Step 1: Identifying a constrained target area within the Celtic Sea
Given the total area of the Celtic Sea, it was necessary to focus operations on a constrained area that is representative of the Celtic Sea, and the UK Shelf as a whole. The rationale for the selection of this broad target area is based on the identification of varied habitats typical of different sediment types (ranging from fine cohesive muds to coarse advective sands) that exhibit: different biogeochemical exchange mechanisms; varied faunal abundance, diversity and function, while staying within a similar hydrodynamic environment. Confounding variables are reduced by adopting a narrow range of depth, temperature and hydrographic variations. To make this selection, a full assessment of the typical conditions within the Celtic Sea is necessary.
The Celtic Sea extends from the shelf-break at approximately 200 m depth, to a narrow, steep coastal zone. The inner shelf (Fig. 1b) comprises depths between 70–120 m (Uncles and Stephens 2007), and is generally featureless, with a more irregular outer shelf deeper than 120 m. Tides are predominantly semi-diurnal (e.g., Robinson 1979), and the mean spring tidal range increases from approximately 3 m close to its South Western boundary near the shelf break to >12 m in the Upper Severn Estuary in the upper reaches of the Bristol Channel (Hydrographic Office 1996). Spring tidal speeds are relatively low, typically 0.2 m s−1 close to the seaward boundary, but increasing to 1.6 m s−1 in the Bristol Channel (Uncles and Stephens 2007). Tidal ellipses tend to be strongly elliptical with a clockwise rotation, apart from a localised region of circular ellipses with anticlockwise rotation west of the Bristol Channel (Robinson 1979; Brown et al. 2003; Simpson and Tinker 2009). Tidal ellipses also become more rectilinear as you approach the English Channel. Highly elliptical tidal currents allow for a constantly elevated bed stress, while their polarity influences the height of the bottom boundary layer (e.g. Simpson and Tinker 2009). Bed shear stresses are typically <0.5 Nm−2 within the central regions (Fig. 2) increasing towards the shallower English and Bristol Channels to the East and the Irish Sea to the North.
Winds are predominantly from the South West or West, and wave conditions change as the sea becomes shallower and more sheltered. 10-year mean significant wave heights vary from 2 m (8 s peak wave period) near the shelf break to 1 m (6 s peak wave period) where the Celtic Sea meets the Irish Sea, while extreme values for a return period of 1 year reach significant wave heights in excess of 8–10 m and peak periods of approximately 15 s (Bricheno et al. 2015).
Water column conditions
Mean winter bottom temperatures are typically 9–10 °C, increasing to 11–16 °C in summer (Uncles and Stephens 2007; Brown et al. 2003). Salinity exceeds 35 near the shelf edge, reducing slightly toward the coast, and varies little seasonally. Winter mixing of the water column in the Celtic Sea leads to a well mixed water column, which is reflected in a homogenous temperature profile between surface and deeper waters. A weak thermocline develops in springtime, which inhibits full water column mixing, providing suitable conditions to initiate a spring bloom (Simpson and Sharples 2012).
Spring blooms in the region are typically dominated by diatoms, which account for up to 80% of primary production during this period (Joint et al. 1986). During the summer months, surface waters become nutrient poor and therefore lacking in phytoplankton. However, the development of a summer deep chlorophyll maximum positioned at the base of the thermocline in the vicinity of the nutricline (Pingree et al. 1977; Hickman et al. 2009) is a well-known phenomenon. Smaller-celled phytoplankton tend to dominate here due to competition for nutrients and include prymnesiophytes, pelagophytes and the cyanobacteria Synechococcus (Hickman et al. 2009).
The wider Celtic Sea area contains sediment types ranging from pure muds to gravels (Fig. 1): sediments typical of a shelf-sea environment (bedrock is excluded from the sediment coverage model presented [Stephens and Diesing 2015], however, this has little impact on the project as it’s contribution to biogeochemical cycling is minimal in the UK shelf setting). To ensure a narrow range of depth, temperature and hydrographic variations, a contiguous target area within the inner shelf region of the Celtic Sea was selected with minimal bathymetric variation (Fig. 3b), high hydrodynamic and water column similarity, but also encompassing the widest possible range of seabed types (Fig. 3a).
Within this selected target area, the sediments are dominated by muddy sands, sand, and gravelly sands (comprising 92% of total sediment coverage; Table 1), which typify the wider Celtic Sea region (88% total sediment coverage). The average water depth across the target area is 95 m below chart datum.
