Molluscs from a shallow-water whale-fall and their affinities with adjacent benthic communities on the Swedish west coast

Abstract We conducted a species-level study of molluscs associated with a 5-m long carcass of a minke whale at a depth of 125 m in the Kosterfjord (North Sea, Sweden). The whale-fall community was quantitatively compared with the community commonly living in the surrounding soft-bottom sediments. Five years after the deployment of the dead whale at the sea floor, the sediments around the carcass were dominated by the bivalve Thyasira sarsi, which is known to contain endosymbiotic sulphur-oxidizing bacteria, while background sediments were dominated by another thyasirid, T. equalis, less dependent on chemosynthesis for its nutrition. The Kosterfjord samples were further compared at the species level with mollusc abundance data derived from the literature, including samples from different marine settings of the west coast of Sweden (active methane seep, fjords, coastal and open marine environments). The results show high similarity between the Kosterfjord whale-fall community and the community that developed in one of the Swedish fjords (Gullmar Fjord) during hypoxic conditions. This study indicates that at shallow-water whale-falls, the sulphophilic stage of the ecological succession is characterized by generalist chemosynthetic bivalves commonly living in organic-rich, sulphidic environments.


Introduction
In the deep sea, the arrival of a whale carcass generates an organic-rich 'island' in an otherwise food-poor deep-sea, supporting a highly specialized and diverse assemblage of animals that exhibit a series of successional stages (Smith & Baco 2003). During the 'mobile scavenger' stage, which lasts months to years, sharks, hagfish and other scavenging organisms remove flesh and soft tissues. Polychaetes, crustaceans and other opportunistic small-sized animals thrive on organic remains during the subsequent 'enrichment opportunist' stage, which can last months to years. A complex community, lasting for decades, relies on the hydrogen sulphide and other chemical compounds produced by microbial consumption of the lipid-rich bones in the final 'sulphophilic' stage. During this stage, chemosynthetic bacteria Á free-living or in symbiosis within vesicomyid clams, bathymodiolin mussels and vestimentiferan tube worms Á are at the base of a food web where organic matter is primarily produced by the oxidation of inorganic compounds. Some of the animals found at whale-falls are restricted to vertebrate carcasses, such as the gutless (although heterotrophic) bone-eating worm Osedax Rouse, Goffredi & Vrijenhoek (Rouse et al. 2004;Rouse et al. 2011), while chemosynthetic taxa are often found at other deep-sea reducing habitats, including hydrothermal vents and hydrocarbon seeps (Smith & Baco 2003;Dubilier et al. 2008).
As they are not restricted to a specific geological setting, whale-falls may have played a key role in the dispersal of chemosynthetic fauna among these habitats over evolutionary time (the stepping stone hypothesis: Smith et al. 1989). Evidence consistent with this hypothesis includes molecular studies suggesting that some obligate taxa, specialist at deep-water extreme habitats, originated from shallow-water ancestors living on organic falls (Distel et al. 2000;Jones et al. 2006). For instance, it has been hypothesized that the ancestors of bathymodiolin mussels were shallow-water species which might have acquired the ability to associate with sulphuroxidizing bacteria, allowing them to first colonize organic habitats such as wood and whale-falls and then moving on to hydrothermal vents at ridges and cold seeps (Duperon 2010). However, natural whale-falls from shallow waters, commonly defined as less than 200 m deep (Dubilier et al. 2008;Dando 2010), are poorly documented both in the modern and fossil record and their faunal composition as well as the course of the ecological succession are poorly known (see Smith 2006). The only report of a fully developed natural (i.e. not artificially sunk) whale-fall community in shallow waters comes from the fossil record, where the discovery of a late Pliocene (about 3 Ma old) baleen whale with large lucinid clams and rare bathymodiolins testifying for the sulphophilic stage, has provided some insights with regard to ecological succession at shelf depths (Dominici et al. 2009;Danise et al. 2010).
