The Stability of Fe-Isotope Signatures During Low Salinity Mixing in Subarctic Estuaries

We have studied iron (Fe)-isotope signals in particles (> 0.22 µm) and the dissolved phase (< 0.22 µm) in two subarctic, boreal rivers, their estuaries and the adjacent sea in northern Sweden. Both rivers, the Råne and the Kalix, are enriched in Fe and organic carbon (up to 29 µmol/L and up to 730 µmol/L, respectively). Observed changes in the particulate and dissolved phase during spring flood in May suggest different sources of Fe to the rivers during different seasons. While particles show a positive Fe-isotope signal during winter, during spring flood, the values are negative. Increased discharge due to snowmelt in the boreal region is most times accompanied by flushing of the organic-rich sub-surface layers. These upper podzol soil layers have been shown to be a source for Fe-organic carbon aggregates with a negative Fe-isotope signal. During winter, the rivers are mostly fed by deep groundwater, where Fe occurs as Fe(oxy)hydroxides, with a positive Fe-isotope signal. Flocculation during initial estuarine mixing does not change the Fe-isotope compositions of the two phases. Data indicate that the two groups of Fe aggregates flocculate diversely in the estuaries due to differences in their surface structure. Within the open sea, the particulate phase showed heavier δ56Fe values than in the estuaries. Our data indicate the flocculation of the negative Fe-isotope signal in a low salinity environment, due to changes in the ionic strength and further the increase of pH.


Introduction
Despite its high abundance in the continental crust (Wedepohl 1995), Fe can be the limiting element in a third of the ocean surface waters (Martin et al. 1991;Morel and Price 2003). Riverine Fe input is the main source of Fe to the oceanic Fe budget, and estuarine mixing controls how much Fe reaches the oceans (Wells et al. 1995;Raiswell and Canfield 2012). Particles and colloids in river water (ranging from 0.01 to 1 µm) flocculate into larger particles and aggregates due to the increased ionic strength within estuaries and then precipitate close to the shoreline (Boyle et al. 1977;Sholkovitz et al. 1978). These Fe particles and colloids consist mainly of two forms, Fe-rich organic carbon (OC) compounds and Fe-rich (oxy)hydroxides (Ingri et al. 2006;Ilina et al. 2013;Kritzberg et al. 2014). The surface properties of these compounds vary as aggregates with OC have fewer mineral phases exposed at their surface (Eusterhues et al. 2008). The amount of mineral phases, in turn, affects their chemical reactivity during estuarine mixing, which for example is defined by size and speciation (Poulton and Raiswell 2005;Tagliabue et al. 2017). Hence, the river input of Fe into the ocean is substantially modified by flocculation processes that occur at the estuarine interface and dependent on the initial Fe phase (Eckert and Sholkovitz 1976;Boyle et al. 1977;Sholkovitz 1978). Various studies showed the ability of OC to keep Fe in solution along the salinity gradient of an estuary (Krachler et al. 2010;Kritzberg et al. 2014). The aggregation, sedimentation and resulting burial of OC associated with Fe in estuarine environments have been described as the "rusty carbon sink" (Lalonde et al. 2012;Shields et al. 2016).
Earlier studies have addressed the possibility of using Fe-isotopes as a tool to identify water sources and to differentiate between Fe phases, i.e. Fe-OC complexes and Fe(oxy) hydroxides (Ingri et al. 2006;Escoube et al. 2009;Ilina et al. 2013;Poitrasson et al. 2014). In low-temperature environments, stable Fe-isotope mass varies about 6.0-8.0‰ in 56 Fe/ 54 Fe ratios (Dauphas et al. 2017;Wu et al. 2018), expressed as δ 56 Fe relative to the reference material IRMM-14 (Dauphas and Rouxel 2006;Rouxel and Auro 2010). The Feisotope values of Fe phases entering the ocean via rivers range between −1.2 and +1.8‰ (Conrad 2019 and references therein), and this variation has been used as a tool to map the contribution of Fe sources to the ocean on temporal and geographical scales (Conway and John 2015). For example, subarctic and temperate rivers supply Fe phases with a negative δ 56 Fe value to seawater (Ingri et al. 2006;Severmann et al. 2006;Escoube et al. 2009;Ilina et al. 2013). It has been shown that Fe-isotope fractionation strongly depends on redox reactions (Wiederhold et al. 2006) and the fractionation can be used to trace the Fe-isotope composition of different Fe phases at their origin (Ingri et al. 2006;Ilina et al. 2013;Dos Santos Pinheiro et al. 2014;Poitrasson et al. 2014). In arctic and subarctic rivers, filtration yields towards a separation of a heavy dissolved (δ 56 Fe of +0.43‰ ± 0.04‰) and a particulate δ 56 Fe between −0.09 and +0.10‰ (Escoube et al. 2009).
