Sediment Records Shed Light on Drivers of Decadal Iron Concentration Increase in a Boreal Lake

Increasing iron (Fe) concentrations are found in lakes on a wide geographical scale but exact causes are still debated. The observed trends might result from increased Fe loading from the terrestrial catchment, but also from changes in how Fe distributes between the water column and the sediments. To get a better understanding of the causes we investigated whether there has been any change in the sediment formation of Fe sulfides (FeS) as an Fe sink in response to declining atmospheric sulfur (S) deposition during recent decades. For our study, we chose Lake Bolmen in southern Sweden, a lake for which we confirmed that Fe concentrations in the water column have strongly increased along with water color during 1966–2018. Our investigations showed that Fe accumulation and speciation varied independently of S accumulation patterns in the Lake Bolmen sediment record. Thus, we were not able to relate the positive trend in Fe concentrations to reduced FeS binding in the sediments. Furthermore, we found that Fe accumulation rates increased along with lake water Fe concentrations, indicating that increased catchment loading rather than a change in the distribution between the sediments and the water column has driven the increase in Fe concentrations. The increased loading may be due to land‐use change in the form of an extensive expansion of coniferous forest during the past century. Altered forest management practices and increased precipitation may have led to enhanced weathering and erosion of organic soil layers under aging coniferous forest.

Fe mobilization from surrounding soils is favored by reducing conditions, which promote dissolution of Fe-(oxy) hydroxides (FeOOH) and formation of mobile Fe(II). Consequently, warmer and wetter conditions with raised groundwater levels enhance Fe and DOM export from land to aquatic systems (Dillon & Molot, 1997;Ekstrom et al., 2016;Heikkinen, 1994;Laudon et al., 2009;Sarkkola et al., 2013). The rates of Fe and DOM export further depend on land-cover, where peatlands and coniferous forests have been shown to be particularly important sources (Dillon & Molot, 1997;Kortelainen et al., 2006;Mattsson et al., 2009). The general transition from agriculturally dominated land-use toward coniferous forestry in Scandinavia during recent decades has recently been put forward as an important driver of browning, since organic matter that accumulates in forest soils can leach DOM to hydrologically connected streams and lakes (Finstad et al., 2016;Kritzberg, 2017). The accumulation of organic matter in coniferous forests may also be a considerable source of Fe, since the production of organic acids and lowered pH promote weathering (Humborg et al., 2004), and the association with DOM enhances solubility and mobilization of Fe (Škerlep et al., 2021). The importance of land-use is also indicated by the fact that the percentage of coniferous forest cover in the catchment was one of the most influential variables explaining variations in Fe trends across 340 European and North American lakes and running waters (Björnerås et al., 2017).
Fe plays a pivotal role in aquatic systems, not only by contributing to browning, but also as an essential micronutrient for aquatic organisms, interacting with the biogeochemical cycling of C, P, N and S (Raiswell & Canfield, 2012). To identify the main drivers of increasing Fe concentrations is thus important for the prediction of future Fe concentrations and for assessment of potential measures needed to mitigate undesirably high Fe concentrations. In this study we investigated the role of declining S deposition as a driver of increasing lake water Fe concentrations in Lake Bolmen in southern Sweden, focusing on the importance of Fe binding by sulfides in sediments. Although Lake Bolmen is a major sink for Fe and DOM (Björnerås et al., 2021), a pronounced browning of the lake water has been observed during the last decades (Kritzberg, 2017). The browning poses a challenge to the production of drinking water from the lake, which provides water to more than half a million people. We hypothesized that the accumulation rates of S and Fe in the sediments of Lake Bolmen co-vary, with peaks in the accumulation of Fe and S during the period of elevated atmospheric S deposition in the 1970s and 1980s. Moreover, we hypothesized that a major fraction of Fe is present as sulfides in sediments deposited during the period of elevated S deposition, while the more recently deposited sediments contain substantially less FeS. Finally, we hypothesized that increasing Fe concentrations in the lake water since the period of elevated S

Sediment Sampling and Radionuclide Dating
Two sediment sequences with lengths of 40 and 34.5 cm were collected in October 2015 at 35 m water depth in the deepest basin of the lake (56°51'40.9"N, 13°41'38.5"E; Figure 1 site A) using an HTH gravity corer (Renberg & Hansson, 2008). One of the cores was subsampled into 0.5-2 cm thick, contiguous sections, while the other one was kept intact and immediately upon sampling placed in a portable freezer. A third sediment core with a length of 27 cm was collected at 22 m depth north-east of the basin with the deepest location (56°57'22.6"N, 13°45'42.1"E; Figure 1 site B) and subsampled into 0.5-2 cm contiguous sections. All sediment samples were stored at −20°C. Samples from the two sectioned cores were freeze-dried and dated by direct radionuclide gamma assay at the Environmental Radiometric Facility, University College London. To generate 210 Pb dates, the constant rate of supply (CRS) model was used (Appleby, 2001), and maxima in 137 Cs and 241 Am concentrations, reflecting the peak in atmospheric fallout in 1963, were used for correction of the age models.

