NatureScot Research Report 1213 - Exploring biological trait analysis and inter-habitat linkages in three seabed habitat PMFs in Scotland

Year of publication: 2023

Authors: James, K., Kamphausen, L., Cunningham, S. & Kent, F.

Cite as: James, K., Kamphausen, L., Cunningham, S. & Kent, F. 2023. Exploring biological trait analysis and inter-habitat linkages in three seabed habitat PMFs in Scotland. NatureScot Research Report No. 1213.

Keywords

Priority Marine Features; habitat linkages; biodiversity; functional traits analysis; Biological Trait Analysis

Background

To date, marine conservation work has focused predominantly on assessing the status of single species or habitats. This approach has obvious limitations, ignoring the interactions between species and habitats which create ecosystems as a whole. While ecosystem functioning has been identified as a more important and more useful indicator of the status of the marine environment, it remains very difficult to quantify and assess.

By exploring the effects of linkages between a small number of Scottish Priority Marine Feature (PMF) habitats this study aimed to explore methodologies which might assist in more systems-based nature conservation work.

Interactions exist between organisms over many trophic levels. Interactions can occur directly, for example, through predation or competition, or indirectly, for example, where the modification of the environment by one species favours the presence of another (Yakovis & Artemieva, 2017; Passarelli et al., 2014; Trussel et al., 2006) or when species move between habitats (for example, the movement of fish between habitats such as kelp forests and adjoining sedimentary habitats). Linkages between habitats can promote biodiversity (Angelini et al., 2015; Tews et al., 2014), enhance ecosystem services (e.g. increase sediment stability (Alteri et al., 2007)) and support fisheries (Jones, 2010; Saintilan et al., 2007).

Understanding if and how habitats are linked is a crucial step in marine management, especially if trying to apply an ecosystem approach, as practices for one habitat may have cascading effects for the whole ecosystem (Bagur et al., 2016; Donadi et al., 2013). To date, research regarding marine inter-habitat linkages has primarily focussed on the tropics (Jones, 2010; Gribben et al., 2009; Saintilan et al., 2007) and there is little information regarding linkages in temperate and Scottish waters.

PMF habitats such as flame shell beds, maerl beds and horse mussel beds are considered to be some of the most biodiverse habitats in the north-east Atlantic (Trigg et al., 2011; Barbera et al., 2003; Farinas-Franco et al., 2014). These habitats can occur together (i.e. ‘linked’), either adjacent to one another (horizontal linkages) or on top of each other (vertical linkages). The implications of such linkages on overall biodiversity are, however, unknown.

The current study progresses research in this area by investigating inter-habitat linkages between maerl beds (formed by Phymatolithon calcareum or Lithothamnion glaciale), flame shell beds (Limaria hians) and horse mussel beds (Modiolus modiolus). Infaunal seabed samples from across Scotland were used to explore the effects of linkages between maerl beds and horse mussel beds, and flame shell beds and horse mussel beds.

Main findings

- This study aimed to investigate the effects of linkages between selected Priority Marine Feature (PMF) habitats in Scotland on their associated biological communities, and trialled a number of different approaches to assess these.

- The study established a useful process to assess linkages between habitats using biodiversity indices and biological trait analysis (BTA).

- In agreement with previous studies (Donadi et al., 2015; Bremner et al., 2006), BTA was found to be an effective way to investigate differences in functional diversity between locations from an ecosystem rather than a species perspective. In contrast, NMDS plots, SIMPER analyses and ANOSIM analyses did not examine differences between locations at an ecosystem level, with reported differences between sites caused by a few dominant taxa.

- Therefore, it is recommended that future studies use BTA to investigate the effects of linkages between habitats to provide a broader ecosystem perspective. The approach is also recommended for general monitoring and assessment. In addition to better reflecting changes in ecosystem functioning, it also greatly facilitates the interpretation of complex datasets and presents results in a format more accessible to external stakeholders or indeed anybody who does not have a good working knowledge of marine infaunal species communities.

- The results of the BTA identified that examples of some habitats used in this study were likely subject to unquantified levels of disturbance (of natural and / or anthropogenic origin) which may have had a role in shaping the benthic communities in these locations and their functional diversity. BTA may thus also be a useful method to assess the effects of pressures on habitats.

- With limited resources, marine management often requires prioritisation of features and sites for monitoring and protection. While results concerning particular effects of linkages between specific PMFs are preliminary and require further research (limited largely by the data available for the study), the results presented here from local scales suggest that habitat linkages are important in terms of increased species and functional diversity. Diverse habitats are recognised for the variety of ecosystem services they provide and for being more resilient to pressures such as climate change.

- Functional diversity appeared to be influenced strongly by environmental factors, and when attempting to compare habitats on a national scale the relatively subtle differences caused by the presence or absence of habitat linkages could not be detected using the samples assessed in this study. Future monitoring would benefit from inclusion of physical factors (e.g. exposure, seabed water temperature, light levels, etc.) and a better understanding of human activities alongside biological sampling, to enable differences in biodiversity in response to linked habitats to be disentangled from man-made pressures and disturbance or environmental factors. Using only samples collected by similar methodologies would further strengthen the power of the analyses.

- Further work is needed to determine if other linked PMFs exhibit similar patterns in terms of elevated diversity and differing community structure. Follow-up studies are needed to explore whether the effects of inter-habitat linkages at the local scale found in this study hold across other locations and habitats. If this is the case it would have important implications for prioritising inter-linked habitats in marine management in the future.

Table 1: Examples of linkages between benthic PMFs in Scotland

Linkage	Description	Functional significance	References
Flame shell beds - Horse mussel beds	Horse mussel beds are sometimes found adjacent to or mixed with flame shell beds. Examples include Loch Alsh, Loch Carron and Loch Creran, where mixed flame shell and horse mussel bed exist, or Loch Linnhe and Loch Leven, where flame shell beds are found close to horse mussel beds.	When the byssal threads of flame shell beds and horse mussel beds intertwine, both PMFs are stabilised. Horse mussels provide a hard substrate for flame shell larvae to settle and increased structural complexity alters hydrodynamics and nutrient cycling across the surface-water interface, influencing the infaunal species found in the sediment.	Connor et al., 2004 Moore et al., 2012 Lancaster et al., 2014 Austen et al., 2002 Trigg & Moore 2009 Moore et al., 2018 Moore et al., 2020
Flame shell beds - Maerl beds	Nests of the flame shell are sometimes found in conjunction with maerl beds. Examples of this linkage can be seen e.g. in the Strome Narrows in the Loch Carron MPA.	Flame shells bind maerl together with their byssal threads, helping to stabilise the maerl bed.	Birkett et al., 1998 Hall-Spencer et al., 2003 Lancaster et al., 2014
Flame shell beds - Tide-swept algal communities	Flame shell beds often support tide-swept algal communities.	Flame shell beds provide a stable substratum for the attachment of seaweeds in an otherwise unsuitable location.	Minchin, 1995 Lancaster et al., 2014
Flame shell beds – Kelp forests	Flame shell beds and kelp forests can be found in close proximity of each other, such as in e.g. the Loch Carron MPA.	The community structure of flame shell beds can be influenced by the presence of a nearby kelp forest.	Trigg et al., 2011 Moore et al., 2018
Horse mussel beds - Maerl beds	Horse mussel and maerl beds can co-exist in a mosaic. Examples of this are within the Fetlar to Haroldswick MPA in Shetland.	Horse mussels bind maerl together with their byssal threads, helping to stabilise the maerl bed.	Birkett et al., 1998 Hirst et al., 2013 Lancaster et al., 2014
Maerl beds - Maerl or coarse shell gravel with burrowing sea cucumbers	Areas of predominantly dead maerl and other coarse gravels with burrowing sea cucumbers often occur on the edge of maerl beds.	This linkage occurs when burrowing cucumbers utilise habitat within and at the edge of a maerl bed.	Lancaster et al., 2014

 

2.2.2 Maerl bed and horse mussel bed communities: local comparisons in Fetlar to Haroldswick

Horizontal linkages between maerl beds and horse mussel beds were investigated within the Fetlar to Haroldswick MPA in Shetland.