Time-series studies carried out so far on modern shallow-water whale-falls in the North Sea show the presence on the shelf of some obligate taxa, such as the siboglinid Osedax mucofloris Glover, Kallstrom, Smith & Dahlgren, 2005(Glover et al. 2005Dahlgren et al. 2006;Schander et al. 2010a), whereas studies of temporal processes at very shallow (B40 m) whale-falls in the Gullmar Fjord, Sweden, show that the carcasses are consumed by generalist mobile scavengers already known from the same area (Glover et al. 2010). The monitoring of a minke whale sunken at 30 m depth in the Sea of Japan reports the exclusive presence of non-selective deposit feeders associated with the carcass (Pavlyuk et al. 2009). Several species of dorvilleid and chrysopetalid polychaetes documented at shelf-depth whalefalls are also present at other shallow sites characterized by high levels of organic carbon flux, such as fish farms (Dahlgren et al. 2004;Wiklund et al. 2009a, b). With regard to molluscs, isolated reports indicate the presence of the bathymodiolin mussel Idas simpsoni (Marshall, 1900) at B300 m depth in the North Sea living attached to whale bones (Marshall 1900;Tebble 1966;Warén 1991). Artificial whale-falls sunk just below the 200 m depth threshold in the northwest Pacific (219Á254 m: Fujiwara et al. 2007) show a general composition of the fauna similar to that of deep-water reducing habitats, with a chemosynthesis-based fauna mainly represented by the bathymodiolin mussel Adipicola pacifica (Dall, Bartsch & Rehder, 1938). In contrast to this finding, no evidence for a sulphophilic stage was found at a whale carcass artificially sunk at 385 m depth in the Monterey Submarine Canyon, northeastern Pacific (Braby et al. 2007).
Given this patchy record, many questions remain open with regard to the faunal composition of shallow-water whale-fall communities. One of the main questions is if they host obligate species, as their deep-sea counterparts, or if they are colonized by a subset of the local fauna tolerant of sulphide, similar to shallow-water seep and vent communities (Sibuet & Olu 1998;Tarasov et al. 2005;Dando 2010). Here we present a species-level study of molluscs associated with a 5-m long minke whale (Balaenoptera acutorostrata Lacépède, 1804) experimentally implanted in October 2003 at a depth of 125 m in the Kosterfjord (Skagerrak, Sweden). Time-series studies previously carried out on the same site have shown that the Atlantic hagfish (Myxine glutinosa Linnaeus, 1758), sharks and other scavenging organisms consumed the flesh and exposed the bones within 5 weeks of implantation, and that the carcass was completely skeletonized after 6 months on the sea floor (Dahlgren et al. 2006). Nine months after sinking the carcass, it was colonized by Osedax mucofloris, the first species of Osedax known from a shelf-depth whale-fall, and the first from the Atlantic Ocean (Glover et al. 2005;Dahlgren et al. 2006). Our sampling was performed 5 years after deployment. Sediment samples were collected at the whale-fall site and at the surrounding soft-bottom sediments, in order to compare the whale-fall mollusc fauna with the organisms commonly living in the area. The Kosterfjord samples were then merged into a larger data set built from literature data on mollusc relative abundances at a regional scale, including samples from active methane seep, fjords, coastal and open marine environments.
Our main aim was to understand whether shallowwater whale-fall communities host obligate taxa or if they are colonized by species commonly living in the surrounding soft bottoms. To answer this question, we (i) analysed the community structure of the mollusc fauna associated with a shallow-water whale-fall and (ii) evaluated the differences in taxonomic composition and community structure between the whale-fall community, the surrounding background community and the macrofaunal community at various marine settings around the west coast of Sweden.

Study area
The Kosterfjord is directly connected with the North Atlantic, and because of the prevailing open-ocean marine conditions (including high salinity) it is not a typical fjord (Palm et al. 2004). It is situated in the northeastern part of the Skagerrak, which is the major gateway between the north Atlantic and the Baltic Sea. It is a 250-m deep, 62-km long submarine trench parallel to the coastline of Sweden to the east and sheltered by the Koster islands to the west (Figure 1). The trench is a fault fissure connected in the northwest to the Norwegian Trough which in turn is connected to the deep North Atlantic.
