Growth of marine microalgae is stimulated by enhanced concentrations of major nutrients like nitrogen and phosphorus in the photic zone (e.g. Sigman and Hain, 2012). Microalgae in the oceans are primary producers, which provide food for consumers at higher trophic levels and oxygen for respiration (e.g. Micheli, 1999). Through photosynthesis and the biological pump, marine primary producers also extract CO2 from the atmosphere. Carbon, nutrients and trace elements, which are fixed by primary producers, are further ingested by higher organisms or sink to the ocean floor where they are stored in the form of organic matter. The organic matter will eventually be remineralised, releasing the carbon, nutrients and trace elements to the bottom water (e.g. Sigman and Hain, 2012). A fraction is also buried in sediments. Supply of carbon to the oceanic ecosystems is an important part of the biogeochemical cycles, which are presently being disturbed by the human impact. Via photosynthesis, primary producers can help to remove CO2 from the atmosphere through the biological pump. However, increasing levels of dissolved CO2 in the oceans associated with so-called “ocean acidification” may negatively impact ocean's carbonate producers (e.g. Doney et al., 2009; Haynert et al., 2012) by influencing their survival and fitness (e.g. Thomsen et al., 2017). Excessive export of organic matter furthermore changes the oxygen condition at the sea floor due to decay of organic matter, which lowers the dissolved oxygen content, in turn negatively influencing the benthic life regime (e.g. Kristiansen and Aas, 2015, and references therein).

Coastal zones are among the most productive marine regions, characterised by high atmospheric CO2 uptake, organic matter accumulation and decomposition (e.g. Hjalmarsson et al., 2010). The Skagerrak, located between the North Sea and the Baltic Sea and in the close proximity to land, has many potential nutrient sources, such as the North Atlantic, Baltic Sea, and North Sea, as well as continental discharge and river runoff (Aure and Dahl, 1994; Andersson, 1996; Gustafsson and Stigebrandt, 1996). Upwelling and precipitation further increase the nutrient supply to the surface waters, additionally stimulating productivity in this region (Pingree et al., 1982; Aure and Dahl, 1994; Fonselius, 1996). The North Sea and the Skagerrak absorb significant quantities of atmospheric CO2 via the biological pump (1.38 and 1.2 mol C m-2 yr-1, respectively) and thus play an important role in the carbon cycle (Thomas et al., 2005; Hjalmarsson et al., 2010). The Skagerrak acts as a main depositional basin for about half of the refractory carbon produced in the North Sea and for a high amount of labile organic matter either imported with waters of the near-bottom current from the Danish coast or produced by intense algal blooms (Boon et al., 1999). The input of nutrients largely regulates food webs, which makes nutrients of great economic importance for the coastal areas worldwide (Micheli, 1999; FAO, 2016). In the Skagerrak nutrients are particularly important for the Nordic fisheries (Hop et al., 1992; Iversen et al., 2002; Olsen et al., 2004; Skogen et al., 2007). Fisheries and aquaculture sectors of the Skagerrak, commercially valuable for the Scandinavian nations, also make a relevant contribution to a growing global demand for food, which will largely rely on coastal regions to host a major part of food production in the future (e.g. FAO, 2016). Effects of increased primary production range from positive impacts on growth rate, size and reproduction of fish and shellfish populations to disruptive alterations in the food webs, thus yielding or reducing the profit rates of fisheries (Hop et al., 1992; Micheli, 1999; Iversen et al., 2002; Olsen et al., 2004, 2005; Breitburg et al., 2009; FAO, 2016). Disruptive changes in trophic levels of the Skagerrak ecosystem have been attributed to overfishing (Cardinale and Svedäng, 2004) but ongoing studies alert about adverse impact of increased nutrient inputs driving heavy phytoplankton blooms and eutrophication in the region (e.g. Baden et al., 1990; Aure et al., 1996; Breitburg et al., 2009). Eutrophication causes high demand and depletion of oxygen in the bottom waters, which affect species diversity, morphology and population growth and forces organisms to migrate (Rosenberg et al., 1990; Conley, 2009). Thus, to better understand the ongoing and possible future productivity changes and associated environmental effects, more historical studies are needed.

