Nitrogen cycling in sediments on the NW African margin inferred from N and O isotopes in benthic chambers

1. Introduction

The Canary upwelling region along the north-western African coast is an important eastern boundary current system with regard to primary production (0.33 Gtons C yr⁻¹), second in magnitude only to the Benguela upwelling (Carr, 2001). Off Mauritania, high primary production is driven by upwelling of nutrient-rich South Atlantic Central Water (SACW) and huge dust inputs from the adjacent Sahara-Sahel region (Fischer et al., 2010). Subsurface waters off Mauritania are better ventilated than the Benguela system and the other large upwelling systems in the Eastern Tropical Pacific, with minimum bottom water dissolved O₂ levels on the margin of several tens of micromoles per liter (Hartmann et al., 1976; Dale et al., 2014; Thomsen et al., 2019). However, the SACW in the Eastern Tropical North Atlantic (ETNA) appears to be undergoing significant deoxygenation of ca. 0.5 µmol kg^–1 yr^–1 (Stramma et al., 2008), possibly due to a warming-induced decrease in solubility and biological O₂ consumption (Schmidtko et al., 2017). At this rate, subsurface waters on the Mauritanian margin are heading towards anoxia by the end of the 21^st century.

In the Eastern Tropical Pacific and South Atlantic upwelling systems, the anoxic water column is a hotspot of microbially-mediated N loss, either by denitrification (nitrate reduction to N₂) and/or anammox (anaerobic oxidation of ammonium by nitrite to N₂) (Devol et al., 2006; Lam and Kuypers, 2011). Along with the Arabian Sea and Bay of Bengal, most fixed N loss in ocean waters occurs in these ‘oxygen minimum zones (OMZ)’ (Gruber, 2008). In sediments, by contrast, and with the possible exception of the oligotrophic deep-sea and euxinic basins, microbial transformation of fixed N to N₂ by microorganisms occurs almost everywhere (Middelburg et al., 1996). In the Mauritanian OMZ, where pelagic O₂ concentrations are consistently above the inhibition threshold for denitrification (ca. 5 μM), benthic denitrification/anammox is the only significant N loss pathway. Benthic denitrification rates measured using ¹⁵N-labelling techniques range from around 4 mmol N m^-2 d^-1 on the shelf to 1 mmol N m^-2 d^-1 on the upper slope (Sokoll et al., 2016), and are toward the higher end for open-ocean continental margins. Anammox was found to be of minor importance (Sokoll et al., 2016).

Current understanding of the fixed N balance in the ocean is dependent on knowledge of the rates of N transformation pathways, both in the water column and in the sediment. To a large extent, this is achieved by balancing the marine inventory of stable N isotopes (¹⁴N and ¹⁵N) under steady state conditions (Brandes and Devol, 2002). This approach is possible because most microbial N transformation processes impart unique degrees of isotopic discrimination, or fractionation, of one isotope over another, expressed as a kinetic isotope effect (KIE), or ϵ (unit: ‰). (In this paper, ϵ = (α – 1)·1000, where α is the fractionation factor defined as ^Lk/^Hk, and where k denotes the reaction rate constant of the heavy (H) and light (L) isotopologues). Studies in the water column and in the laboratory have revealed a wealth of information concerning N and O effects of microbial N cycling (e.g. Sigman et al., 2006; Casciotti and McIlvin, 2007; Granger et al., 2008; Buchwald and Casciotti, 2013; Gaye et al., 2013; Bourbonnais et al., 2015 and many others). For instance, the N and O isotope effects during the dissimilatory reduction of nitrate (NO₃^-) to nitrite (NO₂^-) are similar at ca. 15 – 30 ‰ in favor of the lighter isotopes (Voss et al., 2001; Kritee et al., 2012; Bourbonnais et al., 2015). Oxidation of NO₂^- to NO₃^-, in contrast, has an inverse isotope effect due to fractionation at the enzymatic level during N–O bond formation by nitrite oxidoreductase (Casciotti, 2009). NO₂^- oxidation leads to enrichment of ¹⁴N and ¹⁶O in the residual NO₂^- and an enrichment of ¹⁵N and ¹⁸O in the NO₃^- product (Casciotti, 2009). This divergence has led numerous investigators to conclude that a significant fraction of NO₂^- produced by reduction of NO₃^- in oxygen deficient waters is reoxidized to NO₃^- (Casciotti and McIlvin, 2007; Buchwald and Casciotti, 2013; Casciotti et al., 2013; Gaye et al., 2013; Bourbonnais et al., 2015; Casciotti, 2016; Granger and Wankel, 2016). Fewer studies have been carried out in sediments, although so far it appears that the microbial isotope effects observed in the water column are largely transferrable to the benthic environment (Lehmann et al., 2004; Lehmann et al., 2007; Alkhatib et al., 2012; Dale et al., 2014; Dale et al., 2019). The major difference between the water column and sediment rests with the much lower isotopic fractionation of NO₃^- uptake by sediments (hereafter benthic fractionation), arising from transport limitation of NO₃^- diffusion within the sediment (< 3 ‰; Brandes and Devol, 1997; Lehmann et al., 2004). Based on this assumption, it is understood that around two-thirds of the global oceanic fixed N loss takes place in the sediment and one-third in OMZ waters (Brandes and Devol, 2002; Somes et al., 2013). However, more recent studies suggest that the benthic N isotope effect may be higher, especially in sediments underlying oxygen-deficient bottom waters (Lehmann et al., 2007; Alkhatib et al., 2012; Dähnke and Thamdrup, 2013; Dale et al., 2014; Dale et al., 2019). Furthermore, it can be enhanced by the efflux of ¹⁵N-enriched NH₄⁺ from the sediment (Lehmann et al., 2007; Granger et al., 2011; Alkhatib et al., 2012). Accurate estimates of the total benthic N isotope effect that also includes NH₄⁺ and other N species, and thus more relevant for the global N budget, require a deeper understanding of the rate of benthic N loss and associated N isotope transformations.

In this study, we deployed in situ benthic incubation chambers on the Mauritanian margin in 2014 during the transition from the strong to weak upwelling season (Thomsen et al., 2019). Fieldwork was conducted along a transect from the shallow shelf to the upper slope at 18 °N. Our objective was to determine the recycling rates and exchange fluxes of dissolved inorganic nitrogen (DIN: NO₃^-, NO₂^-, NH₄⁺) between the sediment and water column and compare them to results from an earlier campaign in 2011. We also measured the N and O isotopic composition of NO₃^- (δ¹⁵N-NO₃^- and δ¹⁸O-NO₃^-) and NH₄⁺ (δ¹⁵N-NH₄⁺) inside the chambers. With this information, and with the help of a numerical model, we were able to calculate the total the benthic N isotope effect along with further information on benthic N turnover rates, sources and transformations.

2. Oceanographic setting

The study area lies within the Mauritanian upwelling region in the ETNA. The prominent water mass that upwells onto the margin at 18 °N is poleward-flowing and nutrient-rich SACW. Upwelling is strongest between January and April when the Intertropical Convergence Zone (ITCZ) migrates southwards (Mittelstaedt, 1983). It weakens when the ITZC moves northwards, leading to a decline in primary production on the shelf. Primary production nonetheless remains relatively high year-round at 80 – 200 mmol m^–2 d^–1 of C (Huntsman and Barber, 1977). Oceanographic data indicate that upwelling was active during the fieldwork (Thomsen et al., 2019). Dissolved O₂ levels below the surface mixed layer are dictated by SACW ventilation, vertical mixing, and respiration rates of organic carbon in the water column (Mittelstaedt, 1983; Schafstall et al., 2010; Thomsen et al., 2019). They present a deep minimum of 40 – 50 µM at 400 – 500 m water depth and a shallow minimum of 30 – 50 µM on the shelf at 50 - 100 m (Karstensen et al., 2008; Schroller-Lomnitz et al., 2019; Thomsen et al., 2019). The lowest O₂ levels on the shelf are usually encountered during upwelling relaxation in July and August due to combined effects of stratification and benthic respiration (Klenz et al., 2018). A detailed oceanographic overview of the study area during the fieldwork is described by Thomsen et al. (2019) and water column O₂ measurements at the sampling stations are presented by Schroller-Lomnitz et al. (2019).

The sediments on the margin at 18°N are classified as slightly sandy mud or muddy sand (Dale et al., 2014). They comprise 10 - 20% carbonates with organic C increasing from ~1% on the shelf to 3% at 1000 m water depth (Hartmann et al., 1976; Gier et al., 2017). Sediment reworking by turbulent mixing exports fine-grained particles down the slope (Fütterer, 1983; Schafstall et al., 2010). Consequently, sedimentation rates on the shelf determined by ²¹⁰Pb accumulation are low, and increase to 350 cm kyr^-1 below the shelf break (Dale et al., 2014). Porewaters in the upper 20 cm are mostly ferruginous, becoming sulfidic on the shallow shelf (Dale et al., 2014; Gier et al., 2017; Schroller-Lomnitz et al., 2019). Benthic respiration rates on the shelf range from approximately 10 to 20 mmol m^-2 d^-1 of C (Dale et al., 2014; Schroller-Lomnitz et al., 2019). These values are typical for the average margin, but far smaller than those in the carbon-rich shelf sediments in the Peruvian and Benguelan upwelling systems (Brüchert et al., 2003; Dale et al., 2015).

