Our study is based on three seasonal sampling campaigns (September 2018, February and June 2019) in Gullmar Fjord, Swedish west coast (Table 1). During each campaign two sites were sampled, GF 71 (water depth ∼70 m b.s.l.) and GF 117 (∼116 m b.s.l.). A CTD with O2-probe and Niskin water-collection system was deployed to record temperature, dissolved oxygen concentration ([O2]), and salinity (reported using the practical salinity scale) throughout the water column and to collect bottom water. Bottom water was sampled in bottles of 25–40 mL and refrigerated until further analyses. Sediment cores were collected with a GEMAX^® twin-barrel corer (modified Gemini corer, 9 cm diameter, from Oy Kart AB, Finland) in two casts. For foraminiferal analyses, the top 1 cm of two duplicate cores from the first casts was sliced in 0.5 cm sections (ca. 32 mL sediment; 0–0.5, 0.5–1.0 cm). Each sample was placed into an HDPE (high-density polyethylene) bottle with roughly equal amounts of ambient bottom water and 36 µL of CellTracker™ Green (CTG) CMFDA dye (5-chloromethylfluorescein diacetate; with dimethyl sulfoxide (DMSO); both ThermoFisher Scientific) to reach a final CTG concentration of 1 µM. The samples were incubated at 6 ∘C (in situ bottom-water temperature 6.5–7.1 ∘C) for 10 h for living cells to take up the CTG epifluorescent tracer for later observation (e.g., Bernhard et al., 2006), and thereafter preserved in 40 mL of ethanol.

At site GF 117, one set of duplicate cores was sampled for bottom-, pore-water, and solid-phase geochemical analyses. During each campaign, the first core was used for bottom-water and pore-water sampling using Rhizons™ at 2 cm vertical resolution. The samples were collected into 10 mL polyethylene syringes through pre-drilled holes (diameter 4 mm; e.g., Jokinen et al., 2020) immediately after core retrieval. Within a few hours after retrieval, all water samples were transferred into 15 mL polypropylene centrifuge tubes and acidified to 1 M HNO3 for elemental analyses.

The second core from each pair was sampled for solid-phase parameters. In September 2018, the second core was immediately sampled on deck after core retrieval. In February and June 2019, the second core was sampled a few hours after core retrieval inside a cool room. Each core was sliced at 1 cm intervals (0–10 cm) and 2 cm intervals (10–30 cm). Each sediment slice was transferred into plastic bags, which were carefully sealed under water, and transferred into gas-tight glass jars. The jars were flushed with N2 and stored in the dark at 4 ∘C prior to further processing to avoid oxidation of the samples. In preparation for solid-phases analyses, each wet sediment slice was subsampled inside a N2-flushed glove bag and frozen for 24 h at -20 ∘C. Subsequently, each sample was freeze-dried for 48 h, homogenized, and weighed in between each step to determine the water [g] and salt contents [g], and porosity [cm3 cm-3] using the bottom-water salinity and the assumed solid-phase density of 2.65 g cm-3 (Burdige, 2006). The water and salt contents were used to determine the salt-free weight of the dry sediment, which is needed to correct the solid-phase elemental concentrations for salt dilution.

Living (CTG-labeled) specimens of Bulimina marginata d'Orbigny, 1826 and Nonionellina labradorica (Dawson, 1960) were selected for laser-ablation (LA) ICP-MS trace element analysis, according to their availability in each sample (i.e., abundance). To extract specimens the samples were washed over a set of sieves (63, 100 µm) and the >100 µm fraction was wet-picked under epifluorescent light (Nikon SMZ 1500 with Nikon Intensilight C-HGFI). When a single core produced an insufficient number of specimens for analyses, specimens from the duplicate cores were also analyzed. An average of 10 living specimens – identified by their CTG-label – per sediment interval and site were selected for geochemical analyses (Table 1). The tests were treated with NaOCl (5 %) for 2–5 h to remove organic materials including cytoplasm (e.g., Mashiotta et al., 1999), following a triple rinse with Milli-Q (Milli-Q Integral, EMD Millipore Corporation, Billerica, MA, USA) water and dried at room temperature.