The Skagerrak bottom is characterized by muddy sediments and a high content of organic matter (about 2% of total organic carbon in the sediments), with sedimentation rates of 0.20 cm/year in the Northern sector (Josefson 1985;van Weering et al. 1987). The overall oceanographic regime is driven by an anticlockwise circulation pattern, where dense, saline (30Á 35 psu) and oxygenated oceanic water underflows the more brackish (8Á30 psu) surface water outflow of the Baltic Sea. The main surface currents entering the area are the Jutland Current from the North Sea (southwest) and the Baltic Current from the southeast. The mixing between these two currents forms the Norwegian Coastal Current, with a predominating northern heading, which flows out of the Skagerrak on the Norwegian side. This surface circulation is compensated by a deep counter current that brings the saline Atlantic water through the 700-m deep Norwegian Trench into the Skagerrak (Saetre 2007). However, the temperature and salinity of the surface waters are subject to strong seasonal fluctuations; in deeper waters the fluctuation is present with lower amplitude. Measurements of bottom water tempera-ture at 125 m depth in the Kosterfjord indicate variations during the year of 4.8Á7.58C, with salinity ranges of 34.3Á34.7 psu.
The Gullmar Fjord is a 27-km long real fjord on the west coast of Sweden, about 70 km south of the Kosterfjord. It has a sill at 42 m water depth which restricts water flow to the deep basin of 115 m water depth. The organic carbon content of the sediments is on average higher than in the Kosterfjord, with values between 2.2% and 3.8% (Dando & Spiro 1993). Periodically the water in the bottom of the basin has low oxygen levels. During the 1979Á1980 winter, the bottom of the basin became azoic, due to oxygen deficiency (Josefson & Widbom 1988).

Sampling
For the present study four sediment samples were collected and analysed for their mollusc composition. Sample W1 was collected in May 2008 from the minke whale skeleton at 125 m depth. Samples B1, B2 and B3 (background samples) were collected in January 2009 at a distance from whale bones, respectively, 18 m south, 13 and 55 m north from the whale, at 125Á126 m depth. The sampling at the whale-fall was conducted with a small Sperre ROV (Remotely Operated Vehicle) equipped with a forward-mounted sampling scoop 16 cm long with a diameter of 8.4 cm. Sediments (4,420 cm 3 ) were collected with three ROV scoops close to the whale bones (W1), stored in a sample basket (size 34 ) 26.5 )25 cm) and retrieved. Due to its size and shape, the scoop was able to collect only surface  Dando et al. (1991) and the experimental whale-fall site, next to the Sven Lovén Centre for Marine Sciences, Tjärnö, are highlighted. Modified from Palm et al. (2004). sediments (max sampled depth Â5 cm). Each of the three background sediment samples (B1, B2 and B3) was collected using a Van Veen grab with a sampling area of 0.1 m 2 which would have penetrated to an average depth of 7Á10 cm, and up to a maximum of 20 cm. The total volume of collected sediments was about 15,000 cm 3 for each background sample ( Table I).
The sediment samples were wet sieved through a 0.5 mm screen and preserved in ethanol (:80%) before identification. The residue was washed with hydrogen peroxide and sorted under a binocular microscope for all recognizable hard-shelled biogenic components. The latter include molluscs, serpulids, echinoids, bryozoans, decapods, ostracods, brachiopods, fishes and whale bone fragments. Molluscs were determined at the species level and used for quantitative comparisons. Both shells from live and dead specimens were counted. The total number of bivalve individuals was counted as the highest number of right or left valves and half of the remaining, the latter roughly corresponding to the number of unmatchable valves (i.e. 50 left and 48 right valves: 50'(48/2) 074 individuals). Gastropods were equated to the number of apices.