In spite of their importance, the knowledge about the fate of different Fe aggregates during estuarine transport is small. The Fe-isotope composition of the Fe-OC aggregates has the potential to be used as a tracer for the cycling of Fe-OC aggregates during transport from land to the open ocean.
Over the last decades, Fe concentrations in boreal rivers draining into the Baltic Sea have been increasing continuously (Kritzberg and Ekström 2012;Sarkkola et al. 2013). Changing redox conditions in organic-rich soils are most likely responsible for the elevated Fe concentrations (Kritzberg and Ekström 2012;Knorr 2013). Higher Fe concentrations lead to increasing coastal export and may have important implications for the receiving aquatic systems.
In this study, we combine the Fe-isotope data from the river mouth, estuary and open sea in a subarctic, boreal environment, to access which Fe phase is present at the varying locations. We show the seasonal variation of dissolved and particulate Fe-isotope ratios in two boreal rivers and their estuaries. Our study provides further information on the seasonal availability of different Fe complexes by the use of their Fe-isotope composition. The 1 3 hypothesis was, can the Fe-isotope ratio be used to trace different Fe complexes from the river mouths into the open ocean. Therefore, we address the following research questions: (1) quantify the variation of Fe-isotope values at the river mouths and along their estuaries; (2) identify processes leading to temporal changes; and (3) reveal the effect of salinityinduced mixing on the Fe-isotope ratio.

Sampling Site
For this study, water samples were collected in two rivers, their estuaries, and the adjacent sea in Northern Sweden (Fig. 1). The boreal rivers Kalix and Råne have their origin in the Caledonian Mountains and the Råne Träsk, respectively. Both river catchments consist of forest, wetlands and lakes (Online resources ESM_1; SMHI 2018). The main difference is the size of their catchment with the Kalix's being approximately seven times larger than the Råne's catchment. Additionally, the catchment of the Kalix River contains about 11% mountainous area. Both rivers drain into the northernmost part of the Baltic Sea, the so-called Bothnian Bay (Fig. 1b). The Råne estuary is a semi-closed water-body, partly isolated by several islands. The Kalix estuary is more open to the Bothnian Bay and is a so-called salt wedge estuary (Holliday and Liss 1976), where freshwater floats on top of the denser seawater, with the freshwater layer getting thinner towards the sea. The Bothnian Bay and its catchment area (including the Kalix and Råne River) are ice-covered up to 6 months per year (Vihma and Haapala 2009  and Råne (R1-4) drain into the Bothnian Bay (GOB1, GOB2, A5 and A13). Each station was sampled at three depths (0.5 m, 5 m and 10 m) if station depths allowed it. The legend on the right side of the maps shows metre above and below sea level (masl and mbsl, respectively). The Ocean Data View program, version 4.7.10 was used to generate the maps (Schlitzer, R., Ocean Data View, http://odv.awi. de, 2018) and the high freshwater input and the low tide-influence lead to salinities below 4 (Kautsky and Kautsky 2000).
The boreal region has a subarctic climate, with extreme seasonal temperature variations, with long, cold winters (− 20 to − 40 °C) and short, mild summers (20-30 °C), (Peel et al. 2007). Between 40 and 50% of the precipitation is snow (Ingri et al. 2005). With 5-7 months of winter, the moisture in the soils and subsoils freezes between 25 and 79 cm deep (Öquist and Laudon 2008;Haei et al. 2010). The river discharge has significant seasonal variations, and the main hydrological event is occurring in spring due to snow and ice melt (Fig. 2). There is an extended period of high discharge in the Kalix River, compared to the Råne River, resulting from snow and ice melt in mountainous regions (Fig. 2). Discharge for both rivers was measured about 50 and 25 km upstream, respectively, at the stations Räktfors (Kalix River) and Niemisel (Råne River) by the Swedish Meteorological and Hydrological Institute (SMHI 2018).
