Seasonal redox cycling of dFe and sFe in Celtic Sea shelf sediments
Oxygen penetration depths were most varied but, on average, shallowest in late spring (2.2–5.8 mm, Fig. 2), which is consistent with wider spatio-temporal assessments of OPD in the Celtic Sea (Hicks et al. submitted). Most likely, this reflects the enhanced supply of organic carbon to the sediment and metabolic consumption of O2, combined with enhanced macro benthic faunal activity at our study site. Higher rates of oxygen consumption and organic carbon oxidation are supported by modelled diffusion-consumption of O2 (Fig. 2). In addition, this hypothesis is supported by peak chlorophyll abundance, observed via MODIS satellites two weeks prior to our sediment sampling in late spring (Thompson et al. submitted), and particulate organic carbon concentrations in the surface layers (0–10 cm) that are highest in late spring (1.25%), and decrease by late summer (1.14%) as bloom-derived carbon is decomposed (Figure S1).
Beneath the sediment surface, dissolved macronutrients (NO3
+) and metals (Fe, Mn) in porewaters follow their anticipated biogeochemical depth-distributions during early diagenesis (Froelich et al. 1979; Burdige 2006). Following the consumption of O2, NO3
− is reduced and transformed to NH4
+. The benthic N-cycle is in reality complex but, briefly, the remineralisation of organic matter leads to the release of nitrogen in the form of NH4
+, which is immediately oxidised to NO3
− in the presence of oxygen, and then removed via denitrification and anammox in anoxic sediments below (e.g., Devol et al. 2015). Nitrate reduction is followed by the reduction of solid Mn and Fe oxides down-core and the release of their soluble reduced forms to porewaters (Fig. 2).
To date, there is little detailed knowledge of the redox state and size distributions of Fe in shelf sediment porewaters. Here we demonstrate that the proportion of porewater dFe as dFe(II) at our study site is >85% in the upper 3 cm and ~100% in the ferruginous zone below (Fig. 2). Thus, Fe(II) is supplied to porewater in the sub-surface (from the dFe(II) maxima, between 3 and 8 cm depth) and is prone to oxidative-removal towards the sediment surface, and most likely to sulphide-mineral trapping at deeper depths, resulting in the curvature typical of porewater Fe(II) profiles (Froelich et al. 1979; Severmann et al. 2006). Porewater profiles of Fe and Mn and their seasonal variations are similar to those in depositional areas of the southern and eastern North Sea (Slomp et al. 1997). Porewater dFe maxima are isotopically light (δ56Fe −1.7 to −0.9‰), consistent with supply of dFe(II) from the dissimilatory reduction of Fe(III) oxide by bacteria (Severmann et al. 2006; Homoky et al. 2009; Henkel et al. 2016). In addition, our observations of dFe, δ56dFe, dMn, nitrate and POC content in the Celtic Sea are similar to observed ranges in sediments from low-oxygen California-Oregon shelves (Severmann et al. 2006; Homoky et al. 2009, 2012) where benthic fluxes of isotopically light dFe to the ocean are observed (Severmann et al. 2010).
For the first time we have combined physicochemical Fe observations in porewater with speciation measurements. We show that porewater dFe(II) is almost entirely (>85%) in the soluble (<0.02 µm) size range, and colloidal forms of Fe are largely absent under these conditions (Fig. 4). Similarly, porewater dMn(II) is also found in the truly soluble phase (100% of dMn is <0.02 μm). The sFe forms may be simple ionic species, nano-particulate forms or may contain small ferrozine-reactive Fe(II) organic complexes. This contrasts with findings from deep sea sediments in the Crozet region, where on average 80% of Fe and 61% of Mn was in the colloidal size fraction (0.02–0.2 μm) (Homoky et al. 2009, 2011). Porewater dFe mainly being in the reduced form in the shelf sediments studied here also means that there is no evidence for any significant concentrations of Fe(III) organic complexes as reported for anoxic porewaters in estuarine systems (Jones et al. 2011; Beckler et al. 2015).
