratio showed large variability between stations across the different depositional environments with values < 100 on the shelf, in the OMZ and on the upper slope (station 2), while values of 100–200 were observed at the deepest slope stations 7 and 1. The lowest values were measured in sediment at station 10: average 28 ± 19, range 6–78. The molar Fe_tot/S_tot ratio was relatively constant down-core at most stations and showed an overall increase from the shelf (< 0.5) to the slope (1–1.5). Station 6 stood out with a strong down-core increase in Fe_tot/S_tot from < 0.1 to a maximum of 2.3 at depth.

The sedimentary Fe pool was dominated by poorly reactive and unreactive Fe (PR-Fe, e.g. silicate-bound Fe) (Fig. 4A), which made up 73 ± 11 % of total solid-phase Fe. Only at sediment depth > 5 cm at station 7 was reactive Fe quantitatively more important than PR-Fe. The major reactive Fe pools were HCl-Fe(II), CBD-Fe and HNO₃-Fe that all accounted for 5 %–20 % of Fe at the selected stations. At Station 4, a small but persistent pool of HCl-Fe(III) was detected.

The solid-phase S pool consisted predominantly of organic S (Fig. 4B), which accounted for 61 ± 17 % of total S. The deeper sediment at station 7, with relatively high reactive Fe, contained relatively large pools of AVS and particularly CRS, together making up more than 80 % of the total S pool. In sediments of the slope stations, reduced inorganic S (RIS) was dominated by AVS and CRS, while AVS and elemental sulfur (ES) made up the RIS pool in sediment of the shelf stations, where CRS was largely absent.

The deepest slope station selected for sequential P extraction (station 2; 750 mbss) showed roughly equally sized pools of Fe-associated, Ca-associated, detrital and organic P, with only minor exchangeable P (Fig. 4C). The sedimentary P pool at slope station 7 showed a similar P sediment chemistry, except that detrital P was nearly absent. Shelf station 4 showed concentrations and composition of P pools similar to station 7, but with a large exchangeable P pool in the surface sediment < 5 cm depth. Shelf stations 6 and 10 showed markedly different sediment P chemistry: total P contents were much higher (values up to 1500 µmolg-1 rather than 50–150 µmolg-1 as found at the other stations), particularly at station 10. Furthermore, P was present predominantly as Ca-associated P with minor contributions of exchangeable and/or Fe-associated P. Note that organic P was not determined for station 10 because of analytical issues; however, mass balance (i.e. total P from sediment digest compared to cumulative P from SEDEX extraction steps) indicated that this was a minor P pool.

Micro-profiles of dissolved O₂ at the sediment-water interface showed an increase in oxygen penetration from < 0.2 mm at shelf stations to 1–2 mm on the slope (Table 3). Pore-water profiles reflected the distinct chemistry of shelf stations compared to OMZ and slope stations (Fig. 5). All DIC profiles showed down-core increases, with much higher concentrations on the shelf stations (up to 15 mmol L⁻¹) compared to the slope (up to 2.5 mmol L⁻¹). Similarly, pore-water NH4+ concentrations reached much higher concentration at depth in the sediment on the shelf (up to 2 mmol L⁻¹) than on the slope (up to 60 µmolL-1). Dissolved nitrate and nitrite concentrations were about an order of magnitude lower than NH4+ (up to 80 and 7 µmolL-1, respectively), with marked peaks at depth in the sediment at the shelf stations, particularly stations 5, 6 and 10. Dissolved HPO42- showed maxima of 100 and 700 µmolL-1 on the shelf and slope, respectively, while dissolved H₄SiO₄ increased downcore to 800 µmolL-1 on the shelf and 400 µmolL-1 on the slope. The shelf stations showed down-core decreases in pore-water F⁻ concentrations (up to 75 %; F⁻ from ∼ 100 to 25 µmolL-1), while dissolved F⁻ was more stable in slope sediments (< 10 % down-core decrease). Dissolved sulfide (HS⁻) was absent or very low in the pore-water at slope stations, where dissolved SO42- concentrations were stable at ∼ 28 mmol L⁻¹. By contrast, HS⁻ concentrations up to 10 mmol L⁻¹ developed at depth in the shelf sediment, with associated decreases in dissolved SO42-. Dissolved Fe and Mn concentrations were low, at most 30 and 2.5 µmolL-1, respectively. Dissolved Fe and Mn showed a pattern of sub-surface peaks that increased in magnitude with increasing station water depth, and Fe additionally showed marked enrichments at depth in the sediment of shelf stations, coinciding with NO3- peaks. Of the shelf stations, station 6 stood out with pore-water peaks in NH4+, HPO42-, HS⁻ and SiO44- in the top sediment rather than at depth. Dissolved organic N was a significant pore-water N pool (up to 50 % of dissolved N) while DOP was a minor P pool, < 10 % (see the Supplement).