3. Material and methods

Data presented here were mostly measured within the framework of the project “Climate – Biogeochemistry Interactions in the Tropical Ocean” (www.sfb754.de) on research campaign M107 (RV Meteor) from 30 May to 3 July 2014. A complete overview of sediment operations can be found in the cruise report (Sommer et al., 2015). Digital links to the data and a general overview of the project are provided by Krahmann et al. (2021).

3.1. Benthic flux measurements

In situ fluxes of O₂, dissolved inorganic carbon (DIC), NO₃^–, NO₂^–, NH₄⁺, silicic acid (H₄SiO₄) and N/O isotopes of N species were measured in benthic chambers along an E-W transect at 18.1°N using autonomous Biogeochemical Observatories (BIGO; Sommer et al., 2008). A total of nine BIGO deployments were made at nine stations along a latitudinal transect across the shelf/slope covering a horizontal distance of 50 km from ~50 to ~1110 m water depth (Figure 1). Seven of the stations were revisited from an earlier cruise during March/April 2011, which coincided with the transition between the strong upwelling and weak upwelling seasons (Dale et al., 2014). Geographical coordinates for each station are listed in Schroller-Lomnitz et al. (2019). These workers reported mean PO₄^3- and DIC fluxes at each site for the same lander deployments.

Figure 1 (A) Benthic fluxes of DIC (black symbols) and TOU (open symbols) versus water depth. Two data points per water depth correspond to the two benthic chambers in each BIGO lander. Cross plots of (B) organic carbon degradation versus measured DIC flux, and (C) denitrification versus organic carbon degradation, as determined by the mass balance. Denitrification rates (Table 2) have been converted to mmol C m^-2 d^-1 by dividing by r_NO3. The trendline in (C) has been forced through the origin so that R_DEN = 0 if RPOC_TOT = 0. For (B, C), Pearson correlation coefficients and p values are given.

The design and implementation of the BIGO landers has been discussed in detail previously (Krahmann et al., 2021). Each BIGO (I and II) contained two circular flux chambers (C1 and C2) with an internal diameter of 28.8 cm where the sediment was incubated for ~30 h. Discrete samples were taken using glass syringes (eight per chamber) at pre-programmed time intervals for chemical analysis. After BIGO recovery, the syringes were immediately transferred to the onboard cool room (4°C) for filtering (0.2 µm) and sub-sampling. Another six water samples were taken from inside each benthic chamber into quartz tubes for analysis of DIC.

Solute fluxes were calculated from the slope of the linear regression of the concentration data versus sampling time and the height of the water inside each chamber. Fluxes represent the net flux by diffusion/advection in addition to solute pumping by animals (bioirrigation). A total of 29 nutrient data points from the whole data set (n = 720) set were visually assigned as outliers and omitted from the calculations. Twelve DIC data were removed from 90 data points, mainly corresponding to the last syringe samples taken from the chambers on the shelf. For the calculation of the O₂ flux, equal to total oxygen uptake (TOU), the linear initial part of the O₂ time series recorded by optodes (see below) was used. Negative fluxes denote net uptake by the sediment and positive fluxes net release.

3.2. Sediment sampling

Sediment cores at each station were retrieved using a multiple-corer (MUC). Porewaters and particulates were extracted from the sediment as described by Schroller-Lomnitz et al. (2019) and Gier et al. (2017). Further details are provided in those publications.

3.3. Analytical methods

Nutrients in the benthic landers were analyzed on board with a Quaatro Autoanalyzer (Seal Analytic) using standard photometric approaches. Porewaters were analyzed for ferrous iron (Fe²⁺), NO₃^-, NO₂^-, NH₄⁺, PO₄^3-, H₄SiO₄, total alkalinity (TA) and hydrogen sulfide (H₂S). Analytical details can be found in Gier et al. (2017); Schroller-Lomnitz et al. (2019) and Sommer et al. (2015). Sediment samples were also stored refrigerated for analysis of solid phase constituents and physical properties at the home laboratory.

The O₂ concentration and temperature inside each chamber were measured using optodes. Bottom water O₂ values were taken from additional optodes mounted externally on the landers. DIC measurements were performed using a quadrupole membrane inlet mass spectrometer (MIMS, GAM200, In Process Instruments) following Sommer et al. (2015). The instrument was equipped with inline sample acidification to convert DIC to dissolved CO₂ gas for measurement at a mass-to-charge ratio of 44.

The particulate organic carbon (POC) content of freeze-dried and milled sediment samples was determined by flash combustion in a Carlo Erba Elemental Analyzer (NA 1500) with an analytical precision and detection limit of 0.04% (dry weight percent) and 0.05%, respectively. POC contents have been published previously (Schroller-Lomnitz et al., 2019). Samples were decarbonated with 0.25 N HCl prior to analysis. Carbonate (assumed to be CaCO₃) was determined by weight difference. The precision and detection limit of the carbonate analysis was 2% and 0.1%, respectively. Total particulate nitrogen was determined on the elemental analyzer with the same precision and detection limit as carbonate. We assumed that this is equal to particulate organic N (PON). Porosity was determined from the weight difference of the wet and freeze–dried sediment, assuming a sediment density of 2.5 g cm^-3 and a seawater density of 1.023 g cm^-3.

During a separate field campaign in July 2019 (M156, Sommer et al., 2020), surface sediment close to St. 5 (182 m depth) was analyzed for δ¹⁵N and δ¹³C using a high sensitivity elemental analyzer (HSEA) connected to an isotope ratio mass spectrometer (DeltaPlus Advantage, Thermo Fisher Scientific) as described by Hansen and Sommer (2007). δ¹⁵N and δ¹³C values of the bulk particulate material (after removing carbonates) were calculated as:

where δ^HX denotes the isotopic composition (‰) of X (C or N) and H and L are the heavy and light isotopologue of C (13) or N (15). The reference standards were V-PDB for C and N₂ in air for N. System calibration was implemented by the combustion of International Atomic Energy Agency (IAEA-N-1, IAEA-N-2, IAEA-NO-3) and National Institute of Standards and Technology (NBS-22 and NBS-600) compounds. Acetanilide was used as an internal standard after every sixth sample within each sample run. The overall standard deviation (SD) for the low measurement range 2.5 - 8 µg N and 5.0 - 80 µg C was ± 0.25 ‰ and ± 0.2 ‰, respectively. The overall SD for the higher measurement range 3 - 15 µg N and 10 - 140 µg C was ± 0.2 ‰ and ± 0.15‰, respectively.

Benthic lander samples were analyzed for the isotopic composition of nitrate (δ¹⁵N-NO₃^- and δ¹⁸O-NO₃^-) and ammonium (δ¹⁵N-NH₄⁺). Most of the samples were analyzed for δ¹⁵N-NO₃^- and δ¹⁸O-NO₃^- at GEOMAR (Kiel, Germany). At three stations (St. 1, 2 and 4), δ¹⁵N-NO₃^- and δ¹⁸O-NO₃^- were measured at the Helmholtz Center Hereon (Geesthacht, Germany). In both labs, nitrate dual isotopes were analyzed using the denitrifier method (Sigman et al., 2001; Casciotti et al., 2002). All samples for δ¹⁵N-NH₄⁺ were analyzed at GEOMAR using the hypobromite/azide-method (Mcilvin and Altabet, 2005; Zhang et al., 2007). δ¹⁵N-NH₄⁺ was only measured in benthic lander samples on the shelf where NH₄⁺ concentrations were above the detection limit for a reliable isotope analysis (1 µM). Due to limited sample availability and low NO₂^-concentrations, δ¹⁵N-NO₂^- could not be measured.

The denitrifier method and the hypobromite/azide-method are based on the isotopic analysis of nitrous oxide (N₂O). In the denitrifier method, NO₃^- and NO₂^- are quantitatively converted to N₂O by Pseudomonas aureofaciens (ATTC 13985). The hypobromite/azide-method is based on the chemical conversion of NH₄⁺ to N₂O by a subsequent addition of a hypobromite and azide solution. For both methods, the sample volume was adjusted to a sample size of 10 nmol of N₂O. N₂O was extracted from the sample vials by purging with helium and measured with a GasBench II, coupled to an isotope ratio mass spectrometer (Delta Plus, Thermo Fisher Scientific, Germany). Each batch of samples using the denitrifier method included two international standards (USGS34: δ¹⁵N = -1.8‰ vs N₂, δ¹⁸O = -27.9‰ vs. VSMOW; IAEA-NO-3: δ¹⁵N = 4.7‰ vs N₂, δ¹⁸O = +25.6‰ vs. VSMOW) (Böhlke et al., 2003) and an internal standard. A bracketing correction (Sigman et al., 2009) was applied to the nitrate isotope data to correct for exchange with water. To calibrate δ¹⁵N-NH₄⁺ measurements, three international standards (IAEA-N-1: δ¹⁵N = 0.4‰, USGS25: δ¹⁵N = -30.4‰, USGS26: δ¹⁵N = 53.7‰) were used with each sample run. All measurements were replicated, with a typical reproducibility for δ¹⁵N-NO₃^-, δ¹⁸O-NO₃^- and δ¹⁵N-NH₄⁺ better than 0.2‰, 0.6‰ and 0.9‰, respectively. Since NO₂^- comprised on average 2% of the combined NO₂^- + NO₃^- pool, the contribution of NO₂^- interference to the reported δ¹⁵N and δ¹⁸O nitrate values was considered negligible.