Laser ablation ICP-MS analyses were carried out at the Department of Geology, Lund University, Sweden, using a Bruker Aurora Elite (quadrupole) ICP-MS and a 193 nm Cetac Analyte G2 excimer laser installed with a two-volume HelEx2 sample cell. Helium was used as carrier gas (approximately 0.8 L min-1) in combination with a down-stream addition of argon (approximately 0.95 L min-1) after the sample chamber. The counts of the isotope masses 138Ba, as well as 27Al, 43Ca, 55Mn and 66Zn, were acquired at 5 Hz (i.e., laser shots per second). To obtain a stable isotope signal a “squid” was inserted downstream of the sample chamber. The scan time of the ICP-MS was synchronized with the laser repetition rate to minimize artificial fluctuations of the isotope signals. Spot sizes were adapted to chamber sizes of the investigated species (75×75, 55×75, 50×70 µm) but kept constant within individual sessions and applied to both samples and standard material. Previous studies indicate that using variable spot sizes does not influence elemental fractionation during LA measurements on foraminiferal tests (e.g., Eggins et al., 1998). The samples were analyzed at an energy density of 1 J cm-2 to avoid breaking of foraminiferal tests and to obtain a longer ablation signal. The conditions were mimicked for the secondary carbonate standards. Glass standards, US National Institute of Standards and Technology SRM NIST610 and 612 (see below), were analyzed at 3 J cm-2 for optimized ablation efficiency. The LA ICP-MS was operated in manual mode for sample analyses. Elemental baseline levels before each spot analysis were measured for at least 30 s.

NIST610 was used for instrument tuning to ascertain high and stable signal counts on the studied elements, low oxide production (<0.5 % monitoring 238U / 238U16O and 232Th / 232Th16O), and low elemental fractionation (232Th / 238U ratios close to 1). NIST610 was used as external calibration material with GeoReM (Geological and Environmental Reference Materials, 2021; Jochum et al., 2005) composition values (via http://georem.mpch-mainz.gwdg.de, last access: 30 April 2021), giving an average internal relative precision for Ba of 4.4 % (based on the relative standard deviation of raw data (cps) per analysis session; Table S1), and to correct for instrumental drift. Primary standard NIST610 and secondary standards, NIST612, MACS-3 (MicroAnalytical Carbonate Standard, United States Geological Survey, 2012; Jochum et al., 2012), and JCt-1 and JCp-1 (AIST Japan calcium carbonate pellets), were run at the beginning and end of each session and the data used to monitor the quality of the analyses (Table S1). A selection of these – NIST610, MACS-3, JCp-1 – were additionally run after each 10–15 sample spots.

For each specimen, the final (n), penultimate (n-1), and antepenultimate (n-2) chambers (as far as possible) were targeted and ablated from the outside to inside of the wall. For B. marginata also the signal of the initial chambers was measured with a laser-spot covering proloculus and following chambers and ablating through these chambers (referred to as “p”). We consider this to represent a bulk signal of an extended growth period, including the primary signal of the juvenile chambers as well subsequently added layers of calcite with each new chamber addition (e.g., Erez, 2003).

Raw counts were converted to elemental concentrations (ppm) with the software package Igor Pro 6.37 with Iolite v3.5 (Paton et al., 2011). The element ratios were calculated based on the average concentration measured during the ablation period of the primary test calcium carbonate. This signal interval was determined through manual examination of each trace element profile. Constant counts of 43Ca were used to find integration intervals of calcite ablation. Intervals of enriched signals of 27Al, 55Mn, and/or 66Zn, indicative of secondary surface coatings and/or contaminations, were excluded. Such were typically measured at the beginning and/or end of each profile (corresponding to the outer and inner test wall surfaces, respectively), with relatively constant raw elemental counts intermediate (see also e.g., Ni et al., 2020). Trace elemental ratios (TE/Ca) were calculated with the assumption of the TEs being bound as carbonate (assuming 40 % wt Ca in the CaCO3 matrix) and normalized to calcium.