Elemental Composition and Accumulation Rates
All elemental analyses were performed on sediments from site A. Freeze-dried subsamples were digested in 7M HNO 3 and measured according to SS 28311 (2017) for total concentrations of Fe, Mn, Al, Si and S with ICP-OES (Perkin Elmer Optima 8300). Total C and N contents were determined by combustion using a Costech ESC 4010 elemental analyzer. A subset of freeze-dried samples were analyzed for biogenic silica (BSi) concentrations according to Conley and Schelske (2002) following digestion in 0.1 M Na 2 CO 3 and heating in a shaking bath to 85°C. The digestion was then added to 0.012 M HCl at 3, 4 and 5 hr for extraction of BSi. BSi concentrations were determined with the automated molybdate-blue method (Grasshoff, 1983) using a Smartchem 200 (AMS System) discrete analyzer.
Element accumulation rates (mg cm −2 yr −1 ) were calculated from sediment accumulation rates (Sed AR) based on the 210 Pb dating (g cm −2 yr −1 ) multiplied with sediment concentrations (mg g −1 ) of Fe, Mn, Al, Si, BSi, S, C and N. The elemental composition of the sediment record from site A was also analyzed with X-ray fluorescence spectrometry (XRF), yielding relative intensity counts of 35 different inorganic elements. The intact sediment core with its natural variations in water content was scanned with a Thermo Scientific Niton XL3t Goldd + XRF analyzer in mining mode (Cu/Zn) for 180 s at 0.5 cm intervals in a freezer compartment. We used Si and K as lithogenic elements indicative of mineral inputs from the catchment to the lake, as well as indicators of grain size since Si is commonly elevated in coarser sand fractions and K in clay (Kylander et al., 2011). The Si, K, Fe, and Mn data obtained from the XRF scanning were normalized to Ti as this element is considered conservative in sediment records based on its limited participation in biogeochemical processes (Boes et al., 2011;Chawchai et al., 2016).

Fe Speciation
The intact sediment core from the deepest location in the lake (site A) that had been kept at-20°C since collection was sliced into 0.5 cm sections in a freezer compartment. Sediment material was collected from a subset of sediment sections using a Ti-blade to avoid Fe contamination. The collected material was stored frozen and in an argon gas environment prior to the XAS analyses to prevent oxidation. The selected sediment samples represent the period before, during, and after peak atmospheric S deposition in the 1970s and 1980s based on the 210 Pb chronology.
Fe K-edge X-ray absorption spectra were collected at the Stanford Synchrotron Radiation Lightsource (SSRL), at beamline 4-1, California, USA. SSRL was operated in top-up mode at 3.0 GeV beam energy and ca. 500 mA ring current. The beamline was equipped with three consecutive ion chambers to monitor the transmitted beam and one solid-state passivated implanted planar silicon (PIPS) detector for fluorescence measurements. The ion chambers were filled with nitrogen gas. A pair of silicon crystals (Si [2 2 0], Φ = 90) were used to monochromatize the beam and the monochromator was detuned 35% to reduce higher order harmonics. A 3 μx Mn filter and Soller slits were used to reduce unwanted scattering and fluorescence contributions. The samples were mounted on a liquid nitrogen cryostat at around 80 K to prevent beam-induced damage, and aligned at 45° with respect to the incident beam. The sample slots were 3 × 7 mm and the hutch slit was set to 1 × 3 mm to prevent the beam from hitting the sample holder. The spectrum of a Fe foil was recorded simultaneously during all scans to allow for internal energy calibration. For each sample, 1-2 spectra were recorded, depending on Fe concentration, and data were collected to k = 13 Å −1 resulting in an acquisition time of approximately 20 min per spectrum. All individual scans were checked for possible beam damage by monitoring the first derivative of the absorption edge. However, no beam-induced changes were detected.
The individual XAS spectra were energy calibrated (setting the first inflection point of Fe(0) to 7111.08 eV), quality controlled with regard to beam damage, and averaged using SixPack (Webb, 2005). These spectra were further treated using Viper (Klementiev, 2001). A smoothing spline function was applied to the normalized spectra above the absorption edge to remove the background, and these spline-fitted spectra were k 3 -weighted in order to enhance the oscillations at higher k-values. A linear combination fitting (LCF) analysis was applied to the sediment samples, again using SixPack. LCF was performed on k 3 -weighted EXAFS spectra from k 2.8-12 Å −1 . Reference spectra of FeOOH (ferrihydrite, lepidocrocite, goethite), organically complexed Fe (Fe(III)-fulvic acid complex), Fe-bearing silicate (biotite, hornblende) and clay (vermiculite and nontronite) minerals, Fe phosphate (vivianite), Fe carbonate (siderite), and Fe sulfides (amorphous Fe(II) sulfide, pyrrhotite, pyrite) were used as model compounds. Non-negative boundary condition was applied and the E0 was allowed to float. Furthermore, the sum was not forced to equal 1, since it is likely that the model compounds cannot fully explain the sample spectrum. Components contributing less than 5% were excluded from the models.
Since freeze-drying appears to have affected Fe speciation, we base our interpretations on analyses of frozen samples. When the freeze-dried samples were analyzed instead of frozen, sulfides were not contributing to the fits. Instead, FeOOH and Fe-bearing silicate and clay fractions dominated, while organically complexed Fe was present at all depths.