N and O isotope ratios are reported in per mil (‰), relative to the analytical standards (N₂ in air for δ¹⁵N, and Vienna Standard Mean Ocean Water (VSMOW) for O) using Eq. (1).

3.4. C and N recycling rates

Depth-integrated rates (total rate per unit area) of POC oxidation to DIC (RPOC_TOT, mmol C m^-2 d^-1), ammonification of particulate organic N (R_AMF, mmol N m^-2 d^-1), nitrification of NH₄⁺ to NO₃^- (R_NIT, mmol N m^-2 d^-1) and denitrification of NO₃^- to N₂ (R_DEN, mmol N m^-2 d^-1) were calculated using a mass balance of the fluxes (F) of O₂, NO₃^- and NH₄⁺ measured in the benthic chambers (Dale et al., 2014):

where r_CN is the atomic C:N ratio in organic matter undergoing degradation, and r_O2 and r_NO3 are the molar ratios of O₂:POC and NO₃^-:POC, respectively, during aerobic respiration and denitrification of organic carbon. r_O2 and r_NO3 were previously defined as 118/106 and 94.4/106 (respectively) for organic matter with an oxidation state of -0.45 (Dale et al., 2014). These equations are only approximations because they ignore NO₂^- fluxes, yet which are equivalent to only a few percent of NO₃^– fluxes.

3.5. Isotope calculations

The loss of NO₃^- inside the chambers due to benthic respiration takes place within a closed system where there is no exchange of NO₃^- with the surrounding waters. Under these conditions, the apparent benthic fractionation of N (¹⁵ϵ_app) or O (¹⁸ϵ_app) was calculated using a closed system Rayleigh model (Marioitti et al., 1981; Prokopenko et al., 2013):

where δ^HX and δ^HX₀ denote the isotopic composition (‰) of X (N or O) during the incubation and at the start of the incubation, respectively, and f is the remaining fraction of NO₃^– in the chamber. ^Hϵ_app is equal to the slope of the linear regression of a plot of δ^HX versus ln(f). Values of ϵ_app at each station were determined by pooling the data from both BIGO chambers. Only slopes that were significant (p value< 0.05) are reported, which eliminates the ¹⁵ϵ_app and ¹⁸ϵ_app data from the two deepest stations and also ¹⁸ϵ_app from St. 3 and 7. At these sites, the change in δ^HX is too small or scattered to calculate reliable ϵ_app.

^Hϵ_app is related to the δ^HX of the NO₃^- flux relative to the δ^HX of bottom water NO₃^-, δ^HX_NO3BW (Alkhatib et al., 2012):

Where

and F_HNO3 and F_LNO3 are the benthic fluxes of the heavy and light isotopologues at the sediment surface, respectively (see Supplement).

4. Results

4.1. Particulate geochemistry across the margin

Mean POC content in the sediment cores retrieved by the MUC increased from < 1% on the shelf (< 200 m) to 2.9% at the deepest station (Table 1). CaCO₃ showed maximum values at St. 4 (37%) and decreased offshore. POC content was significantly and positively correlated with the particulate mud fraction (MF, %) reported by Dale et al. (2014), where POC = 28.1 MF – 13.3 (p< 0.05, data not shown) whereas CaCO₃ was not. The mean δ¹³C and δ¹⁵N of the particulate material close to St. 5 was -20.8 ± 0.2 ‰ and 3.7 ± 0.4 ‰, respectively (see Supplement). These δ¹³C_POC fall within the range of -16 to -22 ‰ characteristic of marine organic matter (Meyers, 1994). The mean atomic ratio of POC to PON (r_CN, mol C (mol N)^-1) in each sediment core was mainly constant over the transect (~9 – 10), with a lower ratio at St. 4 of 7.6. These values are typical for marine algae (Burdige, 2006).

Table 1 Mean contents of POC, CaCO₃ and the C:N ratio in the sediment cores recovered from each station along with mean bottom water O₂ concentrations recorded on the BIGO landers.

4.2. Seafloor recycling of biogenic matter

NO₃^- fluxes were directed into the sediment with highest values on the shelf (-2.9 mmol m^-2 d^-1) (Table 2). NO₂^- effluxes were highest on the shelf (0.13 mmol m^-2 d^-1) and at least a factor of 20 lower than the NO₃^- flux. NH₄⁺ and H₄SiO₄ fluxes were also highest on the shelf and directed from the sediment to the bottom water (Figure S1). The ratio of Si fluxes relative to DIC was 0.5 ± 0.2 and higher than the diatomic Si:C ratio of 0.13 (Brzezinski, 1985), indicating preferential dissolution of opal compared to POC.

In situ fluxes of O₂, DIC and nutrients were highest on the shelf and lowest offshore (Table 2 and Figure 1A). DIC fluxes (F_DIC) decreased from 20.7 mmol m^-2 d^-1 at St. 1 to 3.5 mmol m^-2 d^-1 at St. 9. O₂ fluxes (= TOU) were on average 25% lower than F_DIC. The offshore decrease in F_DIC and TOU are consistent with a decrease in remineralization of labile organic matter with increasing water depth. RPOC_TOT decreased from 18.2 mmol m^-2 d^-1 on the shelf to ~2 mmol m^-2 d^-1 at St. 9. RPOC_TOT was equal to 82% of F_DIC for the margin as a whole (Figure 1B). The excess DIC flux over TOU is likely to be driven by carbonate dissolution induced by metabolically-produced CO₂ in surface sediment (Jahnke and Jahnke, 2004). This is further indicated by the respiratory quotient (RQ), determined as the ratio between DIC outflux and O₂ influx across the sediment surface. RQ was 1.5 ± 0.6 across the margin (Table 2). Whilst similar to other estimates using benthic chambers (Jørgensen et al., 2022), it is much higher than the expected value of 0.7 based on complete aerobic mineralization of organic matter with an average phytoplankton composition (Anderson, 1995).

Table 2 Benthic fluxes (F) measured in the individual benthic chambers (mmol m^–2 d^–1) across the transect, the respiratory quotient (RQ, see text) and depth-integrated reaction rates, R [Eq. (2)-(5)].

The sediments acted as a net sink for DIN [NO₃^-+NO₂^-+NH₄⁺)]. Denitrification rates (R_DEN) decreased from 3.6 mmol N m^-2 d^-1 on the shelf where RPOC_TOT was highest, to < 1 mmol N m^-2 d^-1 at the deepest stations. These are consistent with those determined from ¹⁵N labelling experiments (Sokoll et al., 2016). R_DEN accounted for 24% of RPOC_TOT (Figure 1C). Nitrification (R_NIT) was apparently active at every station with rates that were 17% of R_DEN on average.

The loss of fixed N was further evident in the flux of the tracer N* (F_N*) (Gruber and Sarmiento, 1997; Lehmann et al., 2004):

where F_PO4 is the PO₄^3- flux (Table 2) and the factor 16 stems from the Redfield composition of organic matter (N:P = 16:1). If organic N and P are completely remineralized, F_N* = 0. F_N* was negative at all sites because the sediments acted as a source of PO₄^3- and a sink of DIN, illustrating preferential loss of DIN relative to PO₄^3-. The most negative N* flux of -5.5 mmol m^-2 d^-1 was determined at St. 1 where R_DEN was highest. The highest F_N* of -1 mmol m^-2 d^-1 was calculated for St. 9. Across the transect, the mean F_N* was -2.8 ± 1.3 mmol m^-2 d^-1.

4.3. N isotopes in benthic chambers

δ¹⁵N-NO₃ and δ¹⁸O-NO₃ inside the benthic chambers are illustrated in Figure 2. Bottom water δ¹⁵N-NO₃, taken as the first time point immediately after the chambers were closed, was approximately 5.0 ± 0.5 ‰. Enrichments in ¹⁵N over time were generally observed in all chambers on the shelf and upper slope. δ¹⁵N-NO₃ increased by roughly 1 ‰ on the shelf and by <1 ‰ on the upper slope over the course of the incubations. At the two deepest sites, variations in δ¹⁵N-NO₃ and δ¹⁸O-NO₃ were within analytical precision. Bottom water δ¹⁸O-NO₃ was more variable than δ¹⁵N-NO₃. Enrichments in ¹⁸O were stronger on the shelf compared to the slope.

Figure 2 Benthic chamber incubation results for NO₃^- on the shelf and upper slope. (A, B) NO₃^- concentrations with linear regression curves from which the fluxes are calculated, (C, D) δ¹⁵N-NO₃ and (E, F) δ¹⁸O-NO₃. Filled and open symbols correspond to chamber 1 and 2 during each BIGO lander deployment, respectively.