Maerl beds and horse mussel beds are considered to be linked horizontally in this location as they co-exist next to one another. The habitat is considered a mosaic with horse mussel beds changing into maerl beds over a small area (e.g. 200 m²) (Hirst et al., 2013). Maerl and horse mussel bed cores from mosaic habitat in Uyea Sound, Hascosay, North Hascosay and Basta Voe were used in this study (see Figure 2).

2.3 Maerl bed communities: national comparison

On a national scale (Figure 3), the maerl beds linked with horse mussel beds in Fetlar to Haroldswick MPA were compared to non-linked maerl beds in Orkney (Figure 4) and the Sound of Barra SCI (Figure 5) with regards to their community structure. Maerl beds in Orkney and the Sound of Barra were selected as both beds are ‘wave formed’ rather than ‘tidally-formed’ so have a similar structure to those in Fetlar to Haroldswick.

Functional diversity was also compared on a local and national level; however, to simplify the analysis, infaunal data from the maerl beds in Fetlar to Haroldswick were only compared with infaunal data from grabs in Barra, and infaunal data from cores in Wyre Sound, Orkney. Wyre Sound was selected as a site because it had the most similar degree of exposure to Barra and Fetlar to Haroldswick.

As sample type (core vs grab) was found to have a large effect on species diversity measurements, the effects of linkages on biodiversity were not explored for the national scale maerl bed analysis.

2.4 Data analysis

2.4.1 Community structure (aim 1)

All analysis was carried out in R Studio (R Core Team, 2019). As a first step, non-metric multidimensional scaling (NMDS) was used to investigate if samples showed differences in community structure, which might be caused by habitat linkages. NMDS uses Bray-Curtis dissimilarity to measure how similar sites are. Sites are then plotted in an NMDS plot which places sites that have a similar community structure closer together. NMDS plots were created using the ‘vegan’ package in R (Oksanen et al., 2019).

Initial data exploration also attempted using cluster analysis (‘fpc’ package; Hennig, 2020) to illustrate which sites were similar to each other. However, ANOSIM analyses showed that differences within clusters were often as large as differences between clusters, potentially because of high numbers of 1s and 0s within the dataset, or because there were more variables (species) than samples. Cluster analysis was therefore not an appropriate technique to be used to analyse similarities between samples.

To investigate differences in community structure via NMDS plots, a minimum species matrix was used. This matrix excluded taxa that were not identified to species level (e.g. nematodes) and combined records of the same species (e.g. juvenile and adult records). A minimum species matrix was used so that differences in community structure picked up by the analyses would be due to the species present, rather than due to larger variations in phylum or age structure. As Bray-Curtis dissimilarity uses the abundance of species to measure dissimilarity, species data from Sound of Barra and species data from the western flame shell bed at Loch Alsh  (grab samples) were divided by 9. This was because the sample volume of the grabs collected at these sites was 9 times larger than that of the cores from Orkney, Fetlar to Haroldswick and the other sites in Loch Alsh. This volume-controlled minimum species matrix was used for the rest of the community structure and species diversity analysis.

Following the creation of NMDS plots, a non-parametric two-way analysis of variance (ANOSIM; Oksanen et al., 2019) was used to quantify the differences between bed types, locations, latitudes and depth. A similarity percentages analysis (SIMPER; Oksanen et al., 2019) was then run for each comparison to determine which 10 species accounted for the largest difference between bed type, location, latitude and depths. As sites in Loch Alsh were at the same latitude, and in a similar depth range, ANOSIM and SIMPER analyses were not used to investigate differences between latitude and depth. NMDS plots, ANOSIM and SIMPER analyses were performed both on log transformed and non-transformed data. In all cases, transforming the data did not change the outputs, indicating that the results were not skewed by dominant species. The report therefore only presents the analyses performed on non-transform the data, which show the true differences between samples.

The effects of inter-habitat linkages on community structure were investigated on a local scale in Loch Alsh and Fetlar to Haroldswick MPA, and on a national scale between maerl beds from Fetlar to Haroldswick, Wyre Sound (Orkney) and the Sound of Barra.

2.4.2 Species diversity (aim 2)

To investigate if inter-habitat linkages influence species diversity, the Shannon diversity index, Pielou’s evenness, species richness and abundance were calculated for each sample (vegan package; Oksanen et al., 2019). For this analysis, samples were only compared with other samples on a local scale.

Linear models were then used to investigate if the diversity indices varied between locations and bed type. Individual indices were calculated from non-transformed raw data. In general, the index-dataset was normally distributed and did not need to be transformed before using a linear model (Shapiro-Wilk test, p>0.05). For the Fetlar to Haroldswick local study, the abundance data was not normally distributed (Shapiro-Wilk test, p<0.05), and was log transformed before using a linear model. Likelihood ratio tests (using the ‘lmtest’ package; Zeileis and Hothorn, 2002), were used to test the statistical significance of the impact of location and bed type on diversity indices.

As grab samples are 9 times larger in volume than core samples, the grabs consistently had a higher Shannon Diversity Index and Species Richness Index. The differences in diversity could not be controlled for by dividing the infaunal abundance data from the grab samples by 9. Therefore, the effects of inter-habitat linkages on species diversity on a national scale (which would have required comparing grab samples to core samples) was not considered further in this study.

2.4.3 Functional diversity - Biological Trait Analysis (aim 3)

Creating a dataset for BTA analysis

Biological trait analysis (BTA) was used to investigate differences in functional diversity between beds that were independent and beds that were linked. Following recommendations by Bremner et al. (2006), species were only included in BTA if they were present in 50 % of the sites or the top 90 % of abundance. For the Fetlar to Haroldswick analysis, species were only included if they were in the top 75 % of abundance, as SIMPER analysis showed that the sites were more dominated by few taxa such as Jasmineira elegans, compared to sites at Loch Alsh. Taxa that were identified only to family or phylum level were included if species within the taxonomic category were known to have similar functional traits.

Biological traits where chosen to reflect lifestyle and behaviour characteristics of the infaunal taxa (Paganelli et al., 2012). Following Donadi et al. (2015), four trait categories - bioturbation, size, longevity, and feeding mode - were selected (Table 2). Bioturbation was chosen as a behavioural trait and refers to the movement of sediment by organisms which plays an important role in many geochemical processes including carbon burial (Queiros et al., 2013). Feeding mode was also selected as a behavioural trait, with feeding type indicating, for example, how stressed an environment is and the sediment type (Paganelli et al., 2012). Size and longevity were chosen as lifestyle characteristics, with each trait able to indicate if a site is disturbed. A fifth category, substratum affinity, was included in this analysis as a proxy to investigate changes in the substrate between the beds (Table 2; Paganelli et al., 2012). Each trait category was separated into several modalities, for example, bioturbation was split into biodiffuser, upward-conveyor, downward-conveyor, bio-irrigator and surficial modifier (Queiros et al., 2013).

Table 2: Functional traits and categories used in the fuzzy coding of taxa (adapted from Donadi et al., 2015 and Jones and Frid, 2009).