In total, 486 ablation profiles (>15 data points of the ICP-MS per signal) were performed on chambers (n, n-1, n-2 and p) of B. marginata and 346 on chambers (n, n-1, n-2) of N. labradorica specimens and evaluated according to the following criteria (i.e., quality control): (1) [Ba] higher than limit of quantification (LOQ = baseline signal + 10×σ; σ = standard deviation of baseline signal) (9 signals), (2) exclusion of ablation signals with potential clay contamination (i.e., Al/Ca >0.4 mmol mol-1; Wit et al., 2010) (30 signals), (3) application of a 95 % confidence limit on Ba/Ca, for each species, site and season, to remove outliers (i.e., >2× standard deviation) (39 signals). In total, 78 of 832 (9.4 %) ablation profiles were rejected from further analyses.

From the benthic foraminiferal Ba/Ca values the median and MAD were calculated per species, season, and site, to limit the influence of extreme outliers. As no systematic differences between sediment intervals were observed (Welch ANOVA, Table S2), Ba/Ca data were pooled. If not indicated otherwise, only data on n-chambers were considered, assuming these were precipitated in close temporal proximity to the time of sampling.

The acidified bottom-water and pore-water samples were analyzed for elemental concentrations of Ba by inductively coupled plasma-mass spectrometry (ICP-MS; Thermo Scientific XSeries 2; Utrecht University, Department of Earth Sciences) and of manganese (Mn), iron (Fe), and sulfur (S) by inductively coupled plasma-optical emission spectrometry (ICP-OES; Thermo Scientific iCAP 6000; Helsinki University, Ecosystems and Environment Research Program). For the determination of total solid-phase Ba, Mn, and Fe concentrations, an aliquot of 100–125 mg of freeze-dried and powdered sediment was digested in 2.5 mL solution of 3:2 HClO4 (70 %) and HNO3 (65 %) and HF (40 %). The Teflon vessels were left with closed lids overnight under a fume hood with water trap on a hotplate at 90 ∘C. After 12 h, the lids were removed, and the extracts were heated to 140 ∘C to evaporate the acids. The evaporate was re-dissolved in 25 mL HNO3 (4.5 %) and left overnight on a hotplate at 90 ∘C. The next day, the Teflon vessels were weighed again to determine the dilution of the final solutions. Subsequently, the sample residues were analyzed by ICP-MS (Ba) at Utrecht University and by ICP-OES (Mn, Fe) at the University of Helsinki. Based on sediment sample duplicates the average analytical uncertainties were 6 ppm (Ba), 3.5 ppm (Mn), and 18.6 ppm (Fe).

All meteorological, hydrographical, and hydrological data of the Swedish west coast, the sill fjord, and its catchment area are publicly available from SMHI (SMHI, 2021). We used the SHARK hydrographic database (https://sharkweb.smhi.se/hamta-data/, last access: 10 February 2021; SMHI, 2021) to obtain monthly data on salinity, [O2], and/or chlorophyll a across 2018 and 2019 from three fjord stations Släggö (58∘25.98′ N, 11∘43.57′ E, 70 m b.s.l.), Alsbäck (58∘19.40′ N, 11∘32.80′ E, 120 m b.s.l.), and Björkholmen (58∘23.26′ N, 11∘37.60′ E, 50 m b.s.l.) (Fig. 1), of which Alsbäck is closely located to GF 117. Daily runoff and sediment load data were collected from the hydrological database “vattenweb” (https://www.smhi.se/data/hydrologi/vattenwebb, last access: 10 February 2021), partly based on the S-HYPE (Hydrological Predictions for the Environment) model maintained by SMHI. For both hydrological and meteorological data, the averages were calculated for 30, 60, 90, and 120 d periods preceding the sampling occasions directly or by 1 month, respectively (i.e., four periods ending on and four periods ending 1 month prior to the sampling dates).