Finally, we compared measured HPO42- profiles to calculated “expected” HPO42- profiles based on DIN and DIP release from organic matter breakdown with Redfield stoichiometry (N / P = 16 mol mol⁻¹; see the Supplement for calculation). The measured HPO42- accumulation strongly exceeded the calculated expected HPO42- concentration in the top 5–15 cm of the sediment at all stations except Station 3 (Fig. 6). At greater depth, measured HPO42- profiles show a sharp decline, generally to values below expected values. Notable exceptions are the OMZ stations, where measured HPO42- were low (5–15 µmolL-1) but consistently higher than calculated values.

Calculated diffusive fluxes for O₂, SO42-, N and P across the sediment-water interface show a shelf-to-slope transition from sulfate reduction to oxic respiration as the dominant OM degradation process (Fig. 8, left panels). Fluxes for DIC, TAlk and HS⁻ can be found in the Supplement. The calculated diffusive fluxes of NH4+ and HPO42- were all positive, up to 11 and 4 mmol m⁻² d⁻¹, respectively, suggesting efflux of these nutrients from the sediment into the overlying water at all stations, with relatively high values on the shelf. A small flux of NO3- into the sediment of up to -1 mmol m⁻² d⁻¹ was calculated for all stations except at station 1 where a small efflux from the sediment (0.1 mmol m⁻² d⁻¹) was calculated. The molar N / P ratio of the benthic efflux was generally well below the Redfield N / P ratio of 16 (Redfield C : N : P ≈ 106 : 16 : 1), ranging between 1–7 : mol : mol⁻¹ except for station 3 (19 mol mol⁻¹). Stoichiometry further showed that the ratio of diffusive DIC and NH4+ fluxes (∼ 3) was relatively low compared to Redfield (∼ 6.5). Lastly, station 6 showed a strong efflux of HS⁻ from the sediment into the overlying water (∼ 100 mmol m⁻² d⁻¹), in line with the steep increase in dissolved HS⁻ in the uppermost sediment (SI).

The measured benthic exchange rates generally showed much higher variability than the calculated diffusive fluxes (Fig. 7, right panels). Where they were both obtained (station 2, 7 and 8), diffusive and whole-core O₂ fluxes were similar, ranging between 2 and 6 mmol m⁻² d⁻¹, but no consistent difference between diffusive and whole-core fluxes. Whole-core NH4+ fluxes were similar but relatively high compared to calculated diffusive fluxes, while whole-core NO3- fluxes were relatively high. The trends were similar: NO3- uptake into the sediment at all stations and NH4+ efflux concentrated on the shelf under low-oxygen bottom waters. In contrast to calculated diffusive fluxes, the measured HPO42- fluxes did not show a trend from high to low fluxes from the shelf to the slope, but rather strong variation between stations, with negative (uptake) and positive (efflux) fluxes measured on the shelf.

In addition to pore-water analysis, the bottom water overlying the sediment collected with the multi-corer was sampled and analysed. The results showed that with decreasing O₂, NO3- decreased while NO2-, NH4+, HPO42- and SiO44- increased (Fig. 8). The opposite trends in bottom-water NO3- and HPO42- resulted in a marked decline of the DIN : DIP ratio with decreasing oxygen, from near-Redfield (16) to as low as 3. Moreover, accumulation of dissolved HS⁻ was detected in the bottom water at the stations experiencing strong bottom-water oxygen depletion (Station 4, 6, 10). For context, Fig. 8 also shows the strong increasing trend in surface-sediment TOC content from the slope to the shelf in the bottom left panel.