Data analyses
The Kosterfjord data set, including 1,575 specimens belonging to 45 mollusc species, formed the basis for the analyses of sample diversity and trophic structure. Rarefaction curves (Hurlbert 1971) were calculated to compare mollusc sample species richness of the whale-fall sample (W1) with species richness of the background samples (B1, B2, B3). The height of a rarefaction curve is a function of community species richness, and its curve steepness is a function of species evenness, allowing a comparison of diversity in samples of different sizes (Hayek & Buzas 1997;Gray 2000). Alpha diversity was measured using the Simpson index of diversity (D), which is an appropriate measure of diversity for species abundance data when sample size is not homogeneous (Clarke & Warwick 2001). The Simpson index of diversity ranges from 1 (one taxon dominates the community completely) to 0 (all taxa are equally present), and can be considered a measure of dominance.
The four Kosterfjord samples (n01,575) were also used for trophic analysis. Seven trophic categories were distinguished consistently following the Molluscan Life Habits Databases (Todd 2000). Abbreviations appropriate for the present study were used: chemosymbiotic deposit feeders (DC); suspension feeders (SU); subsurface deposit feeders (DU); surface deposit feeder (DS); herbivores, including herbivores on fine-grained substrates, herbivores on rock, rubble or coral substrates and herbivores on plant or algal substrates (HE); and predatory carnivores, including scavengers (CP). Comparisons were expressed through percent of number of specimens (n, abundance) and number of species (S, richness) for each category.
The larger data set, made by merging the Kosterfjord data with literature data (Supplementary Material Tables SIÁSII), includes (i) five samples from a North Sea pockmark with active methane seeps and three from the surrounding sediments at 150Á166 m depth (Dando et al. 1991), (ii) three samples collected in the Gullmar Fjord during periods of low oxygen conditions (Josefson 1986(Josefson , 1987(Josefson , 1988, and (iii) 82 samples from the west coast of Sweden ranging from 26 to 106 m depth (Agrenius 2001(Agrenius , 2002(Agrenius , 2003(Agrenius , 2005. The latter were collected in the Kattegat and the Skagerrak and are subdivided into samples from fjords (25), coastal areas (35) and open sea settings (22). The total data set is comprised of 97 samples, for a total of 105 mollusc species (gastropods, bivalves, scaphopods) and 26,298 individuals.
After removing species occurring only in one sample (singletons), multivariate analysis was performed on a data set with 68 species and 26,174 individuals (99.5% of the original data set). To overcome problems connected with comparing samples of different size, abundances were transformed into percentages. Percentages were then square-root transformed to de-emphasize the influence of the most abundant taxa and increase the effect of rare species (Clarke & Warwick 2001), allowing for a stronger correspondence with known environmental gradients (see Tomašových & Kidwell 2009).
Hierarchical agglomerative CLUSTER analysis was performed using the paired group method and the BrayÁCurtis algorithm (Q mode CLUSTER). Data were elaborated through detrended correspondence analysis (DCA), a multivariate statistical technique widely used with ecological data to ordinate taxa along underlying ecological gradients (Hill & Gauch 1980). In a DCA plot, axis 1 reflects the primary source of ecological variation in the composition of fauna and axis 2 the additional sources of variation beyond the principal gradient. A similarity percentage analysis (SIMPER; see Clarke & Warwick 2001) was performed to determine which species were responsible for similarity within groups of samples. Those species for which the ratio of mean similarity to standard deviation of similarity is !1 typify the sample group, and were listed in the comparisons. Diversity indices, CLUS-TER analysis and DCA analysis were performed with the software PAST (Hammer et al. 2001). SIMPER analysis was performed with the software PRIMER (Clarke & Gorley 2006).