Large scale commercial fisheries expanded comparatively recently in the Celtic Sea, but have had a relatively large and consistent impact on the area (Blanchard et al. 2005). Fishing activities tend to focus on specific areas (Sharples et al. 2013), targeting the Celtic Deep, shelf edge, and to a lesser extent the central Celtic Sea region (Fig. 4), where trawlers target the Norway lobster Nephrops norvegicus on muddy grounds. Fishing occurs year-round at the Celtic Deep (with a slight reduction in Jan-March), although a seasonal pattern is seen in more central regions, with the bulk of activities taking place in spring and summer (Sharples et al. 2013). Vessel Monitoring System (VMS) data from between 2009–2014 suggests a differing trend in fishing ground preferences within the Celtic region when split by UK and non-UK vessels (Fig. 4), likely driven by differences in gear preference, target species, regulations, and fuel prices (Jennings et al. 2012).
Step 1 summary
The selected target area provides a constrained region on the inner shelf of approximately 87 × 95 km (8265 km2) within which to limit long-term observational measurements, cruise operations and in situ experimentation. This restricts sampling to an area of minimal topographic and depth variation, away from the shallower coastal regions where bed stresses are higher, and increasingly varied, and away from freshwater inputs which would affect salinity and temperature. The area contains a wide range of sediment and therefore habitat types, and minimises variations in depth and regional hydrodynamics. To further limit potential depth and hydrodynamic variations, an approximately 20 km wide transect running from the south-west to the north-east across this region (following the tidal flow and predominant wave directions) was identified. The same selection conditions were met, but the required coverage was reduced to an area of approximately 2500 km2. The next step was to make a full assessment of the spatial heterogeneity within this new, limited, target area and select discrete sampling sites suitable for repeat seasonal sampling, and representative of the dominant habitat types and biogeochemical exchange mechanisms of the shelf.
Step 2: assessments of spatial and temporal heterogeneity within the target area and implications for benthic habitats
The main observational and experimental work for the Shelf Seas Biogeochemistry programme was carried out during 2014–2015. At the start of this cruise programme, a series of benthic landers and SmartBuoys were deployed within the target area to measure long-term hydrodynamic conditions during the survey period (Fig. 5; Online Resource 2).
Four benthic Landers were deployed at The Celtic Deep 2 (CD2L) and East of the Celtic Deep (ECD) both to the North of the region, Nymph Bank (NB) in the central region and East of Haig Fras (EHF) to the South in areas which have similar hydrodynamic regimes (depth, temperature, current direction), but a range of bed types. Consideration was also made to existing infrastructure: a SmartBuoy has been located at the Celtic Deep (CD) site since 2009, and was moved to Celtic Deep 2 (CD2) in 2012. In addition, a SmartBuoy was located at the shelf edge (Candyfloss) for assessments of shelf exchanges and links to the pelagic component of the SSB programme (http://www.uk-ssb.org/science_components/work_package_1/).
Measured tides in the target area (Fig. 6) are dominated by the M2 tidal constituent, followed by S2 and N2 constituents resulting in semi-diurnal tides with significant spring-neap variations (Robinson 1979). Total spring and neap amplitudes reach 3.1 and 1 m, respectively, at CD2L (Fig. 6a), reducing in the south to 2.9 m springs at EHF, and increasing to the east to 3.4 m springs at ECD consistent with the wider shelf area. Measured near-bed currents are also summarised in Fig. 6(2). While there is little difference in the lowpass current magnitude, the maximum spring currents are strongest at EHF (mean maximum spring current approximately 0.4 m s−1), followed by CD2L and ECD (0.36 m s−1) and weakest at NB (0.32 m s−1). There is a similar behaviour for the maximum bed shear stress (mean spring maximum value of 0.60 Nm−2 at ECD, 0.48 Nm−2 at ECD and CD2L, and 0.37 Nm−2 at NB), but the minimum bed shear stress is significantly higher at ECD (0.02 Nm−2 vs. zero at the other three locations) resulting in an increase of the mean bed shear stress. The tidal ellipses also vary from near circular ellipses at ECD to near rectilinear at EHF matching the expected behaviour of the wider Celtic Sea region, with the polarity of the ellipse anti-clockwise for ECD, CD2L and NB, but clockwise for EHF.