For this study, the rivers, their estuaries and the Bothnian Bay were sampled during pre-flood, spring flood and post-flood (Table 1)

Sampling and sample processing
All sampling stations were sampled at depths of 0.5, 5 and 10 m if they were deep enough (all except station K1 and R1). Water samples from the Kalix estuary, the Råne estuary and the Bothnian Bay were collected from the coast guard vessel KBV005 in May and June 2013. Four stations in each estuary (R1-R4 and K1-K4) and two stations in the Bothnian Bay (A5 and A13) were sampled. During the sampling in March 2014, samples were  Bauer et al. (2018). Briefly, the samples in 2013 were collected from the coast guard vessel KBV005 with 2 × 2 L Niskin bottles (Hydros-Bios, Kiel, Germany which had been custommade trace metal clean containing only Ti screws, on a Dyneema rope, acid-cleaned 2-L LDPE bottles all parts acid cleaned) on a line in depths of 0.5, 5 and 10 m. In 2014, the samples were conducted through a hole in the ice, drilled with an ice auger and pumped up from the depths of 0.5, 5 and 10 m. During July 2014, the Råne estuary (R2) and two stations in the Bothnian Bay (GOB1 and GOB2) were sampled from the M/S Fyrbyggaren following the procedures and methods by (Cutter et al. 2010). Water samples were obtained with 5-L Niskin bottles (Hydros-Bios, Kiel, Germany) on a line in depths of 0.5, 5 and 10 m. Once retrieved, the bottles were emptied into sample acid-cleaned LDPE bottles, wrapped in polyethylene bags. Salinity, temperature, pH, alkalinity and dissolved oxygen were monitored with a Hydrolab Multisonde 5 (Online resources ESM_1). The sampling stations in the Kalix River, Kamlunge, and in the Råne River, Orrbyn, were partly icecovered until early May 2016. Samples were collected from a bridge in the middle of each river. Specific conductivity, temperature, pH and dissolved oxygen were monitored with a Hydrolab Multisonde 5 (Online resources ESM_1).
All materials in contact with samples consisted of LDPE and were acid cleaned (1.1 M HNO 3 ; Merck Millipore pro analysis EMSURE), before and between samplings (Ödman et al. 1999). All water samples were filtered within 6 h through 0.22 -µm pore  Råne Estuary KR4e-KR9e KR10e-KR13e size membrane filters (Merck MF-Millipore Membrane Filter, mixed cellulose esters, hydrophilic, diameter 142 mm), which were locked in polycarbonate filter holders under a clean bench with a peristaltic pump, yielding into a particulate (PFe > 0.22 µm) and a dissolved (DFe < 0.22 µm) fraction. Membrane filters were cleaned with ultrapure acetic acid (99-100%) and stored in Milli-Q ultrapure water (18.2 MΩ cm at 25 °C) before they were used. The filtered water was acidified with ultrapure distilled 10% HNO 3 to a pH below 2 and stored at 4 °C. The filters were stored in petri dishes at −18 °C until analyses. Furthermore, water samples from 2016 were filtered through 0.7-µm glass fibre filters (GF/F Whatman ® ) for dissolved and particulate organic carbon analyses (DOC and POC). The filters were pre-combusted for four hours to limit the C blanks (Brodie et al. 2011). The filters for POC analyses were stored at − 18 °C before analyses by UC Davis Stable Isotope Facility (USA).

Analytical Methods
All samples were analysed for elemental composition and Fe-isotope composition in collaboration with ALS (Australian Laboratory Services) Scandinavia AB, Luleå, Sweden. All sample manipulations were performed in a clean laboratory (Class 10,000) by personnel wearing clean room gear and following all general precautions to reduce contamination (Rodushkin et al. 2010). High-purity Suprapure ® acids were used throughout sample treatment and analyses. Accredited DOC analyses were determined at the Umeå Marine Sciences Centre (Samples 2013 and 2014) and ALS Scandinavia (Samples 2016).

Element Concentration by Inductively Coupled Plasma Sector Field Mass Spectrometry (ICP-SFMS)
For the elemental composition, the dissolved samples were diluted (2-200 fold) in 2.2 M HNO 3 ; the degree of dilution depended on the salinity of the sample. The filters were treated with 10 mL of a 1000:1 mixture of HNO 3 /HF overnight followed by closed-vessel digestion in a microwave oven (600 W, 1 h). An aliquot of the digests was further diluted in 2.2.M HNO 3 by the factors of 5 and 50 (2016 and 2013/2014, respectively) for the determination of Fe concentrations. Multi-elemental analyses (i.a. Fe, Rb, Sr) were performed in the dissolved and particulate phases of the samples by ICP-SFMS (ELEMENT XR, Thermo Scientific, Bremen, Germany). A combination of internal standardization (indium added at 2 µgL −1 to all measurement solutions) and external calibration verified the measurements. All measured data fall within a 4-point calibration curve. Details of the analytical procedure as well as instrument parameters and analysis conditions can be found elsewhere (Rodushkin and Ruth 1997;Rodushkin et al. 2005). The analytical procedure was validated with SLRS-4 River Water CRM for Trace Metals, SLEW-2 Estuarine Water CRM for Trace Metals and NASS-4 open ocean water (supplied by National Research Council, Ottawa, Canada), (Rodushkin et al. 2005(Rodushkin et al. , 2016. Fe was analysed on three different instrument runs with an average detection limit (LOD) of 14 nM for the dissolved samples and 21 nM for the particulate samples taken in 2013 and 2014 (LOD = X bl + 3SD bl , X bl = mean concentration of Fe in blanks; SD bl = Standard deviation of blanks). Replicated measurements showed a precision of ± 1.3% (n = 8) for the particulate phase and ± 3.0% (n = 3) for the dissolved phase.    The average detection limit for both phases in 2016 was 9 nM and replicated measurements showed a precision of ± 4.5% (n = 41).