Trait	Category	Abbreviation
Bioturbation	None	none
Bioturbation	Biodiffuser	biodiff
Bioturbation	Upward-conveyor	upw
Bioturbation	Downward-conveyor	down
Bioturbation	Bio-irrigator	bioirr
Bioturbation	Surface modifier	surf_mod
Size	<5mm	s_5
Size	5-10mm	s5_10
Size	10-20mm	s10_20
Size	20-40mm	s20_40
Size	40-80mm	s40_80
Size	>80mm	s80
Feeding mode	Deposit feeder	dep
Feeding mode	Suspension feeder	susp
Feeding mode	Detritus feeder	detritus
Feeding mode	Opportunist/Scavenger	opp
Feeding mode	Grazer	graz
Feeding mode	Predator	pred
Longevity	<1 year	<1yr
Longevity	1-2 years	1_2y
Longevity	3-5 years	3_5y
Longevity	6-10 years	6_10y
Longevity	>10 years	>10yr
Substratum affinity	Rock	rock
Substratum affinity	Gravel	gravel
Substratum affinity	Coarse Sand	coarse sand
Substratum affinity	Sand	sand
Substratum affinity	Muddy Sand	muddy sand
Substratum affinity	Mud	mud
Substratum affinity	Other species	other_sp

 

When carrying out BTA, first a trait-by-taxon matrix is created, where each species is given a code for each trait category. A fuzzy coding procedure was used to assign functional trait codes to each species (Chevenet et al., 1994). The fuzzy coding procedure allows for intra-specific variability in traits, as some taxa may exhibit multiple traits within the same category. Species were scored from 0 to 3 for each trait category, with 0 indicating that the taxa have no affinity for the trait and 3 indicating that the taxa have total affinity for the trait. For example, with regards to bioturbation, the polychaete Aonides paucibranchiata is both an upward-conveyor and a downward-conveyor. Therefore, this species has a code of 1.5 for both the upward-conveyor and downward-conveyor trait, and a 0 for all other bioturbation traits. If no information was available for a species for a specific trait category, then a code of 0 was given for all traits within that category. In some cases, if trait information was unavailable, species were given the same code as other species within the same genus, or family. This only occurred if the species shared the same codes for other trait categories indicating it had a similar function. Trait information was obtained from biological trait databases (WoRMS, Polytraits, BIOTIC), and primary and secondary literature (Appendix 1; Fish and Fish 2002).

Following the creation of the trait-by-taxon matrix, the trait category scores for each taxon were multiplied by the abundance of the taxon in each sample. As before, the abundances for samples from Sound of Barra, and for the western flame shell bed at Loch Alsh were divided by 9 to account for differences in the sample volume. Trait categories were then summed for each sample and added to a sample-by-trait matrix containing total scores for each trait category for each sample. For the Loch Alsh analysis, the matrix included 16 samples and 28 traits. For the Fetlar to Haroldswick analysis, the matrix included 36 samples and 28 traits.

To investigate which traits were dominant in each sample, an additional BTA data-frame was created containing proportion data. In this matrix, each trait value was divided by the total sum of trait values within that category. This resulted in a matrix that had a relative proportion of each trait for each sample. For example, for the western flame shell bed 1, the trait proportion values were: 0.12 (none), 0.23 (bio-diffuser), 0.07 (upward-conveyor), 0.01 (downward-conveyor), 0.02 (upward-conveyor) and 0.55 (surface modifier), giving a total sum of 1.

Maerl bed samples from Wyre Sound were strongly dominated by the caprellid Caprella acanthifera. As this species had a dominant effect on the BTA analysis, even when the dataset was transformed, it was removed from the analysis for the national-scale maerl bed comparisons.

The above procedure resulted in the creation of two datasets for each analysis. The first, a sample-by-trait matrix from abundance data, was used to show how functionally rich sites were. The second, a sample-by-trait matrix from proportion data, was used to show which functional categories were associated with which locations.

Principle component analysis (PCA) and regression

Principle component analysis (PCA) and regression were used to explore differences in functional diversity between sites. PCA analysis was first used to visualise differences in functional diversity between sites with PC1 and PC2 scores corresponding to the x and y-axis (using the ‘factoextra’ package; Kassambara and Mundt, 2019). Following plot visualisation, PC rotation scores were obtained, showing the correlation between PC1 and PC2 values and specific traits. PC rotation scores were plotted using the package ‘corrplot’ (Weio and Simko, 2017). This information was then used to determine which traits were more abundant in certain samples. For example, if PC1 was positively correlated with suspension feeders and short-lived species, and samples from Basta Voe had a positive PC1 value, then this indicated that Basta Voe was characterised by suspension feeders and short-lived species. If the direction of all, or most functional trait vectors pointed away from a sample, then it indicated that the sample was not functionally diverse.

Following PCA visualisation, PCA regression was used to quantify the differences in PC1 and PC2 axis values on location and bed type. Linear models were then run, with PC1 and PC2 axis values as dependent variables and location and bed type as independent variables. Likelihood ratio tests were used to determine if the relationship was significant, and post hoc analysis was used to investigate the specific sample to sample differences.

PCA analysis and regression were run on abundance and proportion sample-by-trait matrices for both the local and national scale analyses.

3. Results

3.1 Local scale comparisons: Loch Alsh study - vertical linkages – flame shell beds and horse mussel beds

3.1.1 Community structure

When analysing differences in community structure between flame shell (FS) beds that were linked with horse mussel (HM) beds, and flame shells beds that were independent, there were distinct locational patterns in community composition (Figure 6). Locational differences in community composition were confirmed by ANOSIM analysis with cores from the same site having a more similar community structure compared to cores from different sites (p<0.001, R²=0.89). There was also a significant difference in bed type with linked beds (FS & HM 1 and 2) having a more similar community structure than the independent beds (W FS and E FS) (p<0.001), although the R² was much lower for this comparison (R²=0.66). The results are not shown for the logged dataset as transforming the data had no impact on the NMDS plots. The western flame shell bed samples were grabs while all other samples from Loch Alsh were cores. The mean abundance in the western flame shell bed was divided by 9 to account for differences in sample volume (other metrics were calculated off full original sample volume), but nevertheless sample volume and collection techniques may be driving some of the differences seen here.

A similarity percentages (SIMPER) analysis on the data identified which species contributed to differences between the flame shell and horse mussel beds and the independent flame shell beds (Tables 3, 4 and 5). Differences between beds were driven by only a few taxa. When comparing the eastern flame shell bed to the linked flame shell / horse mussel beds, the top 10 species contributed to ~30.3 % of the average dissimilarity between the bed types, with the polychaete Jasmineira elegans contributing the greatest amount (7.9 %; Table 3). When comparing the western flame shell bed to the linked flame shell / horse mussel beds, the top 10 species contributed to ~48.3 % of the average dissimilarity between the bed types, with Jasmineira elegans again contributing the greatest amount (15.4 %; Table 4). The two independent flame shell beds (E FS and W FS) were also distinguished most strongly by Jasmineira elegans though less strongly so (3.4 %, Table 5). Horse mussel Modiolus modiolus abundances made the second largest contribution to dissimilarity here (2.8 %, Table 5).

In all three comparisons, dissimilarities were dominated by Polychaeta (Jasmineira elegans, Nereimyra punctata, Pholoe inornata, Aonides oxycephala, Spirobranchus triqueter, and Psamathe fusca) which were more abundant in the linked flame shell / horse mussel bed than in the independent flame shell beds.

Table 3: Similarity percentages analysis showing dissimilarity between communities in the FS & HM beds and the eastern flame shell bed. Numbers in bold denote higher abundance.

Species	Class	Abundance in E FS	Abundance in FS & HM	Contribution (%)	Cumulative Contribution (%)
Jasmineira elegans	Polychaeta	61	112	7.9	7.9
Nereimyra punctata	Polychaeta	9	73	7.7	15.6
Aonides oxycephala	Polychaeta	4	26	2.6	18.2
Spirobranchus triqueter	Polychaeta	0	22	2.5	20.7
Ophiopholis aculeata	Ophiuroidea	0	17	2.2	22.9
Pholoe inornata	Polychaeta	28	40	2.0	24.9
Psamathe fusca	Polychaeta	9	21	1.5	26.4
Janira maculosa	Malacostraca	1	10	1.4	27.8
Ophiothrix fragilis	Ophiuroidea	0	10	1.3	29.1
Flabelligera affinis	Polychaeta	7	17	1.2	30.3

 

Table 4: Similarity percentages analysis showing dissimilarity between communities in the FS & HM beds and the western flame shell bed. Numbers in bold denote higher abundance.