Dissolved Fe generated in sediments and supplied to porewater will diffuse along its concentration gradient towards regions of reactive consumption or transport loss. Iron(II) oxidation in porewaters may be coupled to O2 or NO3
− reduction (Laufer et al. 2016). There are good empirical basis’ to understand Fe(II) oxidation kinetics attributed to O2 (Millero et al. 1987) and in the presence of NO3
− (e.g., González et al. 2010), previously explored in a study of benthic Fe flux (Homoky et al. 2012). However, the impact of enzymatic Fe(II) oxidation via NO3
− reduction is still unclear. Near-surface gradients in dFe and dMn clearly indicate diffusion towards reaction in the surface oxic-layer, and potentially to the overlying water-column. A concomitant increase in sediment-leachable Fe and Mn is seen towards the sediment-water interface (Fig. 3), and accounts for an important fraction of dissolved Fe and Mn removal in surface sediments. A similar inverse correlation between porewater dFe(II) and hydroxylamine-HCl leachable Fe has been observed in surface sediment from the North Sea, where the leachable Fe pool was also suggested to originate from sub-surface DIR (Henkel et al. 2016). The ascorbic acid leach extracts easily reducible ferrihydrite, which is the first amorphous Fe oxyhydroxide phase precipitated due to Fe(II) oxidation (Raiswell et al. 2010). The hydroxylamine-HCl leach extracts ferrihydrite as well as other reactive Fe phases that have been argued to be bioavailable (Berger et al. 2008). Manganese seems to be trapped preferentially by materials released by the reducing hydroxylamine-HCl leach.
No sulphide was detected at any depths in our porewaters, and we observe negligible down-core enrichment of heavy Fe isotopes in porewater that would be indicative of removal to sulphides (Severmann et al. 2006). However, the gradual decrease in dFe(II) below its maximum (Fig. 3) most likely reflects downward diffusion as dFe(II) is converted to FeS in an underlying sulphate-reducing zone (Froelich et al. 1979).
The near-surface oxidation of Fe(II) to Fe(III) and the subsequent formation of Fe(III) oxides, is understood to preferentially incorporate heavier isotopes into authigenic Fe(III) phases, leaving behind lighter Fe(II) (e.g., Welch et al. 2003). Accordingly, a trend towards lower porewater δ56Fe is observed from ~6 cm depth (~ −1.0‰) towards the sediment surface (~ −3.0‰) during both seasons, indicative of oxidative Fe(II) removal, and recycling during DIR (Severmann et al. 2006; Homoky et al. 2009). A return to higher porewater δ56Fe in the uppermost sediment layer was observed during late spring and similar trends have been observed in sediment cores collected from shelf-slope sediments in the South East Atlantic (Homoky et al. 2013) and in the North Sea (Henkel et al. 2016). Henkel et al. (2016) reason that oxidative precipitation of Fe preferentially removes light isotopes, as proposed by Staubwasser et al. (2013), due to environmental variances in kinetic and equilibrium isotope fractionation processes compared to experiments (e.g. Welch et al. 2003). Whereas Homoky et al. (2013) reasoned that a transition to higher δ56Fe towards the sediment surface resulted from mixing with an isotopically heavier and more stable Fe source, that has a relatively low dFe concentration. In both scenarios, a potential role for organic complexion of Fe exists. If such organic complexes were to stabilise a fraction of dFe across the surface oxidising zone of porewaters, the isotopic composition of the dFe pool might shift towards heavier isotopic compositions (Dideriksen et al. 2008; Morgan et al. 2010), and would resist authigenic precipitation. Accordingly, our observed trend towards higher δ56Fe extended from surface porewater into oxygenated core-top water and bottom water samples, while dFe concentrations steadily decreased (Fig. 3).
Porewater dFe(II) was elevated in the surface (0–1 cm) during the late spring (4.5–13.4 μM, n = 3) compared to late summer (0.3–1.2 μM, n = 3, Fig. 5; Table S3), coincident with shoaling of the OPD, linked to the deposition of organic matter during the bloom. Such seasonal variations in dFe and dMn were also reported in the water column of the North Sea and attributed to bloom-promoted release from sediments (Schoemann et al. 1998). Unexpectedly, our observed late spring surface porewater dFe(II) values are in the same range as surface sediment dFe concentrations (~7 μM) from the high-carbon accumulating and low-oxygen Oregon Shelf (120 m water depth, Homoky et al. 2009), where substantial benthic fluxes of dFe have been measured. Albeit, our reported values remain lower than surface sediments underlying near-anoxic waters (e.g. ~100 μM dFe(II), (Severmann et al. 2006, 2010). Dissolved Fe concentrations in overlying bottom waters (20 nM, with 15 nM as dFe(II); Table S4) were 1–5 orders of magnitude lower than the upper centimetre of porewater, and even lower in bottom water 10 m above the seafloor (5–10 nM; Fig. 3). The presence of Fe(II) ranged from ~70% of dFe in directly overlying bottom water at our study site (Table S4) to ~4% at 10 m above sea floor (Fig. 6; Table S5), indicating that a significant fraction of upward diffusing Fe(II) is able to escape the oxidative trap in the surface sediments and enter the water column.