All statistical analyses were performed with the software package PAST 4.05 (Hammer et al., 2001), considering a p value of <0.05 as significant. Since parts of the data were non-normally distributed (Shapiro–Wilk's test) and heteroscedastic (Bartlett's test), we performed ANOVA Welch or Kruskal–Wallis tests associated with Mann–Whitney post hoc tests to compare Ba/Ca between seasons, sediment intervals, species, and chambers (non-parametric tests). To determine strength and direction of monotonic relationships between categorial and continuous variables (e.g., chamber number and Ba/Ca) Spearman rank correlations were tested. We used Pearson's r to determine the relationship between two continuous variables (e.g., Ba/Ca and river discharge; dissolved TE concentrations), using normalized data if testing data across several sites and species (0–1 scaling).

In the year 2018 north-western Europe experienced severe heat and drought, ascribed to an interplaying Rossby Wave-7 circulation and positive North Atlantic Oscillation index (NAO) across the Northern Hemisphere (e.g., Kornhuber et al., 2019). The occurrence of this large-scale synoptic weather pattern has increased within the last three decades, being responsible also for the 2003, 2006, 2012, and 2015 drought events, in response to anthropogenic global warming (e.g., Kornhuber et al., 2019).

In Sweden, the drought was expressed as above-normal mean air temperatures and below-normal precipitation and distinctly imprinted in the seasonal runoff conditions across the country (Fig. 2a; SMHI).

While the hydrological cycle naturally experiences seasonal variability – high runoff in winter, spring, and autumn (150 mm each), low in summer (50 mm) – in 2018 already the spring was exceptionally dry. The very dry spring conditions likely contributed to amplifying the following summer drought (Bastos et al., 2020b), with summer and autumn precipitation being 50 %–75 % of the 1961–1990 average (southwestern Sweden).

On the contrary, the year 2019 registered above-average annual precipitation (120 %–130 % in southwestern Sweden), although strongly varying monthly (e.g., record-low in April, highest precipitation in May), and was warmer than average both in Sweden and globally (Dunn et al., 2020).

In the study area the contrasting hydroclimate conditions of 2018 and 2019 is expressed in annual average discharges of Örekilsälven to GF of 18.5 and 29.0 m3 s-1, respectively (Fig. 2). Coinciding, the benthic foraminiferal Ba/Ca shifted significantly in most samples, corresponding to an increase of up to 5 times (Table 1; Figs. 2b, 3, S1).

Correlations of foraminiferal Ba/Ca with the shift in hydroclimatic conditions of the consecutive years imply Ba/Ca representing the discharge 60–90 d prior to sampling (r=0.36–0.40) or discharge conditions preceding the Ba/Ca record by 1 month (r=0.46–0.48; testing Ba/Ca across species and sites; Table S3). On the one hand, a potential lag time in the transport of riverine Ba to coastal sediments is conceivable, determined by marine processes influencing the sinking and deposition of Ba. This may be influenced by environmental conditions including the degree of primary productivity and/or bottom-water oxygenation state and thus temporally variable (further discussed in Sect. 4.2). On the other hand, the lagging aims to account for the uncertainties in the exact timing of foraminiferal growth. In this respect, the lag is probably most applicable to the foraminiferal samples collected in February 2019, assuming the majority of foraminiferal calcite was precipitated in connection to the preceding autumn phytoplankton bloom (e.g., Gustafsson and Nordberg, 2001; Fig. S2). Temporal differences in reproduction and growth of the species could explain the differing strength of correlation to the observed hydroclimatic conditions, as well as species-specific intrinsic factors that can influence benthic foraminiferal Ba/Ca (further discussed in Sect. 4.3). While riverine Ba concentrations may also be temporally variable and thus could conceivably affect foraminiferal Ba/Ca signals, we consider such variations negligible in comparison to the contrasting Ba supply in response to the hydrological conditions of 2018 and 2019.