Water-column profiles of key physical and chemical parameters (Fig. 9) showed rapid oxygen depletion and near-anoxic waters in the bottom 25–50 m on the shelf, except at station 3 where oxygen decline was more gradual. The OMZ stations were located where the core of the perennial OMZ intersected the slope, with oxygen concentrations around the threshold for hypoxic conditions (63 µmolL-1). Below the OMZ, bottom-water dissolved oxygen increased to 179 µmolL-1 (54 % saturation) at the deepest station (1, 1500 mbss). Salinity profiles indicated the presence of a distinct, low-salinity water mass between 500–1000 mbss. The turbidity profiles showed increased turbidity in bottom waters, most prominently at stations 2, 4 and 6.

Dissolved carbon and nutrients (DIC, NH4+, NO2-, NO3-, HPO42-, SiO44-; Fig. 10) showed relatively high DIC concentrations in the top 100 m of shelf waters. Dissolved NH4+ was generally low (< 1 µmolL-1) with small sub-surface (25–50 mbss) peaks of 2–3 µmolL-1 at several OMZ and shelf stations. Furthermore, there was a marked bottom-water NH4+ accumulation up to 10 µmolL-1 at station 6. At this station, a strong enrichment in NO2- was observed in the waters above the deep NH4+ peak. For the rest, NO2- concentrations were < 1 µmolL-1 with small sub-surface peaks coinciding with peaks in NH4+. Generally, the NO3- profiles showed surface-water depletion and downward increases to 30–40 µmolL-1. Notable exception was the NO3--depleted bottom water at station 6, where NO3- decreased from 25 to 0.3 µmolL-1 between 50 and 100 mbss The HPO42- profiles generally showed downward increasing concentrations that stabilized at 2–3 µmolL-1. Station 6 showed an increase in HPO42- in the bottom 25 m to 4 µmolL-1, concurring with the NH4+ peak. The molar DIN : DIP ratio was close to Redfield (14–16) throughout the water column at the slope and OMZ stations, except for lower ratios down to 6 in surface waters. At the shelf stations, by contrast, the DIN : DIP ratio was much more variable and showed decreases both in surface and bottom waters, most prominently at station 6 where the ratio declined to 3–4. Dissolved silica profiles showed down-ward increases up to 50 µmolL-1, with relatively sharp increases in deeper waters on the shelf. Also notable was the decline below 1250 mbss at Station 1, co-occurring with an increase in salinity (Fig. 9).

The measurement of various N and P species in selected water-column samples from stations 2 and 6 suggested small DON (3–6 µmolL-1) and DOP (0.1–0.4 µmolL-1) pools (SI). Furthermore, the difference between total N and total dissolved N in the deeper waters at station 6 (where turbidity was high; Fig. 9) suggested a particulate N pool that was roughly the same size as the DON pool. There was no evidence in the data for a particulate P pool.

Dissolved phosphate concentrations in the shelf sediment (stations 3, 4, 5, 6, 10) were about an order of magnitude higher than on the slope, reaching values up to 600 µmolL-1 (Fig. 5) with 200–400 µmolL-1 of excess HPO42- with respect to stoichiometric N and P release from OM degradation in the top 10 cm of sediment (Fig. 6). The low solid-phase TOC / P ratios in the strongly reducing shelf sediment (Fig. 3; TOC / P as low as 5) in combination with high excess HPO42- indicates the presence of P sources other than OM. This suggests additional P cycling processes that drive the development of strong HPO42- gradients and associated fluxes across the SWI. The high relative abundance of ex-P, Fe-P and Ca-P in the shelf sediments investigated in this study (Fig. 4), together accounting for > 90 % of total solid-phase P, reflects the predominance of labile P pools that can easily transform and thereby drive intense P cycling. Reductive dissolution of Fe(III) (oxyhydr)oxides and release of sorbed P (Ruttenberg, 2014) likely plays a minor role in this shelf setting; the pool of Fe(III) (oxyhydr)oxides (FeOx; CDB-Fe and Ox-Fe in Fig. 4) is too small: < 20 µm mol g⁻¹ (Fig. 4). The limited availability of reducible solid-phase Fe is further expressed in the low dissolved Fe²⁺ concentrations (Fig. 5). Assuming an Fe / P stoichiometry of 5–15 for highly reactive Fe(III) (oxyhydr)oxides with sorbed or coprecipitated P (Dzombak and Morel, 1990; Jensen et al., 1992; Sundby et al., 1992; Anschutz et al., 1998; Kraal et al., 2019), dissolution of reactive Fe and associated P would result in only a few µm mol P L⁻¹, an insignificant fraction of the observed excess HPO42-. Considering the minor role that Fe-P plays in the shelf sediment, it is likely that other labile P pools, e.g. CaCO₃-associated P and labile Ca-P, dissolved in the extraction step targeting Fe-associated P (Kraal et al., 2017). In general, caution in required when interpreting chemical P fractionation results for the investigated shelf sediments with abundant, highly labile P.

