Epibenthic and infaunal bivalves were sampled in a 15 km² archipelago area in the Kosterhavet National Park on the west coast of Sweden ( Figure 1 ). The area is divided by the ≈ 2.7 km wide and 200 m deep Kosterfjord. The eastern part of the area consists of a mosaic of small islands and skerries around the main islands Öddö, Tjärnö and Rossö (from here on “Tjärnö”). The inner parts of this area are generally more sheltered compared to the areas close to the fjord. West of the Kosterfjord, samples were taken around the Koster islands (from here on “Koster”). This area is a similarly diverse mosaic of exposed and sheltered, rocky and sedimentary habitats, but overall, the area is more exposed and rugged than the sites east of the fjord.

The tidal range in this area is small (± 0.3 m), but due to differences in atmospheric pressure and winds the water level may vary by up to 1.0 m. In general, surface salinity varies over the year between 20-30 PSU, and temperature varies among seasons between 0-21°C, but strong spatial gradients in both temperature and salinity are expected to occur within the area. Shallow waters habitats (<2 m deep) in this area are usually near or within small bays and beaches, with a wide variety of substrate conditions ranging from bare rock to silty bottoms. The prevailing wind direction comes from the south-west, which determines the level of wave exposure experienced by any particular stretch of coastline together with its orientation and fetch.

To estimate representatively, abundance and biomass of bivalves in the area, and to evaluate the importance of spatial variability and of selected environmental factors, a total of 20 sites were sampled in the archipelago from the 13^th of July to 20^th of August 2021. The sites were stratified with respect to area (“Tjärnö” vs “Koster”), wave-exposure (“Sheltered” vs “Exposed”) and depth (“Shallow” vs “Deep”). Potential sampling sites were selected by checking for areas that were shallower than 2 meters with a length of at least 100 meters, using digital nautical charts and satellite images. Wave-exposure was determined from available digital maps of modelled, classified exposure (Albertsson et al., 2006). In total 69 “Sheltered” (ultra-sheltered to very sheltered) and 57 “Exposed” (sheltered – moderately exposed) sites in the “Tjärnö” area and 15 sheltered and 95 exposed sites in the “Koster” area were identified (see Figure 1 ). Five sites for each level of area and exposure were randomly selected for sampling. Each site was also stratified by depth with “Shallow” samples taken at depths of 0-0.5 m and “Deep” samples between 0.5-2 m. The start- and endpoints of each 100-meter stretch where a site would be sampled were predetermined on the map to avoid selection bias in the field. Working from one point of the 100-meter stretch to the other, one infaunal and one epifaunal sample were taken at about equal intervals (≈10 m) along the shore (n=10 each for epi- and infauna). The exact depth of each sample was assigned using random numbers in order to get representative samples within the defined depth strata. Local tidal level was checked regularly during field work and sampling depth was adjusted accordingly.

At each site, ten epibenthic samples were taken per depth stratum using a 1x1 m metal quadrat, which was laid on the substrate surface, with the middle of the square at the randomly selected sampling depths (n=10). All epibenthic bivalves within the quadrat were collected. “Shallow” samples were collected while wading and “Deep” samples were collected by snorkelling. In each quadrat the percental cover of substrate categories mud, sand, gravel, shell hash and rock were recorded. In a few cases of very dense epibenthic bivalve cover, subsamples of the quadrat were taken instead using smaller quadrats positioned randomly.

Infaunal bivalves were sampled using sediment cores. “Shallow” cores were taken using a PVC tube with a diameter of 30.15 cm (0.07 m²). The cores were pushed ≈20 cm into the sediment or as far as it would go, and the content was dug out and sieved in the field using a sieve with a 4 mm mesh size. “Deep” cores were taken from a small boat using a steel pole corer with a diameter of 9.8 cm (0.008 m²). In order to obtain comparable sampling areas in both depth strata, 9 pole cores (0.068 m²) were taken within a radius of ±1 m at each of the 10 sampling positions. Note that infaunal samples could only be taken at sites where sedimentary habitats were found and therefore these data are partially unbalanced among sites. Bivalve samples were in labelled containers and transported to the Tjärnö Marine Laboratory where they were kept frozen at -20°C until further analysis.