The range of Ba/Ca across 2018/19 varies between the inner-fjord and the outer-fjord site in all species (1–4 times vs. 2–5 times, respectively). The Ba/Ca signal of the outer-fjord site is generally lower than in the inner fjord in B. marginata, although the same is not resolved by N. labradorica (Fig. 3). To explain the trend in B. marginata, we hypothesize that the outer-fjord site experiences larger differences between periods of contrasting influence of Ba-rich river water versus low-Ba seawater of Kattegat and Skagerrak. This is plausible as the influence of Ba addition by Örekilsälven discharge diminishes, and the frequency of water exchange events increases, from the inner to outer fjord (Fig. S3). The fjord topography could be an additional factor, with the deep basin acting as a Ba sink. While the effects are minor compared to those of prolonged drought periods, these spatial factors should be of consideration when inferring hydroclimatic gradients from foraminiferal Ba/Ca.

The Ba/Ca signals of 2018 compared well to values expected for environments with negligible terrestrial influence (e.g., Anholt site in Kattegat; see Fig. 1, compilation of data from this study and Groeneveld et al., 2018 in Fig. 3); thus they are anomalously low considering the close spatial proximity to a river mouth (<25 km; Fig. 1) and seasonally restricted water exchanges with the open sea (Fig. S3). On the other hand, Ba/Ca of the 2019 fjord samples ranges distinctly higher, consistent with the expectation for coastal environments such as the Baltic Sea (Fig. 1 for location; Fig. 3).

We deduce that the drought prevailing in spring and summer 2018 caused a distinct lack of riverine input, interrupting terrestrial Ba addition to the fjord. The results imply that benthic foraminiferal Ba/Ca in the coastal environment responds on short timescales to the prevailing hydrological conditions on the adjacent continent.

Barium is incorporated into foraminiferal tests proportionally to ambient concentrations (e.g., de Nooijer et al., 2017). Hence, contrasting Ba2+ availability in the foraminifera's microenvironment is a prerequisite for the observed benthic Ba/Ca signals. For planktonic foraminifera it has been shown that Ba incorporation is temperature- and salinity-independent (e.g., Hönisch et al., 2011). As both parameters were stable in the fjord (T=6.7±0.2 ∘C, S=34.4±0.1), an influence on benthic Ba/Ca can be excluded. Bottom waters stayed oxygenated, albeit over a large gradient (ambient [O2] during sampling = 70–217 µmol L-1; Fig. S3), and [O2] shows no correlation to benthic Ba/Ca. This implies that Ba2+ concentrations close to the sediment–water interface must have changed between the sampling moments.

The expected difference in Ba2+ availability between 2018 and 2019, however, is not clearly supported by the Ba sediment profiles obtained during sampling (Fig. 4; based on GF 117). Sedimentary Ba contents range from 3.11–3.53 µmol g-1, indicating a strong enrichment relative to sediments of the Skagerrak (ca. 0.41 µmol g-1 particulate Ba; Lepland et al., 2000), confirming a near-shore accumulation due to terrigenous inputs. However, no significant changes were recorded in the surface sediment Ba concentration. Meanwhile, pore-water Ba profiles indicate diagenetic remobilization of Ba at depth in the sediment column. Values in excess of 1 µmol L-1 Ba2+ at 20 cm depth – distinctly higher than marine water column values (Kattegat [Ba] = 0.02–0.06 µmol L-1, Groeneveld et al., 2018; global open ocean surface [Ba] = 0.04–0.05 µmol L-1, Hsieh and Henderson, 2017) – support the concept of Ba release into pore waters fueling potential uptake into benthic foraminifera. However, the pore-water profiles also do not show significant changes between sampling years and most of the enrichment of Ba2+ in pore waters is observed deeper in the sediment column.