Monitoring Data
Water chemistry data (Fe and SO 4 concentrations, water color, pH, TOC and KMnO 4 concentrations) for the main lake tributaries and the lake outlet were provided by the company Sydvatten, which produces drinking water from the lake, and by the Swedish University of Agricultural Sciences (SLU), which maintains the Swedish national lake inventory program (http://miljodata.slu.se/mvm/). Data series from Sydvatten (1968-1984for Storån and Lillån, and 1968 for Unnen) were combined with data from SLU (1985for Storån and Lillån, and 2010 for Unnen) to receive a complete time series lasting from 1968 to 2018. To take data from two different sources was possible since both monitoring programmes were based on similar sampling and analyzing protocols following standard limnological methods. All data are from water samples taken from 0.5 m below the surface.
For S deposition rates in the Lake Bolmen area we used modeled data from the European Monitoring and Evaluation Programme (EMEP), which provides data on long-range transmission of air pollutants in Europe (https:// www.emep.int/). Climatic data (annual precipitation and mean annual air temperature) were obtained from the Swedish Meteorological and Hydrological Institute (SMHI) database for weather stations at Ljungby within 20 km from the lake (SMHI, 2019). As a measure of changes in land-use, information on coniferous forest cover (as Norway spruce volume) in the Lake Bolmen catchment was provided by SLU and estimated using data from the Swedish National Forest Inventory (NFI), which is a sample-based inventory of forest resources designed to assess status and change at national and regional level. A polygon layer from SMHI was used to extract NFI transects and plots surveyed in 1926 and between 1953 and 2016 within the catchment from the NFI databases. From 1953 and onwards calculations of total volume of Norway spruce within the watershed were performed as 5-year running means. Southern Sweden has gone from a largely open, more agriculturally influenced landscape with dominance of grasses and deciduous vegetation, toward a landscape characterized by intensive forestry and a dominance of conifers, especially Norway spruce (Picea abies Karst) during the last century, reaching peak volumes and coverage in recent decades (Lindbladh et al., 2014;Mazier et al., 2015). Total spruce volume integrates both areal coverage and stand age, and thereby corresponds to the available soil organic matter pool (Rosenqvist et al., 2010).

Statistical Analysis
Changes over time in lake water Fe concentrations and water color, and in potential environmental drivers (S deposition, annual precipitation, yearly mean air temperature, and spruce volume) were analyzed using the non-parametric Mann-Kendall trend test (MKT). For the test yearly median values during 1966-2018 were used. Absolute rates of change were determined from the Theil slope of the MKT test (Dery et al., 2009;Stahl et al., 2010;Theil, 1950). To generate relative rates of change (% yr −1 ) the Theil slope was divided by the longterm median (covering the whole time period) and multiplied by 100. In addition to trend analyses we made predictions on variations in the annual mean Fe concentration in the lake outlet Bolmån during 1966-2018 using a standard least squares model with S deposition, annual precipitation, annual mean air temperature and yearly variations in spruce volume as explanatory variables. The standard least squares analysis was carried out in the JMP package version 15.