Mean daily wind speeds between 2012 and 2015 were 8.1 m s−1, with a maximum of 22.9 m s−1. There is a strong seasonal signal, with daily mean values of 6.5 m s−1 during the summer, and 10.3 m s−1 in winter. The M5 Wexford coast wave buoy shows winter waves have a mean height of 2.3 m with a maximum recorded height of 8.1 m in January, and summer mean wave height of 1.4 m.
Water column conditions
Measured surface temperatures since 2009 ranged between 8.06–19.73 °C (mean 13 °C). Stratification formed in early April in both 2014 and 2015, with re-mixing in mid-December in 2014. This is in keeping with prior observations (Brown et al. 2003). CTD data indicate that the mixed layer depth was shallowest in August (~25 m), deepening from September. Surface temperatures during the sampling period were typical of the overall temperature range in the Celtic Sea, with bottom temperatures limited to ~12 °C (Fig. 7a), reaching a maximum following re-mixing during the winter months, and also closely following the trend for the wider Celtic Sea region. Salinity has a narrow range between 34.8 and 35.3 as expected for this inner region of the shelf. Riverine input from the southern coast of Ireland is relatively minor. Freshening during winter and spring is thus primarily attributable to input from the River Severn (Brown et al. 2003). Profiles of PAR allow calculation of vertical attenuation coefficients (Kd; Kirk 2003) between 0.1 and 0.25 m−1 in Summer and Autumn, also typical of offshore shelf waters (Foden et al. 2008). Water clarity reaches higher values in summer (ranging from 0.13 and 0.9 m−1) and is limited in range in winter (0.2 and 0.4 m−1).
The timing of the thermal stratification observed was supported by water column macronutrient profiles collected from CTD deployments over the course of both pelagic and benthic SSB field campaigns (Fig. 7b). During winter months the water column is completely mixed with total oxidised nitrogen (TOxN) concentrations between 6.3 and 6.8 μM at all water depths (March 2015). Similarly, profiles of silicate (range 4.6–5.2 μM) and phosphate (0.56–0.77 μM) demonstrate the homogeneity of the water column at that time. In early April 2015 the onset of stratification and assimilation of nutrients is witnessed with surface concentrations of TOxN depleting to 4.9 μM while bottom water concentrations increased to 7.4 μM. Silicate and phosphate followed suit but depletion was not as pronounced, with surface concentrations at 4.3 and 0.4 μM, and bottom concentrations at 5.1 and 0.6 μM, respectively. By the end of April 2015 once the bloom had successfully established, a strong nutricline is observed between 20 and 30 m. Here, nitrate concentrations have been significantly depleted in surface waters to 0.01 μM, whilst bottom water concentrations have increased further to 10.6 μM. Depletion of surface silicate (0.3 μM) and phosphate (0.01 μM) is also witnessed with elevated concentrations of 5.7 and 0.8 μM, respectively, found at depth. These nutrient conditions are observed throughout the late spring/summer period until the nitrate and phosphate surface water concentrations are further depleted, falling below detection limits (Woodward and Rees 2001). This highlights the biological drawdown of nutrients from the surface waters and probable remineralisation of organic matter at depth, combined with the absence of water column mixing during this period.
Data from SmartBuoys show that phytoplankton blooms are variable in both timing and magnitude in the region, usually occurring in March or April. In 2011, peak Chlorophyll concentrations occurred in March, reaching 16 μg L−1. During the SSB survey period, maximum Chlorophyll peaks were lower (3–4 μg L−1) and occurred later in the season. Moderate Resolution Imaging Spectroradiometer (MODIS; NASA) satellite data demonstrate that the spring bloom was initiated in early April 2015 coinciding with the onset of stratification, with full bloom conditions observed by mid-April 2015 (Fig. 7c). The bloom lasted for approximately four weeks before crashing by mid-May. During the summer months when surface waters were nutrient poor, the phytoplankton population was reduced.
During March 2015, a broad-scale spatial benthic survey was completed to assess the heterogeneity of the sediments within the previously defined target area (Fig. 8). At each sampling location NIOZ box cores were collected and subsampled for particle size, bulk sediment characteristics (bulk density, porosity, permeability and organic content), oxygen and pH profiles, pore-water nutrient concentration profiles and meio- and macro- faunal assessment (see Online Resource 1 for full methodologies). SMBA cores were taken for measurements of megafaunal abundance and assemblage. SPI images were collected for visual determination of sediment type, zone of mixing (previously the apparent redox potential discontinuity [aRPD]; Teal et al. 2010) and bed roughness.