Species	Class	Abundance in W FS	Abundance in FS & HM	Contribution (%)	Cumulative Contribution (%)
Jasmineira elegans	Polychaeta	3	112	15.4	15.4
Nereimyra punctata	Polychaeta	2	73	10.7	26.1
Pholoe inornata	Polychaeta	10	40	4.3	30.4
Aonides oxycephala	Polychaeta	1	26	3.8	34.2
Spirobranchus triqueter	Polychaeta	0	22	3.1	37.3
Psamathe fusca	Polychaeta	2	21	2.9	40.2
Ophiopholis aculeata	Ophiuroidea	0	17	2.8	42.0
Eumida sanguinea	Polychaeta	0	16	2.2	44.2
Flabelligera affinis	Polychaeta	3	17	2.1	46.3
Aurospio banyulensis	Polychaeta	0	13	2.0	48.3

 

Table 5: Similarity percentages analysis showing dissimilarity between communities in the eastern and western flame shell beds. Numbers in bold denote higher abundance

Species	Class	Abundance in W FS	Abundance in E FS	Contribution (%)	Cumulative Contribution (%)
Jasmineira elegans	Polychaeta	1	4	3.4	3.4
Modiolus modiolus/ Modiolula phaseolina	Bivalvia	0	2	2.8	6.2
Aurospio banyulensis	Polychaeta	0	2	2.5	8.7
Eumida sanguinea	Polychaeta	0	2	1.2	9.9
Golfingia (golfingia) elongata	Sipunculidea	0	2	1.8	11.7
Psamathe fusca	Polychaeta	1	2	1.7	13.4
Thracia distorta	Bivalvia	0	1	1.7	15.1
Nereimyra punctata	Polychaeta	1	2	1.6	16.7
Metaphoxus fultoni	Malacostraca	0	2	1.5	18.2
Polyophthalmus pictus	Polychaeta	0	1	1.5	19.7

 

3.1.2 Species diversity

In general, inter-habitat linkages between flame shell beds and horse mussel beds had a positive effect on diversity indices (Table 6 and Table 7). Shannon’s diversity index, species richness and abundance were all significantly higher in the linked flame shell and horse mussel beds compared to the independent flame shell beds (Table 7). Pielou’s evenness did not vary significantly between linked and independent beds, with both bed types being relatively even (p>0.05 for both comparisons).

Table 6: Mean diversity measures for each site. *The mean abundance in the western flame shell bed was divided by 9 to account for differences in sample volume.

Location	Shannon Diversity Index	Pielou's Evenness	Species Richness	Abundance
W FS	2.99	0.777	54	46.3*
E FS	2.92	0.774	44	212
FS & HM 1	3.46	0.732	113	728
FS & HM 2	3.44	0.794	77	522

 

Table 7: Results of linear models showing the effects of bed type (linked vs independent) on diversity measures. Likelihood ratio test statistics (LRT) and p-values are shown.

-	Bed Type	Bed Type
Diversity Index	LRT	p-value
Shannon Diversity Index	13.11	<0.001
Species Richness	18.64	<0.001
Species Abundance	23.65	<0.001

 

3.1.3 Functional richness - abundance data

Principle Component Analysis suggested that linked flame shell and horse mussel beds were more functionally rich than the independent flame shell beds (Figure 7 and Figure 8). The first two PCA axes explained 97.9% of the total variance in function traits, with the first principal component axis (PC1) accounting for 91.7% of the variance, and the second principle component axis (PC2) accounting for 6.2%. The linked beds and individual flame shell beds were separated along the PC1 axis, which suggests that differences in functional diversity were driven by inter-habitat linkages. Differences in functional diversity due to bed type were confirmed by statistical modelling, with linked beds having a significantly higher PC1 score than independent flame shell beds (LRT: χ(1) =20.2, p<0.001). Sites also had different PC2 values, with FS & HM 2 having a significantly higher PC2 value than all other sites (LRT: χ(3) =20.56, p<0.001; post hoc analysis, p <0.02 for all FS & HM 2 - other site comparisons). Interestingly, samples from individual flame shell beds (E FS and W FS) were more closely clustered than samples from linked flame shell beds (FS & HM 1 and 2), suggesting that functional diversity within the linked flame shell bed is more heterogeneous than within the independent flame shell beds.

The direction of functional traits plot PC1 and PC2 rotation scores (Figure 8) show that linked flame shell and horse mussel beds are more functionally rich than the individual flame shell beds. With regards to the PC1 rotation scores, scores were mainly positive. In general, samples from the linked flame shell beds were also positive indicating that traits were found in a higher abundance in these beds. In particular, organisms that were suspension feeders with a short life span (1-2 years) were found in abundance in FS & HM bed 1, and organisms that were deposit feeders and biodiffusers were found in abundance in FS & HM bed 2. Furthermore, whilst the FS & HM 1 bed tended to support organisms that favoured rocky substrate, the FS & HM 2 bed supported organisms that favoured both gravel and muddy sand substrates.

3.1.4 Functional evenness - proportion data

PCA analysis showed that flame beds within Loch Alsh (E FS, FS & HM 1 and FS & HM 2) had distinct functional differences to the bed at the mouth of Loch Alsh (W FS) (Figure 9 and Figure 10). The first two PCA axes explained 80.5% of the total variance in function traits, with the first principle component axis (PC1) accounting for 54.9% of the variance, and the second principal component axis (PC2) accounting for 25.6%. As beds within Loch Alsh and beds at the mouth of Loch Alsh (W FS) were separated along the PC1 axis, differences in functional diversity were driven by locational patterns. Differences in functional diversity due to the location were confirmed by statistical modelling, with beds within Loch Alsh having a significantly higher PC1 score than the bed at the mouth of Loch Alsh (LRT: χ(3) =32.01, p<0.001, post hoc analysis, p <0.001 for all FS & HM 2 - site comparisons). PC2 values were dependent on site, with FS & HM bed 2 having a significantly lower PC2 value than the other beds (LRT: χ(3) =18.4, p<0.001; post hoc analysis, p <0.04 for all FS & HM 2 - site comparisons).

The direction of functional traits plot and PC1 and PC2 rotation scores (Figure 10) show that beds within Loch Alsh have a different functional community than the flame shell bed at the mouth of Loch Alsh (W FS). PCA analysis indicated that the flame shell bed at the mouth of Loch Alsh (W FS) is dominated by predators, medium long lived species (3-5 year life span) and muddy sand. The eastern flame shell bed is dominated by surface modifiers and suspension feeders with a small body size (10-20 mm). Linked flame shell and horse mussel beds did not cluster in the PCA plots, with multiple functions correlating with different samples. For example, whilst FS & HM 1 had a higher proportion of short lived species (1-2 year life span) that preferred to inhabit a rocky substrate, FS & HM 2 had a higher proportion of deposit feeders and biodiffusers that were large (body size <80 mm) and short lived (<1 year life span) and preferred to inhabit a gravel substrate. This shows that these sites are functionally heterozygous, with multiple traits found within each site.

3.2 Local scale comparison: Fetlar to Haroldswick study - horizontal linkages – maerl beds and horse mussel beds

3.2.1 Community structure

As illustrated in the NMDS plots (Figure 11), on a local scale, clustering according to community composition was largely driven by location. Samples grouped into 3 distinct clusters which did not correspond to geographical proximity. Community structure in the linked horse mussel beds in Uyea Sound and Hascosay was similar despite Uyea Sound being at the greatest physically distance from the other samples (cluster 1). The linked horse mussel bed at Basta Voe had a distinctly separate community structure (cluster 2). The linked maerl beds in Hascosay and North Hascosay shared a similar community composition (cluster 3). The community structure of the two Hascosay sites (a horse mussel bed and a maerl bed, HS HM and HS MB) did not overlap. This result suggests that even within the horse mussel maerl bed habitat mosaic community composition is driven by the biogenic bed type, with samples taken from distinct maerl and horse mussel bed areas within the ‘mosaic’ having different communities. Differences in community composition due to location were confirmed by ANOSIM analysis (R²=0.81, p<0.001). ANOSIM analysis also confirmed that difference in community between clusters was statistically significant (R²=0.81, p<0.001).