Iron Isotope Ratio Measurements by MC-ICP-MS
For the Fe-isotope measurements, an aliquot of the dissolved samples and the digested filters was evaporated. Depending on the Fe concentration, 50 and 1000 ml of sample solution was evaporated. The residuals were re-dissolved in 1 mL 8 M HCl. Iron was separated from the matrix elements by using an AG-MP-1 M ion-exchange resin (microporous, 100-200 mm dry mesh size, 75-150 mm wet bead size, Bio-Rad Laboratories AB, Solna, Sweden). After the sample was loaded, the matrix was washed with 9.6 M HCl, and Cu was eluted with 8 ml 5 M Cl. Afterwards, Fe was eluted with 6 ml 2 M HCl and could be used for further steps (Rodushkin et al. 2016). After evaporating to dryness, 50 µL of concentrated HNO 3 was pipetted directly to the residue followed by the addition of 5 mL MQwater. Samples with high Fe content were diluted with 0.2 M HNO 3 to a concentration of 2 mg L −1 in the measurement solutions. Low Fe concentration water samples were further diluted to 0.7-0.9 µmol/L and measured using high-efficiency desolvation nebulizer (Aridus) in a separate analytical sequence. Iron was separated from the matrix by this ion exchange with a yield of > 95%. Fe-isotope ratio measurements were performed on a multi-collector inductively coupled plasma mass spectrometer (MC-ICP-MS, NEPTUNE and NEPTUNE PLUS) in high-resolution static mode relative to IRMM-14 CRM. The 54 (Fe and Cr), 56 Fe, 57 Fe, 58 (Fe and Ni), 60 Ni and 62 Ni were collected by the eight adjustable Faraday cups of the NEPTUNE. The instruments were equipped with a micro-concentric nebulizer and tandem cyclonic/Scott double-pass spray chamber. Instrumental mass biases were corrected by sample-standard bracketing using IRMM-14 CRM, while an internal standard (Ni) was added to all samples and used to correct for instrumental drift. The online data correction considered baseline subtraction (120 s before each measurement), calculation of ion beam intensity ratios and filtering of outliers by a 2σ test. Detailed information on the correction procedures can be found in Baxter et al. (2006). Iron isotope data are expressed as δ 56 Fe, relative to the IRMM-14 standard.
In-house quality control samples (prepared by sequential dilutions of SPECTROSCAN 10,000 mg L −1 Fe element standard for atomic spectroscopy from TEKNOLAB, Drøbak, Norway) were analysed at the beginning and the end of each analytical session to ensure internal consistency of the analytical results. Each analysis was made as a sequence of standard, three samples, standard. All samples and standards were analysed in duplicates, and the internal analytical precision was better than ± 0.01% [± 2 standard deviations (SD)].
Fe-isotope ratios in this material were measured on a regular basis at ALS laboratory from 2003, and a mean δ 56 Fe of − 0.24 ± 0.03 ‰ (n > 120, one sigma) makes it a suitable reproducibility control for Fe-isotope ratio measurements in low fractionated samples . Within the analysed samples, the δ 56 Fe values were reproduced with a standard deviation of 8.03% for the particulate phase and 2.50% for the dissolved phase in 2013/2014. In 2016, iron isotope data in both phases were reproduced with a precision of 2.7%. In the three-isotope plot of δ 56 Fe and δ 57 Fe, all samples plot on a single-mass fractionation line similar to the theoretical kinetic fractionation line (Young et al. 2002;Kavner et al. 2005), which shows the robustness of the measurements (Online resources ESM_2). In this study, we only discuss the δ 56 Fe, δ 57 Fe data which are reported in the online resources (ESM_1) including their errors (2SD).

Spring Flood
The peak discharge in the Kalix River was about 3-4 times higher than in the Råne River (Fig. 2). The pH in the Kalix and the Råne River varied between 5.6 and 6.9 with no clear correlation to the discharge (ESM_2, Fig. S2). Within the estuaries, the pH was increasing with distance to the shoreline in both estuaries and showed little variation with depths (from 6.5 to 7.4). pH in the Bothnian Bay showed little variation between 7.7 and 7.9 (ESM_2, Figs. S3 to S5).
The salinity generally increased with distance to the shore and with depths (ESM_2, Fig. S3 to S5). The salinity at the northernmost station (A5) was about 0.5 points lower than at station A13 during late spring flood (2.5 and 2.9, respectively).
