Shetland is a subarctic archipelago covering a total area of 1466 km2, lying about 170 km north of mainland Scotland, UK (Fig. 1). Several voes dissect Shetland's coastline, providing an extended route for the transport of OC-rich material from land to sea (Cui et al., 2016). The current study focuses on six voes on Shetland's west coast: Clift Sound, Sand Sound, Olna Firth, Aith Voe, Busta Voe and Vaila Sound (Fig. 1, Table 1). These six voes are characterised by different geomorphologies, freshwater inputs, proximity to the open sea and intensity of local currents; together these characteristics determine the amount of carbon deposited in the sediments. Clift Sound lies south of Shetland, with the islands of Trondra and East Burra to the west, and has steep sides and a relatively unrestricted geomorphology (Fig. 1). Two main types of soil are present in the catchment of Clift Sound: humus–iron podzols cover the islands of East Burra and Trondra, while peaty soils dominate the eastern coast of Clift Sound (Fig. S1 in the Supplement). Winds are generally from the southwest and can be channelled, resulting in intensified currents due to the geometry of the sound and the surrounding highlands. Tidal energy can also be channelled into more vigorous currents (Cefas, 2007a).

Sand Sound is somehow T-shaped, located in the southwest of Shetland and comprises three areas: the head of the voe, the inner basin and the outer basin (Fig. 1). The head of the voe, almost perpendicular to the inner and outer basins, is a sheltered area that receives freshwater from a significant number of rivers draining the surrounding land. Three main types of soil characterise this area: peaty gleys, peaty podzols and peaty rankers (Fig. S1). Here, the voe is very shallow and drying can occur in its eastern and western periphery (Cefas, 2007b, 2008b). In contrast, the rest of the voe is much deeper. The inner basin gently slopes away, reaching more than 20 m at its maximum depth; drying can still occur where this basin meets the head of the voe (Cefas, 2007b). The outer basin has steep sides; it is 42 m deep at its deepest point and completely open to sea. A shallow sill divides the inner from the outer basins (Cefas, 2007b).

Busta Voe, Olna Firth and Aith Voe are part of a major inlet on the southern coastline of St Magnus Bay on the west coast of Shetland (Fig. 1). Olna Firth represents the eastern branch of the inlet and is roughly oriented east–west; it is the furthest away from the open sea (Fig. 1). A large area of this voe exceeds 30 m of water depth; the northern coastline gradually slopes into the voe, whereas the south side shows a steeper gradient. Olna Firth is classified as micro-tidal, and due to the generally low level of energy in this system, stratification may occur, especially during summertime (Cefas, 2013).

Aith Voe is the southern component of the inlet, with a north–south orientation (Fig. 1). Due to the hilly landscape (up to 100 m) surrounding the voe, Aith Voe is exposed to winds from the north, which can significantly alter local circulation and prevent stratification of the water column. Two types of soil surround Aith voe: peaty and organic soils (Fig. S1); few streams drain these soils, with the largest of them discharging on the east coast of Aith Voe (Cefas, 2010).

Busta Voe lies in the northern part of the inlet and is oriented north–south, sheltered and with a maximum water depth of 39 m (Fig. 1). Three main types of soil surround this area: peaty gleys, peaty podzols and peaty rankers (Fig. S1); two rivers drain a relatively small catchment area into Busta Voe. Tidal flow is weak in the voe and wind-generated currents are predominant; stratification of the water column, especially during warm periods, can occur (Cefas, 2008a).

Vaila Sound lies on the western coast of Shetland, with the island of Vaila to the southwest and the isle of Linga in the middle (Fig. 1). Three types of soil dominate the catchment of this voe: peaty gleys, peaty podzols and peaty rankers (Fig. S1). Most streams discharge in the northern part of the sound; hence, locations in this area might become more affected by terrigenous inputs. Additionally, Linga offers protection from wind and currents, which could facilitate the localised accumulation of terrigenous material. Shelter from strong winds is also provided by the island of Vaila to the south. Here the tidal range is small and the associated energy weak; however, Vaila Sound is connected to the Atlantic Ocean west and east of the island of Vaila (Cefas, 2009).