The full results of the survey will be reported in detail elsewhere (e.g. McCelland et al. 2016; Silburn et al. in prep), and confirmed that the targeted area contained a range of sediment types from sandy muds, through to gravelly sands, reflecting the wider shelf region (For full details, see Online Resource 3). In summary, coarser sediments dominate the central region, and the percentage of fine sediments (median grain size < 63 μm), which ranges between 1.73 and 86.61% across the entire area, increases towards the Northeast and Southwest corners (Fig. 8). Multivariate statistical analysis of particle size data suggested that the sites could be allocated to one of eight different seabed types that corresponded well to the Folk and Ward (1957) textural group classifications for sediment bed types. The majority of the samples (92%) were poorly to very-poorly sorted, fine to very-fine skewed (80%) and mesokurtic to very leptokurtic (96%). When overlaid on the targeted area it is clear that the sediment coverage map is successful at representing the range and spatial distribution of surface sediments in the Celtic Sea.
Faunal analysis of the spatial survey samples demonstrated that sediment particle size distributions were generally a good predictor of macrobenthic community structure (McClelland et al. in prep). However, there was considerable overlap in community composition between closely related sediment types. This was due principally to many benthic species present having broad habitat preferences occurring in multiple sediment habitats. In addition, despite changes in community composition between sediment types, levels of macrofaunal abundance, biomass and diversity remained largely constant across all the samples with perhaps only a slight reduction in these parameters for the sites with the highest fines percentages to the Northeast (McClelland et al. 2016). Given that these sites were also subjected to the greatest intensity of trawling, this slight reduction may be due to anthropological disturbance rather than to any natural ecological process.
Step 2 summary
The spatial survey demonstrated that the target area contains a wide range of benthic sediment and habitat types typical of the wider Celtic Shelf region, while being exposed to minimal variations in water depth, water column conditions and hydrodynamic forcing spatially, which all fall within the ranges expected of the wider Celtic Sea area, but exhibit clear seasonal changes.
Step 3a: identify and describe exemplar sites; Physical Parameters
Final site selections were made based on the sediment maps and past cruise data presented above, and were further refined using ground-truthing during the first SSB cruise in 2014 (Table 2), and the spatial survey in 2015. Based on the sediment coverage data, four final process sites were selected within the targeted area, which represent the overall range of habitat and sediment types within the region, ranging across the end-member biogeochemical exchange mechanisms (diffusive and advective). Discounting the gravel dominated sediments, due to the practicalities of using the proposed experimental methods on gravels, there are four main sediment types across the target area: mud; sandy mud; muddy sand; and sand. Pure mud is of negligible coverage (0.005%) and so the sites chosen are a sandy mud (with as low a sand fraction as possible) to represent the diffusive end member, a sand sediment to represent the advective end member, and two muddy sand sites in between.
Each process site is represented by a 0.25 km2 box (500 m × 500 m) within which sampling is constrained, minimising local heterogeneity while ensuring sufficient space to resample the sites without on-going impacts from previous sampling efforts. Process site names represent the order in which they were ground-truthed and are presented according to decreasing fines percentage. The boxes with the highest percentages of fines (A) and sand (G) are used to represent the end-members of the observed spectrum, with the sites H and I displaying intermediate values on the continuum.
The full benthic Shelf Seas Biogeochemistry programme visited each site four times, to assess seasonal differences across each of the sites, and assess conditions prior to, during and after the spring bloom (Table 2). Much of this seasonal data is presented in full within the other contributions to this special issue.
These cruises used a combination of in situ observation, sediment and biological sampling and experimentation to make assessments of biogeochemical processes occurring at each of the sites. While site selection was based on data collected in DY008 and DY021, the data presented below represent typical values averaged over all four cruises, to provide baseline ranges throughout the year for each site, providing the most thorough assessment of site representativeness to the wider target area and Celtic Sea region.