As the NMDS analysis indicated that there were 3 distinct communities present in Fetlar to Haroldswick (Figure 11), a similarity percentages (SIMPER) analysis was run to identify which species contributed to differences between these communities (Table 8 to Table 10). When comparing differences between the first two clusters, the top 10 species contributed to 60.3% showing that differences between the clusters was driven by a few key species (Table 8 and Figure 8). Species abundant in cluster 1 contributed the greatest amount of dissimilarity between the communities, with Capitella capitata, a polychaete (18.9%), followed by the oligochaete Tubificoides pseudogaster (14.4%). Although species between cluster 1 and cluster 2 differed, similar functional groups appeared to be present in both bed types. For example, polychaetes, which play a role in the bioturbation of sediment, were present in both bed types. The brittle star Ophiopholis aculeata, a detritus feeder, was more abundant in cluster 1.

When comparing differences between cluster 1 and cluster 3, the top 10 species contributed to 65.7% showing that differences between the clusters was driven by a few key species (Table 9). As before, Capitella capitata, a polychaete, contributed the greatest amount of dissimilarity between the communities (17.3%), followed by oligochaete Tubificoides pseudogaster (15.5%). As before, polychaetes and brittle starts were found in abundance in both beds, however, Malacostraca were more abundant in cluster 3, and oligochaetes were more abundant in cluster 1.

Although differences between the community structure in cluster 2 and cluster 3 were driven by a few key taxa, the top 10 taxa only contributed to 46.8 % of dissimilarity (Table 10). Therefore, compared to comparisons between the other clusters, differences in community structure between the cluster 2 and cluster 3 were driven by smaller changes in taxa. Although both communities had Malacostraca species present, polychaetes and oligochaetes were more abundant in cluster 2, and Ophiuroidea species were more abundant in cluster 3.

Table 8: Similarity percentages analysis showing dissimilarity between communities in the horse mussel beds in Hascosay and Uyea Sound (cluster 1) and the horse mussel bed at Basta Voe (cluster 2). Abundance in bold represents greater abundance

Species	Class	Abundance in Cluster 1	Abundance in Cluster 2	Contribution (%)	Cumulative Contribution (%)
Capitella capitata	Polychaeta	86	1	18.9	18.9
Tubificoides pseudogaster	Oligochaeta	81	12	14.4	33.3
Ophiopholis aculeata	Ophiuroidea	29	2	6.2	39.5
Chaetozone setosa	Polychaeta	1	17	4.9	44.4
Tubificoides benedii	Oligochaeta	23	0	4.7	49.1
Mediomastus fragilis	Polychaeta	24	9	3.5	52.6
Pomatoceros triqueter	Polychaeta	4	11	2.5	55.1
Polydora caeca	Polychaeta	0	8	2.0	57.1
Crassicorophium bonellii	Malacostraca	0	6	1.6	58.7
Hiatella arctica	Bivalvia	6	8	1.6	60.3

 

Table 9: Similarity percentages analysis showing dissimilarity between communities in the horse mussel beds in Hascosay and Uyea Sound (cluster 1) and the maerl beds at Hascosay and North Hascosay (cluster 3). Abundance in bold represents greater abundance.

Species	Class	Abundance in Cluster 1	Abundance in Cluster 3	Contribution (%)	Cumulative Contribution (%)
Capitella capitata	Polychaeta	86	0	17.3	17.3
Tubificoides pseudogaster	Oligochaeta	81	1	15.5	32.8
Ophiopholis aculeata	Ophiuroidea	29	0	6.0	38.8
Socarnes filicornis	Malacostraca	0	25	5.9	44.7
Mediomastus fragilis	Polychaeta	24	1	5.2	49.9
Amphipholis squamata	Ophiuroidea	2	23	4.8	54.7
Tubificoides benedii	Oligochaeta	23	0	4.3	59.0
Vaunthompsonia cristata	Malacostraca	0	14	3.1	62.1
Janira maculosa	Malacostraca	0	8	1.9	64.0
Paradialychone filicaudata	Polychaeta	0	7	1.7	65.7

 

Table 10: Similarity percentages analysis showing dissimilarity between communities in the horse mussel bed in Basta Voe (cluster 2) and the maerl beds at Hascosay and North Hascosay (cluster 3). Abundance in bold represents greater abundance.

Species	Class	Abundance in Cluster 2	Abundance in Cluster 3	Contribution (%)	Cumulative Contribution (%)
Socarnes filicornis	Malacostraca	0	25	8.7	8.7
Amphipholis squamata	Ophiuroidea	0	23	7.4	16.1
Chaetozone setosa	Polychaeta	17	0	6.7	22.8
Vaunthompsonia cristata	Malacostraca	0	14	4.6	27.4
Tubificoides pseudogaster	Oligochaeta	12	1	4.5	31.9
Pomatoceros triqueter	Polychaeta	11	0	3.9	35.8
Mediomastus fragilis	Polychaeta	9	1	3.3	39.1
Polydora caeca	Polychaeta	8	1	2.7	41.8
Janira maculosa	Malacostraca	1	8	2.6	44.4
Crassicorophium bonellii	Malacostraca	6	5	2.4	46.8

 

3.2.2 Species diversity

In general, maerl bed habitat within the maerl and horse mussel mosaic had significantly higher species diversity measures than horse mussel bed habitat (Table 11 and Table 12). The maerl beds had a higher Shannon Diversity Index, a higher Pielou’s index and higher species richness values (Table 11 and Table 12). Abundance was not significantly different between habitat types. 

Table 11: Mean diversity measures for each site.

Location	Shannon Diversity Index	Pielou's evenness	Species Richness	Abundance
HS MB	2.85	0.81	35	168
NH MB	2.92	0.82	36	160
HS HM	2.25	0.68	28	323
US HM	2.03	0.66	22	282
BS HM	2.77	0.85	31	113

 

Table 12: Results of linear models showing the significant effects of bed type (horse mussel vs maerl bed) on diversity measures. Likelihood ratio test statistics (LRT) and p-values are shown.

-	Bed Type	Bed Type
Diversity Index	LRT	p-value
Shannon Diversity Index	8.49	<0.01
Pielou's Evenness	5.57	<0.02
Species Richness	5.76	<0.02

 

3.2.3 Functional richness - abundance data

At Fetlar to Haroldswick, PCA analysis indicated that in general, horse mussel bed areas sampled within the mapped horse mussel and maerl bed mosaic were more functionally rich than the maerl bed areas (Figure 12 and Figure 13). The first two PCA axes explained 91.8 % of the total variance in functional traits, with the first principal component axis (PC1) accounting for 81.4 % of the variance, and the second principle component axis (PC2) accounting for 10.4 %. As samples from the Uyea Sound  horse mussel bed (US HM), and the horse mussel bed at Hascosay (HS HM) were separated from the other sites along the PC1 axis, differences in functional diversity appeared to be driven by biogenic bed type. Differences in functional diversity were confirmed by statistical modelling, with the horse mussel beds at Uyea Sound and Hascosay  having a significantly lower PC1 score than the other locations (LRT: χ(2) = 13.173, p<0.002; post hoc analysis, p<0.02 for all comparisons).

As the majority of functional traits are negatively correlated with PC1 values (Figure 13), the horse mussel bed areas within Uyea Sound and Hascosay are more functionally diverse than the other locations (reflected in the correspondence of the ellipses in Figure 12 with these negative PC1 values). Interestingly, different functional groups were found at the two locations. At Uyea Sound, taxa that were upward-conveyors, short lived (life span of less than 3 years), small (body size of 5-10 cm) with an affinity to muddy sand substrate were abundant. At Hascosay, a diverse range of taxa were abundant in the horse mussel bed including surface modifiers and biodiffusers, deposit feeders and suspension feeders, and species that ranged from medium to large (body size of over 80 mm and from 10-20 mm). Furthermore, taxa found at the Hascosay horse mussel bed had an affinity for a large range of substrate types including rock, coarse sand and sand.