The MV Moder Dy 2015 cruise in west Shetland surveyed the six voes in August 2015 (Table 1). Marine surface sediments were sampled using a Duncan and Associates Van Veen grab with a sampling area of 0.1 m2. A total of 23 surface sediment samples were obtained by scraping the top layer (∼1 cm thick) of each grab with a domestic spoon; samples were then stored in a cold box in sealed plastic bottles.

An earlier field survey of Shetland voes carried out in August 2009 measured bottom water temperature (BWT), salinity (BWS) and oxygen (O2) at the same locations as this study (Fig. 2); however, no data were collected in Vaila Sound. Oceanographic measurements were made using a Sea-Bird SBE 19 plus conductivity–temperature–depth (CTD) profiler with a dissolved oxygen probe. Bottom water temperature varies between 14.14 and 12.20 ∘C, being on average 12.74 ∘C (Fig. 2a). Bottom water salinity remains almost constant across these voes, ranging between 35 and 35.2 psu with the exception of MD15-06, where BWS reached a minimum of 34.7 psu (Fig. 2b). Similarly, bottom waters are mostly well-oxygenated in all the voes (O2>5 mg L-1) with the exception of Olna Firth where lower O2 was observed (Fig. 2c).

Particle size analyses were performed on a subset of surface sediments using a Coulter LS230 particle size analyser to quantify the volume (%) of the various grain sizes in each sample. We identified three categories based on the International Organisation for Standardisation (ISO) scale: “clay” includes particles <2 µm, “silt” particles between 2 and 63 µm, and “sand” particles > 63 µm (upper limit of 2000 µm). Before analysis, sediment samples were digested using 30 % hydrogen peroxide to remove organic matter and 10 % hydrochloric acid to remove carbonate, following a similar methodology to that outlined by Austin and Evans (2010). Digestions were carried out to obtain an insoluble residue useful for comparing sediments having very different carbonate and organic matter content, which otherwise might skew the measurements because of in situ processes.

Loss on ignition measurements were carried out to quantify the percentage of total organic matter (TOM) and the respective amounts of labile (LOM) and refractory organic matter (ROM). The initial combustion temperature was set to 250 ∘C, as significant mass loss of LOM has been recorded at this temperature (Mook and Hoskin, 1982). Sediments were successively heated to 550 ∘C to quantify ROM, as refractory terrestrial and aquatic OM will likely burn off at this temperature (Kristensen, 1990). About 1 g of dried sediment was precisely weighted into crucibles of known weight for each surface sample (M0). These were ashed in a muffle furnace at 250 ∘C for 4 h, cooled and weighted (M250). Samples were then returned to the furnace at 550 ∘C for 4 h, cooled and reweighted (M550). TOM, LOM and ROM were calculated as follows: LOM=[(M0-M250)/MS]⋅100ROM=[(M250-M550)/MS]⋅100TOM=LOM+ROM, where MS is the sample weight minus the crucible weight. Replicate samples were measure at stations MD15-01A, MD15-01B, MD15-05A and MD15-05B, which resulted in a mean relative error of ±0.07 % for LOM, ±0.06 % for ROM and ±0.03 % for TOM, pointing to very good data reproducibility and representation of local conditions.

Surface sediment samples were analysed to determine bulk elemental C and stable isotope (δ13C) values. Each sediment sample was freeze-dried and homogenised, and approximately 12 mg of milled sediment was placed into silver capsules; a further 10 mg was placed into tin capsules.

The samples encapsulated in silver underwent an acid fumigation step (Harris et al., 2001) to remove inorganic carbon (IC). After drying for 24 h at 40 ∘C, both OC and δ13C were measured using an elemental analyser coupled to an isotope ratio mass spectrometer (IRMS) at the NERC Life Science Mass Spectrometer Facility (Lancaster, UK). The standard deviation of δ13C triplicate measurements was 0.07 ‰, and δ13C values are reported in standard delta notation relative to Vienna Pee Dee Belemnite (VPDB).

The samples in tin capsules were analysed for total carbon (TC) using an Elementar elemental analyser (EA) at the School of Geography and Sustainable Development, University of St Andrews (Verardo et al., 1990). The precision of the analysis is calculated based on repeat measurements of standard reference material B2178 (medium organic content standard from Elemental Microanalysis, UK) with C = 0.08 %.