Bivalves were counted and identified to species level when possible. After being allowed to defrost, total wet weight was determined, and soft tissues were removed. The tissue was dried at 40°C for a minimum of 48 hours or until constant in weight. The dry weight was noted and the tissue was then burned at 560°C for 6 hours and weighed again after cooling down in a desiccator. Ash free dry weight (AFDW) was calculated by subtracting the weight of the ash from the dry weight. Occasionally, individuals were too small or broken to remove the flesh reliably. In such situations the tissue was burnt with the shell with the assumption that the weight of the shell would remain constant during the burning procedure and would be subtracted together with the weight of the ash. AFDW and abundance values were standardized to grams and individuals per m² by dividing them by the surface area of the sample method.

The bivalve communities were described using frequency, biomass and abundance of individual species. In addition, the species composition and its reliance on environmental factors (substrate types and wave-exposure) was assessed at the scale of sites using constrained ordination by distance-based redundancy analysis (dbRDA; Legendre & Anderson, 1999). Tests were performed using distance based multiple regression (DistLM; Mcardle and Anderson, 2001) in PRIMER v6 software with PERMANOVA+ (Anderson et al., 2008). Dissimilarities in species composition among sites were estimated using Bray-Curtis dissimilarities calculated on 4^th root transformed abundance data. Three sites (two “Koster-Exposed” and one “Tjärnö-Exposed”) were excluded from this analysis as no bivalves were found at these sites. Quantitative measures of exposure were estimated a posteriori from the mean of the values extracted from the same wave-exposure layer (Albertsson et al., 2006) used to classify the sites at the coordinates of each sample. Exposure, together with the other variables (mean rock, mud, sand, gravel, and hash cover from the quadrat samples) were transformed to X ¯ = 0 ,   s = 1 and checked for collinearity using the variance inflation factor (VIF; Critical threshold =3). The number of permutations for DistLM was set to 9 999 and a marginal test was used to test significance of environmental factors.

The combined effects of the site classifications Area (Koster, Tjärnö), Exposure (Exposed, Shallow) and Depth (Shallow, Deep) on mean biomass and abundance of infaunal and epifaunal bivalves were tested using analysis of variances (ANOVA, SS type II) with four factors:

where the response variable y _ijklm   is abundance or biomass, A _i is the effect of the ith area, E _j is the jth exposure class, D _k is the kth depth stratum, Site(AE) _l(ij) is the lth site within area i and exposure j and ϵ _ijklm is the deviation not explained by any factor or interaction. After inspection of residuals, it was decided to transform biomass and abundance with a log10(y+1) transformation. In the case of the epifauna the total number of samples was 400 (a=2 areas, b=2 exposure levels, c=2 depth strata, d=5 sites, n=10 replicates), but because no infauna samples were taken at locations where the substrate was rocky or in other ways impossible to sample with a core, the data for the infaunal bivalves had less samples (a total of 225) and hence became unbalanced. The same analyses were applied also to the percent cover of the three most important substrate types (Rock, Sand and Mud) that were estimated from the quadra samples. This was done to quantify and test for differences in habitat characteristics among areas, exposures and depths at different sites and the substrate availability for the two groups of bivalves. All statistical analysis was done in R 4.1.3 (R Core Team, 2022) unless otherwise stated.

In order to assess the overall importance of epi- and infaunal bivalves in terms of their biomass and abundance in different environmental contexts (e.g. areas and levels of exposure), estimates of mean bivalve biomass and abundance, y ¯ i j k , proportions of habitat availability, P _ijk , and areal extent of each combination of exposure and area, Ar _ij (in the 0-2 m depth range) were combined. The latter was obtained from a standardized GIS-layer provided by the Swedish authorities (Albertsson et al., 2006). In order to compensate for the fact that shallow and deep samples represented depth ranges of 0-0.5 and 0.5-2 m respectively, estimated abundances were weighted by ¼ and ¾ for the two depths (w _i ). For infauna the available sedimentary habitat, P _ijk , was estimated from quadrat samples ( P i j k = 1 − y ¯ [ R o c k ] i j k ), while for epifauna P _ijk =1 as this faunal component is found on both soft and hard substrates (all data are provided in the Supplementary Table S1 ). The importance of epi- and infaunal bivalves in different environmental contexts were calculated as:

In general, the observations from the quadrats showed that the a priori stratification of exposure- and depth-classes was ecologically meaningful despite being made from GIS-maps ( Figure 2 ). For example, the cover of hard substrata in exposed areas in Tjärnö and Koster ranged from ≈40-80%, with higher values in the shallow stratum as expected. Similarly, mud was exclusively observed at sites classified as sheltered with a predominance at depths between 0.5-2 m (i.e. ≈60 and 30% in Tjärnö and Koster respectively). Sandy sediments were more evenly distributed at levels of 20-40%. These patterns were also largely supported by statistical analyses ( Table 1 ). Thus, significant differences in the cover of rocky, sandy, and muddy substrates between the two exposure levels and depth strata were found ( Table 1 ). Rock was significantly higher in exposed sites and at shallow depth. The patterns of mud opposed that of rock, having more cover in shallow sites and at deeper depth. In general, there were no significant differences of sediment cover between the two areas Koster and Tjärnö, nor interactive effects with exposure and depth, indicating that the quantitative criteria used to define levels of exposure and depth, were in fact comparable in terms of physical impacts in both Tjärnö and Koster. Nevertheless, the significant variability observed among sites for all variables indicates that there were also other uncontrolled variables that influence the substrate characteristics of shallow sites in the study area ( Table 1 ).

In total 17 species of bivalves (11 infaunal and 6 epifaunal) were observed during this study ( Table 2 ). In terms of the number of sites where the species occurred, Mytilus edulis was the most prevalent epifaunal species with a 75% occurrence rate. The most common infaunal species was Polititapes aureus with an occurrence rate of 40% of the sites. While all infaunal and epifaunal species were observed in the Tjärnö area, only 11 species were found in Koster. Overall, the contribution by epifaunal species to the total biomass was considerately larger than that of infauna in both Tjärnö (88%) and Koster (75%) ( Table 2 ). The total biomass of both epifaunal and infaunal bivalves was more than three times higher in Tjärnö compared to Koster (2093 versus 678 grams) and the total abundance, however, was only 1.8 times higher in Tjärnö compared to Koster (3639 versus 2044 individuals).

Visual inspections of the dbRDA-plots did not show any clear separation among areas or exposure classes ( Figure 3 ). If anything, it appeared that the exposed sites in Koster were located in the high end of the second axis (which appears to be associate with large values of exposure and cover of rock; Figure 3B ). A total of 36% of the variability was explained by the two first axes and the importance of the first axis was approximately twice that of the second axis (41 vs 24% of the explained variability). Interestingly, all infaunal species tended to be negatively correlated with the first and second axes (excluding species with a correlation coefficient<0.3 and those observed in less than four sites; Figure 3A ). This correspond with sites dominated by large cover of sand, gravel and shell hash, but with small cover of rock and independently of mud ( Figure 3B ). This pattern was further corroborated by the DistLM showing significant correlation between community structure on one hand and exposure, rock and sand on the other (p<0.05; Table S2 ). Occurrence of epifaunal species were largely correlated by the second axis ( Figure 3A ). In particular, the native species, Ostrea edulis and Mytilus edulis, were associated with the presence of gravel, while the recently introduced Magallana gigas, appeared to be more of a generalist.

Although the patterns of biomass in relation to depth and exposure appear similar in epi- and infauna, it is clear that the main part of the biomass is associated with the epifaunal species ( Figures 4A, C ). On average, the epifaunal biomass was ≈7 times larger per square meter (of suitable habitat) than that of the infauna in both levels of exposure in Tjärnö and Koster. In contrast to analyses of substrate characteristics, detailed analyses using ANOVA revealed that patterns of biomass of epi- and infauna were more complex and unpredictable in relation to area, depth and exposure ( Table 3 ). For both types of fauna there was a substantial and significant variability among sites within area and level of exposure. Despite the fact that both types of fauna generally demonstrated larger means in the deeper stratum, this pattern was only fully consistent and significant for infauna.