The apparent fractionation of N (¹⁵ϵ_app) and O (¹⁸ϵ_app) accompanying NO₃^- loss inside the benthic chambers varied between 0.7 and 2.0 ‰ for ¹⁵ϵ_app and between 1.7 and 2.9 ‰ for ¹⁸ϵ_app with no clear trends across the transect (Figure 3A and Table S4). The average ¹⁵ϵ_app and ¹⁸ϵ_app were 1.5 ± 0.4 ‰ and 2.0 ± 0.5 ‰, respectively. For the shelf (< 100 m), ¹⁸ϵ_app was twice as high as ¹⁵ϵ_app (1.8 versus 0.9 ‰). The ratio of these numbers, ¹⁸ϵ_app:¹⁵ϵ_app, varied from 2.0 to 2.3 at stations 1, 2 and 4. ¹⁸ϵ_app:¹⁵ϵ_app was 0.9 at St. 5 and 6 (Figure 3B). The mean ¹⁸ϵ_app:¹⁵ϵ_app ratio of the shelf stations was 2.1 ± 0.3.

Figure 3 Benthic chamber isotope results versus water depth on the Mauritanian margin. (A) Fractionations of N-NO₃^- (¹⁵ϵ_app) and O-NO₃^- (¹⁸ϵ_app). (B) The ratio ¹⁸ϵ_app: ¹⁵ϵ_app, where the dashed line shows unity. (C) The δ¹⁵N of the NH₄⁺ flux from the sediment. Error bars represent the standard error of the regression slopes (see Methods). Data are tabulated in Table S1 in the Supplement.

δ¹⁵N-NH₄ inside the benthic chambers at St. 1 and 2 was between 9 and 10 ‰ whereas the values at St. 3 were more scattered (Figure S2). The δ¹⁵N of the NH₄⁺ flux out of the sediment to the bottom water (δF_NH4) for the shelf stations was calculated as the intercept of a linear regression of δ¹⁵N-NH₄ versus the inverse NH₄⁺ concentration (Mortazavi and Chanton, 2004; Dale et al., 2019). Values of δF_NH4 were similar and varied from 8 ± 1.8 ‰ at St. 1 to 9.6 ± 1.2 ‰ at St. 2 (Figure 3C and Table S4). The average δF_NH4 of the three shelf sites was 9.0 ± 0.7 ‰.

5. Discussion

5.1. N cycling inferred from benthic chamber isotopes

The increase in δ¹⁵N-NO₃ and δ¹⁸O-NO₃ in the benthic incubation chambers, along with a decrease in NO₃^- concentrations, is attributed to dissimilatory reduction of NO₃^- to NO₂^- (Lehmann et al., 2004; Dale et al., 2014; Dale et al., 2019). Isotope effects for nitrate reduction estimated from δ¹⁵N-NO₃ distributions in the water column are ca. 15 – 30 ‰ (Voss et al., 2001; Kritee et al., 2012; Bourbonnais et al., 2015) for both ¹⁵N and ¹⁸O (Marioitti et al., 1981; Voss et al., 2001; Casciotti et al., 2002; Granger et al., 2008; Kritee et al., 2012). Thus, our data would appear to present two conflicts with the known isotopic fingerprint of dissimilatory NO₃^- reduction: (i) the apparent benthic fractionation factors, ¹⁵ϵ_app and ¹⁸ϵ_app, are much lower than observed in the water column at all sites, and (ii) the ratio ¹⁸ϵ_app:¹⁵ϵ_app is higher than unity at St. 1, 2 and 4, but closer to unity at St. 5 and 6 (Figure 3).

On continental margins, including Mauritania, NO₃^- is typically exhausted within a few centimeters below the sediment surface due to high rates of respiration by microorganisms (Dale et al., 2014). The strong drawdown of NO₃^- leads to diffusive transport limitation of NO₃^- to the subsurface denitrifying layer, leading to complete consumption of NO₃^- that diffuses into the sediment (Brandes and Devol, 1997). This results in an under-expression of the isotope effect in the benthic NO₃^- flux and explains why larger isotope effects are observed in the water column (Brandes and Devol, 1997; Brandes and Devol, 2002; Lehmann et al., 2007; Alkhatib et al., 2012; Dale et al., 2019). The range of ¹⁵ϵ_app determined in this study is similar to that reported for other coastal sediments under oxygenated waters (Brandes and Devol, 2002; Lehmann et al., 2007; Alkhatib et al., 2012). ϵ_app may be larger in oxygen-deficient regions because the distance for molecules to diffuse between the seawater NO₃^- reservoir and the denitrification layer is smaller (Brandes and Devol, 1997; Lehmann et al., 2007). For instance, ¹⁵ϵ_app of 6 ± 2 ‰ were observed in sediments under hypoxic bottom waters in the Bering Sea (Lehmann et al., 2007). Sediments underlying anoxic waters in the Peruvian OMZ displayed even higher ¹⁵ϵ_app of 7.4 ± 0.7 ‰, decreasing to 2.5 ± 0.9 ‰ under oxygenated waters below the OMZ, with similar trends in ¹⁸ϵ_app (Dale et al., 2019). Dähnke and Thamdrup (2013) measured ¹⁵ϵ_app and ¹⁸ϵ_app of 18.9 ‰ and 15.8 ‰, respectively, in anoxic ex-situ incubations of organic-rich sediment. These results confirm observations and model predictions that nitrate isotope effects inferred from benthic flux data are under-expressed under oxic waters (Brandes and Devol, 1997; Lehmann et al., 2004; Lehmann et al., 2007). However, the assumption of a universally low or negligible benthic N isotope effect, as originally predicated in earlier studies, is not fully applicable to oxygen-deficient settings (Brandes and Devol, 1997; Lehmann et al., 2004).

¹⁵ϵ_app and ¹⁸ϵ_app were substantially lower than the values of 8 and 14 ‰, respectively, determined in a single chamber at St. 1 in 2011 (Dale et al., 2014). While some temporal variability is to be expected, the difference between 2011 and 2014 is as much striking as it is puzzling, because the redox conditions and fluxes were not radically different. Lower O₂ levels in 2011 are unlikely to be the cause (see Section 5.3). δ¹⁵N-NO₃ and δ¹⁸O-NO₃ were measured in 2011 using the Cd-reduction/azide method with addition of NaCl (Ryabenko et al., 2009) and verified against the same international standards USGS34 and IAEA-NO-3. In 2011, the BIGO samples for isotope analysis were taken in quartz tubes whereas in 2014 the syringe samples were used (see Methods). The quartz tube samples were first analyzed on board for dissolved gases by passing the sample (without air contact) through the membrane inlet of a mass spectrometer prior to filtering and freezing. It is currently unclear whether this procedure significantly altered the isotopic composition of NO₃^-. Unfortunately, additional samples from the syringes were not analyzed during the 2011 campaign for comparison, and we are presently unable to resolve the inter-annual differences in ¹⁵ϵ_app and ¹⁸ϵ_app.

A more realistic estimate of the total benthic N isotope effect on the oceanic N reservoir can be made from the flux of all fixed N species across the sediment-water interface, termed ϵ_sed by Lehmann et al. (2007). ¹⁵ϵ_sed can be higher than ¹⁵ϵ_app in continental margins if the sediment acts as a source of ¹⁵N-enriched NH₄⁺ to the water column, which in our setting has a δ¹⁵N of ~9.0 ‰ (Figure 3C). Vertical mixing and oxidation of ¹⁵N-enriched NH₄⁺ higher up in the water column will then increase the δ¹⁵N of the ocean NO₃^- reservoir. Mathematically, ¹⁵ϵ_sed can be calculated as:

where δ_JNreac is the δ¹⁵N of the net flux of reactive N species (NO₃^-, NO₂^-, NH₄⁺, PON) and δ_JN2 is the δ¹⁵N of all fixed N lost to N₂ (δ_JN2 ). This expression ignores dissolved organic nitrogen, which may be important in some settings (Alkhatib et al., 2012). Alkhatib et al. (2012) used a similar approach to determine ¹⁵ϵ_app based on NO₃^- fluxes (Eq. 7). Accurate estimations of ¹⁵ϵ_sed are hampered by the fact that δ_JNreac requires accurate isotope values for the fixed N fluxes including PON remineralization (Dale et al., 2019). Similarly, precise analytical determination of δ_JN2 is difficult due to the large background reservoir of N₂ in seawater (Brandes and Devol, 1997). In the following section, ¹⁵ϵ_sed is estimated using a numerical model of the benthic chamber incubation.