In upwelling areas, high primary production supports large fish populations and associated fluxes of organic and mineral fish remains to the seafloor (Suess, 1981; Froelich et al., 1988; Schenau and De Lange, 2000). Fish bone consists for ∼ 50 wt % of hydroxyapatite (HAP), a poorly crystalline Ca-P mineral with the general formula Ca₁₀(PO4)6(OH)2 (Nriagu, 1982). Fish debris are characterized by TOC / P ratios well below 100 (Logesh et al., 2012; Wijayanti et al., 2021; Løes et al., 2022) Biogenic HAP is efficiently buried in rapidly accumulating, anoxic sediments: sedimentation rates in the research area range from ∼ 10 cm kyr⁻¹ on the slope to 30–120 cm kyr⁻¹ on the shelf in our research area (Calvert and Price, 1983; Summerhayes et al., 1995; Küster-Heins et al., 2010a). Subsequent dissolution of labile biogenic HAP in the sediment boosts pore-water HPO42- concentrations and fuels authigenic precipitation of more stable carbonate fluorapatite (CFA), Ca₁₀(CO₃, PO4)6F₂ (Suess, 1981; Froelich et al., 1988; Schenau et al., 2000; Oxmann and Schwendenmann, 2015). Sediments from the two stations with the highest and most variable P content (stations 6 and 10) contained visible, mm- to cm-scale fish debris (large fragments were removed prior to chemical analyses). The heterogeneous distribution of fish debris in the sediment likely resulted in the marked peaks in sediment P content and the relatively large discrepancy between total P determined by sediment acid digestion (Fig. 2) and the sum of the sequentially extracted P pools (Fig. 4). The values of the TOC / P at stations 6 and 10, as low as 5 with a mean of 28 ± 15 compared to Redfield (106), suggest that biogenic Ca-P may be an important P pool in shelf sediment, accounting for the majority of P in specific sediment intervals. In line with this, a chemical extraction method used to estimate the contribution of fish debris to the solid-phase P pool (Schenau and De Lange, 2000) indicated that HAP in fish bones constituted 20 %–50 % of total P in sediments on the continental slope (∼ 1300 mbss) off Namibia in the BUS (Küster-Heins et al., 2010b). Fish debris were also suggested to play a central role in P cycling and benthic P fluxes on the Peruvian shelf (Suess, 1981; Froelich et al., 1988). Heterogeneous distribution of fish debris in sediment may introduce bias; additional analyses such as bone counts could help constrain the role of fish debris in the P budget. However, a more recent detailed study into P cycling on the Peruvian shelf and slope (74–1000 mbss) by Lomnitz et al. (2016) provided no evidence for an important role for fish debris as additional P source, instead highlighting the potential role of bacteria is supplying additional P to the sediment.

Phosphorus cycling in sediments from upwelling regions is known to be impacted by large sulfur-oxidizing bacteria (SOB) (Schulz and Schulz, 2005; Cosmidis et al., 2013; Lomnitz et al., 2016). Off Namibia, previous work has established the presence of SOB genera Thiomargarita, Beggiatoa and Thioploca, their presence and abundance varying with availability of the oxidants O₂ and NO3- in bottom waters (Gallardo et al., 1998; Brüchert et al., 2003; Schulz and Schulz, 2005; Flood et al., 2021; Chuang et al., 2022). Sulfur-oxidizing bacteria accumulate cellular P in the form of polyphosphates, releasing dissolved HPO42- to the pore-water under reducing conditions and triggering formation of authigenic Ca-P and even phosphorite. The chemical P extraction procedure used here does not allow for quantification of polyphosphate-P, which could be extracted in the first few extraction steps as part of the labile P pool. While trial DNA analysis and ³¹P-NMR did not provide conclusive evidence for SOB in the sediment of shelf stations (SI), we visually observed white filaments, likely SOB genera, in the surface sediments from shelf stations 4, 5 and 6 (see the Supplement). Furthermore, chemical signatures indicative of SOB were observed in shelf sediment, even at stations where there was no visual confirmation of SOB presence: at stations 3, 4, 5 and 10, dissolved HS⁻ accumulation only occurred at depth (> 5 cm) in the sediment (Fig. 5), while the sulfide oxidation product elemental sulfur was detected in the uppermost sediment < 5 cm (Fig. 4). In addition, unexpected peaks in pore-water NO3- and NO2- at depth in the reducing shelf sediment (Fig. 5) suggest the presence of SOB that concentrate NO3- in their vacuoles to use as electron acceptor for sulfide oxidation (McHatton et al., 1996; Jørgensen and Gallardo, 1999; Schulz, 2006). Lastly, the sharp peaks in excess pore-water HPO42- (Fig. 6) in surface sediment were similar in shape and magnitude to those associated with SOB in the work on the BUS shelf by Schulz and Schulz (2005). We note that these lines of evidence for SOB are indirect. The role of SOB can be further tested with cell counts, metagenomics and poly-P assays which lay beyond the scope of this study.