The DOC and POC concentrations in both rivers increased with increasing discharge (ESM_2, Fig. S2). DOC concentration ranged from 500 to 740 µmol/L, while POC ranged from 71 to 134 µmol/L. The high DOC concentrations were also found in the estuaries. Overall, the DOC concentration decreased with distance to the shore, and the variations δ 56 Fe(‰) = 56 Fe∕ 56 Fe sample 56 Fe∕ 56 Fe IRMM-14 − 1 * 10 3 with depths were negligible (411-718 µmol/L) (ESM_2, Fg S3 and S4). In the Kalix estuary, during spring flood, the DOC concentration increased with distance to the shore (K1-K3) and then decreased towards station K4. During late spring flood, in the Kalix estuary, there is an overall increase towards the sea. In the Råne estuary, the DOC concentration was decreasing from land to sea (ESM_2, Fig. S4). The DOC concentration in the Bothnian Bay was homogenous between 348 and 396 µmol/L. Station A5 had the highest DOC concentrations within the Bothnian Bay during late spring flood (ESM_2, Fig. S5). Both rivers showed a large increase in DFe and PFe concentration during the spring flood ( Fig. 3; Table 2). The main increase in DFe occurred before the maximum discharge of the season, increasing from 5 and 14 µmol/L. Dissolved Fe concentrations measured in the Råne River were slightly higher than in the Kalix River. Particulate Fe increased later,  Table 3). At most stations, the DFe concentration also decreased with depths as salinity increased. During spring flood, DFe in the Kalix estuary varied between 18.7 and 1.4 µmol/L. In the Råne estuary, the DFe concentration varied between 1.7 and 16.3 µmol/L (Fig. 5) with the lowest concentrations at depths of 10 m, where the salinities were comparably high. During late spring flood, the DFe concentration was slightly lower, but the decrease in DFe with depths was less distinct (6.4-14 µmol/L). Within the Bothnian Bay at salinities around 3 g/kg, the DFe concentration ranged from 0.1 to 0.8 µmol/L with little variation in depth (Fig. 6). The highest concentrations occurred at station A5 in June. Along both estuaries, the depths profiles suggest decreasing PFe concentrations from land to sea, but the trend was not as clear as in the dissolved fraction. In general, the PFe concentration decreased with depths. In the Kalix estuary, the PFe concentration ranged from 0.4 to 5.8 µmol/L during the spring flood. In the Råne estuary, the measured PFe concentration ranged from 0.4 to 6.5 µmol/L, whereas station R1 showed exceptional high PFe concentrations of 15.5 µmol/L. Particulate Fe in the open Bothnian Bay showed little variation between 0.1 and 0.7 µmol/L, whereas the highest concentrations occurred at station A5 during June (Fig. 6).

Pre-flood and Post-flood
During pre-flood, discharge in the Kalix River is four to five times higher than in the Råne River. During post-flood on the other hand, the discharge of the Kalix River is about 23 times higher, due to the delayed snowmelt from the mountains in the catchment of the Kalix River (Fig. 2). The salinity in the estuaries during pre-flood varied between 0 and 2.6 g/kg (ESM_1 and _2), whereas just the surface water samples had salinities around 0 g/ kg, the deeper samples had values between 1.7 and 2.6 g/kg. During post-flood, the water column in the Råne estuary (R2) was well mixed and showed an average salinity of 2.4 g/ kg (ESM_1 and _2). The water column in the Bothnian Bay was well mixed with an average salinity of 3.2 g/kg. Dissolved OC concentrations were lowest in the rivers and estuaries during base flow conditions (242-552 µmol/L) (ESM_2). During pre-flood, the DOC concentrations in the Kalix estuary were lowest in the surface samples (293-303 µmol/L), and values increased with depths to 415 µmol/L. In the Råne estuary, DOC concentrations During pre-flood, the DFe concentration varied between 0.2 and 13.2 µmol/L for both estuaries (Figs. 4,5). The surface samples had very high concentrations (average 8.5 µmol/L) compared to the underlying water column (average 1.3 µmol/L). During postflood, the DFe concentrations in the Råne estuary showed little variation with an average of 0.2 µmol/L (Fig. 6). The concentrations in the Bothnian Bay showed little variation around 0.1 µmol/L (Fig. 6). The surface waters sampled during pre-flood showed very high PFe concentrations (up to 13.3 µmol/L) values compared to the underlying water (on average 0.8 µmol/L). During post-flood, the PFe concentration in the Råne estuary ranged from 0.5 to 0.9 µmol/L. Samples from the Bothnian Bay after spring flood showed little variation between 0.1 and 0.3 µmol/L.