Deviations of ¹⁸ϵ_app:¹⁵ϵ_app above 1 are generally interpreted as the superimposition of the kinetic isotope effects of nitrification of NH₄⁺ and denitrification (e.g. Lehmann et al., 2004; Dale et al., 2014). NH₄⁺ is initially released to the porewater by the ammonification of PON with a small KIE of 1 – 2 ‰ (Möbius, 2013). The δ¹⁵N of PON in the upper 30 cm of sediment on the margin is ~ 5 ‰ (Table S1). Hence, the δ¹⁵N of NH₄⁺ from ammonification should be close to this value. By contrast, nitrification of NH₄⁺ to NO₂^- is associated with a large kinetic N isotope effect of 14 – 38 ‰ (Marioitti et al., 1981; Casciotti et al., 2003). The mass balance (R_NIT/R_AMF< 1) and the ¹⁵N-enriched NH₄⁺ effluxes indicate that NH₄⁺ is only partially nitrified. If newly nitrified and ¹⁵N-depleted NO₂^- is quantitatively oxidized to NO₃^-, the ¹⁵N-depleted NO₃^- will dilute the ambient porewater NO₃^- pool with ¹⁴NO₃^- and counteract the loss of ¹⁴NO₃^- incurred through dissimilatory reduction (Lehmann et al., 2004; Lehmann et al., 2007). Nitrification does not lead to a similar behavior of ¹⁸ϵ_app because the δ¹⁸O of newly nitrified NO₃^- is largely set by the δ¹⁸O of seawater and is equal to ca. 0 ‰ (Buchwald and Casciotti, 2010; Casciotti et al., 2010; Buchwald et al., 2012). In other words, the δ¹⁵N of NO₃^- in porewater subject to denitrification and partial nitrification will increase less quickly than its δ¹⁸O signature, as long as the isotope effects for N and O by nitrate reduction are equivalent, leading to ¹⁸ϵ_app:¹⁵ϵ_app.

The isotopic fingerprints of partial nitrification must be communicated back to the overlying chamber water in order to bring about a change in observed ¹⁸ϵ_app:¹⁵ϵ_app. This could take place directly by the injection of NO₃^- from the porewater into the chamber through flushing of animal burrows (Lehmann et al., 2004; Dale et al., 2014; Rooze and Meile, 2016), or indirectly by the alteration of the concentration gradients of NO₃^- isotopes at the sediment-water interface and subsequent changes in diffusive fluxes of individual isotopes (Bender, 1990; Brandes and Devol, 1997). All other things being equal, the combined effects of denitrification and partial nitrification ought to lead to ¹⁸ϵ_app:¹⁵ϵ_app values above 1 (Lehmann et al., 2004; Dale et al., 2014).

The turnover of NO₂^- adds an additional dimension to the N cycle that has received far more attention in the water column compared to sediments (Casciotti and McIlvin, 2007; Casciotti et al., 2013; Granger and Wankel, 2016). Reduction of NO₂^- occurs with a normal KIE (Casciotti et al., 2002), whereas NO₂^- oxidation occurs with an inverse KIE in both N and O, leaving residual NO₂^- depleted in ¹⁵N and ¹⁸O (Casciotti, 2009; Buchwald and Casciotti, 2010). Concurrent nitrate reduction and nitrite oxidation can lead to an enrichment of ¹⁵N in the ambient NO₃^- pool and an enrichment of ¹⁴N in NO₂^-. Several lines of evidence based on NO₂^- isotopes suggest that a significant fraction of NO₂^- produced by nitrate-reducing bacteria in oxygen deficient waters is reoxidized to NO₃^-, resulting in the δ¹⁵N of NO₂^- being up to 60 ‰ lower than coexisting NO₃^- (Casciotti and McIlvin, 2007; Casciotti et al., 2013). A comprehensive box model of N and O isotope dynamics in the water column has since shown that the extent to which NO₃^- is produced relative to the fraction that is reduced plays a central role in determining the trajectories of δ¹⁵N-NO₃^- and δ¹⁸O-NO₃^- and hence ¹⁸ϵ_app:¹⁵ϵ_app (Granger and Wankel, 2016). The relevance of NO₂^- for benthic N fractionation and ¹⁸ϵ_app:¹⁵ϵ_app trajectories has been largely unexplored in benthic N isotope studies. In the following, we examine NO₂^- cycling in sediments more closely using a numerical model.

5.2. Insights from model simulations of a benthic chamber incubation experiment

5.2.1. Model set-up and calibration

In this section, we set up a multi-process numerical model of dual N and O isotope dynamics. The model consists of a 1-D, vertically-resolved, reaction-transport model of the surface sediment (upper 12 cm) coupled to a single box representative of a virtual benthic chamber, similar to the one described by Dale et al. (2019) for the Peruvian shelf. In this first step, the model is parameterized using data from St. 2 to constrain key model parameters and fractionations of the benthic N cycle for the Mauritanian shelf. In the second step, we apply the model more generally to explore the sensitivities of ¹⁸ϵ_app:¹⁵ϵ_app ratios toward different pathways of N turnover. We also interrogate the impact of the ¹⁵N-enriched NH₄⁺ efflux on the benthic N isotope effect on water column NO₃^-. Since most of the model structure and reaction network have been described elsewhere, we relegate the bulk of the corresponding model details to the Supplement, and here we only discuss the relevant aspects for this study (Bohlen et al., 2011; Dale et al., 2019).

The degradation of POC in the sediment drives the benthic N cycle, which in turn impacts the concentrations of N species and their isotopes in the benthic chamber. No N cycling is assumed to take place in the benthic chamber water, such that the measured chamber data act as passive tracers for the processes taking place in the sediment. The model includes most of the processes in the water column box model of Granger and Wankel (2016), including aerobic NH₄⁺ oxidation (AMO), NO₃^- reduction (NAR), aerobic NO₂^- oxidation (NXR), and NO₂^- reduction to N₂ (NIR). In our model, NAR and NIR are coupled to the degradation of POC. Other electron acceptors for POC remineralization include O₂ and sulfate (SO₄^2-). The transition from one metabolic pathway to another during POC respiration depends on the relative concentrations of the terminal electron acceptors (TEA) and user-defined TEA threshold concentrations akin to half-saturation constants (Jourabchi et al., 2005). The TEAs are consumed in the order O₂: NO₂^-: NO₃^-: SO₄^2-, such that a high concentration of a TEA will inhibit less-favorable metabolic pathways. We also include ammonification of PON (AMF) that is linked to POC degradation through the atomic C:N ratio of organic matter. NH₄⁺ and NO₂^- oxidation during AMO and NXR are defined as second-order reactions with rate constants k_AMO and k_NXR, respectively. At this stage, we do not consider anammox as it is a minor N loss pathway on the margin (Sokoll et al., 2016). We later relax this assumption for the sensitivity analysis. Solutes are able to diffuse through the porewater and across the sediment-water interface into and out of the benthic chamber depending on the reactions taking place in the sediment. Solute transport is also affected by bioirrigation, expressed mathematically as a simple non-local mixing function, and by bioturbation, expressed as a diffusion-like process.

The rate of organic matter degradation typically decreases with depth in the sediment as it becomes more recalcitrant over time (Middelburg, 1989). This can be constrained by the corresponding increase in NH₄⁺ and total alkalinity concentrations (e.g. Berner, 1980). However, these analytes are of limited use to determine POC degradation on the Mauritanian shelf because porewaters are intensively irrigated with seawater by tube-dwelling animals (Gier et al., 2017). Therefore, following Stolpovsky et al. (2015), we used an empirical power law function to parameterize POC mineralization as function of sediment depth, x, termed RPOC(x):

where RRPOC is the rain rate of POC to the seafloor (in mmol m^-2 d^-1). Comparisons can be drawn between Eq. (11) and the power model proposed by Middelburg (1989) where the rate constant for POC remineralization depends on the age of POC being degraded. If RRPOC provides an upper limit of the total amount of POC available for degradation, B1 can be defined as:

The value of B2 was determined empirically by simulating a global dataset of benthic O₂ and NO₃^- fluxes (Stolpovsky et al., 2015):

We assigned a value of 16.7 mmol m^-2 d^-1 to RRPOC, based partly on the measured TOU and partly on achieving a good fit to the benthic chamber NH₄⁺ and NO₃^- fluxes. This gives B₁ and B₂ values of 0.76 cm and -2.3 (dimensionless), respectively.

The KIEs associated with nitrogen cycling (ϵ_AMO, ϵ_NAR, ϵ_NIR, ϵ_NXR), the TEA threshold NO₃^- and NO₂^- concentrations during NAR and NIR (K_NAR, K_NIR) and the rate constants k_AMO and k_NXR were mainly constrained by fitting the model to the benthic chamber data and, to a lesser extent, to the porewater data (see Supplement Figure S3). All other model parameters were fixed at reasonable values for fine grained shelf sediments, or values derived from field observations. Following Granger and Wankel (2016), we assumed that ¹⁵ϵ_NAR = ¹⁸ϵ_NAR, ¹⁵ϵ_NIR = ¹⁸ϵ_NIR, and ¹⁵ϵ_NXR ≠ ¹⁸ϵ_NXR. The δ¹⁵N of NH₄⁺ produced from PON was assigned a value of 5 ‰, which is close to the mean δ¹⁵N of oceanic NO₃^- (Brandes and Devol, 2002). The δ¹⁸O of the NO₂^- produced by AMO was fixed (-2.3 ‰, see Supplement) and calculated as a function of the δ¹⁸O of ambient H₂O and O₂, and the fractionation factors associated with their incorporation into NO₂^- (Casciotti et al., 2010). Similarly, the δ¹⁸O of the NO₃^- produced by NXR was set by the inverse O isotope effect of NO₂^- oxidation (¹⁸ϵ_NXR) and the isotope effect of O incorporation into NO₂^- from water (Buchwald and Casciotti, 2010). In addition to NXR and NIR, the δ¹⁸O of the NO₂^- pool is affected by the branching isotope effect of O atom extraction during NAR (¹⁸ϵ_NARBR), which was assigned a value of 25 ‰ (Granger and Wankel, 2016). In contrast to observations in the water column, we assumed that the O atoms in NO₂^- are not equilibrated with water (Buchwald and Casciotti, 2013). This is because the abiotic equilibration time is days to weeks (Buchwald and Casciotti, 2013), whereas the residence time of NO₂^- in shelf porewaters in our simulations is typically on the order of minutes. In deep-sea sediments, where NO₂^- turnover rates are far lower, abiotic equilibration must be considered (Buchwald et al., 2018).