The maerl beds at North Hascosay and Hascosay were not as functionally rich as the horse mussel beds. The positions of both maerl bed locations in Figure 12 overlap, indicating that they are functionally similar.

The horse mussel bed at Basta Voe was also not functionally rich with samples clustering tightly in Figure 12. This suggests that the habitat at Basta Voe may be relatively homozygous compared to the other sites, supporting a smaller range of functional groups.

3.3 National scale analysis: maerl beds

The maerl beds horizontally-linked with horse mussel beds in Fetlar to Haroldswick considered in Section 3.2 are compared here with non-linked  / independent examples of maerl bed habitat from two areas >100 miles apart (no data was available for this study from independent maerl beds nearby Fetlar to Haroldswick).

3.3.1 Community structure

There were distinct locational patterns in community composition (Figure 16). Locational differences in community composition were confirmed by ANOSIM analysis (p<0.01, R²=0.84). The Sound of Barra and Orkney samples showed larger within site dissimilarity than the Fetlar to Haroldswick samples. Results for the log-transformed dataset are not shown, as transforming the data had no impact on the NMDS plots.

A similarity percentages (SIMPER) analysis on the data identified which species contributed to differences between Fetlar to Haroldswick and the other maerl bed locations (Table 13 and Table 14). Differences between maerl beds were dominated by a few particular taxa, especially between Orkney and Fetlar to Haroldswick. When comparing Orkney and Fetlar to Haroldswick, the top 10 species contributed ~47.5 % of the average dissimilarity between the bed types, with Caprella acanthifera contributing the greatest amount (13.6 %; Table 13). When comparing Barra and Fetlar to Haroldswick, the top 10 species contributed ~49.9 % of the average dissimilarity between the bed types, with Socarnes filicornis contributing the greatest amount (12.4%; Table 14). In both comparisons, dissimilarities were dominated by Socarnes filicornis, Amphipholis squamata, Vaunthompsonia cristata, Janira maculosa, Maera othonis, and Paradialychone filicaudata, which were more abundant in the Fetlar to Haroldswick site.

 

3.3.2 Functional richness - abundance data

Functional diversity varied between sites (Figure 17 and Figure 18). The first two PCA axes explained 66.9% of the total variance in functional traits, with the first principle component axis (PC1) accounting for 48.1% of the variance, and the second principal component axis (PC2) accounting for 18.8%. Differences in PC1 values and PC2 values were driven by site (LRT: χ(3) = 30.56, p<0.001 and LRT: χ(3) = 38.07, p<0.001 respectively), with Wyre Sound having a significantly lower PC1 value, and Fetlar to Haroldswick having a significantly higher PC2 value (post hoc analysis, p<0.001 for both comparisons). As PC1 values accounted for a larger amount of variance than PC2 values, maerl beds in Wyre are more dominated by different functional groups compared to maerl beds in Barra and Fetlar to Haroldswick.

The direction of functional traits plot and PC1 and PC2 rotation scores (Figure 18) show that Wyre maerl beds had a high abundance of organisms that were surface modifiers, detritus feeders, and short-lived (less than 1-year life span). In comparison, beds at Fetlar to Haroldswick had a higher abundance of traits such as upward-conveyors and biodiffusers, deposit feeders, and organisms with an affinity with coarse sand (Figure 17 and Figure 18). Maerl bed samples at Barra were found to have a low functional richness, with the direction of functional traits often pointing in the opposite direction to the samples (Figure 17 and Figure 18), however, some functional traits (such as long lived species and predators) were abundant in Barra.

3.3.3 Functional evenness - proportion data

Functional diversity also varied between sites when using the proportion dataset (Figure 19 and Figure 20). The first two PCA axes explained 66.7 % of the total variance in function traits, with the first principle component axis (PC1) accounting for 42.8 % of variance, and the second principle component axis (PC2) accounting for 24.0 %. Differences in PC1 values and PC2 values were driven by site (LRT: χ(3) = 35.73, p<0.001 and LRT: χ(3) =  56.83, p<0.001 respectively), with Wyre having a significantly higher PC1 value, and Fetlar to Haroldswick having a significantly lower PC2 value (Figure 19; p<0.0012 for all comparisons). As PC1 values accounted for a larger amount of variance than PC2 values, maerl beds in Wyre are more functionally separate to maerl beds in Barra and Fetlar to Haroldswick.

The direction of functional traits plot and the PC1 and PC2 rotation scores (Figure 20) show similar results as above, with Wyre maerl beds dominated by species that were surface modifiers, detritus feeders and had a short life span of less than 1 year (Figure 19 and Figure 20). The Sound of Barra maerl bed supported small species (body size of less than 5mm) with a long life span (6-10 years and >10 years), as well predators. In Fetlar to Haroldswick, the maerl beds supported more species that were biodiffusers and had a small-medium body size (5-10 mm and 10-20mm) and a medium lifespan (3-5 years).

4. Discussion

This study aimed to investigate the effects of linkages between selected Priority Marine Feature (PMF) habitats in Scotland on their associated biological communities, and trialled a number of different approaches to assess these.

While results concerning particular effects of linkages between specific PMFs are preliminary and require further research (limited largely by the data available for the study), this study has identified a successful framework for investigating the effects of inter-habitat linkages between PMFs.

4.1 Methods review and recommendations

In agreement with previous studies (Donadi et al., 2015; Bremner et al., 2006), BTA was found to be an effective way to investigate differences in functional diversity between locations from an ecosystem rather than a species perspective. In contrast, NMDS plots, SIMPER analyses and ANOSIM analyses were not able to examine differences between locations from an ecosystem perspective, and instead reported differences between sites caused by a few dominant taxa. Although locations may appear to differ greatly in the NMDS plots, SIMPER analysis showed that often species with similar traits were abundant at both locations. For example, in Fetlar to Haroldswick, both horse mussel beds and maerl beds often had a high abundance of brittle stars, with differences in specific species of brittle stars causing the bed types to appear distinct in the analyses. However, whilst NMDS plots may indicate that community composition between locations differs, the locations may be functionally similar, as different species of brittlestar do not fulfil vastly different functional roles.

Therefore, it is recommended that future studies use BTA to investigate the effects of linkages between habitats to provide a broader ecosystem perspective. The approach is also recommended for general monitoring and assessment. In addition to better reflecting changes in ecosystem functioning, it also greatly facilitates the interpretation of complex datasets and presents results in a format more accessible to external stakeholders or indeed anybody less familiar with the species of marine infaunal communities.

For the BTA, trait information was obtained for 94% of taxa in the Fetlar to Haroldswick study and 96.4% of taxa in the Loch Alsh study. Similar to previous studies, information on life span was the main category of missing data (Paganelli et al., 2012). Also similar to previous research, this study found that searching for literature regarding biological traits was time-consuming, and the available information on life-cycle traits was limited. As the number and type of biological traits selected can influence the results (Paganelli et al., 2011), future studies should aim to include as many traits as possible. However, because information regarding  the functional traits of many species is limited there must be a trade-off between time spent researching trait data and the number of traits included in the analysis. Future studies could ascertain that functional traits are not having a dominant effect on the result by removing traits and re-running the analysis. Dominant functional trait identification can be an informative process in its own right as well.

A key limitation in the study was the variation in infaunal sample data collection methods which influenced the results. This highlights the importance of using data collected by the same methods for comparisons. While one uniform sampling method and strategy for all habitats would be ideal statistically, practical considerations dictate that different methods will continue to be used, not least because no one method is suitable for every habitat. Future sampling programmes should however aim to use the same methods as used for existing data. 