Although the Skagerrak is a well-investigated area, retrospective studies from this region have focused on climate change instead of productivity and, hence, past primary productivity changes are not well known (Polovodova Asteman et al., 2018). Previous studies suggested that among the processes driving primary productivity changes in the Skagerrak are (1) the North Atlantic Oscillation (NAO), (2) an input of anthropogenic nutrients from the Skagerrak catchment area and (3) nutrients transported from the Baltic Sea. Thus, while the NAO influences the inflow of the nutrient-rich Atlantic water mass into the Skagerrak (Gil et al., 2005; Brückner and Mackensen 2008), anthropogenic activities in the Skagerrak catchment area, such as land-use changes, also cause an increased nutrient input and high organic matter flux to the basin (e.g. Filipsson and Nordberg, 2010). In addition, eutrophication of the Baltic Sea and the resulting transport of nutrient-rich water with the Baltic Current through surface water exchange processes potentially represents an additional nutrient source (Andersson, 1996; Hjalmarsson et al., 2010; Krossa et al., 2015; Filipsson et al., 2017; Polovodova Asteman et al., 2018). In its turn, a nutrient-overloaded ecosystem in the Skagerrak may further supply dissolved inorganic nitrogen (DIN) to the Kattegat bottom waters (Carstensen et al., 2006) and increase the nutrient level in the coastal waters of Norway (Rosenberg et al., 1987; Rydberg et al., 2006) (Fig. 1). The last two processes involve human-induced influences on the ecosystem, while the NAO relates to effects following variability in the climate system. Over the last 1100 years the interaction between these processes has changed, as human influence has increased. In perspective of the last 4500 years, it is seen that productivity and its variability has increased from 50 years before the Common Era (BCE) until the present day (Polovodova Asteman et al., 2018). The increased productivity occurred as a response to enhanced local runoff coinciding with high winter rainfall and general cooling in Scandinavia, as well as intensified Baltic outflow, which may have significantly contributed to the nutrient supply in the Skagerrak in the past (Polovodova Asteman et al., 2018).

Our study aims to detect the variability and causes of marine primary productivity changes in the Skagerrak over the last 1100 years. We address this aim through a comprehensive multi-proxy study of two sediment cores from the central and the western Skagerrak, integrating records of the total organic carbon (TOC), foraminiferal assemblage data, stable carbon and oxygen isotopes (δ13C, δ18O) and trace element ratio (Mg/Ca). In addition, we evaluate our results in the context of the long-term productivity changes described previously in Polovodova Asteman et al. (2018).

The Skagerrak basin is located in the north-eastern part of the North Sea, connected to the Baltic Sea through the Kattegat (Fig. 1). The basin has a mean water depth of 210 m and a sill depth of 270 m. With a maximum depth of 700 m, the Skagerrak represents the deepest part of the Norwegian Trench (Rodhe, 1996). The area is characterised by an anticlockwise circulation and complex hydrography. The surface circulation (< 30 m) is to a large extent dominated by a surface current consisting of inflowing saline water from the southern North Sea and the North Atlantic and outflowing less saline water from the Baltic Sea (Danielssen et al., 1997). The inflowing nutrient-rich surface water flows along the Danish coast driven by the Southern Jutland Current (SJC) and the Northern Jutland Current (NJC) while the outflowing Baltic Sea water (BW ∼20–30 psu) flows as the Baltic Current (BC) along the Swedish west coast towards the north-eastern Skagerrak where it merges with the NJC and turns to the north-west as the low-salinity Norwegian Coastal Current (NCC) (Rodhe, 1996; Rydberg et al., 1996). The water flowing from the Skagerrak towards the Norwegian Sea (NCC) partly recirculates to the western Skagerrak (Rodhe, 1996). The surface water has a high nutrient concentration mostly due to the freshwater input via rivers draining from the Norwegian south coast, German and Danish east coasts, and the Baltic catchment area, but the upwelling of the underlying nutrient-rich Atlantic water is also considered to be an additional nutrient supply (Gustafsson and Stigebrandt, 1996; Rodhe, 1996). As a consequence of the mixing of different water types and the high freshwater input enhanced by precipitation, the upper layer of the surface water has low salinity (25–32 psu) and is determined as the Skagerrak Coastal Water (4.5–10 ∘C). The intermediate water layer (30–270 m) is referred to as Skagerrak Water (32–35 psu, 4.5–10 ∘C) and is driven by the subsurface circulation (Andersson, 1996). The deep-water layer below sill depth (> 270 m) is dominated by Atlantic Water and is recognised as the Skagerrak Basin Water (> 35 psu, 5.5–6.5 ∘C) (Aure and Dahl, 1994).