Impact of water column oxygen on release of benthic Fe(II)
Large benthic fluxes of dFe to the water column are widely reported in oxygen deficient zones and are on the order of 100–1000 μmol m−2 day−2 (Homoky et al. 2016 and references therein). These observations enable an empirical assessment of the impact of oxygen concentration on the release of Fe(II) from seafloor sediments (Dale et al. 2015). Parts of the UK shelf, other than our study site, seasonally undergo modest periods of reduced-oxygen concentration (e.g., 160–200 μmol l−1, compared to 280–310 μmol l−1 at other times of the year; Greenwood et al. 2010). To examine the likely impact of such changes in bottom water oxygen on the release of dFe from our study site, sediment cores and bottom water were sealed from the atmosphere, so that benthic respiration processes would draw down oxygen from the overlying water into the sediment. This resulted in similarly reduced dissolved oxygen concentrations at t = 0 h of ~150 and ~120 µM for late spring and late summer, respectively, but the accumulation of dFe in bottom waters was substantially different (Fig. 7). High dFe concentrations (up to 240 nM) were measured in the late spring experiment, while in the late summer dFe concentrations only reached ~25 nM. This indicates that seasonal differences in near surface pore water dFe concentration (Fig. 5) and OPD (Fig. 2) are important controls on the release of dFe to bottom waters.
During the late spring, aeration of the incubated core top water induced rapid oxidation of Fe(II) and removal of dFe from solution (Fig. 7). However, for both seasons the residual concentration of dFe is in the range 25–30 nM, of which 30–50% is present as Fe(II) despite reaching saturated oxygen concentrations—roughly 10 times greater than dFe concentrations reported for bottom waters at this site (5.4–10 nM, ~10 m above seafloor; Figs. 3, 6; Tables S4, S5). Seawater Fe(II) oxidation kinetics predict nearly all Fe(II) should be oxidised to Fe(III) in our experiments in just a few minutes (Millero et al. 1987) and for this reason it has been generally assumed that oxic shelves are not a significant source of Fe to the overlying water column. However, the observation of Fe(II) present in oxic waters over a period of >2 days suggests that rapid oxidation and fallout of Fe oxides is inhibited, due to some sort of Fe(II) and Fe(III) stabilisation. It is possible that organic carbon present during the late spring period, not only enhances the release of dFe, but also enhances the formation of organic ligands that are able to bind with Fe(II) and Fe(III) and serve to reduce the oxidative removal of dFe.
Fe(II) oxidation kinetics in core-top and water column seawater
The oxidation kinetics of Fe(II) were investigated in water column samples and in seawater overlying sediment cores (Fig. 8). Oxidation rates in core top water were nearly twice as slow as in bottom waters, which themselves were more than 5 times slower than our empirical predictions (respective [Fe(II)] half-lives were 41, 23 and 3.8 min after Millero et al. 1987). Further, the observed Fe(II) concentrations do not rigorously follow first order kinetics. Such behaviour has been observed in hydrothermal vent plumes, suggesting some stabilisation of the reduced form of Fe, possibly through organic complexation (Statham et al. 2005). Evidence for stabilisation of dFe(II) in marine systems by organic ligands has been observed in estuarine waters (Hopwood et al. 2015), in previous shelf sediment incubation experiments (Homoky et al. 2012), and in bottom waters adjacent to the continental margin (Bundy et al. 2014). Thus, most likely, the complexation of Fe to organic ligands plays an important role in stabilising sediment-derived Fe delivered to the water column. Laboratory studies have shown that a range of simple organic molecules can impact Fe(II) oxidation rates, and whilst some had no effect, others directly or indirectly slowed the net oxidation rate (Santana-Casiano et al. 2000). It is also possible that inorganic complexes such as sulphides could stabilize Fe(II) in solution in the form of Fe sulphide nanoparticles (Yücel et al. 2011), but evidence for this in shelf systems has not been demonstrated.