Water column conditions
The long-term Lander data can be used to assess the hydrodynamic conditions occurring at the process sites (Table 3), to confirm whether the confounding variables were well constrained. The average water depth of the four sites is 106 m, and between site variation less than 10% of the total average water depth. This is confirmed by Autosub3 collected bathymetry data (Online Resource 4). Bottom temperatures over the sampling period average 9.76 °C, varying within 5% between sites; salinity was 35.2 (<1% variation between sites). Significantly different spatial variations in turbidity (standard error of the mean; p < 0.0001) and O2 saturation (p < 0.0001) are apparent which, given the water column similarities between the sites, likely result from differences in the bed sediment or habitat type. Turbidity is highest at ECD, which also corresponds to the highest O2 saturation.
Underway and Lander measured Chlorophyll concentrations indicate that the spring bloom occurred concurrently across the sites, were in agreement with the MODIS satellite data for the Celtic Sea in 2015, and closely correlates with the onset of stratification. The bloom results in similar drawdowns of CO2 (Fig. 9b) at each site.
Sidescan surveys were undertaken as part of DY034 using Autosub3 (Fig. 10; Online Resource 5). These encompass the immediate process sites (500 × 500 m black boxes), plus the surrounding areas. High backscatter (light tones) likely represents area of coarser or more mixed sediments, whereas low backscatter (dark tones) finer or more homogeneous sediments. The presence of bedforms at Site G is clear, reducing in wavelength towards the north of the region (from ~130 to ~25 m). These also appear in the bathymetry data collected at site G (Online Resource 4). Presumed ‘trawl marks’ are particularly evident at Site A, but also present at sites I and H.
SPI images (Fig. 11) from the four process sites show clear visual differences in grain size, surface roughness and sediment colour indicative of different sediment and habitat types. Photographs from the Autosub3 survey were used to visually distinguish between habitat types and could be divided into three broad categories: hard (Fig. 11a: >50% of the photograph covered by cobbles or boulders); intermediate (Fig. 11b: 1–49% coverage of granules, cobbles or boulders); and soft (Fig. 11c: 100% coverage by sand or mud). Particle Size Analysis (PSA) of multiple sediment samples taken from NIOZ box cores over the 4 cruises (Table 4) confirms that the differences between mean values at each site are statistically significant.
The four sites exhibit statistically different averaged median grain sizes (standard error of the mean; p < 0.005), although H and I fall into the same textural classification (Table 4). In summary: site A is a very poorly sorted, very fine skewed, mesokurtic, very coarse silt, classified according to the Folk classification scheme as a sandy mud; site I is a very poorly sorted, very fine skewed, leptokurtic very fine sand, classified as a muddy sand; site H is a very poorly sorted, very fine skewed, leptokurtic fine sand, also classified as a muddy sand; and, site G is a poorly sorted, fine-very fine skewed, very leptokurtic medium sand.
The structure of the near-bed sediment (top 5 cm) was also assessed for each of the sites (Table 4). Depth averaged dry bulk densities are statistically different between sites (p < 0.005), with the exception of H and I (p = 0.48). Porosity and permeability are significantly different in all cases (p < 0.020 and p < 0.001 respectively). As expected, bulk density and specific permeability both increase with median grain size, while porosity decreases.
Small-scale seabed topography is provided from acoustic images of the bed measured by the 3D Acoustic Ripple Profiler (ARP) on the miniSTABLE intra-tidal monitoring lander. Results for the four sites show a variation in bed height of up to 4 cm (Fig. 12). Bed structures at the more cohesive sites (A, H and I) appear to be dominated by circular depressions, probably caused by benthic fauna. Ripples are observed at the sandy site with little if any migration in all cases. These ripples are predominantly two-dimensional in March and May with ripple height approximately 2–3 cm and ripple wavelength approximately 20-30 cm, and three-dimensional in August with height approximately 1 cm and wavelength approximately 15 cm. The footprint of the ARP is too small to capture the larger scale (~30 m) bedforms seen in the sidescan data. Surface roughness (measured from SPI images; e.g. Fig. 11) is similar at all the muddy sites, and only significantly different at G (p < 0.05), as confirmed from the acoustic bed roughness measurements presented above (Fig. 12).
Step 3a summary
The analysis described confirms that the four process sites can be considered as statistically different from each other in terms of the sedimentary characteristics (a key scientific variable of the SSB programme), showing a clear and concurrently occurring seasonal signal (key variable), while being similar in terms of hydrodynamic parameters (confounding variables).