No dissolved Fe-isotope data are available for pre-and post-flood. The δ 56 Fe composition of the PFe in the estuaries showed no clear trend during pre-flood (Figs. 4, 5). In the Kalix estuary, the Pδ 56 Fe ranged from −0.09 ± 0.09 to +0.21 ± 0.07‰, whereas just the surface water samples had a positive Fe-isotope composition. In the Råne estuary, the Pδ 56 Fe ranged from +0.02 ± 0.05 to +0.14 ± 0.05‰ pre-flood and from +0.32 ± 0.06 to +0.38 ± 0.07‰ post-flood conditions. Within the Bothnian Bay during post-flood, the Pδ 56 Fe composition ranged from +0.48 ± 0.11 to +0.67 ± 0.15‰. Fig. 6 Bothnian Bay concentration transects for DFe (first row) and PFe (second row), as well as the Dδ 56 Fe (third row) and Pδ 56 Fe (fourth row) isotope composition. The stations were sampled during spring flood and late spring flood and post-flood. For reference data, the transects start at station R2 in the Råne estuary. The Ocean Data View program, version 4.7.10, was used to generate the transects (Schlitzer, R., Ocean Data View, http://odv.awi. de, 2018) 1 3

Temporal Variations of Fe
Seasonal dynamics in the boreal region suggest varying hydro-geological pathways throughout the year (Dahlqvist et al. 2007;Rosenberg and Schroth 2017). During pre-flood, low Fe and DOC concentrations, as well as high pH values, indicate groundwater as source for riverine input (Dahlqvist et al. 2007;Wortberg et al. 2017). In boreal regions, up to 2/3 of the annual discharge is transported during the spring flood (Pontér et al. 1990;Lidman et al. 2011). Increased discharge has a significant influence on the chemical composition of rivers, estuaries and the open sea. During spring flood, increasing DOC, Fe and decreasing pH suggest the inflow from organic-rich soil layers into the rivers (Seibert et al. 2009;Ledesma et al. 2015;Ingri et al. 2018). The riparian zone and its organic-rich soil layers are the main source for the increased DOC and Fe in subarctic, boreal forest rivers during the snow melt and the associated spring flood (Laudon and Bishop 1999;Bishop et al. 2004;Grabs et al. 2012). Recently, the flushing of the upper soil horizons in the catchment of the Kalix River has been validated with the use of the Rb/Sr ratio and Sr-isotope ratio by Wortberg et al. (2017). Our data show a clear correlation between the Pδ 56 Fe composition and the Rb/Sr ratio (R 2 = 0.75), with positive Pδ 56 Fe values corresponding with low Rb/ Sr ratios, whereas a negative Pδ 56 Fe composition corresponds with the high Rb/Sr ratiothe identified flushing component. This correlation implies that the origin of a negative δ 56 Fe composition is lying within the upper soil layers, available due to flushing events. The decrease in Pδ 56 Fe values from +0.24 to −0.06‰ in the Kalix River and from +.30 to −0.07‰ in the Råne River reflects the changing hydro-geological pathways between pre-flood and spring flood for PFe. The Dδ 56 Fe composition, on the other hand, is almost stable with an average composition of +0.53 ± 0.05‰ throughout pre-flood, spring flood and post-flood, indicating no or little variation of the DFe source.
In boreal regions, Fe is transported in two primary forms in the rivers: Fe-OC complexes and Fe (oxy)hydroxides (Ilina et al. 2013;Neubauer et al. 2013;Sundman et al. 2014). In arctic and subarctic rivers, the high DOC concentration leads to the building of organo-mineral colloidal status for most metals, i.e. Fe (Pokrovsky et al. 2010;Schroth et al. 2011;Ingri et al. 2018). Organic carbon can stabilize Fe aggregates during transport (Kritzberg et al. 2014;Herzog et al. 2017). Salinity-induced aggregation experiments in both rivers resulted in an aggregated phase dominated by Fe(oxy)hydroxides with a lower Fe-isotope composition than the remaining suspended phase, with a higher amount of Fe-OM (Herzog et al. 2019). Therefore, we assume that DFe is dominated by Fe-OM complexes with positive Fe-isotope values, while PFe seems to be dominated by Fe(oxy)hydroxides with lower δ 56 Fe. Evidence suggests that the decrease in Pδ 56 Fe is caused by the input of negative Fe-OC complexes from organicrich surface soils, flushed during spring flood. The comparable heavier values during pre-and post-flood origin from Fe(oxy)hydroxides precipitated from anoxic groundwaters with high amounts of dissolved Fe(II) concentrations (Neubauer et al. 2013). Boreal first-order streams have mainly Fe-OC complexes, where the Fe occurs in the reduced Fe(II) or a mixed Fe(II)/ Fe(III) oxidation states (Emmenegger et al. 1998;Rose and Waite 2003;Sundman et al. 2013), the higher-order streams consist of Fe-OC complexes and nanoparticulate Fe(oxy)hydroxides, which aggregate and can be found in the > 0.22 µm fraction (Neubauer et al. 2013). With the decreasing amount of Fe-OC complexes, the amount of Fe(oxy)hydroxides with Fe(III) increases (Lofts et al. 2008;Neubauer et al. 2013).