Changes Over Time in Lake Water Fe Concentrations and Potential Drivers
During the period of 1966-2018, there was a significant increase in Fe concentration in the outlet of Lake Bolmen (Bolmån) by 97% or 0.35 mg L −1 (Figure 2). Large gaps in the time series of Fe concentration in the three main tributaries made trend analyses unreliable. However, water color increased significantly at all three sites (MKT Storån = 5.3, p Storån < 0.001; MKT Lillån = 5.4, p Lillån < 0.001; MKT Unnen = 6.5, p Unnen < 0.001). The relative increase in water color was larger at the outlet of Lake Bolmen (170%) than at the tributaries Storån, Lillån, and the connected Lake Unnen (75%, 84%, and 142% respectively). After 1999, we observed a shift in Fe concentration and color, resulting in that mean Fe concentration in the lake outlet were 149% higher during 1999-2018 than during 1966-1990 (130% for color when compared to the time period 1966-1998; Figure 2).
Among the potential drivers for lake water Fe concentrations in Lake Bolmen, atmospheric S deposition in the region showed the strongest changes with a rapidly increasing deposition rate exceeding 2,500 mg S m 2 yr −1 around 1980 and a rapidly declining rate thereafter (Figure 3a). The significant decrease in S deposition between 1966 and 2018 (MKT = −8.7, p < 0.001) corresponded with declining sulphate concentrations in the lake tributaries and outlet, and with a concurrent increase in pH from ∼6 in the 1960's to ∼7 in recent years ( Figure S1 in Supporting Information S1). Annual precipitation and mean annual air temperature were variable, ranging between 541 and 1,061 mm and between 4.9 and 8.3°C, respectively (Figures 3b and 3c). During the study period 1966-2018 we observed significant increases in annual precipitation (23%, MKT = 2.5, p < 0.05) and in the annual mean air temperature (13%, MKT = 2.3, p < 0.05). Also vegetation cover has changed in the catchment area with a major increase in coniferous forest volume since the first half of the 1900's ( Figure 3d). The largest spruce volume in the catchment was recorded before the severe storm Gudrun causing substantial forest losses in 2005, when spruce volume was 366% higher than in 1926.
Using a standard least squares model, 80% of the variation in annual mean Fe concentrations in the lake outlet from 1966 to 2018 could be explained by variations in annual S deposition, annual spruce volume, annual precipitation and annual mean air temperatures. From the four model input variables, S deposition and spruce volume were significant for the model performance (p < 0.001 and < 0.05, respectively), while annual precipitation and annual mean air temperature had little explanatory power (p = 0.92 and 0.98, respectively).

Sediment Stratigraphic Properties and Chronology
The sediment sequence at Lake Bolmen's central site A (Figure 1) consisted of rather homogenous, dark gray, clayey, organic-rich sediments throughout the upper 30 cm. There is a transition to light gray, more clay-rich sediments below 30 cm, followed by more coarse-grained sediments (slightly organic gray clay with some silt and sand) in the bottom 5 cm of the core retrieved. For a detailed lithostratigraphic description of Lake Bolmen's sediments, with a focus on the Holocene and Late Weichselian strata below the gravity-core sequence described here, we refer to Ising (2001).
The sediment core from site A extends back to 1943 (±11 years; Figure 4a), and Sed AR have been relatively uniform (range 0.04-0.13 g cm −2 yr −1 ; Figure 5c), especially between 4.25 and 28 cm, as indicated by the generally exponential decline in unsupported 210 Pb activity with depth in this interval (Figure 4b). The unsupported 210 Pb activity exhibits a maximum at 4.25 cm, which suggests an increase in sedimentation rate in the last few years of deposition. Abrupt shifts in unsupported 210 Pb activity at 29 and 37 cm may be related to decreases in sedimentation rate with time at these levels. Interpretation of the 210 Pb record is complicated by the fact that equilibrium between total and supported 210 Pb activity is reached around 39 cm depth, near the bottom of the core (Figure 4b). However, the peak in 137 Cs activity at 34 cm, which reflects the maximum fallout from atmospheric testing of nuclear weapons in 1963 (Appleby, 2001), as confirmed by the distinct 241 Am peak at the same depth ( Figure 4c), enabled correction of the CRS model and substantial reduction of the dating uncertainty. The core from site B shows sediment characteristics similar to the upper part of the site A sediment sequence, as well as similar accumulation rates (range 0.03-0.11 g cm −2 yr −1 ).

Fe and S Accumulation Rates and Fe Speciation in the Sediments
The average accumulation rate of Fe (Fe AR) in the top 38 cm of the sediment record, corresponding to the period of 1945-2015 was 6.9 mg cm −2 yr −1 (Figure 5b). For the period of 1968-2015 there was a significant increase in Fe AR of 86% (MKT = 4.1, p < 0.001), which is in the same range as the increase in Fe concentration in the lake outlet (+97%, Figure 5a). The increase in Fe AR is significant also if excluding the surface sample (+78%, MKT = 3.8, p < 0.001). Moreover, the sulfur concentration and accumulation rate (S AR) was elevated in sediments deposited from the late 1950's to the late 1980's ( Figures 5d and S2 in Supporting Information S1), corresponding well with higher sulphate concentrations in the lake tributaries and outlet in the 1960's -1980's ( Figure S1 in Supporting Information S1). However, the Fe AR was not elevated during the period of high S accumulation, and rather followed dry weight Sed AR throughout the sediment profile. Furthermore, Fe was present in excess in relation to S throughout the sediment profile (mean molar Fe/S of 35.4), and although the molar Fe/S ratio was lower during the period of elevated S accumulation it never fell below 3.9.
Fe sulfides (FeS) were present in sediments deposited during the period of elevated atmospheric S deposition, but equally so in more recent samples ( Figure 6, Table S1 in Supporting Information S1). Overall, the Fe speciation was dominated by FeOOH and Fe-bearing silicate and clay fractions, and was rather similar throughout the core ( Figure 6, Table S1 in Supporting Information S1). The presence of FeS in some of the samples was further supported by the back-filtered first shell peak of the Fourier transformed EXAFS spectra ( Figure S3 in Supporting Information S1). The sulfide-containing samples showed different amplitudes and a significantly shorter phase as compared to samples with purely oxygen-containing minerals, which is consistent with back-scattering from S at longer distances than the typical Fe-O distances.