The SOB contribute to efficient reoxidation of HS⁻ produced by sulfate-reducing bacteria and drive rapid sulfur redox cycling in shelf sediments off Namibia which further impacts C, N and P cycling (Dale et al., 2009; Chuang et al., 2022). Furthermore, by inhibiting HS⁻ buildup in the uppermost sediment, SOB maintain weak HS⁻ gradients around the sediment-water interface and thereby limit benthic efflux of toxic HS⁻ (Ohde and Dadou, 2018; Callbeck et al., 2021; Flood et al., 2021). At the same time, steep pore-water HPO42- gradients develop which result in high benthic HPO42- fluxes, particularly at stations 5, 6 and 10. By contrast, shelf stations 3 and 4, with similar general sediment properties and environmental conditions (Table 3), showed a much smaller pool of excess HPO42- (Fig. 6), much smaller labile P pools (Fig. 4) and relatively small diffusive P fluxes (Fig. 7). These contrasting results for apparently similar shelf stations highlight that P cycling processes on the shelf in the NBUS are highly variable not only temporally (seasonality of BWO; growth and decline of microbial communities) but also spatially, the latter perhaps revealing a high sensitivity of the P cycle to small differences in depositional conditions. The observed differences between diffusive and whole-core HPO42- fluxes (Fig. 7) further underscore the highly variable nature of P release.

Overall, we find abundant pools of labile P, low TOC / P and high excess pore-water HPO42- in NBUS shelf sediments, indicating P cycling mechanisms other than degradation of algal material, such as biogenic Ca-P (fish debris) and/or P accumulation and release by SOB. At depth in the shelf sediment, HPO42- accumulation triggers phosphogenesis, i.e. formation of calcium fluorapatite (CFA) reflected in concurrent decreases in pore-water HPO42- and F⁻ (Figs. 5, 6). We observe not only a stark difference in P cycling processes on the slope and shelf but also between shelf stations. The temporal and spatial variability in the P cycle, expressed on both the short-term (pore-water HPO42- profiles, benthic fluxes) and the long-term (solid-phase P profiles), is significant. Therefore, (i) robust P budgets for upwelling systems can only be made with insight into seasonal variability and (ii) the functioning of the sediment as sink or source of P can change dramatically over time in response to even small changes in depositional conditions. Considering that upwelling systems are areas of (periodically) efficient and extensive P removal from the water column, the latter is a crucial consideration when forecasting the impact of environmental change (specifically deoxygenation) on the global marine P cycle.