The subsurface circulation (below 30 m water depth) consists of nutrient-rich Atlantic deep water (AW > 35 psu, 5.5–8.5 ∘C) flowing through the northern North Sea, and the water from the central and southern North Sea (NSW ∼31–35 psu) (Rodhe, 1996). The inflowing water follows the southern side of the Norwegian Trench, entering the Skagerrak in its central part where this water is mixed with fresh surface water and flows out as the NCC (Winther and Johannessen, 2006).

Large-scale atmospheric systems and regional meteorological factors (e.g. precipitation and storms) influence the flow regime, creating a highly dynamic system in the upper layer of the water column where water mixing is largely caused by the south-westerly winds (Gustafsson and Stigebrandt, 1996). At the same time, calmer hydrographic conditions are typical for the intermediate layer and the deep water down to ∼400 m, with a maximum water residence time of 3 months (Andersson, 1996). This is in contrast to the renewal of the deepest water mass below ∼400 m, which is replenished every 1 to 3 years depending on the strength of the Atlantic water inflows (Aure and Dahl, 1994; Rodhe 1996), closely correlating with the NAO index (Brückner and Mackensen, 2006).

Two gravity cores (GCs) were retrieved from the Skagerrak during a R/V Elisabeth Mann-Borgese cruise in May 2013. Core EMB046/20-3GC (4.8 m long) was taken from the central Skagerrak (south-east Norwegian Trench; 58∘31.75′ N, 09∘29.13′ E; 533 m water depth), while core EMB046/10-4GC (4.62 m) comes from the western Skagerrak (south-west Norwegian Trench; 57∘49.73′ N, 07∘17.62′ E; 457 m water depth) (Fig. 1). Both cores were taken below sill depth within the deep waters of the Norwegian Trench. The cores were cut in 1 m sections on board before being split and subsampled every 1 cm ashore. This study is based on results from the upper 170.5 and 164.5 cm of cores EMB046/20-3GC and EMB046/10-4GC respectively, which corresponds to the last 1100 years. CTD measurements are available in the Supplement (Fig. S1). A part of the foraminiferal dataset and the TOC data were previously published in Polovodova Asteman et al. (2018). Here we present new stable isotopes (δ13C, δ18O), and trace element ratio (Mg/Ca) data covering the last 1100 years, in combination with foraminiferal assemblage data and multivariate statistics. Both cores consist of mostly homogeneous soft organic-rich clay, have olive-grey colour and show no significant changes in grain size and lithology throughout the studied intervals. The TOC content was determined using “Rapid CS cube –Elementar” analyser (Department of Geology and Paleogeography, University of Szczecin, Poland) with a measurement accuracy of 0.01 % (95 % confidence level (CL) at 99.5 % detection limit (DL)). For detailed methodology of geochemistry measurements (TOC) see Polovodova Asteman et al. (2018).