Trends of Other Elements in the Sediments
Si AR increased from the early 1980's and onwards, but was strongly elevated in the sediments deposited during the late 1960's and early 1970's (Figures 7a and S2 in Supporting Information S1). Since less than 25% was biogenic Si-a commonly used proxy for diatom abundance and productivity (Conley & Schelske, 2002)-the sediment Si content was mainly of clastic origin. However, there was a higher percentage of biogenic Si in the sediment layers with elevated Si accumulation in the lower part of the record (Figure 7a). Fe exhibited similar dynamics as the clastic element K, and to some extent Si, with enrichments in Fe and K relative to Ti often, but not always, coinciding with wet years (e.g., year 2007; Figure 8). Fe/Ti, Si/Ti, K/Ti and Mn/Ti ratios were also elevated near the bottom of the core (below 30 cm; Figure 8).
There was a significant increase in both C and N accumulation rates during the period of 1968-2015 (MKT C = 4.4, p C < 0.001, MKT N = 4.2, p N < 0.001; Figure 7b), and the C AR was around four times higher than Fe AR throughout the sediment profile (mean molar C/Fe of 3.9; Figures 5b and 7b). Al and Mn AR were similar to Fe AR and generally followed sediment AR, although the molar Fe/Mn ratios decreased toward the top of the sediment profile (Figures 7c and 7d).