The abundance of labile sedimentary P pools not derived from algal OM profoundly disrupts the common relationship between redox conditions and TOC / P. The P retention capacity of the sediment generally decreases as redox conditions become more reducing, resulting in TOC / P ratios well above Redfield under oxygen-depleted bottom waters (Slomp et al., 2002; Algeo and Ingall, 2007; Kraal et al., 2010; Kraal et al., 2012). Our results differ markedly from this paradigm: TOC / P well below Redfield were also recorded in shelf sediments underlying strongly oxygen-depleted waters, while highest TOC / P values were measured in sediment on the slope where bottom waters were reasonably well-oxygenated (> 100 µmolL-1) (Fig. 11). This suggests that the supply of additional P by SOB and fish debris can disrupt TOC/ TP as redox proxy in upwelling areas. Comparison to other OMZs may provide further insight into the key driver of this disruption. Similar to our results, Lomnitz et al. (2016) found TOC / P around and below Redfield in organic-rich sediments underlying oxygen-depleted bottom waters in the Peruvian OMZ, with evidence for SOB but not fish debris as additional sedimentary P pool. Conversely, Kraal et al. observed a more straightforward positive correlation between degree of bottom-water oxygen depletion and TOC / P in the upwelling area of the Arabian Sea, with values exceeding Redfield in the core of the OMZ. In the latter study area, fish debris play an important role in benthic P cycling (Schenau and De Lange, 2001) but pore-water sulfide accumulation hardly occurs in surface sediments that therefore do not provide a suitable niche for SOB. This suggests that while biogenic P in fish debris can affect the TOC / TP ratio, SOB may be predominantly responsible for actual disruption of the positive correlation between TOC / TP and BWO in upwelling regions. As such, careful consideration of depositional conditions and P supply mechanisms is required when interpreting TOC / P ratios in the context of redox conditions. Considering that redox conditions and seafloor colonization by SOB are seasonal in the NBUS, it is noteworthy that the impact of SOB seems to be recorded in bulk sediment chemistry. This indicates that while SOB presence is transient, it has a lasting effect on P burial. Furthermore, we observe that the N / P ratio mirrors the trend in TOC / P along the BWO transect (Fig. 11), emphasizing that excess P supply and cycling rather than C dynamics control the TOC / P relationship.

Besides excess supply of P in biogenic Ca-P and by SOB, other factors also play a role in shaping TOC / P ratios, such as shelf-to slope sediment transport and high biological activity and bioturbation in the OMZ. Station 7 (1027 mbss) was located in a “depo centre” where OM, which has been resuspended from the shelf, is concentrated and buried (Inthorn et al., 2006a; Inthorn et al., 2006b). Resuspended OM undergoes repetitive redox sequences and oxic respiration (Aller, 1998), which selectively preserves relatively recalcitrant, N- and P-poor OM and thus enhances TOC / P. Contrasting with previous work that mentions the refractory nature of OM in the depo centre, we note that relative high TOC content and benthic O₂ and DIC fluxes suggest rapid breakdown of relatively fresh OM at station 7 during our visit. The peaks in bottom-water turbidity observed at most stations during our cruise suggest active seafloor mass transport in a bottom nepheloid layer (Inthorn et al., 2006a) throughout the research area (Fig. 9). In and below the OMZ, we observed abundant benthic faunal activity and burrows, suggesting active bioturbation and -irrigation that can enhance OM breakdown in surface sediment by intensifying and prolonging exposure to high-energy-yield oxidants such as O₂ and NO3- (Aller, 2014; Middelburg, 2018). Such faunal activity would add to an already relatively long oxygen exposure time (OET) on the slope (5–30 yr) compared to the shelf (< 0.5 yr), calculated using the OPD (Table 3) and conservative estimates of the sedimentation rate (0.3 mm yr⁻¹ on the shelf; 0.1 mm yr⁻¹ on the slope (Calvert and Price, 1983; Summerhayes et al., 1995)). In the sediment, prolonged OET can lead to low TOC / P ratios (below Redfield) as the C is lost as CO₂ or HCO3- while P is retained, likely by scavenging onto Fe (oxyhydr)oxides.

Overall, our results indicate that minor changes in environmental (redox) conditions in upwelling areas can strongly alter benthic P burial mechanisms and efficiency. This is important because the extent, duration and intensity of oxygen depletion in upwelling areas is changing as a result of anthropogenic global warming (Stramma et al., 2008; Wright et al., 2012; Breitburg et al., 2018; Oschlies et al., 2018). Moreover, distinct P deposition and cycling mechanisms in high-productivity areas can introduce alternative, counteracting feedback mechanisms that disrupt the BWO–TOC / P relationship. With the majority of global OM and P burial occurring on the shelf and slope in productive areas, such modulations of the redox dependence of P burial can have marked implications for our understanding of P cycling and bioavailability under eutrophication and intensifying ocean deoxygenation in the (ancient) past and the future.