As with most ecological models, ours is underdetermined with respect to the number of field observations available to constrain the unknown parameter values. The lack of δ¹⁵N and δ¹⁸O data for NO₂^- from the benthic chambers unfortunately compromises the accuracy of the model simulations. To reduce the margin of error, we ran the model using a Monte Carlo procedure to find the optimum fit to the chamber data by tuning the nine unknown parameters to the benthic chamber data. The model was run over an ensemble of 20 k iterations, in each case with a random set of parameter values chosen from typical ranges reported in the literature (Dale et al., 2014). The parameter values from simulations meeting a least squares criterion for the fit between the model and the benthic chamber data were then averaged. These are listed in Table 3 along with other model outputs of interest including ratios of depth-integrated rates and benthic isotope effects.

Table 3 Key model results for the Mauritanian shelf ^a.

The optimized isotope effects for ϵ_AMO (35 ± 3 ‰), ϵ_NAR (9 ± 3 ‰), ϵ_NIR (14 ± 6 ‰), ¹⁵ϵ_NXR (-8 ± 5 ‰) and ¹⁸ϵ_NXR (-7 ± 5 ‰) are well-within ranges of previously-determined isotope effects in the water column (reviewed by Granger and Wankel, 2016) and sediments (Lehmann et al., 2007; Alkhatib et al., 2012; Dale et al., 2014; Dale et al., 2019). The best-fit model solutions (n = 54) of the chamber data agree well with the measured data (Figure 4). Modeled δ¹⁵N-NO₂^- and δ¹⁸O-NO₂^- are clearly sensitive to the set of best-fit parameters. This suggests that NO₂^- isotope data would help to narrow the parameter uncertainty in Table 3. It is interesting that the δ¹⁸O-NO₂^- flux is always positive (mean = 27 ± 6 ‰). This is explained by (i) the δ¹⁸O of newly nitrified NO₂^- is set by the δ¹⁸O of seawater, (ii) ¹⁸ϵ_NIR > ¹⁸ϵ_NXR, which enriches porewater NO₂^- in ¹⁸O, and (iii) only a small fraction of NO₂^- is reoxidized to NO₃^- (NXR/NAR = 9%). In contrast, the δ¹⁵N of newly nitrified NO₂^- (mean = -3 ± 5 ‰) can be positive or negative because it depends on ϵ_AMO and the fraction of NH₄⁺ that is oxidized. The δ¹⁵N of the NO₂^- flux is more sensitive to ϵ_AMO than any other model parameter (not shown), with higher ϵ_AMO leading to more ¹⁵N-depleted NO₂^- fluxes.

Figure 4 Model simulation results (curves, n = 54) of the benthic chamber incubation experiment at St. 2 on the shelf (BIGO 2-5) including (A–C) N concentrations, (D–H) N and O isotopes, and (I) O₂ concentrations (see text). Open and filled symbols correspond to chamber 1 and 2, respectively. For the O₂ optode data, chamber 1 and 2 are denoted by the thin and thick black lines, respectively, and have been normalized to the initial values.

The model predicts that 14% of NH₄⁺ produced from PON is oxidized to NO₂^- (AMO/AMF). This is higher than predicted by the mass balance (RNIT/RAMF ~7%, Table 2), yet necessary in order to reproduce the efflux of ¹⁵N-enriched NH₄⁺. The disagreement with the model and mass balance may be rooted in the fact that the latter does not explicitly consider NO₂^-. A better agreement between the model and the mass balance was found for denitrification, with the model predicting a fixed N loss of 2.0 ± 0.1 mmol m^-2 d^-1.

The N and O isotope effects for NO₃^- reduction (¹⁵ϵ_app = 0.8 ± 0.3 ‰, ¹⁸ϵ_app = 1.2 ± 0.4 ‰) and the ratio ¹⁸ϵ_app:¹⁵ϵ_app (1.8 ± 0.8 ‰) are within the uncertainty of the observed values (Figure 3). In sediments, high ¹⁸ϵ_app:¹⁵ϵ_app are thus possible at low NXR/NAR, whereas in the water column elevated ¹⁸ϵ_app:¹⁵ϵ_app are associated with NXR/NAR of >50% (Casciotti et al., 2013; Granger and Wankel, 2016). This discrepancy probably arises from the abundance of NH₄⁺ in porewaters produced by anaerobic remineralization pathways, mainly sulfate reduction, such that only a small amount of NH₄⁺ oxidation with a high ϵ_AMO is needed to produce large ¹⁸ϵ_app:¹⁵ϵ_app trajectories. In the water column, NH₄⁺ production is stoichiometrically constrained by organic matter remineralization using O₂, NO₂^- and NO₃^-, and elevated ¹⁸ϵ_app:¹⁵ϵ_app are only feasible with tightly coupled NO₂^- oxidation and NO₃^- reduction.

It is noteworthy that ϵ_NAR is much higher than ¹⁵ϵ_app, confirming that the intrinsic cellular fractionation of NO₃^- reduction is not fully expressed in the benthic NO₃^- flux (Lehmann et al., 2007). ϵ_NAR is similar to 10 - 15 ‰ determined in pure cultures of denitrifying bacteria under typical marine conditions (Kritee et al., 2012) and to estimates from our previous box model of shelf sediments (~13 ‰, Dale et al., 2014). It also fits well with modelled data from the North Atlantic and Eastern Tropical North Pacific water column (14 ± 2 ‰, Granger and Wankel, 2016). Others have reported comparable values using model approaches that account for both closed and open system effects of NO₃^- reduction on ϵ_NAR (Deutsch et al., 2004; Altabet, 2007).

The total benthic N isotope effect, ¹⁵ϵ_sed, calculated using Eq. (10), equals 3.6 ± 0.5 ‰. As expected, it is higher than ¹⁵ϵ_app because of the efflux of ¹⁵N-enriched NH₄⁺. The NO₂^- flux, although acting to decrease ¹⁵ε_sed, hardly affects ¹⁵ϵ_sed in this setting since the flux is relatively small (Table 2). The modelled ¹⁵ϵ_sed is consistent with values of 2 – 6 ‰ determined from porewater gradients in the upper St. Lawrence Estuary where O₂ levels are below 100 µM (Alkhatib et al., 2012). It is also close to estimates of 4 – 7 ‰ reported by Lehmann et al. (2007) for the upper continental margin and with results from a global N isotope model (2 – 4 ‰; Somes et al., 2013).

The environmental controls on ¹⁵ϵ_sed are not well understood presently, and correlations between ¹⁵ϵ_sed and organic matter quality and bottom water O₂ levels have been discussed (Lehmann et al., 2007; Alkhatib et al., 2012; Dale et al., 2019). However, changes in organic matter quality and bottom water redox conditions are often intertwined (Cowie et al., 2014), and their individual contribution to ¹⁵ϵ_sed has not yet been investigated under laboratory conditions. Furthermore, the relationship between ¹⁵ϵ_sed and bioirrigation is highly non-linear and far more intricated than the simple mathematical expression of bioirrigation used in our model (Rooze and Meile, 2016). The contribution of dissolved organic N fluxes to ¹⁵ϵ_sed also deserves to be explored in future studies due to its significance in other organic-rich environments (Alkhatib et al., 2012). Despite these caveats, our results provide supporting evidence that benthic N fractionation on the continental margin exceeds previously assumed values of 0 – 2 ‰ for marine sediments (Brandes and Devol, 2002; Altabet, 2007).

5.2.2. Sensitivity of ¹⁵ϵ_app, ¹⁸ϵ_app:¹⁵ϵ_app and ϵ_sed to N cycling pathways

We now apply the model to explore the sensitivity of ϵ_app, ¹⁸ϵ_app:¹⁵ϵ_app and ¹⁵ϵ_Sed to a broader range of N turnover rates, using the best-fit ϵ values in Table 3. To keep the procedure manageable, parameters associated with the physical environment (e.g. compaction, bioirrigation, boundary conditions) and RPOC(x) were left unchanged. Our primary interest here is on the internal interplay of the N cycle pathways in a continental shelf setting like the Mauritanian margin. Anammox (AMX) is now included to make the model more generic and to test the impact of AMX/AMF on the benthic isotope effect (see Supplement). Fractionations of 24 ‰ (¹⁵ϵ_AMXNH4) and 16 ‰ (¹⁵ϵ_AMXNO2) were assigned to NH₄⁺ oxidation and NO₂^- reduction, respectively (Brunner et al., 2013). The ¹⁸O isotope for NO₂^- reduction (¹⁸ϵ_AMXNO2) was set equal to ¹⁸ϵ_NIR (Granger and Wankel, 2016). During AMX, an additional 0.3 moles of NO₂^- are oxidized to NO₃^- per mole of NO₂^- reduced, which accounts for the reduction of inorganic carbon during production of biomass (Strous et al., 1999). NO₂^- oxidation by AMX has an inverse N isotope effect (¹⁵ϵ_AMXNO2ox) of around -30 ‰ and produces ¹⁵N-enriched NO₃^- (Brunner et al., 2013). The δ¹⁸O of the NO₃^- was defined analogously as described above for NXR, assuming that the O isotope effect of NO₂^- oxidation by AMX (¹⁸ϵ_AMXNO2ox) is equal to ¹⁸ϵ_NXR (Granger and Wankel, 2016).