These results were combined together to calculate the quantity of IC in each sample; the percentage of OC was subtracted from the percentage of TC to obtain the percentage of IC.

To discriminate between marine-sourced and terrestrially sourced OC, we used a two-endmember (binary) mixing model assuming a two-point source of OC, marine and terrestrial, following Thornton and McManus (1994). We used δ13C measurements in the sediment samples as a tracer of the source of OC in west Shetland voes and calculated the fraction of terrestrially sourced OC (OCterr) as follows: OCterr=(δ13Cmar-δ13Cterr)/(δ13Cmar-δ13Csample)OCterr+OCmar=1, where the endmember value for marine-sourced OC (δ13Cmar) was taken from Smeaton and Austin (2017) based on phytoplankton, zooplankton, macroalgae and benthic microalgae carbon stable isotope values, while the endmember for terrestrially sourced OC (δ13Cterr) was based upon the work of Thornton et al. (2015), who studied the distribution of carbon stable isotope in Scotland's topsoil; OCmar indicates the fraction of marine-derived OC. We assumed no isotope discrimination between the source and the surface sediment samples.

Aliquots of surface sediments were stained with rose bengal to allow for the identification of living foraminifera, as detailed in Schönfeld et al. (2012). However, it should be noted that foraminiferal counts are “total” (live + dead) because the main objective of this study is to provide a tool for the interpretation of fossil foraminiferal assemblages and their relationship with changes in OM and OC content in sediments over time (e.g. Conradsen, 1993). Our priority is to understand how benthic foraminifera respond to prevailing (long-term) environmental conditions rather than to seasonal variability. Additionally, scraping the top layer (∼1 cm thick) of each grab with a domestic spoon may lead to a potential underestimation of living fauna if mixing of the top layer occurred during sampling. In this scenario, the number of living foraminifera at the sediment surface will be “diluted” due to the presence of dead and/or fossil specimens from deeper sediments. Having said this, a recent study by Rillo et al. (2019) reported that historical sediment samples collected in a way that could have caused disturbance of the sediment surface (sounding and dredge) are still representative of surface conditions, and their foraminiferal assemblages can be used to reconstruct environmental changes reliably. At two sites, MD15-01 and MD15-05, replicate samples were taken (MD15-01A and MD15-01B; MD15-05A and MD15-05B, respectively) to check measurement reproducibility and seafloor heterogeneity.

As part of an unpublished work, rose-bengal-stained surface sediments were wet-sieved over a 63 µm mesh; the residues were oven-dried at <60 ∘C, then weighed and dry-sieved into 63–150 µm and >150 µm size fractions. As part of an associated work aiming to understand the influence of different size fractions on the composition of benthic foraminiferal assemblages (Lo Giudice Cappelli and Austin, 2019), both size fractions were analysed independently, and the results demonstrated that a more holistic picture of environmental change is obtained when benthic foraminiferal assemblages of both size fractions are compiled together. This improves the representation of the entire assemblage and results in more statistically robust environmental reconstructions. Similar findings were also reached by Weinkauf and Milker (2018, and references therein). Depending on sample volume, each sample was divided into a number of splits using a standard dry splitter and, when possible, at least 300 specimens were dry-picked to ensure statistical significance when discussing benthic foraminiferal assemblage diversity and composition (Schönfeld et al., 2012). Species were identified following Austin (1991) and counted using a tally sheet; notes on tests preservations were made when signs of etching were observed.

Relative benthic foraminiferal abundances were calculated for each sample. It should be noted that we grouped under the name E. excavatum both forma selseyense and forma clavata despite them being recognised as genetically different species (Darling et al., 2016). It was not possible to consistently identify and discriminate between these two species; hence, we used a common identifier.

The Palaeontological Statistics software package PAST (version 3.16; Hammer et al., 2001) was used to analyse relationships between benthic foraminiferal assemblages through cluster analysis, canonical component analysis (CCA) and non-metric multidimensional scaling (MDS) based on the Bray–Curtis similarity index and relative abundance data. Cluster analysis showed good reproducibility of benthic foraminiferal assemblages (Fig. 4b) with the two pairs of replicate stations exhibiting assemblage reproducibility of 90 % (MD15-01A and MD15-01B) and 94 % (MD15-05A and MD15-05B).