Summarizing, we suggest that the DFe reaching the estuaries has a relatively stable δ 56 Fe (+0.53 ± 0.05‰) throughout the season, whereas the PFe reaching the estuaries is mainly Fe(oxy)hydroxides with a lower δ 56 Fe during pre-and post-flood and a negative δ 56 Fe signal during spring flood, caused by Fe-OC complexes from the riparian zone.

Fe in Low Salinity Mixing
The mixing of freshwater and seawater influences the salinity in the estuaries (Dyer 1973). During spring flood, the salinity within the estuaries is generally lower than during pre-and post-flood, due to the increased input of freshwater. The temporal and spatial measurements of salinity within the estuaries and the Bothnian Bay showed a fast-changing and responding system (ESM_2, Figs. S3 to 5). During spring flood, salinity measurements showed, (1) at peak discharge the estuaries are freshwater dominated; (2) the influence of freshwater vanishes within 2 weeks; (3) peak discharge can be traced out to the open Bothnian Bay. During pre-flood, salinity measurements show a layering in the Kalix and Råne estuaries (ESM_2, Figs. S3 and 4), the high salinities in 5 and 10 m depths suggest seawater, with generally higher salinities, are the primary influencer of the salinity in these areas. The low surface salinities might be caused by the ice on top of the water column. By drilling into the ice, the ice melts and dilutes the surface water. Salinity has a significant effect on the Fe and DOC interaction and concentration in the estuaries (Boyle et al. 1974;Holliday and Liss 1976). During spring flood, we observed about 50% removal of DOC along the estuaries, which is in accordance with data found for the Bothnian Bay and to other authors for the arctic region (Alling et al. 2010;Letscher et al. 2011;Deutsch et al. 2012).
During spring flood, Fe is transported out to the Bothnian Bay in a freshwater layer on top of the denser saltwater, which reaches as far as station K3 (Fig. 4). While the Bothnian Bay (A5 and A13) shows constant Fe values during spring flood, and during the late spring flood the freshwater input reached as far as station A5, where Fe concentration increased compared to earlier in the year (Fig. 6). Both DFe and PFe concentrations showed a non-conservative behaviour in the Kalix and Råne estuary, which has been found in other estuaries before (Boyle et al. 1977;Sholkovitz et al. 1978;Gustafsson et al. 2000). Dissolved Fe often contains a significant amount of Fe colloids (Schroth et al. 2011;Pokrovsky et al. 2012;Ilina et al. 2013;Conrad et al. 2019). The destabilisation of Fe-rich colloids and particles by seawater cations are one of the major factors for the flocculation (Mosley et al. 2003;Gerringa et al. 2007;Escoube et al. 2009). These flocculates might sink to the surface sediment or might be transported out of the estuaries (Daneshvar 2015). The amount of Fe flocculating at a certain salinity can be calculated. The flocculation factor F represents the fraction of DFe, which is removed by colloidal flocculation, with 1 equals 100% removal of DFe (Boyle et al. 1974;Escoube et al. 2009). The flocculation factor is calculated using two system end members, DFe and the salinity (S) of each sample. The samples R1a_05 (16.3 µmol/L DFe) and K1a_05 (17.3 µmol/L) are defined as the river end members, as they have relatively high Fe values and low salinity (< 0.02 g kg −1 ). The seawater end member should have less than 1% of the initial Fe concentration, we defined sample A13b_05 (0.1 µmol/L; 3 g kg −1 ), as it showed the highest salinity concentration.
The flocculation factor was calculated for the spring flood estuary data (Fig. 7), about 75% of the DFe flocculated at salinities below 1 g/kg and about 90% at salinities below 3 g/ kg. In comparison, Escoube et al. (2009) showed removal of DFe by colloidal flocculation of 50% below a salinity of 5 g/kg in the North River, which increased to 70% until a salinity of 10 g/kg. At salinities > 10 g/kg, the flocculation is stable between 70 and 80%.