Step 3b: identify and describe exemplar sites; biological and biogeochemical parameters
Assessments were made of key biogeochemical and biological parameters (Tables 5, 6), measured over all four cruises, providing typical ranges found at each site.
Sediment total organic carbon and total organic nitrogen content are both highest at site A, intermediate at H and I, and lowest at site G. These differences are significant (standard error of the mean; p < 0.05) in all cases, except for organic nitrogen between H and G. These trends are maintained with similar magnitudes when considered seasonally, except for site G, where the core used for analysis had a much higher fines content than typical for this site. Oxygen penetration depths are significantly different only between I and G, although total oxygen consumption rates are significantly different in all cases except between I and H. It should be noted, however, that total oxygen consumption rates are calculated based on the combination of data from three different analytical methods providing total oxygen uptake rates, diffusive oxygen uptake rates and oxygen penetration depths, and are discussed in more detail in Hicks et al. (2017) and Smith et al. (in prep). There are both site and seasonal differences, with more noticeable changes in the cohesive sites, and greatest O2 consumption nearest the spring bloom. These seasonal signals are discussed further in Hicks et al. (2017). Chlorophyll measured in the surface sediments at A is significantly higher than the other three sites (p < 0.001), and significantly lower at G than at I (p < 0.05). The zone of mixing is significantly different at all sites (p < 0.05) being lowest at H, and highest at A.
Pore water nutrient concentrations were measured in triplicate usually down to 20 cm using a depth variable resolution. Data averaged for 0–10 cm are presented (Table 5). The concentration of NH4
+ ranges between 0.23 and 145 μM across all sites and cruises. The concentrations at Sites A, H and I generally increase from the sediment surface to 10 cm depth, and are relatively stable below 10 cm (Fig. 13). At Site G, increases do not occur until below 3–4 cm depth. Silicate profiles show similar trends as the NH4
+ with higher concentrations (3–368 μM).
TOxN is usually at a maximum in the top 2 cm except at Site G where values at depth are occasionally higher than at the surface, with a maximum value of 16.6 μM. Nitrite ranged between 0.07 and 8.27 μM and is generally evenly distributed throughout the top 20 cm. The differences between sites are not statistically significant, however, this is likely due in part to large ranges resulting from measurements averaged over the different seasons (e.g. Fig. 13a). Ranges were similar to those measured over the spatial survey described in step 2 above (Fig. 13b) and therefore considered representative of the region as a whole, and the inherent variability in the profile shapes, likely due to high variability in the vertical sediment structure, should be noted.
Typically, porewater Fe concentration maxima occur in the shallow subsurface (up to >100 μM at approx. 5 cm depth) and decrease sharply across the oxic surface layer (profiles not shown, see Klar et al. this issue). Average surface (0–2 cm depth) porewater Fe concentrations are highest at site I, lowest at site H and intermediate at site A (Table 5). Most of the porewater Fe is in its reduced and soluble Fe(II) form, and our data suggests that oxygen penetration depths (which can be related to e.g. advective transport, bioirrigation or bioturbation) exert a strong influence on pore water Fe contents across the study sites (Klar et al. this issue). Seasonal variations are discussed in detail in Klar et al. (this issue).
Diffusive nutrient fluxes
Ten centimetre diameter sediment sub-cores were collected from the NIOZ cores and incubated with overlying bottom water to assess fluxes of TOxN and nitrite, ammonia, silicate and phosphate in the absence of direct flow forcing (herein termed ‘diffusive’) using two similar sampling methods (Trimmer et al. 2005; Mayor et al. 2012; Main et al. 2015). Sub-samples taken from the overlying water provide a time-series of nutrient exchange, and data presented here are combined from between 5 and 11 cores spanning all three SSB cruises that took place in 2015 (Table 5, Online Resource 1, Online Resource 6). Fluxes are stated with reference to the sediments (i.e. a negative result indicates removal from the water column overlying the sediment). Where there is no measurable change in nutrient concentrations, the flux is quoted as zero. Data are not corrected for water column controls (overlying bottom water in the absence of sediments).