For the stable carbon and oxygen isotopes (δ13C, δ18O) and trace element analyses, well-preserved shells of benthic foraminifera Melonis barleeanus were picked from the dried sediment (fraction > 150 µm). Melonis barleeanus was selected for the analyses due to its relatively high abundance throughout the investigated intervals at both sites and its well-known potential for geochemical palaeoreconstructions (Mackensen et al., 2000; Kristjánsdóttir et al., 2007; Brückner and Mackensen, 2008; Butruille et al., 2017). Stable isotope measurements were performed at 1 cm intervals and the trace element analyses were done at 1–2 cm intervals from both gravity cores down to 170 cm (EMB046/20-3GC) and 164.5 cm (EMB046/10-4GC). No stable isotope or trace element analyses were done between 4.5 and 7.5 cm in EMB046/10-4GC, due to lack of material, because most of the foraminifera from that interval where used for 14C AMS dates.

Stable isotope analyses were run on a Finnigan MAT 253 mass spectrometer equipped with an automatic Kiel device at FARLAB of the University of Bergen. Prior to measurement the tests of M. barleeanus were lightly crushed, cleaned with methanol (≥98.8 %) using an ultrasonic bath and dried at 60 ∘C. For each measurement 2–4 specimens were used. All results are reported in ‰ versus Vienna Pee Dee Belemnite standard (V-PDB), using the National Bureau of Standards (NBS) 19 and 18, in combination with the internal lab standard CM12. The long-term analytical uncertainty is ±0.08 ‰ and ±0.03 ‰ (95 % CL) for oxygen and carbon isotopes, respectively.

Shells of M. barleeanus were also cleaned and analysed for trace elements at the Trace Element Lab (TELab) at Uni Research Climate, Bergen (Norway). For each analysis, approximately 15–20 specimens were gently crushed between two glass plates under a microscope to allow the contaminants to be removed. The samples were cleaned following the procedure described by Barker et al. (2003). The cleaning method includes clay removal steps, oxidation of the organic matter and surface leaching. Samples containing enough material were mixed and split into two subsamples to allow duplicate analysis. All samples were dissolved in trace metal pure 0.1 M HNO3 (prepared from HNO3 TraceSELECT^®) and diluted to a final concentration of 40 ppm of calcium (Ca).

The trace elements were measured on an Agilent 720 inductively coupled plasma optical emission spectrometer (ICP-OES). Six standards have been prepared at the TELab and they have a composition similar to foraminiferal carbonate (0.50–7.66 mmol mol-1). Fe/Ca and Al/Ca values have been checked and showed no significant correlation with the measured Mg/Ca values. The correlation coefficients (R2) between the Mg/Ca ratio and the Fe/Ca and Al/Ca ratios are 0.012 and 0.095 for the core EMB046/20-3GC and 0.020 and 0.067 for the core EMB046/10-4GC, indicating no systematic contamination due to insufficient cleaning. The Mn/Ca values in our samples are higher than the recommended maximum (< 105 µmol mol-1) (Boyle, 1983), indicating that diagenetic coatings might also affect our results. The Mn/Ca values, however, show no significant correlation with the measured Mg/Ca values (R2=0 for EMB046/20-3GC and R2=0.021 for EMB046/10-4GC). A standard solution with Mg/Ca of 5.076 mmol mol-1 is analysed after every eight samples to correct for instrumental biases and analytical drift of the instrument. The long-term Mg/Ca analytical precision, based on the standard solution, is ±0.016 mmol mol-1 (1σ standard deviation, SD) or 3.11 % (relative SD). The average reproducibility of duplicate measurements (pooled SD, dof = 41) is equivalent to an overall average precision of 4.09 %. The average Mg/Ca of long-term international limestone standard (ECRM752-1) measurements was 3.76 mmol mol-1 (1σ= 0.07 mmol mol-1) with the average published value of 3.75 mmol mol-1 (Greaves et al., 2008).