Also, for the spatial patterns of total abundance, there was partial consistency between epi- and infauna ( Figures 4B, D ). However, the relative importance of the two types of fauna in terms of abundance was reversed compared to that of the biomass. Thus, the average abundance of infauna, in areas where sedimentary habitats were available, was on average ≈3 times that of epifauna. Similarly to biomass, there was substantial variability among sites but there were also complex interactions among the main effects for both epi- and infauna ( Table 3 ). The numerically more important infauna showed a different response to depth in exposed and sheltered areas; infauna was virtually absent at shallow depths in exposed sites. Epifauna on the other hand, was absent all the way down to 2 m in exposed sites in Koster, while in Tjärnö epifauna was most abundant in exposed areas ( Figures 4B, D ).

While the previous analyses describes the average amount of biomass and abundance of epi- and infauna in different geographic areas and levels of exposure where suitable habitats were present, a more realistic comparison of their importance require that (a) the relative occurrence of sedimentary and rocky habitats (P _ijk ) and (b) the areal extent of exposure and geographic strata (ArP _ij ) are accounted for ( Figure 1 ). These analyses show that the exposed areas in Koster deviate from all other areas with on average small total bivalve biomass and abundance (<1 g and ≈1 # m² respectively; Figures 5A, B ). In all other areas (sheltered areas in Koster and exposed and sheltered areas in Tjärnö), total biomass was on average 8-18 g AFDW/m² for both depth strata combined. The majority (80-90%) of this biomass consisted of epifauna. This pattern was partly reversed in comparisons of abundance, where sheltered sites in both Tjärnö and Koster were dominated by infauna (≈75%). In exposed areas in Tjärnö, presumably low availability of sedimentary habitats mean that abundances are dominated by epifauna. Moreover, both total biomass and abundance were highest in exposed parts in Tjärnö and sheltered parts in Koster, both of which represents intermediate positions in the longitudinal inshore-offshore gradient in the study area.

Finally, in order to assess the potential importance of epi- and infauna, to overall population and ecosystem processes in different parts of this coastal gradient, information on the areal extent of 0-2 m habitats in sheltered and exposed areas in Tjärnö and Koster as quantified ( Figures 5C, D ). Note that the estimates of areal extent correspond to the somewhat arbitrary boundaries in Figure 1 , from which the sampling sites were randomly selected (data on areal extent are given in Table S1 ). The calculations show that the contribution of the areas to the total bivalve biomass and abundance decrease in a similar way: Tjärnö (sheltered)> Tjärnö (exposed)> Koster (sheltered)> Koster (exposed) ( Figures 5C, D ). This pattern was of course partly explained by the fact that Tjärnö (sheltered) contributed with ≈54% of the area, which offsets the slightly lower estimates of total biomass and abundances compared to when data were expressed per square m ( Figure 5 ). Note also that the second largest area (Koster (exposed) at ≈29%) had a negligible contribution to bivalve total biomass and abundance while the intermediately located areas, which where the smallest (≈9% each), contributed substantially.

Depth was found to be of significant importance for the distribution of infaunal biomass. However, contrary to what was predicted, there were no significant differences for depth for epifauna and exposure and area for both in- and epifauna, even though there were remarkable differences between then in terms of absolute numbers. This is a result of the significant unexplained variation between sites, which seems to indicate that there are more complex patterns and processes that shape the bivalve communities in this area that were not captured in this study. One striking result of this study is that epifaunal bivalves were the dominant group in terms of average biomass per unit area in both Tjärnö and Koster. At the time of the study, there was ≈100 tonnes of bivalve biomass in the whole area (100 AFDW/0.8≈124 tonnes of shell free dry weight or 100/0.07≈1 400 tonnes of wet weight; conversion factors from Rumohr et al., 1987). 70% of the biomass was epifauna and 30% was infauna. The amount of biomass was highly variable among sites, but in general larger biomass was found at depths between 0.5-2 m both for infauna and epifauna. Accounting for the uneven distribution of infaunal habitat (measured as 100% rock), the highest bivalve biomass (10-15 g·m^-2) occurs in the intermediate parts of the coastal gradient, i.e. in the exposed parts of Tjärnö and in the sheltered part of Koster areas, followed by the inner, sheltered parts of Tjärnö (≈8 g·m^-2). Finally, scaling up these results using GIS-estimates of the extent of sheltered and exposed areas at 0-2 m depth, show that sheltered habitats in the inner parts of the archipelago contribute with ≈50% of the biomass in this region. Despite the fact that >90% of the deeper and ≈80% of the shallow parts of this area is covered by sediments, infaunal bivalves only contribute to 10-15% of the biomass.