Whale-fall and background community structure
During sampling at the whale-fall site the skull, one mandible and some ribs were still visible on the sea floor. Exposed bones were covered in a mixture of bacterial mats (associated with blackened bone regions, indicative of sulphide release) and muddy sediments. No molluscs were seen lying directly on or right next to the bones during the survey (nor in more than 20 bones brought to the laboratory and analysed over the year the carcass has been studied). Algal debris was trapped within the bones (e.g. Fucus serratus Linnaeus, 1753) and the decapod Hyas araneus Linnaeus, 1758 was frequently observed close to the skeleton (Figure 2). Bones were highly bioeroded and specimens of the bone-eating worm Osedax mucofloris were recorded living on collected bone samples, 5 years after carcass deployment. The sieving residue included molluscs, regular and irregular echinoids (Brissopsis lyrifera (Forbes, 1841) and Spatangus purpureus Mü ller, 1776), brachiopods (Novocrania sp. and terebratulids), benthic foraminiferans, ostracods, serpulids, bryozoans, decapods, fish fragments and teeth and myxinid dental plates. Sample W1 was dominated by the bivalve Thyasira sarsi (Philippi, 1845) (51% of the total), followed by Abra nitida (O.F. Mü ller, 1776) (16.2%), Tellimya ferruginosa (Montagu, 1808) (8%), Mytilus edulis Linnaeus, 1758 (4.9%) and the nuculanid Ennucula tenuis (Montagu, 1808) (4.7%) (Figure 3) (Figure 3). Also A. nitida and Parvicardium minimum (Philippi, 1836) were represented in significant quantities in the background sediments.
The background samples B2 and B3, which are located north of the whale-fall site, exhibit higher species richness and a more even distribution than W1 (Figure 4). B1, which is to the south of the minke whale, yielded fewer individuals than all the other samples. Although W1 derives from a smaller volume of sediments compared with the background samples (see Table I), it contains a larger number of individuals, and its rarefaction curve reaches an asymptotic shape. This indicates that if a larger volume of similar sediments had been collected from the whale-fall, no further taxa would have been added. The Simpson index of Dominance (D) helps to interpret the results from rarefaction curves (Table I). W1 has the highest value of D, being dominated by a few species. Also B1, which is downcurrent to the whale-fall site, has a higher value of D with respect to the samples located north of the carcass.

Trophism
The two predominant thyasirids, Thyasira sarsi (dominant at W1) and Thyasira equalis (dominant at B1, B2, and B3), are infaunal chemosymbiotic deposit feeders containing symbiotic sulphuroxidizing bacteria in their gill tissue (Southward 1986). Both of them are mixotrophic and can derive  part of their nutrition heterotrophically by particulate feeding (Dufour & Felbeck 2006). In particular, studies on the nutritional dependence of the two bivalves on chemoautotrophic symbiotic bacteria show that T. equalis has fewer symbiotic bacteria in its gills compared to T. sarsi, indicating that the nutritional importance of carbon fixed by the bacteria is less in T. equalis (Dando & Spiro 1993;Dufour 2005). Thyasira sarsi instead derives 50Á100% of its tissue carbon from carbon fixed by bacteria (Spiro et al. 1986;Schmaljohann et al. 1990). The chemosymbiotic trophic group has the highest abundance in all the four samples, but the lowest species richness ( Figure 5). Like chemosymbiotic deposit feeders, surface deposit feeders have a high overall abundance but low species richness, being represented only by the semelid Abra nitida, more abundant in W1 than in the background community. Subsurface deposit feeders (nuculids, nuculanids, yoldiids and dentaliids) have both high abundance and high diversity in B1, B2 and B3 (38%, 46% and 47.9%, respectively). The same subsurface deposit feeders are present in all samples, but their abundance in W1 is the lowest (8.4%).

Suspension feeders have high species richness, both in whale-fall and background fauna. The mytilids
Mytilus edulis and Musculus discors (Linnaeus, 1767) and the montacutid Tellimya ferruginosa (Montagu, 1808) are found in sample W1, whereas pectinids, anomiids and cardiids are typical of B1, B2 and B3. Herbivores are diverse but rare in all samples. Those associated with the whale-fall, such as the rissoids Rissoa lilacina Récluz, 1843 and Pusillina sarsii, are typical of shallower settings where they are associated to algae (Laminaria spp.) or seagrass (Zostera marina Linnaeus, 1753) (Warén 1996). The rissoid species Onoba cf. tumidula Sars, 1878 was recently found also at relatively shallow water vents (557Á713 m) in the North Atlantic (Schander et al. 2010b), suggesting a rather wide environmental range for the whole family. Carnivores are the least represented among the trophic categories; only the burrowing Cylichna cylindracea was present in the whale-fall sample.