The high residual dFe(II) concentrations at the end of incubation experiments show that something—most likely organic complexation—must routinely inhibit the oxidation of Fe(II) and maintain a fraction of dFe(II) in solution. Core top waters sampled throughout our study had consistently elevated dFe(II) and dFe concentrations (up to 14 and 21 nM, respectively, Table S4). The diffusion experiment in the late summer showed no significant increase or decrease in dFe(II) (0.8 ± 0.5 nM, n = 6) or dFe (2.8 ± 0.6 nM, n = 6) over a period of 6 days in core-top water (Figure S2). Therefore, elevated dFe(II) and dFe concentrations found in sampled core top waters must reflect an effectively stable form of Fe. Most importantly, dFe(II) concentrations in core top water were substantially higher than dFe concentrations in the overlying water column, indicating dFe(II) most likely originates from the sediments and provides a source of dFe to the water column even in the late summer.
Organic complexation is also able to keep Fe(III) in solution above solubility-controlled values, within the available ligand capacity. Evidence for Fe(III) ligand production in sediments has been provided for an estuarine system (Jones et al. 2011). As the Fe(II) in the proposed organic ligands is oxidised, it may remain associated with the ligand complex, converting to Fe(III)–L complexes, which may be much stronger than Fe(II)–L complexes. Alternatively, Fe(II) precipitation to Fe(III)-oxyhydroxide nanoparticles may constitute a colloidal fraction of dFe with or without organic complexes (e.g., Raiswell and Canfield 2012).
Modelling of Fe(II) fluxes from sediments to an oxic water column
Organic complexes may inhibit the oxidative precipitation of dFe(II), and could therefore increase the diffusive flux of dFe(II) across the oxic surface layer of shelf sediments to the overlying water column. We consider the impact of organic complexes using a 1-dimensional, steady-state, transport-reaction model to calculate diffusive fluxes of Fe(II) from porewater to the water column. Our approach follows Raiswell and Anderson (2005), which is used elsewhere to evaluate pore water fluxes of Fe(II) (Homoky et al. 2012, 2013; Wehrmann et al. 2014). To simulate the presence of organic ligands we simply use a fraction (f) between 0 and 1 of the Fe(II) oxidation rate constant (k), to calculate the diffusive fluxes of Fe(II) based on site A sediment characteristics (Fig. 9, see Supplementary Information).
In the absence of any Fe(II)-stabilising ligands (f
= 1) a diffusive flux of 24 µmol Fe(II) m−2 day−1 is calculated from Site A under late spring conditions, where OPD was 3.3 mm, pH 7.25 and near-surface Fe(II) concentration was 6.1 μM (averaged at 0.5 cm, n = 3). A smaller flux of 3.6 µmol m−2 day−1 is calculated for late summer conditions (OPD = 4.1 mm, pH 6.88, Fe(II) = 0.9 μM, n = 3). If Fe(II) oxidation was prevented in the surface sediment (f
→ 0) these fluxes would increase by 30 and 8% to 31 and 3.9 µmol m−2 day−1, respectively. Although we only have ionic diffusion coefficients available for our treatment of Fe(II)–Ligand complexes, Fe(II)–stabilizing ligands at Site A have a clear potential to impact diffusive fluxes. However, by stabilising dissolved species of Fe, their impact in the overlying water-column will likely be even more significant for benthic inputs. Diffusive fluxes of Fe(II) in the late spring period would provide up to 0.3 nmol l−1 day−1 throughout an evenly-mixed 110 m water column, compared to <0.04 nmol l−1 day−1 in the late summer period. Our theoretical approach considers only diffusive transport, yet transfer of Fe(II) from pore waters could be further enhanced by advective transport due to physical mixing in the water-column, bio-turbation and bio-irrigation or anthropogenic disturbance to surface sediments.