The calculated benthic diffusive NH4+ and HPO42- fluxes show that release of N and P from the sediment is highest from organic-rich shelf sediments with low-oxygen bottom waters. The calculated diffusive HPO42- fluxes on the shelf (0.5–4 mmol m⁻² d⁻¹) were in the same range as measured and calculated fluxes from the Arabian (0.2–6 mmol m⁻² d⁻¹), Mauritanian (0.1–0.2 mmol m⁻² d⁻¹) and Peruvian (0.2–1.5 mmol m⁻² d⁻¹) OMZs (Schenau and De Lange, 2001; Lomnitz et al., 2016; Schroller Lomnitz et al., 2019). The maximum measured whole-core HPO42- flux (∼ 1 mmol m⁻² d⁻¹) was similar to measured in-situ HPO42- flux from Namibian shelf sediment (140 mbss) from the same area (∼ 2 mmol m⁻² d⁻¹ (Chuang et al., 2022)) and to the maximum efflux measured under strongly oxygen-depleted bottom waters in the Peruvian OMZ using in-situ sediment incubation chambers (Noffke et al., 2012). In line with our results, previous work on the Namibian shelf also showed variability in the magnitude and direction of the HPO42- flux, ranging between -0.9–1.1 (this study) and -0.2–2 mmol m⁻² d⁻¹ (Chuang et al., 2022). Our results show low values for the NH4+ / HPO42- ratio of diffusive fluxes at most stations (1–8 mol mol⁻¹), well below the Redfield molar N / P ratio of 16, except at station 3 (NH4+ / HPO42-= 19). These low ratios result from high HPO42- gradients around the sediment-water interface, driven by excess release of HPO42- relative to NH4+ (Fig. 6). Additionally, N release can be suppressed by anaerobic nitrogen-loss processes such as denitrification (also by SOB) and anammox, which may constitute an efficient N sink in surface sediments with co-occurring NO₃, NO2- and NH4+ (Voss et al., 2013; Chuang et al., 2022; Pascal et al., 2025). We found a strong correlation coefficient (Pearson's r= 0.9, see the Supplement) between NH4+ and DIC diffusive fluxes, while the correlation coefficient between HPO42- and DIC fluxes was much lower (Pearson's r= 0.6), further supporting a decoupling between OM degradation and sedimentary P effluxes.

The calculated and measured HPO42- fluxes were in the same range but the trends between diffusive and net fluxes were very different. Measured net HPO42- fluxes on the shelf were high and highly variable, with positive (sediment to water) and negative (water to sediment) values. This variability is likely linked to spatially and temporally dynamic benthic P cycling processes. Deposition and breakdown of biogenic Ca-P as well as P sequestration and release by SOB can occur on short timescales, restricted to the uppermost sediment. Furthermore, bio-irrigation by sediment macro-fauna, particularly under more oxygenated slope sediments, may have affected benthic fluxes and thereby contributed to the difference between diffusive and whole-core fluxes. Overall, the role of NBUS sediments as sink or source of essential elements such as P can only be determined with long-term data of high temporal resolution, which was beyond the scope of this research expedition. The results from the diffusive calculations, whole-core flux measurements and bottom-water chemistry (more detail below) highlight that (i) benthic-pelagic coupling in the NBUS is not at steady state and (ii) benthic C, N and P fluxes become (temporarily) decoupled.

Using bottom-water chemistry, we can further evaluate benthic-pelagic coupling and the impact of sediment biogeochemistry on water-column chemistry. This requires some caution, as pressure effects during coring can force pore-water into the overlying bottom water to an unknown extent. The results show chemical trends as function of surface-sediment OM enrichment and BWO (Fig. 8). Organic matter enrichment and BWO are strongly correlated (Pearson's r= -0.72; see the Supplement). With decreasing BWO and increasing TOC, nutrient fluxes increase, and bottom waters become increasingly enriched in NH4+, HPO42- and H₄SiO₄. The high Si concentration in bottom waters at slope station 7 may reflect dissolution of redeposited diatom frustules. With decreasing BWO and increasing sediment TOC enrichment, the strong decrease in the bottom-water dissolved N / P ratio reflects decoupling of fluxes. Anaerobic N loss is reflected in a strong decrease in bottom-water NO3- with decreasing BWO. Together with nutrients, dissolved HS⁻ also accumulated in the bottom waters at stations with oxygen-depleted bottom waters, most notably at stations 6 and 10 for which positive diffusive HS⁻ fluxes were calculated (Fig. 7). The calculated fluxes for the latter stations varied strongly, while bottom-water HS⁻ concentrations were similar, suggesting that the pore-water chemistry from which the fluxes were calculated is a snapshot that may not reflect benthic exchange on longer timescales. In general, we note that samples were taken during late austral summer (early February) during which seasonal anoxia develops on the NBUS shelf and thus HS⁻ efflux is most likely to occur. Similarly, the seasonally highly variable redox conditions will affect fluxes of C, N and P as well and as such, measured fluxes cannot easily be extrapolated: long-term time series of benthic fluxes are needed to quantify nutrient budgets and monitor the response to environmental change on seasonal to decadal timescales.