Extreme vs. normal benthic environments
The samples collected at the Kosterfjord whale-fall and in the surrounding sediments were compared with samples from a shallow-water North Sea methane seep area, from soft sediment samples collected in the Gullmar Fjord at 115 m depth at times of hypoxic bottom water conditions and with samples collected in normal marine bottoms along Molluscs from a shallow-water whale-fall 9 the Swedish west coast. The CLUSTER agglomerative diagram shows that samples group in four main clusters at rather high value of similarity (around 0.4: Figure 6). Cluster 1 includes samples from 'organic-rich sediments', i.e. the whale-fall sample (W1) and the three samples from the Gullmar Fjord (GLF1, 2, 3), the latter with an organic carbon content between 2.2% and 3.8% (Dando & Spiro 1993). Cluster 2 groups samples from the methane seep area and comprises all samples related to the North Sea pockmark with active methane seepage, whether they were collected from the side of the pockmark (R5, R8, S1, R1, S2) or from the surrounding bottom sediments not directly related to methane seepage (S4, S5, S6). Cluster 3 contains samples from the Swedish west coast with average depth B50 m, comprising those from fjords and those from onshore settings. Cluster 4 groups all offshore soft bottom samples, i.e. with an average depth !50 m. The Kosterfjord background samples (B1, B2, B3) became included in the latter group, in particular with samples collected in the same area at 91Á102 m depth (SK13 and SK14: Figure 6; Supplementary Material Table SII). This result confirms that the distribution of species abundances in samples is relatively unaffected by the heterogeneity of the data set, which includes data collected with different sampling and processing methods (see Supplementary Material Table SII). Consistent with the CLUSTER analysis, in the DCA diagram, the four main sample clusters show practically no overlap (Figure 7). The first two axes of the ordination, DC1 and DC2, are representative of the full distribution of data, explaining 90.7% of the variance (DC1 065.7%, DC2 025%). A small overlap occurs between onshore and offshore samples, consistently with a gradual depth-related transition between samples. If in the CLUSTER analysis, fjords and onshore samples were grouped together, in the DCA they are well separated and form two distinct subclusters where fjord samples have low DC2 values and onshore samples have high DC2 values. The ordination of samples along DC1 follows a depth gradient (see Supplementary Material Figure S1). Onshore and fjord sample scores have the lowest DC1, onshore samples ranging 28Á 59 m depth, fjord samples 21Á47 m.  SIMPER analysis allowed us to highlight which taxa are responsible for the similarity within samples forming the five main groups (Table II). The whalefall and the Gullmar Fjord samples have the highest similarity among the five groups. In particular, their similarity is given by the occurrence in all of them of the bivalves Thyasira sarsi and Abra nitida, with a cumulative contribution of 76.7%. Even though T. sarsi occurs in two of the samples collected inside of the North Sea pockmark (R8 and S1), the largest contribution to the similarity between samples from the methane seep area is given by the thyasirids T. equalis and T. obsoleta. Onshore, offshore and fjord samples are dominated by the bivalves Kurtiella bidentata (Montagu, 1803), A. nitida, nuculids like Ennucula tenuis and Nucula nitidosa and the gastropod Hyala vitrea (Montagu, 1803), which contribute with different percentages within each group.

