Coupled dynamics of iron, manganese, and phosphorus in brackish coastal sediments populated by cable bacteria

Abstract Coastal waters worldwide suffer from increased eutrophication and seasonal bottom water hypoxia. Here, we assess the dynamics of iron (Fe), manganese (Mn), and phosphorus (P) in sediments of the eutrophic, brackish Gulf of Finland populated by cable bacteria. At sites where bottom waters are oxic in spring, surface enrichments of Fe and Mn oxides and high abundances of cable bacteria were observed in sediments upon sampling in early summer. At one site, Fe and P were enriched in a thin layer (~ 3 mm) just below the sediment–water interface. X‐ray absorption near edge structure and micro X‐ray fluorescence analyses indicate that two‐thirds of the P in this layer was associated with poorly crystalline Fe oxides, with an additional contribution of Mn(II) phosphates. The Fe enriched layer was directly overlain by a Mn oxide‐rich surface layer (~ 2 mm). The Fe oxide layer was likely of diagenetic origin, formed through dissolution of Fe monosulfides and carbonates, potentially induced by cable bacteria in the preceding months when bottom waters were oxic. Most of the Mn oxides were likely deposited from the water column as part of a cycle of repeated deposition and remobilization. Further research is required to confirm whether cable bacteria activity in spring indeed promotes the formation of distinct layers enriched in Fe, Mn, and P minerals in Gulf of Finland sediments. The temporal variations in biogeochemical cycling in this seasonally hypoxic coastal system, potentially controlled by cable bacteria activity, have little impact on permanent sedimentary Fe, Mn, and P burial.

where J is the diffusive flux [mmol m -2 yr -1 ], ϕ represents the porosity, D s is the diffusion coefficient [m 2 yr -1 ], C is the concentration and z represents the sediment depth [m]. The diffusion coefficient was calculated as a function of ambient tortuosity, pressure, salinity and temperature using the R package marelac (Soetaert et al. 2010), which implements the constitutive relations described in Boudreau (1997). The concentration gradient between the bottom water and the first sediment depth interval was used for the calculations.
where, in this case, P total represents the average concentration of total P (mol g -1 ) in the 20 to 30 cm sediment depth interval, SR is the sedimentation rate (cm yr -1 ), ϕ is the porosity within the same depth interval and ρ represents the density of dry sediment 2.65 g cm -3 (Burdige 2006). Total Fe and Mn burial were calculated in a similar way.

Iron, Manganese and Phosphorus Burial Calculations
The diffusive fluxes of dissolved NH 4 + , HPO 4 2-, Fe and Mn were calculated using Fick's first law as described in Berner (1980): The total Fe, Mn and P burial rates in mmol m -2 yr -1 were calculated as a function of the sedimentation rate, total Fe, Mn and P in the deeper sediment and the porosity using the following equation:

Manganese XAS Analysis of Suspended Matter Samples
Suspended matter samples from the water column from sites GOF5 and LL3A, were analyzed for their Mn content and mineralogy. High-resolution µXRF maps (1.6-2.5 mm 2 ) were analyzed using a horizontal and vertical step size of 0.5 to 0.75 µm, respectively. Mn XANES spectra were retrieved at 12 to 14 spots that were enriched in Mn. At each spot only one Mn XANES spectrum was collected to avoid photo-induced reduction of Mn during X-ray analyses. Mn XANES spectra, were subjected to component analysis using the Iterative Transformation Factor Analysis (ITFA) software package (Rossberg et al. 2003).
The indicator function from the component analysis revealed that three components are the optimal number for reproducing the Mn XANES spectra of suspended material at the three locations. Spectra from three spots were identified based on the Varimax analysis, each having a high loading on one of the components and only a minor loading on the others. None of the three Mn XANES spectra showed a close resemblance with one of the spectra from reference materials including several Mn oxides, Mn carbonates, Mn phosphates and Mn sulphides. The position of the absorption maximum of all spectra was close to that of Mn in birnessite. The spectra differ, in particular, regarding the shape and position of the edge which is shifted to lower energies compared to birnessite. This indicates that the fraction of Mn with a lower oxidation state than IV (e.g. Mn(II) or Mn(III)) at the analyzed spots is higher than that of Mn in birnessite. Birnessite contains predominately Mn(IV) but can also contain Mn(III) which leads to an excess negative charge in the layers which is compensated by accommodating cations in the interlayer. 3

Iron Enrichment Classification
Spots in the synchotron-based µXRF map were classified into three categories based on their total Fe content as reflected in the relative count intensities. (Table S1).
Based on the results from the component analysis, the maximum number of spectra for LCF was constrained to three. For LCF, the normalized Mn XANES spectra in the range 6530 to 6580 eV was used. When searching the best combination of three Mn XANES spectra out of a set of spectra from various materials for reproducing the three spectra by LCF, consistently birnessite was selected in combination with hausmannite (Mn 3 O 4 ), Mn(II) phosphates or dissolved Mn 2+ . The contribution of aqueous Mn 2+ to one of the spectra was only very minor. Hence, the Mn XANES spectra can be well reproduced when combining the spectra of birnessite, hausmannite and Mn(II)phosphate. The R-factor (sum((fit-data) 2 )/sum(data 2 )) was on average 0.00167 and always below 0.005.
The quality of the spectra was insufficient for investigating their extended X-ray absorption fine structure (EXAFS). Hence, confirmation of the results from LCF of XANES spectra by investigating the EXAFS was not possible. Furthermore, due to the relatively short energy range, the results of LCF are strongly influenced by the position and shape of the edge. Consequently, the assigned fractions of hausmannite and Mn(II) phosphates might not necessarily reflect the concentrations of these two phases but, instead, account for the presence of Mn(II) and Mn(III) including, for example, also adsorbed Mn(II) or birnessite with a higher Mn(III) content compared to the reference material in the LCF.