Model results show linear and non-linear dependencies of ϵ_app, ¹⁵ϵ_Sed and ¹⁸ϵ_app:¹⁵ϵ_app on NXR, AMO and AMX (Figure 5). Over the ranges tested, ¹⁵ϵ_app is most sensitive to the proportion of NH₄⁺ oxidized by AMO and AMX (Figure 5A). ¹⁵ϵ_app increases steadily with AMX (grey curve), which is due to the production of ¹⁵N-enriched NO₃^- by the negative N isotope effect of NO₂^- oxidation. On the contrary, ¹⁵ϵ_app barely changes when NO₂^- oxidation by AMX is turned off (green curve). The production of ¹⁵N-enriched NO₃^- by AMX enhances ¹⁵ϵ_app in the following way. Addition of ¹⁵NO₃^- to the porewater slightly weakens the ¹⁵NO₃^- concentration gradient at the sediment surface relative to ¹⁴NO₃^-, leading to an increase in porewater δ¹⁵N-NO₃^- (Figure 6A). The flux of ¹⁵NO₃^- into the sediment is reduced (relative to ¹⁴NO₃^-) and brings about a lowering of the δ¹⁵N of the NO₃^- flux into the sediment (Figure 6B). As a consequence, ¹⁵ϵ_app increases (Eq. (7)). When NO₂^- oxidation by AMX is turned off, the substrates for AMX (NH₄⁺ and NO₂^-) are permanently lost as N₂ with no effect on NO₃^-, which curtails any impact on ¹⁵ϵ_app. A similar effect of AMX is observed for ¹⁸ϵ_app due to the negative O isotope effect of NO₂^- oxidation (grey curve, Figure 5B), which together with ¹⁵ϵ_app produces very little change in ¹⁸ϵ_app:¹⁵ϵ_app over the ranges of AMX tested (grey curve, Figure 5C). This finding is tentative, but agrees with studies in the water column where AMX fails to explain ¹⁸ϵ_app:¹⁵ϵ_app trajectories >1 (Granger and Wankel, 2016). Overall, the model suggests that observed ¹⁸ϵ_app:¹⁵ϵ_app trajectories > 1 on the Mauritanian shelf are not easily explainable by anammox alone.

Figure 5 Model sensitivity analysis of (A) ¹⁵ϵ_app, (B) ¹⁸ϵ_app, (C) ¹⁸ϵ_app:¹⁵ϵ_app, and (D) ¹⁵ϵ_sed towards AMO, AMX and NXR, given as percentage contributions to AMF or NAR. Blue, red and grey curves were generated by varying the rate constant for AMO (k_AMO), NXR (k_NXR), and AMX (k_AMX) stepwise between 0 and 2.6 × 10⁷ cm³ µmol^-1 yr^-1. The green curves are the results for AMX/AMF without oxidation of NO₂^- to NO₃^-. All other parameter values in Table 3 and Table S2 are held constant.

Figure 6 The effect on increasing AMX/AMF on (A) the δ¹⁵N-NO₃ of porewater NO₃^- at the sediment surface and on (B) the δ¹⁵N of the NO₃^- flux into the sediment. In (A), an increase in AMX/AMF leads to a steepening of the δ¹⁵N-NO₃ gradient and consequently, in (B), to a decrease in δ_FNO3. Each curve in (A) has a corresponding symbol in (B). The green lines in (A) are almost superimposed, reflecting a negligible change in δ¹⁵N-NO₃ when NO₂^- oxidation by AMX is turned off. See Figure 5 for more details.

AMO has a different impact on ¹⁵ϵ_app since NH₄⁺ is recycled to NO₂^-, part of which can be oxidized to NO₃^-. At low AMO rates, nitrified NO₂^- is strongly depleted in ¹⁵N due to the large N isotope effect, ϵ_AMO. The portion of this newly nitrified ¹⁴NO₂^- that is oxidized to ¹⁴NO₃^- causes a weakening of the ¹⁴NO₃^- concentration gradient and a decrease in ¹⁵ϵ_app (blue curve, Figure 5A). As NH₄⁺ becomes more quantitatively consumed (AMO/AMF > ~30%), newly nitrified NO₂^- and NO₃^- are increasingly ¹⁵N-enriched, reversing the impact on the NO₃^- concentration gradients and producing an increase in ¹⁵ϵ_app. By contrast, ¹⁸ϵ_app decreases by less than 0.5 ‰ for the full range of AMO/AMF because the δ¹⁸O of newly nitrified NO₂^- is fixed (-2.3 ‰). The N and O effects of AMO combine to give a trajectory in ¹⁸ϵ_app:¹⁵ϵ_app that is highest at intermediate values of AMO/AMF (blue curve, Figure 5C). If AMO/AMF exceeds ~60%, ¹⁸ϵ_app:¹⁵ϵ_app even falls below 1. Thus, for the present model configuration, AMO can explain a wide range of ¹⁸ϵ_app:¹⁵ϵ_app ratios observed in the field.

In contrast to AMO/AMF, the fraction of NO₂^- that is oxidized relative to NO₃^- reduction (NXR/NAR) is important for ¹⁸ϵ_app but not for ¹⁵ϵ_app (red curves, Figures 5A, B). Following the reasoning based on diffusive gradients, the increase in ¹⁸ϵ_app with NXR can be explained by preferential production of N¹⁸O₃^- in the oxidized surface layers. This is conceivable because the δ¹⁸O of newly nitrified NO₃^- is determined by the δ¹⁸O of O₂ and H₂O and the inverse O isotope effect of NXR (ϵ_NXR = -7 ‰). Since newly produced NO₂^- has a δ¹⁸O of -2.3 ‰ and is oxidized with an inverse isotope effect, the δ¹⁸O of newly nitrified NO₃^- will always be higher than the δ¹⁸O of NO₃^- reduced (ϵ_NAR = 9 ‰). ¹⁸ϵ_app rises sharply to 1.8 ‰ for NXR/NAR = 20%. (Higher ratios could not be simulated in the model due to diffusive constraints on the O₂ flux). This trend is carried through to the ¹⁸ϵ_app:¹⁵ϵ_app trajectory that also shows a steep increase (red curve, Figure 5C). NXR, therefore, is also a viable candidate for producing ¹⁸ϵ_app:¹⁵ϵ_app > 1 in benthic chambers, as proposed for the water column (Granger and Wankel, 2016).

The model results show that ¹⁵ϵ_Sed is more sensitive to AMO and AMX compared to NXR, largely due to escape of residual ¹⁵NH₄⁺ from the sediment (Lehmann et al., 2007; Alkhatib et al., 2012). Although anammox also has a large N isotope effect of 24 ‰ for NH₄⁺ oxidation, it does not present as similarly strong of a dynamic as AMO. Reasons for this are difficult to disentangle due to the tight coupling of the N turnover pathways, but the depth in the sediment where AMO and AMX take place may be important. Ammonium oxidation is restricted to the oxidized surface sediment that is only a few millimeters thick, which allows residual ¹⁵NH₄⁺ to quickly escape to the bottom water. Anammox, in contrast, occurs over a greater depth interval where NO₃^- and NO₂^- are available (Dalsgaard et al., 2005), implying a greater path length for residual ¹⁵NH₄⁺ to diffuse into the bottom water. We suggest that this raises the likelihood that residual ¹⁵NH₄⁺ can be oxidized by AMO or AMX, thereby muting the intrinsic isotope fingerprint of AMX on ¹⁵ϵ_Sed. High rates of AMX (and AMO) even tend to decrease ¹⁵ϵ_Sed, presumably because the total flux of NH₄⁺ leaving the sediment is reduced at higher consumption rates.

A final point of significance is that ¹⁵ϵ_Sed and ¹⁵ϵ_app display opposing trends with regards to AMO. At low rates, AMO is associated with a decrease in ¹⁵ϵ_app, whilst simultaneously increasing ¹⁵ϵ_Sed, and vice-versa at high rates (blue curves Figures 5A, D). ¹⁵ϵ_app is not, therefore, a reliable indicator of the total sediment N isotope effect, ¹⁵ϵ_Sed. The interplay of the N cycle leads to a difference between ¹⁵ϵ_Sed and ¹⁵ϵ_app that can be several ‰ at intermediate AMO/AMF values. This is a non-trivial amount considering that the oceanic N isotope balance is sensitive to small changes in the benthic N isotope effect (Brandes and Devol, 2002). For the continental margin, where most benthic denitrification takes place, the contribution of the NH₄⁺ flux to the total benthic N isotope effect may be as important as NO₃^-. Consequently, δ¹⁵N-NH₄ measurements in benthic chambers should be prioritized alongside δ¹⁵N-NO₃.