On average, the fluxes of all macronutrients are positive, indicating a general release of macronutrients from the sediments into the water column. However, both negative and positive nutrient fluxes are measured at all sites, except for silicate fluxes at site A, which were consistently positive (0.206–3.741 mmol m−2 d−1). The range of fluxes measured at each site for all nutrients was such that there was no significant difference when considered spatially between sites. Both nitrite and TOxN fluxes are lowest on average at site A and increased through sites I and H, with the highest average fluxes at site G. The greatest range in nitrite and TOxN fluxes are at site H (−0.035 to 0.132 and −0.586 to 0.649 mmol m−2 d−1 respectively). The fluxes of ammonium are highly variable at all four sites, and site I is the only one to be negative overall with an average flux −0.003 mmol m−2 d−1. Sites G and H have the highest fluxes of ammonium (>0.04 mmol m−2 d−1) with the greatest range at site H. Silicate fluxes are on average highest at site A (1.212 mmol m−2 d−1) almost double that of the other sites. Site H and I silicate fluxes are very similar with the lowest fluxes at site G (0.531 mmol m−2 d−1). Phosphate fluxes are highest at Site A, which has a negative flux (into the sediment) on average (−0.018 mmol m−2 d−1) and has the smallest range of fluxes compared to the other three sites.
Diffusive iron (Fe) fluxes are positive at all sites ranging from 0.01 to 54.4 × 10−3 mmol m−2 d−1. Averaged across the year, diffusive Fe fluxes are highest at site A (14.4 ± 19.7 × 10−3 mmol m−2 d−1), and 3-times lower at the site with the coarsest sediments, site H (2.70 ± 5.54 × 10−3 mmol m−2 d−1). However, the range in Fe flux calculations is also greatest at site A, and equal to the range across all sites, while the range is smallest at site H. It is important to note that our assessment of diffusive Fe flux requires a simplification of benthic exchange processes. For example, the roles of advection and bioturbation/bioirrigation at these sites are not accounted for directly in the presented results, and yet they can serve to enhance the transport of Fe (e.g. Reynolds et al. in prep).
Variability in biological abundance, biomass and diversity
Large mobile epifauna
Note that some shallow burrowing infauna were also collected, but for clarity all fauna collected in the trawls will be termed as epifauna.
At all sites, epifaunal organisms are rather sparsely distributed (Table 6). Average abundance was highest at site G, although differences between sites are not statistically significant. Average blotted wet weight biomass values are lowest at sites I and H, slightly higher at the site G and highest of all at the site A, with significant pair-wise differences between all sites (p < 0.01) except between A and H or G. Diversity is highest at H, with site G being just a little less diverse. Sites A and I has the lowest epifaunal diversity.
Autosub3 seabed photographs were also analysed to estimate faunal density and biomass during DY034. At the time of survey, near-bottom water column turbidity at Site A prevented the acquisition of useful seabed photographs. All megabenthos and demersal fish were counted, measured and identified to the lowest taxonomic level possible (Table 6; Example images can be found in Online Resource 7). For comparability with trawl-caught megabenthos biomass data, our estimates are scaled to a sampling unit equivalent to trawl catch data (500 m2). Three phyla dominated the three sites: (1) Cnidaria are the most dominant at Site I and H and the third dominant at Site G; (2) Arthropoda is the second dominant at all sites; and (3) Echinodermata is the dominant at Site G and the third dominant at Site H and I.
Mega-infauna (>1 cm)
All sites contain very few large infaunal species with no single sample containing more than a couple of individuals. It is concluded that, due to their low densities, large (>1 cm) infaunal organisms are not a substantial part of the benthic fauna in the study area and that adequate sampling of the benthic fauna is provided by the Jennings trawl (large epifauna) and the 0.08 m2 NIOZ boxcorer (macrofauna).
Macro-infauna (>1 mm)
Macrofaunal abundance is highest at sites I and H. Site A has slightly lower average abundance, significantly lower than H and G (p < 0.05) whilst site (G) has less than 50% of the abundance of the other three sites (p < 0.0001).
In direct contrast to abundance, wet weight biomass (g m−2) is considerably (2–3×) higher at site A than at the other three sites. This indicates that the average body size of macrofauna is larger at site A than at the other three sites.
The average number of species per 0.08 m2 core (a measure of α-diversity) is highest in the intermediate sites H and I, with significantly lower diversity seen at sites A (p < 0.001) and G (p < 0.0001). However, the cores taken at site G are much more variable in terms of species composition and this higher variability in species between replicate samples (β-diversity) meant that the total number of species identified at site G is the same as site I and only a little less than site H. Site A displays relatively low diversity compared to the other sites.