The Role of S Deposition as a Driver of Increasing Lake Water Fe Concentrations
The doubling of Fe concentration in the outlet from Lake Bolmen between 1966 and 2018 conforms with the observations of Björnerås et al. (2017) who reported a median increase in Fe concentration of 61% between 1990 and 2013 for freshwaters distributed in northern Europe and North America. Water color also increased significantly in water flowing in and out of Lake Bolmen, showing that the lake water is getting browner and that increasing lake water Fe concentrations likely contribute to the ongoing browning process. Fe-(oxy)hydroxides (FeOOH), organically complexed Fe (Fe-OM), Fe-bearing silicate (Fe-Si) and clay (Fe-clay) minerals, FeS, Fe phosphate (Fe-P), and Fe carbonate (FeCO 3 ) were used as reference compounds and their relative contribution to the LCF models are expressed as percentages of the depth specific sediment Fe accumulation rate. Only reference compounds contributing >5% to the LCF models were considered significant in explaining Fe speciation in the sediments. Fe-OM, Fe-P, and FeCO 3 contributions to the LCF models were below this threshold, and these fractions do not appear in the graph. The estimated age of the sediment samples based on the 210 Pb chronology (±2-5 years uncertainty) is shown on the upper x-axis.
The period of elevated S accumulation rate in the sediment record is highlighted.
We found that atmospheric S deposition was a strong predictor of Fe concentration in the lake outlet. This finding corroborates our hypothesis and confirms previously observed relationships between increasing Fe concentrations and decreasing sulphate concentrations in lakes across northern Europe and North America (Björnerås et al., 2017). We further observed clearly elevated S AR in the sediment record of Lake Bolmen during the period of elevated atmospheric S deposition in the region, and the presence of FeS in the sediments, demonstrating that conditions suitable for sulfide precipitation existed in the lake. Based on those results, and considering that around 70% of the Fe entering the lake is lost to the sediments (Björnerås et al., 2021), we suggest that formation of FeS of low solubility in the sediments is a potentially important Fe sink.
Although FeS can be considered an important Fe sink in Lake Bolmen it is not sufficient to explain the observed lake water Fe concentration increase, since the relative contribution of FeS to total Fe accumulated in the sediments was relatively stable over the entire time period (Figure 6). Furthermore, Fe AR was not elevated when high S AR prevailed in the sediments, implying that the availability of S was not a limiting factor for Fe accumulation. Although FeS was present in the sediments, FeOOH dominated together with Fe-bearing silicates and clay fractions at all studied depths. The dominance of FeOOH and Fe-bearing clays and silicates was supported by recent XAS analyses of surface sediments from several locations in the lake at different water depths (Herzog, S. personal communication). Thus, Fe speciation in the sediments provide no support to the hypothesis that rising Fe concentrations in the lake water took place as a result of declining FeS accumulation in the sediments since peak S deposition in the 1980's. Lake sediment records have previously confirmed a temporal agreement between sediment S accumulation, lake water S concentrations, and atmospheric S deposition . Furthermore, in the annually laminated (varved) sediments of Lake Nylandssjön in northern Sweden, Fe concentrations increased along with S concentrations from the 1950's to 1990's, followed by declining concentrations until the early 2000's (Gälman et al., 2009). However, while FeS was responsible for the black color of the darker varves, high Fe/S ratios (5-11) and an estimated Fe(II)/Fe(III) ratio of 0.25 in a dark varve (Shchukarev et al., 2008), suggest that FeS formation was not the major mechanism behind the corresponding temporal dynamics of Fe and S in Lake Nylandssjön (Gälman et al., 2009).
Although lake water Fe concentrations seemed to be independent on FeS formation and consequent sediment binding, it is possible that the declining atmospheric S deposition during recent decades has affected Fe concentrations in the lake by altering export from the catchment. Rather than declining Fe AR in the sediments with increasing Fe concentrations in the water of Lake Bolmen, an increase in Fe AR since the 1980's was observed in this study, indicating increased Fe loading to the lake. Sulfide precipitation can reduce Fe mobilization in waterlogged organic soils, which are active sites of sulfate reduction (Bottrell et al., 2007). It has also been proposed that increasing pH following the decline in atmospheric S deposition may have enhanced Fe export by increasing solubility of organic matter, which can form mobile complexes with Fe (Neal et al., 2008). In contrast, Fe mobilization from organic soils was enhanced by high sulfate concentrations in incubations of boreal soil, which probably was the result of a pH decrease promoting the solubility of Fe (Björnerås et al., 2019). In the current study, S deposition was approaching its peak in the beginning of the investigated time period, but studies that have explored historical data extending back to the period when atmospheric S loadings were still low in Southern Sweden have shown that rising S deposition had limited effects on water color in lakes and streams (Kritzberg, 2017;Škerlep et al., 2020). In conclusion, although it is possible that reduced atmospheric S deposition is a contributing driver behind increasing lake water Fe concentrations in Lake Bolmen during recent decades, reduced FeS binding in the sediments is an unlikely mechanism.