The biomass data presented here are representative for this region and the specific depth ranges and exposures levels that were included in this study. This type of sampling differs from previous studies done on the Swedish west coast, which had different objectives and have typically focused on smaller geographical areas and usually include either epi- or infaunal bivalves. Perhaps the most comparable data available is presented in Evans and Tallmark, 1977; Möller and Rosenberg, 1983 and Loo and Rosenberg, 1989. Firstly, Evans and Tallmark, 1977 found epifaunal biomass of 7.1 and 3.1 g AFDW·m^-2 at one site in two consecutive years, compared to 5-15 g AFDW·m^-2 in this study. The infaunal portion of the biomass in this study was also lower (3.1 and 0.3 g AFDW·m^-2, 0-3 g AFDW·m^-2 in this study). Secondly, in Möller and Rosenberg, 1983 biomass of two infaunal species varied over a six-year period at two sites depending on the survival of the recruitment of the previous year, but was most often between 1 and 10 g AFDW·m^-2 around the same time of year of this study. It was also found that there was generally more biomass deeper than 0.5 meter than above it. Lastly, Loo and Rosenberg, 1989 reports about 22 and 5.5 g AFDW·m^-2 (converted from WW with factors from Rumohr et al., 1987) of infaunal bivalve biomass at two sites. It must be clear that because of the differences in methods and objectives a comparison between those studies is not entirely correct. Nonetheless, it can be concluded that the finding on biomass per area reported here fall within the ranges that have been reported previously, which supports the validity of the results of this survey. In contrast this study allowed for up-scaling and direct comparison between epi- and infaunal bivalves.

Although the relationships between biomass and filtration-rate or -capacity is non-linear and modulated by environmental factors and behavioral responses of bivalves, and despite the fact that measurement of filtration is complex in laboratory- and field-experiments, it is clear that the biomass of an individual or a bivalve community is a fundamental indicator of its filtration capacity (e.g. Kryger & Riisgård, 1988; and reviewed in Riisgård, 2001). Furthermore, because several critical ecosystem functions and processes are linked to filtration (e.g. water clearance, nutrient uptake and cycling and production), the observed patterns of biomass arguably have important implications also for many ecosystem services. First, the results indicate that overall, epifauna (mainly Magallana gigas, Mytilus edulis and Ostrea edulis) contributes to a larger extent to water clearance and removal of nutrients compared to infaunal species. This is true considering the total biomass per square meter in sampled sites but also when considering the areal extent of sheltered and exposed habitats in Tjärnö and Koster. Nevertheless, in certain sites, particularly at intermediate areas in the in- and offshore gradient, the contribution made by infaunal bivalves such as, Ensis leei, Mya arenaria and Macomangulus tenuis, appeared to match that of the epifauna. Studies targeting both functional groups with suitable sampling regimes in soft and rocky habitats, particularly those accounting for areal extent, are scarce (but see for example Seitz et al., 2006), but it has been found before that the contribution of biomass per unit area is greater for epifauna in Swedish coastal areas (Evans & Tallmark, 1977).