A further limitation of the study was the effects of environmental variables and pressures on the results. When comparing sites on a national scale, both BTA and NMDS plots showed distinct differences between locations, which was likely a result of environmental parameters and / or anthropogenic pressures. While efforts were made to only make comparisons between sites with similar physical attributes (and any that were extremely different were excluded from the analysis), future inter-habitat linkage studies should aim to select samples from beds within a narrow environmental and pressure envelope. A potential study site for this would be Loch Carron, where infaunal core samples were taken from a mixed flame shell and maerl bed as well as independent flame shell and maerl beds. Future research should consider available environmental data (such as e.g. wave energy, current speeds, depth, salinity) to investigate which environmental factors are driving differences between locations.

4.2 Effects of linkages

At a local scale, vertically linked habitats (i.e. flame shell beds overlaying or mixed with horse mussel beds) had a distinct biological community which was associated with higher biodiversity compared to non-linked examples of the habitats nearby.

Horizontally linked habitats (i.e. maerl bed and horse mussel bed patches adjacent to each other in a mosaic) did not share similar biological communities, suggesting that infaunal community structure was particular to biogenic bed type, and not driven by proximity to other PMFs. It was not possible to look at differences in diversity between horizontally linked habitats and independent habitats on a local scale because there is no data from locations which contain both independent as well as linked beds. When trying to compare linked and unlinked habitats on a national scale, community structure of maerl beds varied so strongly between locations that it was impossible to assess the potential impact of linkages. Temperature, depth and salinity executed the strongest influence on functional diversity of the habitats considered.

There was evidence that habitats with vertical linkages (i.e. flame shell beds overlaying or mixed with horse mussel beds) increased functional diversity on a local scale. Patterns in functional diversity were driven by factors such as substrate type and the type of biogenic structure (i.e. the presence of either maerl or horse mussels), and an increase in habitat and niche heterogeneity.

As an incidental finding, the results indicated that disturbance (both natural and / or anthropogenic) likely plays a role in shaping communities and their functional diversity, a finding made when comparing the functional diversity of communities from different sites with different degrees of disturbance. Further research is needed across multiple environmental and pressure gradients to understand this relationship.

The impact of linked habitats and habitat interactions on overall biodiversity and ecosystem functioning at a site is likely to have an important role in enhancing ecosystem resilience in the face of pressures, for example climate change. Previous studies have found that habitats with linkages can act as a buffer to climate stress (Bulleri et al., 2018). For example, in addition to increasing the species diversity (Angelini et al., 2015), the linkage between ribbon mussels and cordgrass (Alteri et al., 2007) also enhances the resistance and resilience of the system to drought (Angelini et al., 2016) and thermal stress (Crotty and Bertness, 2015). As another example, coastal plants can increase the pH in the surrounding water column via photosynthesis, protecting nearby mussel beds and coral reefs from the stress of ocean acidification (Wahl et al., 2017; Unsworth et al., 2012). As interactions between habitats can reduce the environmental stress experienced by individual organisms, understanding existing linkages in the marine environment is crucial for effective conservation in a changing world (Gaines and Roughgarden 1987, Rilov and Schiel 2011).

4.2.1 Community structure

Local-scale comparisons

When comparing community composition in Loch Alsh, habitat linkages between flame shell beds and horse mussel beds appeared to influence community composition. Linked beds had higher abundances of most species, including polychaetes, such as Jasmineira elegans and Nereimyra punctata, and oligochaetes such as Aonides oxycephala. As horse mussels and flame shells co-existed spatially, the higher abundances in the mixed bed may have been a consequence of increased niche availability by effectively providing two habitats in one physical location (Petrosillo and Zurlini, 2016; Donadi et al., 2015).

The community structures of the eastern and western independent flame shell beds in Loch Alsh differed from each other; the eastern flame shell bed had a higher abundance of polychaetes, bivalves and Malacostraca species than the western bed. The western bed was a little shallower than the eastern bed 6 – 10 m and 16.5 m respectively (see Appendix 4 for details). The observed differences between the communities of the independent flame shell beds may be caused by both locational and temporal parameters, as well as differences in sampling methodology. The western flame shell bed infaunal samples were collected using a Day grab and those from the eastern flame shell bed were from diver-collected cores. Grab samples may preferentially sample infaunal taxa in the surface layers of the bed, while core samples may more consistently target infaunal taxa at greater depths in the sediments below. Locational patterns may have driven the differences since the western bed is situated outside the Kyle Akin Narrows, where conditions (e.g. greater exposure and reduced current speeds) may be less favourable for certain species. The effects of locational patterns on community structure has been observed in the literature previously, with environmental parameters, such as organic enrichment, sediment instability and oxygen levels leading to locational differences in community structure (Paganelli et al., 2012; Trigg et al., 2011). Differences between sites may also be due to seasonal patterns, with the western flame shell bed sampled in March 2014 and the eastern flame shell bed in August 2014. Organisms that exhibited seasonal changes were not excluded from this analysis and could have driven changes in community composition between the two sites (Trigg et al., 2011).

In Fetlar to Haroldswick, community structure within the maerl and horse mussel mosaic varied between different patches of biogenic habitat. Horse mussel bed samples from Hascosay and Uyea Sound shared a similar community that was distinct to that from the horse mussel bed at Basta Voe. Differences between horse mussel bed communities were often driven by species associated with organic enrichment, such as the polychaete Capitella capitata and the oligochaete Tubificoides pseudogaster, which suggests some horse mussel beds may have been organically enriched. The pattern was not mirrored in maerl bed communities at the same locations though, so it is unlikely that organic enrichment had been imposed on the wider environment.  The findings of the BTA indicate that community composition within the maerl and horse mussel bed mosaic in Fetlar to Haroldswick may also be influenced by environmental parameters that were not considered in this study.

Whilst mobile species were not included in this analysis, it is likely that habitat mosaics are beneficial for cross-habitat foragers (Lundberg and Moberg 2003). In the absence of data from unlinked examples of either habitat in this location it was, however, not possible to quantify the biodiversity benefits conferred by the horizontal linkages.

National-scale comparisons

During the investigation of differences between linked and independent maerl beds on a national scale, results indicated that species composition seemed to be driven by location-specific environmental differences more strongly than by the presence or absence of habitat linkages.

The differences in community structure between locations were defined to a large degree by a few key species. At Fetlar to Haroldswick, samples had a higher abundance of the oligochaete Tubificoides Pseudogaster, the polychaete Capitella capitata, the crustacean Socarnes filicornis, and the brittle stars Ophiopholis aculeata and Amphipholis squamata, than samples from the other locations. Species that had a higher abundance in the other locations were mainly bivalves (such as Goodallia triangularis, which was more abundant in the Sound of Barra), although samples from Orkney were dominated by the skeleton shrimp species Caprella acanthifera and Phtisica marina.

As an example of location-specific factors which drive differences in community structure, in Wyre Sound the very shallow maerl bed (0-5 m) supported a dense macro-algal coverage in the summer which could explain the high abundance of Caprella acanthifera and Phtisica marina species, which live on fronds of algae (Isaac and Moyse, 1990). In contrast, the sites at Barra were deeper at mainly >20 m and in a more open-coast setting, which could explain the presence of the small robust bivalve Goodallia triangularis, which has a depth range of up to 150m (Dall et al., 1803). Fetlar to Haroldswick is characterised by species that bio-turbate the sediment (such as Tubificoides Pseudogaster and Capitella capitata), suggesting that the sediment at this site may be siltier. This could be a consequence of horizontal linkages between maerl and horse mussel beds, as mussels are known to dissipate wave energy and create accumulations of pseudofaeces, which could lead to a siltier environment (Donadi et al., 2013).

4.2.2 Biodiversity

Findings on the impacts of linkages on biodiversity followed the same pattern as community structure.

Local-scale comparisons

On a local scale, vertical habitat linkages in Loch Alsh appeared to influence biodiversity. The Shannon Wiener Index, species richness, and abundance were all significantly higher in the linked flame shell and horse mussel bed samples than in the independent flame shell beds nearby. As the habitats are linked vertically, it is assumed that the increased biodiversity at the linked sites was a result of increased niche availability (Donadi et al., 2015). For example, organisms may attach to the shell of the horse mussel (Alteri et al., 2007), or bury themselves in the stable mat created by the byssal threads of both the flame shells and the horse mussels.