We observed a total of 64 species of benthic foraminifera at the 21 stations analysed in this study, including the two replicates samples (Fig. S2). Nine of the observed species are agglutinated: Ammoscalaria runiana, E. scaber, Connemarella rudis, Haplophragmoides sp., Reophax fusiformis, Reophax scotii, Spiroplectinella wrightii, Textularia earlandii and Trochammina sp. Of these, only E. scaber and Trochammina sp. were observed at all 21 stations. However, the E. scaber relative abundance varies from 0.4 % in Vaila Sound (MD15-20) to 43.0 % in Sand Sound (MD15-10) (Fig. 5), whereas the Trochammina sp. relative abundance reaches a maximum of 2.3 % in Clift Sound. Other common agglutinated species are Haplophragmoides sp. and S. wrightii, with a relative abundance ranging from 0.1 % (Vaila Sound) to 2.9 % (Busta Voe) and 0.1 % (Olna Firth) to 3.9 % (Clift Sound), respectively. The distribution of the remaining agglutinated species is patchy, and their relative abundance varies around the average value of 1 %.

A total of 6 of the 64 observed species belong to the porcelaneous group Miliolina: Cornuspira sp. Miliolinella subrotunda, Quinqueloculina bicornis, Quinqueloculina seminula, Quinqueloculina sp. and Spiroloculina rotunda. Of these, the most abundant species is Q. seminula, which is present in low numbers in five of the six studied fjords; it does not occur in Aith Voe. The general distribution of Miliolina is patchy, and their relative abundance varies around the average value of 0.5 %, with the highest value of 2.4 % observed at the mouth of Clift Sound (MD15-05B).

The remaining 49 species are hyaline foraminifera of which 9 species have relative abundances of >10 % for at least one of the 21 stations studied: Ammonia spp., Bulimina marginata, Buliminella elegantissima, Cibicides spp., Elphidium gerthi, Elphidium margaritaceum, Elphidium excavatum, Rosalina spp. and Stainforthia fusiformis (Table 2). Additionally, of all the hyaline species, 20 are rare, having relative abundances <1 %, while 18 species are frequent, having a relative abundance between 1 % and 5 %, and only 2 species are common, having relative abundances ranging between 5 % and 10 % (Fig. S2).

Overall, 10 taxa of benthic foraminifera were observed in each of the six voes and had relative abundances of more than 10 % for at least one of the 21 stations studied (Table 2): Ammonia spp., Bulimina marginata, Buliminella elegantissima, Cibicides spp., Eggerelloides scaber, Elphidium gerthi, Elphidium margaritaceum, Elphidium excavatum, Rosalina spp. and Stainforthia fusiformis. These 10 taxa were categorised as dominant and imaged with a scanning electron microscope (SEM) at the Scottish Oceans Institute, University of St Andrews (Fig. 5). Altogether, these 10 dominant taxa account for more than 70 % of the total assemblage at each station (Table 2). In general, the most abundant taxa observed in west Shetland's six voes is the hyaline Cibicides spp., followed by the agglutinated E. scaber (Table 2 and Fig. 5).

