Potentially bioavailable iron produced through benthic cycling in glaciated Arctic fjords of Svalbard

The Arctic has the highest warming rates on Earth. Glaciated fjord ecosystems, which are hotspots of carbon cycling and burial, are extremely sensitive to this warming. Glaciers are important for the transport of iron from land to sea and supply this essential nutrient to phytoplankton in high-latitude marine ecosystems. However, up to 95% of the glacially-sourced iron settles to sediments close to the glacial source. Our data show that while 0.6–12% of the total glacially-sourced iron is potentially bioavailable, biogeochemical cycling in Arctic fjord sediments converts the glacially-derived iron into more labile phases, generating up to a 9-fold increase in the amount of potentially bioavailable iron. Arctic fjord sediments are thus an important source of potentially bioavailable iron. However, our data suggests that as glaciers retreat onto land the flux of iron to the sediment-water interface may be reduced. Glacial retreat therefore likely impacts iron cycling in coastal marine ecosystems.

ordered mineral phase at 77 K as was observed in all other samples. The main components in samples from LF are considered to be paramagnetic Fe(II) and Fe(III) mineral phases.
The difference in the iron mineral composition between the individual samples was also recognizable in the streak color of the individual sample material (Supplementary Figure 5). While the two source (plume & iceberg) samples and samples from KF showed a very dominant red/brownish color, samples from LF do not show an intensive red streak color. This is supports the aforementioned interpretation that the higher crystalline and magnetically-ordered phase (which is likely to be hematite) is only present in samples KF1 0-1 cm, KFa7 0-1 cm, Plume KB a and IB KB b, while samples LF1 0-1 cm and LF5 0-1 cm only contain a composition of paramagnetic (and potentially poorly crystalline) Fe(II) and Fe(III) phases. Paramagnetic relaxation was also detected in the 5 K spectra of samples LF1 0-1 cm and LF5 0-1 cm.

Uncertainty in mineral identities identified by
Furthermore, the absence of a wide sextet with a hyperfine field >50 confirms that hematite was not present in samples from LF sediment. However, the similarity of the 5 K spectra and the same fitting parameters suggest that samples LF1 0-1 cm and LF5 0-1 cm are very similar in terms of their iron mineral composition, with the exception of hematite

Grain size analysis and interpretation
Over 95% of the grain size distributions from surface sediment samples recovered along three transects are characterized by silt and clay (< 63 m or 4 ) (Supplementary Figure 6). Through dynamic processes relating to sediment erosion, deposition and aggregate breakup, silt material 63-10 m (4 -6.6 ) will undergo sorting in response to hydrodynamic processes, whereas, fine silt size particles < 10 m (6.6 ) behave in a cohesive manor in the same way as clay particles (< 2 m or > 9.0 ). Thus, sorting of silt maybe observed through current winnowing or as the distance from areas of higher energy (i.e., rivers and glacial meltwater plumes) increases with the finer grain component accumulating farther away. Detailed grain size analysis of the lithogenic component along three transects demonstrate no systematic relationship between the percent fine grain material (< 63 m or > 4  ) and the distance from glacier source (Supplementary Figure 6). While there are clear differences between grain size distributions from source material, surface sediment from stations do not systematically get finer grained in texture as distance increases. For example, across three transects, sites farthest from the glacier sources nearly overlap and appear indistinguishable from or contain a slightly higher percent (at most ~2-1 %) of coarser silt material, than adjacent sites located closer to the glacier. Thus, the increase of FeR over distance from the glacier cannot be explained as a function of the transport of the fine grain particles.

Changes of SRR over distance in relation to changes TOC and C:N over distance from the glacial source
The increasing amount of TOC and decreasing C:N ratios with greater distance from the fjord heads in all three investigated fjords (Figure 6, Supplementary Table 7), indicate the presence of more fresh and labile organic carbon further out in the fjord. We therefore expected that organic carbon mineralization in the fjord sediment, and thereby also SRR, would follow this pattern and increase with increasing distance from the fjord head. However, in all four investigated fjord transects, SRR at the station furthest away from the head of the fjord were not much higher compared to the station closest to the fjord head ( Figure 6 and S8). In KFa and KFb, SRR peaked mid-fjord but then also deceased to a value close to the starting value again. This pattern indicates a change from a predominance of SR close to the fjord head, where there is only little FeA and FeM, towards a higher proportion of Fe reduction further away from the fjord head. This is likely driven by the higher content  Table 8 and S9).
In Kongsfjorden, highest SRR were measured at station KFb3 (up to 330 nmol cm -3 d -1 ; Supplementary Figure 8), also integrated rates of SRR were highest ( Figure 6) and high TOC values were measured. The high SRR are probably connected to the close by bird cliff, where a colony of kittiwakes are nesting 13,14 . The large bird colony contributes to a higher primary productivity on land around the colony, likely introducing elevated amounts of nutrients and organic carbon to the close by marine environment.
At KF1, KFa3, KFa5, KFb3 SRR peaked directly at the sediment surface ( Supplementary Figure 8). At stations further away from the Kongsfjorden head, the peak in SRR was found deeper down in the sediment. Depthintegrated rates of SR at first increased with increasing distance from the fjord head but then dropped again ( Figure 6). In Lilliehöökfjorden maximum SRR were 80 nmol cm -3 d -1 and no surface peaks were found, but SRR reached a maximum between 3 and 8 cm sediment depth (Supplementary Figure 8). In Dicksonfjorden, the maximum SRR was 6 nmol cm -3 d -1 and no substantial change of depth-integrated SRR was found over distance from the fjord head ( Figure 6).

Supplementary Figures
Supplementary Figure 1 -Mössbauer spectra collected at 77 K. All samples showed a wide doublet (Db1; light blue) that can be attributed to the presence of a high-spin Fe(II) mineral phase in all samples. The narrow doublet (Db2; orange) in all samples can be attributed to the presence of a non-magnetically ordered poorly crystalline Fe(III) mineral phase similar to ferrihydrite. Samples a)-d) showed an additional sextet feature (red) that shows the presence of a magnetically-ordered mineral phase at 77K. The wide hyperfine field (>50 T) suggests hematite as iron(III) oxide likely to be present in these samples.  Supplementary Table 9.

Supplementary Figure 4: Amount (M(0)), reducibility (apparent rate constant, v/a), lability (initial rate), and composition (1+1/v) of FeA and FeM over distance from the glacial source.
Glacial sources are shown in the grey shaded areas: light grey=samples of proglacial plumes and meltwater rivers at the head of the fjord, medium grey= icebergs, dark grey = meltwater rivers at the side of the fjord. Figure 6: Results from particle size analysis of surface sediment and glacial sources. a shows the mode of particle size distribution (PSD) over distance from fjord head for the sediment samples from all three fjords as well as the sources (grey area). b shows mean Φ value of the PSD over distance for the sediment samples from all three fjords as well as the sources (grey area). c-k show the distribution of Φ as cumulative percent in the surface sediments of all three fjords for the entire PSD (2-11 Φ), silt material that undergoes size sorting (4 -6.6 Φ), and fine silt and clay that behaves in a cohesive manor (>6.6 Φ).