A previous study of benthic Fe cycling in depositional areas of the North Sea found no dFe flux from sediments to overlying water using a steady state reaction-diffusion model, but found 20–210 µmol m−2 day−1 when calculating diffusive dFe fluxes from measured porewater profiles modelling simple diffusion that ignored Fe(II) oxidation (Slomp et al. 1997). We can compare our diffusive Fe(II) fluxes from porewaters with those predicted by a recent global assessment of benthic Fe flux measurements from benthic chambers. Dale et al (2015) describes the dependence of benthic Fe flux on the rate of organic carbon oxidation in sediments and bottom water oxygen concentrations based on a compilation and regression of all known determinations. Where our organic C oxidation rates are calculated to be 8.6–11.9 mmol m−2 day−1 and bottom water oxygen is 267–252 µM, benthic Fe fluxes for our study site are estimated to be 8.0 µmol m−2 day−1 in late spring and 5.5 µmol m−2 day−1 in late summer—slightly higher than our late summer determination of 3.6–3.9 µmol m−2 day−1, but less than our late spring determination of 24–31 µmol m−2 day−1. To a first approximation, this is a favourable comparison, and it is not unreasonable that individual study sites will have benthic Fe fluxes that deviate from the averaged relationships described by Dale et al. (2015). However, it is also clear that well-oxygenated ocean margins have been largely absent from the compiled benthic Fe flux data used to parameterise ocean biogeochemical models to date (Homoky et al. 2016), hence there is potential for an underestimated contribution of dFe from oxic ocean margins.
Implications of benthic Fe(II) fluxes to an oxic water column
The dFe(II) stabilisation outlined above may enhance and maintain dFe(II) fluxes to the overlying water column. A study of the shelf and slope in the Bay of Biscay, south-west of the coast of Brittany found that labile dFe(II) species account for >8% of dFe species in bottom waters of the shelf break, and suggested that benthic processes (resuspension and diagenesis) represent important sources of dFe(II) and dFe, increasing the availability of Fe to microorganisms (Ussher et al. 2007). Elevated dFe(II) near the sea floor at Site A was also observed in July 2015 (Fig. 6) and represents ~4% of the dFe pool (Birchill, personal communication). The steep increase in dFe(II) concentrations towards the seafloor is consistent with our evaluation of a sedimentary source, although release of dFe(II) from the degradation of organic matter in the water column may also contribute to bottom water dFe(II) maxima.
Whilst there appears to be a low background diffusive input of Fe from the sediments throughout the year (Fig. 9; Figure S2), the deposition of C in biogenic debris from the upper water column after the spring bloom is a major driver for additional inputs of Fe to porewaters, as was previously hypothesised (Schoemann et al. 1998). The degradation and turn-over of this organic material at the seafloor appears to happen within a 2-month period. Following release of Fe in the late spring, sediments gradually reset to pre-bloom conditions. This view of the carbon flux to sediments being a major driver for iron release supports predictions from Elrod et al. (2004) and the revaluations by Dale et al. (2015). It is shown that the Fe flux is dependent on the position of the redoxcline within the sediment and the availability of organic material at the seafloor, with the stabilisation of dFe(II) by ligands acting as a further mechanism to enhance dFe transfer to the water column (Fig. 10). The findings discussed here apply to cohesive shelf sediments. However, sandy mud only covers 0.8% of the seafloor in the Celtic Sea. Sandy sediments, on the other hand, cover a large fraction of the seabed (16.5%, Thompson et al. submitted) and are also affected by seasonal inputs of organic matter, where organic Fe complexation could mediate benthic exchanges of Fe. This unconsolidated coarse sediment contains less organic carbon, but is much more permeable and so may host important advection-dominated exchange of dFe in shelf settings.
Based on our findings, we suggest that temperate shelf seas equivalent to the Celtic Sea need to be more explicitly represented in future ocean biogeochemical models, where ligand-mediated benthic exchanges of Fe occur in response to seasonal phytoplankton blooms. In such environments, Fe flux predictions based on previous benthic chamber studies (e.g., Dale et al. 2015) might underestimate the true magnitude of dissolved Fe input to the shelf seas. Although many sediment types across the Celtic Sea receive seasonal inputs of organic matter, benthic Fe fluxes in cohesive and non-cohesive sediments will be controlled by distinct diffusion-advection regimes for porewater solutes. Animal activity, waves, tides and human-induced disturbances of shelf sediments will all impact transport processes, but they are also ill quantified. An appraisal of exchange processes and rates across coarser and more permeable sediments will support a more rigorous scaling-up of our findings across the Celtic Sea, to quantify the impact of ligand-mediated benthic fluxes to oxic shelf seas. Such a result, once seasonal perturbations to benthic oxygen and to carbon dynamics and ligand-sustained fluxes of Fe are properly accounted for, is likely to reveal that a larger amount of Fe is released from oxic shelf sediments, which are estimated to typify most of the ocean-continent boundary, especially in large areas of the Atlantic, Arctic and Southern Oceans (Homoky et al. 2016), than previously assumed.