The particle size data from the voes of Shetland exhibit a large range with the silt and sand fractions but little variation in the proportion of clay (Fig. 3a). These samples suggest a significant size sorting, which is consistent with the tidal and estuarine dynamics of these voes. In Vaila Sound, intensified bottom currents and winnowing of finer sediments owing to the influx of Atlantic Ocean waters (Cefas, 2009) appear to control local grain size distribution (Fig. 3a). Here, stations exposed to the open sea and intensified bottom currents are characterised by coarser sediments (MD15-20 and MD15-21), whereas more sheltered and calm locations (MD15-18 and MD15-19) are characterised by finer sediments (Fig. 3a). In contrast, sediments in Clift Sound are silt-dominated (Fig. 3a) despite the unrestricted geomorphology (Fig. 1) and vigorous bottom currents (Cefas, 2007a). It appears that a mechanism other than winnowing of finer sediments by intensified bottom currents drives particle size distribution here. In Clift Sound, the overall low percentage of OM, together with the relatively high IC content and δ13C values, points to a reduction in terrigenous input as the main driver of particle size distribution (Fig. 3). Conversely, in Sand Sound, an increased terrigenous input appears to be responsible for the coarser sediments found at stations MD15-06, MD15-07 and MD15-09 (Fig. 3). The head of the voe is a sheltered area that receives freshwater from a significant number of rivers draining the surrounding land (Cefas, 2007b, 2008b), resulting in an overall high percentage of OM and OC, low δ13C values (Fig. 3c and d), and coarser sediments (Fig. 3a). In contrast, station MD15-08 at the head of Sand Sound is characterised by finer sediments (Fig. 3a) despite the high terrigenous supply at this location (Fig. 3c and d). It is possible that, due to the restricted geomorphology of Sand Sound (Fig. 1), weak currents are unable to transport the coarse material to station MD15-08, the deepest station at the head of the voe (11 m vs. 6 m). In the Sand Sound inner basin (MD15-10), silty sediments dominate (Fig. 3a). Here, a combination of reduced terrigenous input and/or weaker currents could be driving particle size distribution. Nørgaard-Pedersen et al. (2006) found similar drivers of particle size distribution in Loch Etive, a sea loch in mainland western Scotland. Vigorous bottom currents and/or enhanced terrigenous supply would result in locally coarse sediments, whereas weaker currents and/or limited terrigenous supply would result in finer sediments. Due to the duality of drivers determining particle size distribution in west Shetland voes (bottom current intensity vs. terrigenous supply), it is challenging to consistently group locations based on granulometry data (Fig. 3a) or identify a relationship between grain size distribution and OM and OC content in the sediments (Fig. S3).

A clear pattern is evident with regard to the quantity, quality and source of organic matter and carbon measured in the six west Shetland voes (Fig. 3). Typically, carbon is present in the sediments as IC and OC, and the distributions of these two forms of carbon tend to mirror one another, with low IC and high OC percentages towards land, while the reverse occurs close to the open sea (Fig. 3d). Organic matter follows the same pattern of OC distribution regardless of its quality (refractory vs. labile); however, ROM content in the sediments is always higher than LOM (Fig. 3c) owing to its resilience to degradation.

About the source of OM and OC, low δ13C and high OCterr values characterise the head of the voes where high ROM and OC are accumulated, indicating that the source of this material is terrestrial (Fig. 3). Terrestrial OM is more stable and resistant to decay than marine OM (Batten, 1996) and it is enriched in 12C from soils, resulting in lower δ13C values than marine OM derived from algae (e.g. Balasse et al., 2005; Ficken et al., 1998; Marconi et al., 2011; Schiener et al., 2014; Schmidt and Gleixner, 2005). Sand Sound is dominated by organic-rich and terrestrially sourced material as it receives freshwater and terrestrial OM from the numerous streams draining the surrounding land (Fig. 3). In contrast, Clift and Vaila Sound are characterised by marine-sourced material and high IC content in their sediments (Fig. 3) as the inflow of Atlantic Ocean waters improves the voes' potential for IC storage (Burrows et al., 2017; Smeaton et al., 2016, 2017). Olna Firth has the highest LOM percentage as this relatively deep and micro-tidal basin (Cefas, 2013) is prone to water mass stratification, which can lower bottom water oxygenation and slow down the reworking or remineralisation of OM, favouring the preservation of the more labile material (Fig. 3).

A slight deviation from this overall trend was recorded at station MD15-05. We speculate that the unusually high TC content at station MD15-05 at the mouth of Clift Sound (Fig. 3d) could relate to the MV Braer oil spill of 1993. It appears that after the spill, due to the adverse weather, the oil became deposited in the sediment of Burra Haaf, just west of Clift Sound (Fisheries Research Services). Despite the fact that oil levels in the contaminated sediments have been decreasing steadily over the intervening years, it is plausible that the OC measurements in this study detected remnants of this event. Nevertheless, the benthic foraminiferal assemblage of this location is dominated by Cibicides spp. (see Sect. 4.3.4 for more details), which does not reflect the oil spill or the high OC content, leaving an open question regarding what drives OC deposition and accumulation at the mouth of Clift Sound.