The Role of Land-Use Change in Lake Water Fe Concentration Dynamics
According to our results the annual spruce volume was a significant predictor of variations in annual mean lake water Fe concentrations. Spruce, in particular Picea abies (Norway spruce) is the dominant tree species in the Lake Bolmen catchment. It spread southwards into this region around 1000 years ago (Giesecke & Bennett, 2004), but expanded drastically in terms of spatial extent and volume only during the twentieth century, mainly as a result of modern forestry (Lindbladh et al., 2014;Mazier et al., 2015). Spruce cover in an adjacent area increased from <20% in 1880 to >50% in 2008 during the transition from traditional to modern land-use, when forest grazing and cultivation on relatively unfavorable soils ceased along with the introduction of silviculture (Fredh et al., 2012). The drastic increase in spruce volume in the catchment of Lake Bolmen, as a result of reduced grazing on former pastures, and modern forest management, is a further illustration of this borealization ( Figure 3). As coniferous forest soils, along with peatlands, are important sources of Fe (Dillon & Molot, 1997;Kortelainen et al., 2006;Maranger et al., 2006), spruce afforestation may be an important driver of increasing Fe concentrations in surface waters.
In a recent study, Fe concentrations were significantly higher in soil solutions under mature (90-year-old) spruce forest stands compared to younger (35-year-old) forests and non-forest soils, primarily because the higher organic matter content and lower pH in the mature forest soils favored formation and leaching of organically complexed Fe (Škerlep, 2021). Expanding coniferous forest cover may have led to enhanced weathering of soil minerals, since decomposition of the thick humus layer in boreal soils and ectomycorrhizal activity provides organic acids, which enhance weathering (Ronchi et al., 2013) and Fe mobilization (Jansen et al., 2004). Moreover, tree growth and harvesting that remove base cations from forest stands may contribute to soil acidification and increased mineral weathering (McGivney et al., 2019). Higher silicate weathering rates with increasing proportions of forest in catchments across Sweden generate higher riverine fluxes of dissolved Si (Humborg et al., 2004), illustrating the importance of land cover for catchment export of weathered elements (Struyf et al., 2010). Hence, the drastic land-use change toward dominance of coniferous forestry during the latter part of the twentieth century may potentially explain the increase in Fe and Si AR in the sediment record of Lake Bolmen through increased weathering and catchment export.
The increasing trends in C and N AR during recent decades (Figure 7) provide additional evidence of increased loading of organic matter from the catchment, which may originate from growing organic carbon stocks in response to forest growth and altered forest management practices Meyers & Lallier-Verges, 1999;Yang et al., 2021). The decline in organic carbon content with depth in the sediments may be due to diagenetic degradation, that is, microbial decomposition, which has had more time to act on older sediments. However, since the microbial decomposition rate can be rapid initially (e.g., 20% of C lost in 5 years), followed by stabilization with time (e.g., 23% loss in 27 years; Gälman et al., 2008), we believe that the more than doubling in C AR since 1966 is largely due to increased catchment supply. Increasing contributions of terrestrial organic carbon to northern boreal lake sediment have previously been attributed to increases in terrestrial DOC inputs (Gudasz et al., 2017). A previous analysis of monitoring data from Lake Bolmen suggests that TOC concentrations have increased in the lake during recent decades and are contributing to lake browning (Klante et al., 2021), but for the longer time period considered in this study, the temporal trend cannot be evaluated due to analytical inconsistencies in the time series ( Figure S4 in Supporting Information S1). However, there is generally a strong positive correlation between trends in TOC and Fe concentrations in freshwaters, as well as between the two variables and water color (Björnerås et al., 2017;Kritzberg & Ekstrom, 2012;Weyhenmeyer et al., 2014).
Afforestation can affect element fluxes from the catchment in several ways. On the one hand, afforestation may lead to a reduction in the accumulation of clastic components in sediments since increased land cover stabilizes the soil against erosion (Engstrom & Hansen, 1985). On the other hand, modern management practices, such as ditching and forest clear-cutting, disrupt the soil stability and may lead to increased soil erosion (Keim & Schoenholtz, 1999;Marttila & Klove, 2010;Nieminen et al., 2018). For example, mobilization of Fe and C from the upper organic soil horizons are markedly enhanced by clearcutting of boreal coniferous forests, partly due to soil mixing and increased mineralization rates, releasing Fe from logging residues (Hughes et al., 1990;Palviainen et al., 2004aPalviainen et al., , 2004b. The elevated sediment Fe and Si AR (Figures 5 and 7), as well as the coarser grain size of the sediments deposited in Lake Bolmen in the 1970's, may thus be a response to ditching, which was a common forestry practice promoted by governmental incentives during this period. Furthermore, the elevated AR of biogenic Si may have been partly caused by terrestrial input (Conley, 2002;Tallberg et al., 2015), such as phytoliths exported from the surrounding peatlands (Karlsson et al., 2016) during ditching activities in the Lake Bolmen catchment.
The Fe-bearing silicate and clay fractions in the sediments, along with the similarity in temporal dynamics of Fe and K, indicate that soil weathering and erosion contribute to Fe loading to the lake (Kylander et al., 2011). Across European forested headwater catchments, fine colloidal fractions dominated by clay minerals contribute substantially to Fe transport, especially in northern Europe where the proportion of coniferous forests and siliceous bedrock are higher (Gottselig et al., 2017).

The Role of Climate Change in Lake Water Fe Concentration Dynamics
Although precipitation was a poor predictor of lake water Fe concentrations in the applied standard least squares model, the significant increase in precipitation during the study period  may have promoted Fe export by enhanced runoff, weathering, redox mobilization (Ekstrom et al., 2016), and erosion, for example, from stream banks (Bjorkvald et al., 2008). The extremely wet year of 2007 (1,038 mm of precipitation) coincided with a peak in lake water Fe concentration, and the concurrent peaks in for example, Fe/Ti and K/Ti in the sediments deposited during this time may indicate that temporarily elevated runoff and groundwater levels favored both weathering and redox mobilization. The decrease in the Fe/Mn ratio since ∼1990 (Figure 7) may be indicative of redox-controlled Fe mobilization and more reducing conditions in the hydrologically connected soils around Lake Bolmen. Since Mn is more readily reduced than Fe, a selective transport of Mn in relation to Fe has been interpreted to reflect reducing conditions in catchment soils (Engstrom & Wright, 1984), although the sedimentary Fe/Mn ratio as a proxy for past redox conditions should be interpreted with care as Fe and Mn accumulation are not solely redox driven (Makri et al., 2021). Moreover, the baseline shifts of Fe and color around 1999 ( Figure 2) were likely triggered by a very wet period, and were observed also in lakes in Norway (Haaland et al., 2010;Riise et al., 2018). Hence, precipitation has likely affected Fe dynamics, and subsequently color, in the lake, but precipitation changes do not seem to be the main driver of Fe trends in Lake Bolmen.
Finally, it is possible that other variables have affected Fe dynamics in ways that cannot be revealed by this type of analysis. For example, although air temperature was not a significant factor explaining Fe trends in the water column, increased annual mean temperature may have contributed to increased catchment export of Fe by extending the duration of growing and runoff seasons (Weyhenmeyer & Karlsson, 2009), and by promoting microbial activity (Dawson et al., 2008) and Fe reduction rates in catchment soils (Curtinrich et al., 2021;Pallud et al., 2020). Likewise, increased rates of atmospheric N deposition during recent decades have likely promoted primary production and vegetation cover (Hyvonen et al., 2007).