5.3. Interannual variability of nutrient fluxes

As with most continental margins, benthic fluxes and N turnover pathways on the Mauritanian margin at 18°N reveal an intensification of nutrient recycling on the shelf and a sharp decrease offshore (Jahnke et al., 1990). This trend is driven by the increasingly recalcitrant nature of bulk sedimentary POC and POC rain rate with water depth (Dale et al., 2014).

The deposition flux of labile POC to the seafloor is positively correlated with benthic denitrification, and high rates of fixed N loss to N₂ are observed in both fine-grained and sandy continental shelves (Middelburg et al., 1996; Cook et al., 2006). The mean fraction of POC respired by denitrification on the Mauritanian margin (24%, Figure 1C) is higher than the global average of ≤10% (Middelburg et al., 1996). This can be explained by the amplification of denitrification in sediments underlying oxygen – deficient bottom waters due to less inhibition by O₂. The impact of denitrification is to raise the RQ since it produces DIC whilst not impacting TOU. If the 24% of POC that is remineralized by denitrification on the margin would instead be channeled through aerobic pathways, we might expect the TOU to increase by a similar amount with a corresponding decrease in RQ from 1.5 to 1.2. This is higher than the theorical value of 0.7 based on complete aerobic mineralization of organic matter with an average phytoplankton composition (Anderson, 1995) and supports the idea that benthic carbonate dissolution is driving an excess DIC flux across the margin.

The loss of fixed N relative to PO₄^3- generates N-depleted recycling fluxes (negative N*) compared to the standard phytoplankton N:P composition of ~16. If the calculated denitrification rates are added to F_N*, the mean F_N* increases from -2.8 to -1 mmol m^-2 d^-1. indicating that N loss accounts for roughly two thirds of the negative N* flux. Enhanced PO₄^3- release under low-oxygen bottom waters probably accounts for the remaining third (Schroller-Lomnitz et al., 2019). Low or negative N* fluxes are characteristic of sediments under oxygen-deficient waters. Outside of these areas, where O₂ levels are > 100 µM, N-depleted fluxes are much more muted (Bohlen et al., 2012), partly due to more efficient sequestration of PO₄^3- by adsorption onto iron (oxyhydr)oxide minerals (Sundby et al., 1986). Mixing of low N:P waters from the benthic boundary layer into oxygen-poor water columns has the potential to enhance N₂ fixation, leading to a positive feedback between primary production, O₂ drawdown and benthic P release (Kemena et al., 2019).

On average, NH₄⁺ and PO₄^3- fluxes on the shelf were 2 to 3 times higher in June 2014 compared to 2011 (Dale et al., 2014; Schroller-Lomnitz et al., 2019). NO₃^- and H₄SiO₄ fluxes were also elevated by 40 and 25%, respectively. Mean TOU in 2011 and 2014 was the same (ca. 10 mmol m^-2 d^-1), and benthic carbon respiration rates on the shelf in 2014 were similar (RPOC_TOT = 10 – 11 mmol m^-2 d^-1). Given that the release of NH₄⁺ and PO₄^3- from sediments is typically regulated by nitrification and adsorption by iron (oxyhydr)oxide minerals, respectively (Sundby et al., 1986; Sloth et al., 1992), the higher NH₄⁺ and PO₄^3- fluxes in 2014 suggest that the bottom waters might have become more reducing since 2011. Indeed, a long-term decline of O₂ concentrations in the ETNA has received much publicity (Stramma et al., 2008; Schmidtko et al., 2017). Our data corroborate these findings, with a decrease in minimum bottom water O₂ concentrations on the shelf from 42 µM in 2011 (Dale et al., 2014) to 25 µM in 2014 (Table 1).

One could conclude that the observed increase in NH₄⁺ and PO₄^3- fluxes between 2011 and 2014 is caused by the long-term decrease in bottom water O₂. Yet, other contributing factors that are not easily quantifiable deserve careful consideration, such as (i) seasonal accumulation and depletion of nutrients in porewaters (e.g. Dale et al., 2013), (ii) changes in the faunal community structure and sediment reworking rates (Dale et al., 2013; Gier et al., 2017), (iii) low frequency changes in O₂ respiration close to the seabed associated with particle resuspension (Thomsen et al., 2019), and (iv) the episodic formation and export of oxygen-poor mesoscale eddies over the shelf on quasi-monthly time-scales (Thomsen et al., 2019). The magnitude of the O₂ decrease between 2011 and 2014 observed on the landers (17 µM) is also much higher than observed deoxygenation rates in the ETNA of ca. 0.5 µmol kg^–1 yr^–1 (Stramma et al., 2008). Long-term monitoring of sediment fluxes is needed to address this issue more precisely, preferably using autonomous methodologies such as bottom crawlers that can operate independently of costly ship-based surveys (Wenzhöfer et al., 2016). Until the technology to do so becomes more accessible, future projections of the nutrient budgets and the economic sustainability of fisheries in tropical upwelling regions will carry larger uncertainties (Wallmann et al., 2022).

6. Conclusions

This study presents in situ fluxes of O₂, DIC, nutrients and stable isotopes of dissolved inorganic nitrogen species in benthic chambers deployed on the Mauritanian margin in 2014. The aim was to quantify the recycling rates of biogenic debris with particular emphasis on the N cycle. The trends in benthic fluxes were generally similar to those measured in an earlier campaign to the region in 2011, and associated with a decrease in organic matter remineralization and denitrification rates with water depth. Notably, though, the fluxes of NH₄⁺ and PO₄^3- on the shelf were 2 – 3 times higher in 2014. We were unable to pinpoint the cause of this increase from only two sets of field data. Long-term, or more frequent, monitoring of sediment-water interactions is needed to determine whether the seafloor is becoming more reducing and to elucidate possible linkages with the observed long-term decline in O₂ concentration in the tropical Atlantic. Elevated release of PO₄^3- from the standing stock of iron-bound P might stimulate the productivity of surface waters and continue to drive down subsurface O₂ levels, leading to a positive feedback between benthic PO₄^3- release and deoxygenation.

The mean isotopic fractionations of N (¹⁵ϵ_app) and O (¹⁸ϵ_app) accompanying benthic NO₃^- uptake were low and in good agreement with previously reported isotope effects for NO₃^- respiration in marine sediments. Analysis of the data using a numerical model showed that the overall benthic N isotope effect (¹⁵ϵ_sed) that includes the efflux of ¹⁵N-enriched NH₄⁺ from the seafloor, and the one that is of relevance to the global N isotope budget, bears no straightforward relationship with ¹⁵ϵ_app. The model predicted ¹⁵ϵ_Sed = 3.6 ‰ for the Mauritanian shelf which, whilst lower than observed isotope effects of pelagic N loss in oxygen-deficient regions, adds to the growing body of evidence that the assumption of a universally low or negligible benthic N isotope effect is not applicable to oxygen-deficient settings.

The isotope model was used to constrain the relative rates of N turnover processes in the sediment. The model showed that observed trajectories in ¹⁸ϵ_app:¹⁵ϵ_app > 1 on the shelf were likely the result of aerobic ammonium oxidation and nitrite oxidation in surface sediments. Anammox has little influence on ¹⁸ϵ_app:¹⁵ϵ_app in this setting. This finding may be more widely applicable because the isotopic impact of anammox is less easily communicated to the water column than ammonium and nitrite oxidation. Anammox, however, tends to increase ¹⁵ϵ_sed if the fraction of NH₄⁺ consumed is not too high. Our model suggests that NO₂^- isotope data from benthic chambers would help to constrain further the rates of N turnover in sediments, provided that the analytical challenge of low sample volumes from benthic chambers and low NO₂^- concentrations can be overcome.

Data availability statement

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://doi.pangaea.de/10.1594/PANGAEA.928206 https://doi.pangaea.de/10.1594/PANGAEA.941569.

Author contributions

AD conceived and wrote the paper. AD, DC, US-L and SS conducted fieldwork and chemical analyses. KD and FK performed isotope measurements. All authors contributed to the article and approved the submitted version.

Funding

This work is supported by the Sonderforschungsbereich 754 “Climate – Biogeochemistry Interactions in the Tropical Ocean” (www.sfb754.de) financed by the Deutsche Forschungsgemeinschaft (DFG), and also by the German Federal Ministry of Education and Research (BMBF, funding ref. no. 03F0815A) through the research project REEBUS (Role of Eddies in the Carbon Pump of Eastern Boundary Upwelling Systems).

Acknowledgments

We thank A. Petersen, M. Türk, and S. Cherednichenko for their assistance in deploying the landers, and B. Domeyer, V. Thönissen, S. Trinkler, S. Kriwanek, A. Bleyer, R. Suhrberg, F. Jung geochemical analyses. We also thank Thomas Hansen (GEOMAR) for measuring the C and N isotopes in sediment samples and to Fabian Jung for assisting with the BIGO isotope measurements. We are grateful to the crew of RV Meteor and RV Maria S. Merian for their enthusiastic support. We express our gratitude to Annie Bourbonnais for editorial handling of the manuscript, and to the reviewers for their expertise and input that improved the finalized version.

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmars.2022.902062/full#supplementary-material

References