In the BUS, benthic N / P stoichiometry is affected by SOB-modulated N loss and P turnover (Chuang et al., 2022) and the resulting release of excess P from the sediment (Bailey, 1987; Flohr et al., 2014; Emeis et al., 2017). In the NBUS region studied here, oxygen depletion in the water column drives (i) pelagic anaerobic N loss processes and (ii) benthic P supply that together result in off-shelf transport of N-depleted waters (Flohr et al., 2014).

These processes drive N / P ratios in the water column to low values, resulting in the so-called “N deficit” determined on the shelf during our research expedition (van Kemenade et al., 2022), calculated as the difference between measured DIN and expected DIN based on HPO42- (DIP) and Redfield stoichiometry (N / P = 16). As such, the N deficit is an operational indicator of the extent to which the measured N / P drops below the Redfield value. Our results illustrate how interacting sedimentary and pelagic biogeochemical processes result in shelf waters with molar N / P as low as 2.6 in bottom waters (station 6, 100 mbss, BWO = 3 µmolL-1).

Here, we use our nutrient data to estimate the impact of excess P supply to bottom waters on the N deficit. The water-column HPO42- profiles for shelf stations 5, 6 and 10 show enrichments in the bottom ∼ 25 m of 2.6–3.6 µmolL-1 (Fig. 5). Stations 3 and 4 do not show this enrichment; HPO42- reaches ∼ 2 µmolL-1. We assume that this would have been the concentration of HPO42- in bottom waters at stations 5, 6 and 10 as well in the hypothetical absence of the sediment-derived bottom-water HPO42- enrichment. Using this conservative estimate, we can then approximate to what extent excess P (likely from sources other than algal OM; see Sect. 4.1) rather than N depletion contributes to the N deficit. The contribution of excess P to the N deficit we call the “P-driven N deficit”, which is the difference between the N-driven N deficit calculated with a bottom-water P concentration of 2 µmolL-1 and the actual N deficit (DIN - 16 × DIP). The results are striking: 25 %–85 % of the N deficit can be contributed to the bottom-water excess P, likely delivered by the sediment (Fig. 12). This highlights how benthic P inputs can shape the N deficit in the NBUS and perhaps more broadly in upwelling regions, as we show an N deficit can occur without extensive anaerobic N loss (see station 10 in Fig. 12). Here, we do not consider in detail the processes underlying the N deficit calculated in surface waters (∼ 0–25 mbss, Fig. 12), but we point out that a relative P excess is in line with global oceanic data showing that N commonly becomes depleted before P in the photic zone (Tyrrell, 1999).

Our results show dramatic changes in the dissolved N / P ratio during low-oxygen conditions on the NBUS shelf. Alteration of nutrient stoichiometry can have a profound impact on ocean biogeochemistry by affecting the microbial community composition (Moore et al., 2013). For instance, low N / P waters are thought to boost the relative abundance of N₂-fixing cyanobacteria in both ancient (Kuypers et al., 2004) and modern marine systems including the NBUS (Vahtera et al., 2007; Flohr et al., 2014; Emeis et al., 2017). The physicochemical water-column data (temperature, density; Fig. 10) do not indicate strong stratification of the bottom waters, suggesting that the enrichment is due to active benthic P efflux and not bottom water restriction. However, our findings are in line with those of Flohr et al. (2014) in suggesting that the wider impact of altered nutrient stoichiometry is minor, at least in the period the research cruise was undertaken. The water-column N / P ratios beyond the shelf break are close to Redfield throughout the water column; the waters transported off the shelf do not have a distinct, P-enriched and/or N-depleted signature. Furthermore, we note that nutrient stoichiometry on bottom water on the shelf is likely a function of the seasonally variable redox conditions. We sampled in austral summer with shelf anoxia, the optimal timeframe to find decoupled N and P cycling. Our results do indicate that the increasing extent and intensity of ocean oxygen depletion in general and OMZs in particular have the potential to shift marine nutrient availability and stoichiometry through interacting, redox-dependent benthic and pelagic biogeochemical N and P cycling processes.