To explain the observed changes in foraminiferal Ba/Ca between 2018 and 2019 in the absence of whole-core sediment or pore-water geochemical evidence, we infer that rapid sub-annual processes at the sediment–water interface must be responsible. With sedimentation rates at the study sites being 0.7–0.9 cm yr-1 (Filipsson and Nordberg, 2004), only millimeters of new sediment material are expected to accumulate between the sampling occasions. Sub-annual changes in reactive Ba supply within this material, however, may dictate pore-water Ba release close to the sediment–water interface and uptake into benthic foraminifera. Such changes are unlikely to be captured in the pore-water profiles obtained at 2 cm resolution on the three sampling occasions. We infer therefore that during wet years such as 2019, reactive Ba is transported into the fjord in short-timescale pulses directed by continental rainfall and riverine discharge events, resulting in transient high fluxes of reactive Ba, which can be recorded by benthic foraminiferal Ba/Ca. Such temporally restricted fluxes may also explain the high relative standard deviation of Ba/Ca (in n-chambers of B. marginata = 46 %–71 %; N. labradorica = 17 %–69 %). While chamber precipitation in foraminifera is generally completed within a few hours (e.g., de Nooijer et al., 2009), population signals represent cumulative seasonal trends in terrestrial Ba input and benthic availability. In the following we explore potential sources of reactive Ba.

Ba/Ca of N. labradorica and B. marginata are significantly different for most seasons at each site (Table S4; Welch ANOVA). Partly, this may be an effect of species-specific differences in trace element incorporation, which has been documented for several taxa of foraminifera including Bulimina and Nonionellina (e.g., Barras et al., 2018; de Nooijer et al., 2017; Koho et al., 2017).

While B. marginata's Ba/Ca reliably records fluctuations in terrestrial-derived Ba input, N. labradorica shows seasonal shifts in Ba/Ca only in the outer-fjord site. Against expectation, Ba/Ca of the inner-fjord site does not deviate substantially from values recorded during the drought period in 2018 in N. labradorica. This might be connected to a deeper effective living depth in this site, exposing the specimens to a constant, low-level supply of Ba2+ (as discussed in Sect. 4.2.2). Factors determining migration and distribution of this species comprise food supply and the bottom-water oxygenation regime (e.g., Alve and Bernhard, 1995), which conceivably could vary between the inner and outer fjord. Nonionellina labradorica was less abundant in the inner fjord. Generally, its abundance has been decreasing within GF since the early 1980s, likely driven by the increase in periods of bottom-water hypoxia ([O2] <64 µmol L-1) (e.g., Filipsson and Nordberg, 2004). This potential sensitivity of N. labradorica's Ba/Ca signal to factors other than terrestrial input-driven Ba variability should be considered in proxy interpretations. Oppositely, B. marginata is particularly promising as a proxy for riverine discharge.

Incorporation of trace elements, including Ba2+, is further known to be a function of ontogeny in some taxa, as calcification rates typically decrease and control of element partitioning increases over the growth of tests (e.g., de Nooijer et al., 2017). Analyses across successive chambers of B. marginata and N. labradorica do show significant differences between chambers for some samples (Mann–Whitney test, Table S5). Particularly the proloculus area of B. marginata exhibits almost consistently higher Ba/Ca ratios than later chambers (i.e., n, n-1, n-2; Fig. S1). Because the strength and direction of the inter-chamber correlations vary (Spearman rank correlation, Table S5) we infer a combined influence of ontogeny and seasonal differences in pore-water [Ba] imprinting in successive chambers. As the sharp trend across the studied years is seen both in most recently formed chambers (n) and prior-formed chambers (n-1, n-2; p for B. marginata; Fig. S1), environmental changes appear to overprint ontogenetic effects on Ba/Ca at least across strong hydroclimatic gradients as experienced between 2018 and 2019.