From the biomass we are able to conclude that the filtration capacity of both epi- and infaunal bivalves is markedly smaller in the exposed parts of Koster. While the scarcity of sedimentary habitat explains the low levels of infauna, the low biomass of epifauna is more surprising. Particularly because of the large densities of epifaunal biomass at similarly exposed areas in Tjärnö, which illustrates that the species can withstand such levels of hydrodynamic forces (note also that the very to extremely exposed areas in Koster were excluded from the study). Thus, in order to explain the lack of filter-feeding biomass in exposed areas in Koster, more complex models including factors correlated with the physical aspects of wave-exposure and the inshore – offshore gradient needs to be invoked. In general, such factors may include supply of food, supply of larvae, fluctuations in salinity and temperature or sedimentation (e.g. Lindegarth & Gamfeldt, 2005; Westerbom & Jattu, 2006; Lindegarth, 2007) but here, low epifaunal bivalve biomass (and abundance) could possibly be explained by waves modulating the effects of ecological interactions (e.g. competition, predation or “algal swiping”) acting on larvae, juveniles or adult bivalves (e.g. Underwood, 1999). Nevertheless, whatever the mechanism, it is clear that the end result of these patterns is a large difference among sheltered and exposed areas in their contribution to bivalve biomass and by extension to regulating services associated with filtration. These results emphasize the need to incorporate knowledge about the distribution and abundance of fauna, availability of habitats and areal extent when assessing relative contributions to ecological processes and ultimately to different types of ecosystem services. It is particularly interesting to note that, while the inner parts of this coastal gradient appear to contribute most both in terms of biomass and abundance, the relative importance of the two functional components are practically reversed.

In general, the overall spatial patterns of density and total abundance resembles that of the biomass, however, in contrast to what was observed for the biomass the main part of abundance was made up of infauna rather than epifauna. While the epifaunal abundance, mainly consisted of two species (Mytilus edulis and Magallana gigas), infaunal abundance was more diverse and variable among habitats. In Koster the abundance was dominated by the tellinid, Macomangulus tenuis, the cockles Cerastoderma edule and C. glaucum. In Tjärnö, the same species and two additional ones (Mya arenaria and Polititapes aureus) were mainly observed, and in general the species abundance was more evenly distributed. Possibly, this can be attributed to the more exposed nature of the Koster islands, causing a thinner margin of suitable sediment and environmental conditions for infaunal species. The multivariate analyses showed that most infaunal species were positively correlated with sandy and gravelly sediments rather than those dominated by mud, which is common to sites in the sheltered parts of Tjärnö. It is possible that here the areal extend of this type of sediment is greater than in Koster, and probably less likely to be subjected to events of extreme wave energy, allowing for a more diverse assemblage of bivalves to occur.

While it can be argued that the contribution of bivalve to certain ecosystem services depends on the filter-feeding capacity and hence mainly on the biomass, others services are more closely linked to abundance and species composition. For example, the habitat forming role of epibenthic mussels and oysters and the importance for associated biodiversity is mainly dependent on their abundance and not primarily on the amount of filter-feeding tissue. Although abundances of epifauna was substantially smaller than that of infauna, particularly when accounting for the large areal extent of sheltered habitats in Tjärnö, they were abundant enough to form unique and particularly valuable biotopes (e.g. both mussel- and oyster-beds are included on the “OSPAR list of threatened and/or declining species and habitats”, OSPAR, 2008), and they provide substrate, shelter and foraging area for many associated invertebrates and fish species (e.g. Humphries et al., 2011; P. Norling et al., 2015; P. Norling & Kautsky, 2007). Note that this role of epifauna, supporting and maintaining biodiversity is potentially important both in rocky and sedimentary habitats. In contrast, infauna performs ecological functions exclusively burrowing in sedimentary habitats. Although less conspicuous than the epibenthic species, infaunal species can also act as ecosystem engineers by reworking the sediment, modifying small-scale topography and oxygen conditions (Rosenberg, 2001; Norkko and Shumway, 2011). Furthermore, many infaunal species in shallow habitats are important food-sources for e.g. wading birds. Large abundances of juvenile or adult infaunal bivalves attract and support a rich variety of birds specialized on feeding on bivalves, supporting terrestrial biodiversity and providing a pathway for energy and nutrient transfer to non-marine systems (van Roomen et al., 2012).