The Melonis spp. Mg/Ca bottom water temperature (BWT) is calculated from the measured Mg/Ca using the new Melonis spp. calibration (Mg/Ca = 0.113 (±0.005)*BWT + 0.792 (±0.036)), based on core top data covering a Mg/Ca range of 0.68–3.66 mmol mol-1 for a temperature range of -0.89–15.58∘C (Hasenfratz et al., 2017). According to the calibration uncertainty, a 1σ temperature error (95 % CL) of ±0.9 to ±1.7 ∘C has to be taken into consideration for the temperature range (4.1–9 ∘C) covered by EMB046/20-3GC. For EMB046/10-4GC the 1σ temperature error (95 % CL) is similar and ranges between ±1 and ±1.6 ∘C for the temperatures 5–8.5 ∘C. Further discussion on Mg/Ca-derived BWT within this article will utilise the Hasenfratz's calibration (Hasenfratz et al., 2017).

The results from Mg/Ca-derived BWT and δ18O were used to calculate the seawater isotopic composition (δw) using the following equation: δw= ((δ18Otop–δ18Odown)–(Ttop–Tdown)*0.23) + 0.3. The δ18Otop is the δ18O value measured on the foraminiferal shells from the uppermost sample (0–1 cm core depth) and the Ttop is the temperature taken from CTD measured at the time of coring. The δ18Otop equals 2.04 and 2.06 (‰) while Ttop was 5.72 and 5.34 ∘C for the EMB046/10-4GC and EMB046/20-3GC, respectively. The δ18Odown and Tdown is the down core δ18O and temperature as measured on the foraminifera samples. Temperature estimates based on δ18O follow Shackleton (1974), where 0.23 ‰ equals 1 ∘C in the temperature interval estimated for these sites. A constant of 0.3 ‰ is used to correct for the difference between V-PDB and δw. No correction for an ice volume effect was applied as this is considered negligible over the last 1000 years.

The foraminiferal analysis was carried out on 5–10 g wet sediment, gently sieved over a 63 µm sieve and wet-counted for foraminifera immediately afterwards. The two Skagerrak records were counted, over the targeted time interval covering the last 1100 years, at 1–3 cm resolution, with the exception of a 7 cm interval between 89 cm and 96 cm (∼ CE 1030–1055) for EMB046/20-3GC, and at 1–2 cm resolution for the EMB046/10-4GC record. In the > 63 µm fraction, at least 300 benthic and 300 planktonic (where possible) specimens were counted under a stereomicroscope and identified to a species level. Both relative (%) and absolute (individuals per gram in wet sediments, denoted as “ind. g-1 wet sed.”) abundances were calculated. The benthic foraminiferal species were categorised depending on their relative abundance in the assemblage as dominant (> 10 %), accessory (5 %–10 %) or rare (< 5 %). Only dominant and accessory species are discussed in this study (Fig. 3, Table 2). Benthic species with a relative abundance of > 5 % in at least two samples were subject to multivariate statistics using simple CABFAC factor analysis with varimax rotation (Table 2), performed by the PAST software (Hammer et al., 2001). This statistical tool provides a reliable method to distinguish the statistically significant foraminiferal units dominated by different species (e.g. Polovodova Asteman et al., 2013). In addition, benthic foraminiferal species indicative of increased organic matter fluxes to the sea floor, and hence, algal blooms, were grouped as “palaeoproductivity fauna” and included Alabaminella weddelensis, Brizalina skagerrakensis, Bulimina marginata, Epistominella spp., Nonionella iridea and Uvigerina spp. (Polovodova Asteman et al., 2018, and references therein). The planktonic foraminifera are presented as total planktonic individuals.