High-resolution Maps of Mn in Suspended Matter in the Water Column
At sites GOF5 and LL3A, Mn XANES spectra of suspended matter in the water column were collected from 14 and 12 spots, respectively (Fig. S3).
6 Figure S3. High-resolution synchrotron-based µXRF maps of suspended matter in the water column at sites GOF5 (60 m water depth) and LL3A (58 m water depth). The colors accentuate the relative count intensities adjusted for brightness and contrast to highlight the Mn enrichments. The open black diamonds indicate the Mn enrichments that were further subjected to Mn XANES analysis to identify their mineralogical composition.     , Fe and Mn in mmol m -2 d -1 determined from the pore water concentrations near the sediment-water interface (PW) and the benthic lander incubations (Lander).

Sedimentation Rates
Sedimentation rates were estimated by fitting a reactive transport model (Soetaert and Herman, 2008) to the 210 Pb depth profiles (Fig. S7) taking the depth dependent changes in porosity into account (Fig. S8). 10 Figure S8. Porosity depth profiles for sites JML, GOF5 and LL3A in vol vol -1 .

Porosity
The porosity (Fig. S8) was calculated from the water loss upon freeze-drying and sediment density following Burdige (2006

Fe XANES Spectra of Resin-embedded Sediment from Site GOF5
In contrast to Mn, the ratio between the reactive and unreactive fraction of Fe (i.e. Fe oxide and Fe sulfide versus clay Fe) was generally low in the sediments, even in enriched layers. As the size of the Fe enrichments was also smaller than the depth of the volume probed by XAS, the obtained Fe spectra also entail a contribution from non-reactive, silicate-bound Fe. In order to investigate the nature of Fe enrichments an attempt was made to isolate the contribution from non-reactive Fe. Based on the relative count intensities derived from synchrotron-based mapping of Fe in the surface sediment at site GOF5, spots containing very little Fe were selected, which are assumed to represent 'background Fe' (Fig. 9B,C). Subsequently, the XANES spectra of background Fe were averaged and compared with the XANES spectra of various Fe mineral reference materials (Fig. S12). Our findings indicate that background Fe did not resemble the XANES spectra of Fe oxides, Fe carbonates, Fe phosphates and Fe sulfides. However, the XANES and EXAFS spectrum of background Fe could be best reproduced by combining the XANES spectra of biotite (an Fe phyllosilicate mineral) and illite (Fe associated with clay) which supports the interpretation that the spectrum reflects silicate-bound Fe (Fig. S12). The isolated XANES spectrum from 'background Fe' was then used in combination with spectra of reactive Fe phases in the LCF to reproduce the XANES and EXAFS spectra collected at the various spots. For the XANES spectra the energy range between about 7110 to 7190 eV was used and for the k 2 weighted EXAFS spectra the k-range between 2 and 8 Å -1 was used in the LCF. Within the set of spectra of reference materials, the combination with the spectra of 6L-ferrihydrite gave the best and consistent results when applied to the XANES and EXAFS spectra. In general, the approach for analyzing the Mn XANES spectra from sediments was similar to that used for suspended matter. In total 32 Mn XANES spectra collected from resin-embedded sediments of site GOF5 were used for the component analysis. The indicator function had a minimum at four components suggesting that the optimum number of endmember spectra for the reproduction of the sample spectra is four. Subsequently, the data set was extended by adding XANES spectra from reference materials to explore the effect on the indicator function and varimax rotation. All attempts to find a set of four spectra of reference materials without the need to increase the number of eigenvectors for maintaining the quality of reproductions did not succeed. This suggests that endmember spectra, which could be assigned to the components, might contain signals from more than one Mn phase. Another possibility is that one or more Mn phases that were present in the sediment are not included in the set of reference materials.
Based on these results, an approach was taken to include more than four spectra from reference materials in the LCF, in particular several spectra of solids containing Mn(II) and Mn(III). After testing several combinations, a good reproduction of all XANES spectra with a R-factor of 0.001 was achieved by using the spectra of the six reference materials: birnessite, manganite, bixbyite, hausmannite, Mn(II) phosphate, and rhodochrosite). However, due to the limitation of unambiguous identification of Mn phases based on Mn XANES spectra only, the results of the LCF were solely used to constrain the fractions of Mn(II) containing phosphates and carbonates, from Mn-containing oxides.     Figure S16. Photograph of the surface sediment at site GOF5, which is characterized by a distinct color zonation. The scale bar denotes a total distance of 2 cm, with 0.5 cm intervals. In contrast to sites GOF5 and LL3A, the surface sediment at site JML was not characterized by an enrichment of total Fe and Mn near the sediment-water interface (Fig. S17). However, total Fe, Mn, P and S do exibit vertical laminations, which is likely a consequence of the contrasting bottom water redox conditions over time.   Table S4. Calculation of the potential change in bottom water P based on the Fe-bound P content of the surface sediment at site GOF5. The increase in HPO 4 2of 7.4 μM was calculated by assuming release of all metal bound P from the upper 2 cm of the sediment at GOF5, to the lower 20 m of the water column, while accounting for the measured porosity in each sediment layer, assuming a sediment density of 2.65 g cm -3 . The total potential P release amounted to 150,000 μmol m -2 . When accounting for a volume of 20 x 1000 liters, this leads to a phosphate concentration of 7.4 μM. The actual change in the water column concentration (~4 μM) was calculated from the minimum and maximum P concentration for GOF5 as given in Fig. S19.