Despite exceeding that of epifauna, the overall infaunal abundances were lower compared to data available from previous studies in the sampled area (e.g. Lindegarth et al., 1995). It is unlikely that this observation was a sampling artefact as the number of cores taken at each site in this study was evaluated to be sufficient to ensure that the probability of false negatives for occurrences of bivalves was acceptably low. For common infaunal species such the two cockle species Cerastoderma edule and C. glaucum, from these previous studies it was expected to find average densities of about 1-10 individuals·m^-2. Taking 20 samples from each site (i.e. 1.4 m²) and assuming a Poisson distribution, this would amount to an average probability of false negatives of 50, 25 and<1% for densities of 0.5, 1 and 10 individuals·m^-2. Average densities of the two Cerastoderma species were, however, found to be on the order of 0.1 individuals·m^-2, i.e. one to two orders of magnitude smaller than expected. Both infauna abundance and biomass in this study appeared much lower compared to studies done in studies done in other areas (e.g. Beukema et al., 2001; Seitz et al., 2006). In the case of Beukema et al., 2001, our data corresponds with the absolute minimums in their area during a sampling period spanning over 25 years. Although these systems differ from the area in this study in for instance tidal ranges and the availability of shallow soft sediment habitat, the comparisons are remarkable and suggest fundamental differences in filter-feeding biomass that are related to these conditions. More data and analyses are necessary to assess whether this is a persistent long-term change in this area or part of natural fluctuation.

Population of infaunal bivalves are known to show strong temporal fluctuations, usually caused by mass die-offs (Burdon et al., 2014), which can be caused by several reasons, and subsequent increases after favorable years of new settlement (Beukema et al., 2001). From the current data, it is unclear whether there is in fact a substantial decline or if the observations made are normal fluctuations in abundances. The cause of such a decline, if unordinary, can be speculated. Earlier monitoring has revealed that this area has had increased occurrences of dense mats of filamentous algae in the genus Cladophora (Pihl et al., 1995). These algae spread quickly in protected bays in the summer months and persist for many weeks. These locations correspond well with habitat that is usually associated with infaunal bivalve species such as Cerastoderma spp. and Mya arenaria. Algae mats like these can theoretically smother infauna bivalves, restricting their access to food and oxygen (Perkins and Abbott, 1972). Another possible cause is an increase in the abundance of mesopredators that feed on juvenile bivalve spat (Möller and Rosenberg, 1983; Flach, 2003).

One interesting and potentially important result from this study is that both epi- and infaunal bivalve communities were dominated by invasive species, Magallana gigas and Ensis leei, particularly in terms of biomass. The total biomass was estimated to ≈100 tonnes AFDW (69 and 31 for epi- and infauna respectively). If M. gigas and E. leei are excluded, only ≈35% of the epifaunal and ≈40% of the infaunal biomass remains. Thus, assuming that these establishments have occurred without any competitive exclusion or otherwise caused decrease of native species, these two species have added 60-65% of filter-feeding biomass to this coastal area. Although these assumptions cannot be automatically accepted (see below), these are substantial contributions to benthic-pelagic coupling processes, particularly in sheltered parts of the Tjärnö area. Consequently, it can be argued that these changes may have had substantial effects on ecosystem functioning in these habitats, as for instance the Pacific oyster that has introduced a new source of habitat formation service to the region.

The Pacific oyster Magallana gigas (Thunberg 1793) is well known invasive species (Ruesink et al., 2005) and since its introduction to western Europe in the 1960’s, it has come to dominate coastal communities that were favored habitat for the native blue mussel Mytilus edulis (Fey et al., 2010; Holm et al., 2016; Reise et al., 2017). This study has shown that also in shallow habitats in the northern west coast of Sweden, it is now the dominating bivalve species. Current status of Pacific oysters on the Swedish west coast has developed comparatively recent, after a wide scale recruitment event observed in 2006 (Wrange et al., 2010). This event coincided with warm summer temperatures (Wrange et al., 2010) and based on DNA evidence (Faust et al., 2017), it is presumed to be a natural dispersal event. Prior reports of declining blue mussel populations in this area are thus unlikely to be attributed to the recent rise of the Pacific oyster (Baden et al., 2021). Pacific oysters now also appears to overlap considerably with the native and endangered flat oyster, Ostrea edulis (Linnaeus 1758) in the study area (Bergström et al., 2021). During this study these three species where often observed growing communally within the same cluster. Other studies report Pacific oysters living in what appears to be stable association with blue mussels (Reise et al., 2017) and flat oysters (Zwerschke et al., 2016; Christianen et al., 2018). This implies that there are interactions between Pacific oysters and the native oyster and mussel species, although the developments are too recent and there is done insufficient research to foresee how the invaders presence will affect the native species in Sweden in the long term.