A common age model has previously been established for the two cores (Polovodova Asteman et al., 2018). The two age models were set at a common depth scale based on 30 available AMS 14C dates, as well as a correlation among total inorganic carbon, relative abundance of B. skagerrakensis and mercury (Hg) records from both cores. All dates were calibrated using Calib 7.10 (Stuiver et al., 2017), Marine13 (Reimer et al., 2013) and ΔR=0±50. The age model of Polovodova Asteman et al. (2018) is reasonable when investigating the longer-term trends over the last 4.5 ka. However, when focusing on the last 1100 years we found that there was a need for an improvement of the age model over this time period. This was achieved by increasing the number of dates and by a fine-tuning of the reservoir age. The new age model of EMB046/20-3GC, reaching back to CE 295, is based on eight 14C AMS dates in addition to the initial Hg increase at CE 1900 (Moros et al., 2017). The core top was set to 2013, the year of coring. A modern core top age is confirmed by a post-bomb age at 5.5 cm, as well as a recording of the Cs-137 signal associated with the nuclear weapons testing period (not shown). Considering all of the information from the upper part of the core in detail makes it clear that the use of a ΔR=200±50 provides a better transfer, avoiding an unlikely jump in sedimentation rate, between the modern ages and the 14C ages than when using a reservoir age of ΔR=0±50. Hence, when establishing the new age model, the 14C AMS dates were calibrated using Calib 7.10 (Stuiver et al., 2017), Marine13 (Reimer et al., 2013) and ΔR=200±50, as well as linear interpolation between the established tie points (Table 1; Fig. 2). The new ages calculated for EMB046/10-4GC are based on the new age model of EMB046/20-3GC and the previously established common depth scale for EMB046/10-4GC and EMB046/20-3GC (Polovodova Asteman et al., 2018) (Fig. 2). Due to the established relationship between the depth scales of the two cores (Polovodova Asteman et al., 2018), the age model for EMB046/20-3GC can be, and are, used to create the new age model for EMB046/10-4GC.

This study provides evidence for changes in primary productivity in the Skagerrak during the last millennium. Our multi-proxy records show that the time interval ∼ CE 900–1700 was characterised by moderate primary production with nutrients largely sustained by warm Atlantic water, as revealed by a high abundance of planktonic foraminifera. The first part of this interval (∼ CE 900–1450) was likely associated with the warm Medieval Climate Anomaly, during which a persistent positive NAO strengthened the westerlies, resulting in more frequent warm Atlantic water inflows. After ∼ CE 1450, continuously moderate productivity at both sites are indicated by a dominance of the P. osloensis factor, high abundance of planktonic foraminifera, relatively stable TOC and overall stable δ13C. This continuously moderate productivity at both sites coincided with the variable climate conditions characteristic of the Little Ice Age. Episodes of negative NAO-triggered deep-water renewal resulted in colder bottom water temperature in the central Skagerrak, while the western Skagerrak seems to be more resistant to this cooling and instead reflects the temperature of warm Atlantic water that occupied this site.

Finally, the high primary productivity period between ∼ CE 1700 and 2000 is documented by an increase in both TOC and the B. skagerrakensis factor, high absolute abundance of palaeoproductivity fauna and decreased δ13C values. Enhanced nutrient availability was likely caused by a stronger Baltic Sea outflow, increased river runoff, intensified inflows of the nutrient-rich Atlantic water, together with agricultural and industrial expansion. Simultaneously, an increase in human-induced CO2 emissions caused a great change in the oceanic carbon isotope budget, indicated by the Suess effect, shown in our records by strongly negative δ13C values since ∼ CE 1800. The most pronounced increase in primary production at ∼ CE 1800–2000 occurred during a warm period with a more positive NAO, wetter and warmer winters in Scandinavia as shown by an increase in BWT, and a decrease in δ18O in our records.

A comparison between the two records show an slight difference in species composition overall. This difference in species composition is likely due to a more favourable habitat in the western Skagerrak, with less exposure to low-salinity nutrient-rich NCC water, while the higher TOC in the central Skagerrak mostly results from an exposure to the Baltic outflow, nutrient-rich water reaching the site via NJC and terrestrial runoff due to a more inland location. The productivity and the fluxes of organic matter to the seafloor appear to not correspond to the temperature and salinity changes recorded in the benthic Melonis barleeanus shells.