Our values indicate a higher removal of DFe at lower salinities in the estuaries compared to the North River. Earlier studies have suggested major removal of Fe between salinities of 5-15 g/kg in the Mullica estuary (USA) and the Amazonas and Para River (Brasil) (Boyle et al. 1974;Sholkovitz et al. 1978). Within boreal estuaries in Finland, significant removal at salinities below 2 g/kg was observed (Asmala et al. 2014). Increasing pH along the estuaries can affect the flocculation and precipitation of Fe oxides and hydroxides. At pH values below 6.5, the oxidation rate normally is low, while at higher pH values [e.g. 7.4-8.6; (Daneshvar 2015)] Fe oxides precipitate as hydrated ferric oxides. In both the Kalix and Råne estuaries, we observed an increase of pH from 6.5 to 7.4 between May and June 2013, which would increase the possibility to form solid-phase ferric iron oxides or hydroxides within the estuaries (Daneshvar 2015). Furthermore, the stability of Fe colloids and particles depends on their composition. The relationship between Fe and OC controls the stability of Fe colloids and aggregates during estuarine mixing and is highly dependent The DFe concentrations in the estuaries during pre-and post-flood are much lower than during spring flood. The only exceptions are the surface waters, which showed elevated DFe concentrations during pre-flood. Studies of DFe and PFe in sea ice showed that the two Fe phases are enriched in the newly built ice (Lannuzel et al. 2010;Janssens et al. 2016). Drilling through the ice and accompanied ice melting might have caused the high DFe and PFe concentrations in the surface waters. Contamination of the water profile by sediment, which got redistributed to the water column, might cause the extreme PFe concentrations at station R1 during late spring flood.
Within both estuaries, we could observe the same two groups of Fe-isotopes that were observed in their rivers during spring flood (Fig. 7). In our study, DFe has a positive Feisotope composition from +0.11 to +0.49‰, within the range of earlier studies, which showed Fe-isotope compositions of various rivers between −1.2 and +1.8‰ (Ilina et al. 2013;Escoube et al. 2015;Zhang et al. 2015;Opfergelt et al. 2017). The PFe phase on the other hand shows lower and mostly negative Fe-isotope values (− 0.13 to 0.09‰) comparable to published values in the range of −0.87 to +0.40‰ (Ingri et al. 2006;Ilina et al. 2016;Cheng et al. 2017;Opfergelt et al. 2017). The Fe-isotope signal in both fractions is stable during spring flood, suggesting that the source of the signal is available over a longer period. The particles formed during spring flood in the organic-rich soils are transported along the river through the river mouth into the estuaries. Therefore, in contrast to the Fe concentrations, the Fe-isotope compositions are stable along the salinity gradient. Within the Bothnian Bay, PFe-isotope values are slightly heavier than in the estuaries. We could not detect the negative Fe-isotope signal in the Bothnian Bay. Hence, the negative particles might be removed from the water column during estuarine mixing. The driver for the removal could be the different surface properties of the Fe-OC and Fe(oxy)hydroxide aggregates, which lead to different flocculation behaviours along the estuaries (Gustafsson et al. 2000).
Particulate Fe-isotope values during pre-and post-flood in the rivers, estuaries and the Bothnian Bay are generally heavier compared to values during spring flood. During preflood, the estuaries show values around +0.2‰, while during post-flood, the measured values are higher around +0.4‰. Suggesting that PFe delivers positive Fe-isotope values during base flow conditions in contrast to the spring flood. The post-flood PFe-isotope values in the Bothnian Bay are as high as +0.6‰ exceeding values found during earlier research [+0.1 to +0.2‰, (Gelting 2009)].

Conclusions
First of all, this study shows the stability of a natural system over decades. Earlier research in the Kalix River [since 1982, e.g. (Pontér et al. 1990;Ingri 1996)] showed similar behaviour of the Fe concentration and Fe-isotopes. The annual spring flood with the accompanying features of increasing Fe concentrations and decreasing Fe-isotope compositions is a regular and periodic event influencing the water geochemistry. Our data set complements earlier research with the dissolved phase, as well as additional data how the Fe-isotope signal is transported along the estuaries into the Bothnian Bay.
1. The merged Fe-isotope signal of a subarctic, forest dominated catchment can be divided into the dissolved and particulate phase. Within the rivers and estuaries, we found positive Fe-isotopes in the dissolved phase and negative Fe-isotopes in the particulate phase during spring discharge. 2. The correlation between different chemical parameters Fe and DOC allows us to conclude that the Fe-isotope composition during spring flood is evolving in the organic-rich soil layers of first-order streams. Therefore, the lighter Fe-isotope signal is correlated to the organic-rich soil layers of the riparian zones. During pre-and post-flood, PFe has a positive Fe-isotope signal. This shows that Fe has different origins throughout the season. 3. Salt-induced flocculation on a small salinity range, up to 1 g/kg, is removing about 80% of the dissolved and particulate Fe during estuarine mixing. Increasing salinity and pH lead to the formation of colloids and particles, which sediment due to small changes in ionic strength. Within the low salinity estuaries, salt-induced flocculation is removing about 90% of the dissolved and particulate Fe. This major flocculation at low salinities might cause an underestimation of riverine Fe flux. Within the estuarine mixing zone, no Fe-isotope fractionation was observed. During spring flood, the Fe-isotope signal is constant along the estuaries, which excludes fractionation processes for example by oxidation. 4. The Fe-isotope signal within the Bothnian Bay was positive showing that different surface properties of Fe-OC and Fe(oxy)hydroxide aggregates lead to the flocculation of negative Fe aggregates.