Macrofauna abundance and biomass data were combined with published trait information describing modes of sediment reworking and mobility (Queirós et al. 2013) to calculate the average community bioturbation potential (BPc) for each of the sites following Solan et al. (2004). Whilst BPc is not a direct measure of the process of bioturbation it does provide a theoretical estimate of the potential of a community to biologically mix the sediment. All of the 4 sites display notably low levels of BPc (mean ± standard deviation) with the highest values of bioturbation predicted for the muddy site A (36.70 ± 22.53), followed by site H (30.31 ± 20.33) and site I (25.01 ± 17.70). The lowest levels of predicted bioturbation are for site G (19.11 ± 13.14). However, the ranges are large.
Macrofaunal bioturbation activity was measured through quantification of redistribution of fluorescent particle tracers and absolute changes in concentrations of the inert tracer sodium bromide respectively (following Hale et al. this issue). Activity levels are very low (Fig. 14) across the Celtic Sea shelf compared to other UK shelf areas (Dauwe et al. 1998; Teal et al. 2008), and similar across all sediment types observed. The median (f-SPILmed, typical short-term depth of mixing), maximum (f-SPILmax, maximum extent of mixing over the long-term) and mean (f-SPILmean, time dependent indication of mixing) mixed depths of particle redistribution are presented in Table 6. In addition, the maximum vertical deviation of the sediment–water interface (upper–lower limit = surface boundary roughness, SBR) provides an indication of surficial activity. Bioturbation is heavily influenced by the presence of mobile active species, such as Nephrops norvegicus and Goneplax rhomboides. Bioturbation activity is observed to peak in August with sediment surface mixing occurring to a depth of approximately 8 mm.
Note that only data from the first two cruises (DY008 and DY021) are presented here.
Meiofauna at site A is most abundant with average densities over 800 × 103 Ind m−2 and maximum values of >1200 × 103 Ind m−2. Sites I, G and H are very similar in terms of meiofauna abundance, with average values lying between 550 and 600 × 103 ind m−2, however the differences are significant (p < 0.05). Muddy sediments are known to harbour greater densities of nematodes (Steyaert et al. 1999), the dominant meiofauna phylum with 85.6% (65.3–97.6%) of total abundance, so the high densities at site A are likely a reflection of sediment composition and related interstitial space (i.e. greater porosity in muddy sediments at site A, Table 5) available to meiofaunal organisms. These values lie within the range of densities commonly found in marine subtidal areas (Heip et al. 1985).
In terms of biomass (based on nematodes) site A and I are very similar (1.13 ± 0.35 and 1.14 ± 0.48 g wet weight m−2, respectively; p = 0.97), and G and H are similar (0.68 ± 0.17 and 0.73 ± 0.39 g wet weight m−2, respectively; p = 0.701). As with abundance values, biomass values lie within the ranges observed for European subtidal areas (Heip et al. 1985) with distinct differences between muddy and sandy sediments. All pairwise comparisons between sites A, I and G, H results in significant biomass differences (p < 0.05).
On the phyla level, multivariate meiofauna community structure data is significantly different between sites and seasons (p ≤ 0.01), and, like abundance and biomass, considerable similarity was found for site pairs A and I (p = 0.635), and G and H (p = 0.054), whilst all other pairwise comparisons show significant differences (p ≤ 0.05).
Porosity (Table 4) is a major determinant of microbial biomass, with the highest measurements at site A and the lowest measurements at site G (Fig. 15). Biomass decreases with sediment depth for all except site G.
Bacterial 16S rRNA genes dominate the total microbial assemblages within coastal sediments, with reports of only 2% of 16S rRNA genes affiliated with archaea (DeLong 1992). Our data suggest a higher abundance of archaea in shelf sediments, in all sediment types examined, with little evidence of differences in the ratio of archaeal:bacterial 16S rRNA genes with depth. At site A, 29.7% (±16.5) of 16S rRNA genes are archaeal, and at site I this figure is 35.8% (±15.9), 38.3% (±20.9) at site H and 22.2% (±14.2) at site G; the differences between sites are significant (p < 0.05).
Step 3b summary
Habitat variations across the four sites echo the differences in sediment variation seen within the constrained target area, and confirmed that the process study sites represent significantly different habitats. These differences were also reflected in the bulk biogeochemical properties of the bed, although seasonal variability in pore water concentrations and nutrient fluxes are sufficient to mask spatial variability between the sites.