The Reliability of Lake Sediments as Chronological Records
The interpretation of sediments as chronological records is challenging, since element compositions are determined by both catchment supply and in-lake processes (Engstrom & Wright, 1984). In addition, redox-active elements such as Fe may migrate from reducing zones, and the concentration and speciation of such elements may thus be sensitive to post-burial alterations (Boyle, 2001). Reductive mobilization of Fe in sediment successions may lead to the sediments acting as a source of Fe if the overlying water is hypoxic, but internal loading of Fe may also occur in lakes with well-aerated water columns if high oxygen consumption cause anoxia to extend to the sediment-water interface (Nürnberg & LaZerte, 2016). If the sediment surface is oxic, upward migration of reductively dissolved Fe results in re-oxidation and Fe accumulation near this zone (Engstrom & Wright, 1984).
In the sediment record of Lake Bolmen, there was a gradual increase in Fe AR over time, and the most superficial sediment sample (0.5 cm) was markedly rich in Fe ( Figure S2 in Supporting Information S1). The sub-surface peak in Fe may indicate accumulation of Fe migrated from deeper sediment horizons. However, since only the most superficial sediment sample was strikingly enriched in Fe, we believe that the gradual increase in Fe AR in the upper part of the record-along with elements which are not redox sensitive, such as Si and Al-is mainly a result of increased loading of Fe from the catchment, rather than substantial upward migration of Fe in the sediments. Lake Bolmen has been demonstrated to be an efficient sink of inflowing Fe, with temporal variations in Fe concentrations mainly determined by catchment loads of Fe (Björnerås et al., 2021). Nevertheless, it is possible that conditions favorable for sediment release of Fe to overlying waters develop occasionally in the lake, and that high availability of organic matter for complexation potentially enables released Fe to be kept in solution in oxic lake water.
It is noteworthy that FeOOH coexists with FeS in the sediments, which indicates strongly reducing conditions. Amorphous FeOOH (ferrihydrite) dominates in the sediments according to linear combination fits, and it is possible that interactions with organic matter or clay particles prevent crystallization into more stable forms (Hirst et al., 2017). Thermodynamic modelling of other lake sediment data has shown that FeOOH and FeS can coexist within a wide range of redox conditions (Gälman et al., 2009).
Interestingly, XAS analyses of freeze-dried and frozen sediment samples from the same depths yielded markedly different results. While FeS were detected throughout the sediment profile in frozen samples, the proportion of Fe sulfides was insignificant in freeze-dried sediment, using the same method and settings (Table S1 in Supporting Information S1). Changes in speciation patterns of major elements such as Fe and S in response to freeze-drying have been shown for lake sediments previously (Hjorth, 2004). The pH may change when carbon dioxide is removed, and the exposure to oxygen may increase during sample preparation. Thus, potential alterations of Fe speciation during sample handling must be taken into account when interpreting Fe records obtained from lake sediment stratigraphies.

Conclusions
We found that Fe concentrations have increased substantially along with water color in Lake Bolmen during the last 50 years. Contrary to our hypothesis, this increase in lake water Fe concentration was not linked to reduced FeS binding in the sediments of the lake in response to reduced atmospheric S deposition during recent decades. Instead, Fe accumulation and speciation varied independently of S accumulation patterns in the sediment record. Since Fe AR increased along with lake water Fe concentrations, increased catchment loading rather than a change in the distribution of Fe between the sediments and the water column appears to have driven the observed Fe trends. However, we cannot rule out that increased TOC concentrations, by promoting anoxia at the sediment-water interface, has allowed for more Fe to be released from the sediments and to remain in solution by being organically complexed, despite oxic conditions in the water column. Afforestation is a likely driver of increasing Fe and organic matter loading to the lake, while short-term changes in Fe dynamics within the lake could be linked to variations in precipitation. An increase in coniferous forest cover and altered forestry practices in the lake catchment during the past century have likely altered both weathering and erosion rates, and thereby Fe mobilization from surrounding soils.