In Fetlar to Haroldswick, where horse mussel beds and maerl beds occur in a habitat mosaic, the maerl bed areas were more diverse than horse mussel bed areas, with horse mussel beds containing more dominant species. However, it was not possible to compare the diversity of the linked habitats in Fetlar to Haroldswick to independent habitats on a local scale because there were no comparable datasets from independent beds in the site.

National-scale comparisons

It was also not possible to compare the diversity of the linked habitats to independent habitats at a national scale because of the limited dataset used in this study. Therefore it cannot be concluded that the high biodiversity of the Fetlar to Haroldswick maerl beds is related to the presence of adjacent horse mussel beds.

4.2.3 Functional diversity

Functional trait analysis was found to be an effective method to assess differences and similarities between communities.

Local-scale comparisons

In Loch Alsh it was possible to demonstrate that functional diversity differed between linked and independent habitats, with vertically linked habitats (flame shell beds and horse mussel beds) having a higher functional diversity compared to the independent flame shell beds.

Biological trait analysis (BTA) indicated that on a local scale in Loch Alsh, habitat linkages increased functional richness. When comparing linked flame shell (FS) and horse mussel beds (HS) to the independent flame shell beds, there was a greater range of functional trait categories in the linked habitats. Unsurprisingly, functional groups were connected to substrate type, with deposit feeders found in muddier sediments, and suspension feeders found in coarser sediments. Fine sediment can clog the feeding appendages of suspension feeders (Paganelli et al., 2012). The results suggest that substrate type varied within the linked bed in Loch Alsh (FS & HM 2 having finer sediment and FS & HM 1 having coarser sediment). Small changes in substrate type may be one of the key drivers of increased functional diversity. Further research is needed, using substrate type as an additional environmental parameter, to explore the link between substrate type and functional heterogeneity and between the key species and the substrate present.

Interestingly, dominant functional traits present within the linked habitat FS & HM 1 were similar to the functional traits found within the independent eastern flame shell bed station, namely suspension feeders and bio-irrigators. In comparison, samples from the western flame shell bed were dominated by different functional traits such as predators and organisms with a medium life span (3-5 years). There was a slight overlap between samples from the western and eastern flame shell beds, although sites shared fewer functional groups compared to the linked beds and the eastern flame shell bed. As the eastern flame shell bed and the two linked beds are located within Loch Alsh, and the western flame shell bed is found just outside the Kyle Akin Narrows, these results indicate that functional traits may be influenced quite strongly by local environmental parameters, such as exposure and current speed. Further research is needed to understand the effects of environmental parameters, such as wave energy, bottom temperature and salinity on driving functional diversity.

Within Fetlar to Haroldswick, functional traits varied considerably between sampling areas within the linked horse mussel and maerl beds. Horse mussel bed samples at Hascosay and Uyea Sound were more functionally rich compared to the other sites, supporting a large range of functional traits. Functional traits varied between biogenic bed type, with maerl beds and horse mussel beds having different dominant traits. The small-scale habitat heterogeneity across the biogenic habitat mosaic in Fetlar to Haroldswick enhances the functional diversity of the overall environment, but no conclusions could be drawn on whether the communities of the individual horse mussel and maerl bed patches were influenced by the proximity of patches of the other habitat in the mosaic.

National-scale comparisons

Biological trait analysis (BTA) indicated that on a national scale, the types of functional traits present varied greatly between locations. The results suggest that on a larger scale, functional diversity may be influenced more strongly by environmental factors and pressure from anthropogenic activities than by the presence or absence of habitat linkages. Assessments of datasets that share similar environmental parameters and pressures would be needed to identify the role of inter-habitat linkages on functional diversity on a national scale.

In Wyre Sound, the maerl bed was dominated by short lived species with a life span of less than 1 year, as well as by detritus feeders. Small, short lived, opportunist species (such as detritus feeders) often characterise stressed benthic habitats, which suggest that the Wyre Sound maerl bed might be exposed to stressors (Paganelli et al., 2012). This high occurrence of opportunist species in Wyre Sound may be a result of the high wave energy experienced by the benthic communities in the very shallow waters.

An interesting difference between the Wyre Sound and Fetlar to Haroldswick beds is that unlike the maerl beds in Fetlar to Haroldswick, the maerl bed in Wyre Sound was dominated by surface modifiers and had a lower abundance of species that work the sediment. Bioturbators play an important role in the breakdown of organic material within the sediment. This could mean that, compared to Fetlar to Haroldswick, carbon in Wyre Sound is broken down less in the top layers of sediment in the bed. This may be driven by the higher levels of wave energy causing mixing of the bed to a deeper level.

The communities in the Sound of Barra maerl beds were less functionally rich than those in Wyre Sound and Fetlar to Haroldswick, and were dominated by species that were small (less than 5 mm body size) but relatively long lived (known maximum life span of over 6 years – though obviously it is unknown whether individuals at the location achieve this age), as well as by predators (such as Psamathe fusca). Like in Wyre Sound, the high abundance of small organisms suggests that the Barra maerl beds may be subject to regular disturbance. Additional work is needed to clarify whether environmental parameters or anthropogenic pressures which were not considered in this study are driving this difference in community composition. The maerl beds in the Sound of Barra are deeper than in Wyre Sound and Fetlar to Haroldswick, potentially more exposed in a more open-coast location, and are made up of mega-ripples, where maerl accumulates in troughs between waves of coarse sand and shell gravel, all of which influence community composition. Dredging pressure, which is present in the Sound of Barra, could also have increased the proportion of smaller species which are not trapped and retained in dredges (Mercer et al., 2018).

5. Conclusions and future work

This study established a useful process to assess linkages between PMFs using biodiversity measurements and biological trait analysis. With limited resources, marine management often requires prioritising of features and sites for monitoring and protection. The results presented here from local scales suggest that linked habitats are important in terms of their increased species and functional diversity. Diverse habitats are recognised for the variety of ecosystem services they provide and being more resilient to pressures such as climate change. Further work is needed to determine if other linked PMFs exhibit similar patterns in terms of elevated diversity and differing community structure. Follow-up studies are needed to explore whether the effect of inter-habitat linkages at the local scale found in this study holds across other locations and habitats. If this is the case it would have important implications for prioritising inter-linked habitats in marine management in the future.

The methodology presented here offers the potential for much better quantification and interpretation of the services provided by PMF habitats, and by systems of multiple habitats. Adoption of such methodologies in wider monitoring and assessment work could make a significant contribution to moving towards a focus on ecosystem functioning in marine nature conservation.

The results from this study also indicated that BTA could be used to explore the effects of anthropogenic and natural disturbances on ecosystems. Functional traits that are prevalent in a site may vary depending on how stressed the system is. By researching the effects of disturbance on functional traits, it may be possible to identify how disturbance affects the ecosystem as a whole, rather than its effects on individual species which may be replaced by species with similar functional traits (Bremner et al., 2006). This research could provide valuable evidence in a marine management context, allowing habitats whose functions and services are being altered through anthropogenic impacts such as demersal fishing to be identified.

There is also potential to investigate the effects of functional diversity on blue carbon sequestration. At certain sites in this study, bioturbators were prevalent, which mix the upper layers of sediment, in turn affecting carbon breakdown and sequestration. The effects of bioturbators on blue carbon sequestration is currently poorly understood, however and remains a vital research gap (Thomson, 2019). By analysing the organic content and infaunal data within sediment cores, future research could use BTA analysis, with organic content as an additional environmental parameter, to explore the effects of bioturbators on carbon sequestration.

In general, this study highlighted the effects of location on community structure, biodiversity and functional diversity. Future work should aim to include additional environmental parameters in the analysis (such as temperature, substrate type and exposure), as variable environmental conditions appeared to override effects of linkages in this study.