Even though the arrival of the Atlantic jackknife clam Ensis leei (Huber 2015) was much earlier than that of the Pacific oyster, i.e. in the early 1980’s, much less is known about its route of invasion, in which habitats it now occurs and what the overall impact has been on the local benthic ecosystems. Similar to the Pacific oyster, the American jackknife clam has a wide salinity tolerance range (7-32 PSU, Maurer et al., 1974), and thus has invaded many varieties of coastal habitats in Europe to the extent where it is also dominating communities along the south-eastern coast of the North Sea and the Wadden Sea (Gollasch et al., 2015). In accordance, these results show this invader contributed to the largest part of the infaunal bivalve biomass in the study area. Like other studies describing its preferred habitat, it was encountered predominantly in clean, fine-grained sand in the sublittoral zone (Armonies and Reise, 1998; Beukema and Dekker, 1995). The species can form very dense assemblages, where it increases the accumulation of silt and organic content in the sediment (Dannheim and Rumohr, 2012), thus causing indirect changes to the benthic community (Armonies and Reise, 1998). Although there are no local extinctions of other species attributed to the arrival of the American jackknife clam, it coincides with a decline in other infaunal species along the Belgian and Dutch North Sea coast (Dekker and Beukema, 2012; Houziax et al., 2009). Additionally, it has been observed that in areas where it has become common, bivalve eating birds have made a switch to this new food source (Cadée, 2000, Caldow et al., 2007, Tulp et al., 2010). Future research should aim to map the distribution of this species further and investigate how it affects the native bivalve communities and coastal ecosystems in this area.

Scaling up spatial patterns of biodiversity and understanding ecological interactions at the landscape level is a vital component for increasing the fundamental understanding of ecological functions and systems (e.g. Duarte et al., 2005; Kutti et al., 2013; Lohrer et al., 2015; Gonzalez et al., 2020), and for monitoring, conserving and restoring marine habitats (e.g. Schoch & Dethier, 1996; Bayraktarov et al., 2020). One manifestation of this is the recent increase in studies using empirical models often linked to remote sensing methods to predict and map the geographic distribution of species or habitats (e.g. He et al., 2015; Breiner et al., 2018). A wide variety of techniques (e.g. species distribution models (SDMs)) are available to predict both categorical (e.g. presence vs absence of species or habitats) and quantitative responses (e.g. abundance, biomass or rates of processes). In order be sufficiently powerful and precise these techniques require large amounts of data and access to suitable GIS-layers of relevant predictor variables.

In this study area, reliable GIS data of sufficient resolution was not available, particularly for substrate characteristics. Furthermore, sampling abundance and biomass of epi- and infaunal bivalves at a sufficient number of sites (i.e. ≈100 sites) for these more sophisticated models to be useful was not feasible. One limitation of this study is that the data collection is limited to a single year, when it is well documented that bivalve population tend to fluctuate annually. This means that the patterns we observed in this study could potentially vary on a yearly basis. Sampling all sites within the area in a short time frame does on the other hand ensure that these temporal changes do not cause uncertainties when combining data from several years, which potentially could obscure some of the patterns that were found in this study. By excluding temporal variation, it proved to be an efficient way to estimate and explore effects of exposure and depth on bivalves and habitats associated to these factors, and allowed us to evaluate random variability within and among sites.

Furthermore, by adopting a “top-down” approach (e.g. Hewitt et al., 2004), it was also possible to scale-up these representative samples and assess the importance of filter-feeding bivalves over the whole archipelago area, which is characterized by strong environmental gradients of wave-exposure, depth and a complex mosaic of soft and rocky habitats. Thus, despite not being able to predict and map biomass and abundance with a detailed resolution (e.g. at a resolution of 5-10 m), instead it was opted to make robust inferences about the importance of epi- and infaunal species at the scale of this coastal landscape, without having to rely on finding powerful and precise models. Key to this was the deliberate strategy to use a random sampling design, within objectively defined strata with known areal extent. Note, however, that data collected in this way can readily be used for more detailed prediction and mapping, in the event that sufficient data and GIS-information becomes available (e.g. Guillera-Arroita et al., 2015; Guisan et al., 2017; Breiner et al., 2018).