Discussion
The Kosterfjord whale-fall The quantitative analysis of the Kosterfjord samples shows that the presence of a minke whale carcass on the sea floor at shelf depths still influences the composition and structure of the benthic community 5 years after its implantation. Although some species are shared between the whale-fall and the background community, the whale-fall community clearly shows a lower diversity in its species composition, a different ranking of species, and dominance of the chemosymbiotic bivalve Thyasira sarsi. Our data, although limited to one sample at the whale-fall site, suggest that the high abundance of T. sarsi in the sediments close to the carcass is connected with the decay of the whale organic matter, which created an ephemeral habitat with high sulphate reduction rates. Previous studies show in fact that the density of T. sarsi is dependent upon the sulphate reduction rate in the sediment, i.e. a certain degree of reliance on chemosynthesis (Dando et al. 2004). Sulphide conditions could have been further locally favoured by the presence of dislocated macroalgae drifting along the seabed and trapped by the bones, which alone can introduce reduced compounds in the sediments .
The species T. sarsi is widely distributed in the NE Atlantic and is generally found in association with organic-rich sediments with high total sulphide concentrations ). In the North Sea and in the Skagerrak, T. sarsi is associated with sewage-polluted fjords, anoxic fjords, fish farms and active methane seeps (Dando et al. 1991(Dando et al. , 1994Figure 7. DCA q-mode diagram. Each point in the diagram corresponds to one sample of the data set (n 097). Samples are grouped according to the clusters recognized after the hierarchical agglomerative CLUSTER analysis. Onshore and fjord samples can be grouped in two distinct subclusters.
Molluscs from a shallow-water whale-fall 11 Dando & Spiro 1993;Rosenberg et al. 2002, Kutti et al. 2008, with a depth range of 50Á340 m (Dufour 2005). Conversely, the dominant species in the background sediments, T. equalis, the most common thyasirid on the North European continental shelf, thrives in sediments with less-organics with respect to T. sarsi, possibly avoiding hydrogen sulphide-rich sediments ).
The opportunist species Abra nitida, common along the northern part of the Swedish west coast, is a density-dependent species unaffected by turbid conditions (Josefson 1982). Its high abundance at the whale-fall site could be linked to the presence of high organic content, as observed in fish farm areas with increased food supply (Kutti et al. 2008). The abundance of Tellimya ferruginosa at the whale-fall site, a small bivalve living symbiotically in the burrow of the echinoid Echinocardium cordatum (Pennant, 1777) (Gillan & De Ridder 1997), is indicative of the occurrence of the echinoid itself. Echinocardium cordatum, a deep burrower, may not have been collected due to the shallow depth of our sampling. The species hosts ectosymbiotic sulphide-oxidizing bacteria, Thiothrix-like, in its intestinal caecum and is known to burrow below or at the level of the oxidized-reduced interface, ingesting both surface and deep reduced sediments. This symbiosis opens an access for E. cordatum to sulphide-rich habitats (Temara et al. 1993;Brigmon & De Ridder 1998) and adds further evidence for the presence of a chemosynthetic ecological niche at this shallowwater whale-fall site (Bromley et al. 1995). As for E. cordatum, the presence of other deeper burrowers may have been overlooked in this study because of the shallow sampling depth within the sediment. Among these missing taxa may be the lucinid bivalves, which host sulphur-oxidizing bacteria in their gill tissue and live in burrows of up to 20 cm in depth ).

Taphonomic and sampling bias
One potential bias in this study is that both dead and alive specimens were counted; hence some considerations need to be made on the implications for the faunal composition of the analysed samples. In fact, as a consequence of the adopted methodology, (i) dead shells may have increased species richness, and (ii) in situ and transported taxa may have been mixed, affecting the ecological interpretation of the encountered communities.
T. equalis shells were not found at the whale-fall, which was unexpected given that T. equalis was living at the whale-fall site prior to implantation, as in the background sediments. One possible explanation could be linked to a taphonomic bias. During the decay of the carcass, sulphide-rich sediments at the whale-fall site might have become acidic due to the oxidation of sulphide and completely dissolved dead shells. The microbial oxidation of organic matter and reduced species like H 2 S can in fact decrease porewater pH immediately below the sedimentÁwater interface and produce a strong carbonate undersaturation (Cai et al. 2006). In this scenario most of the background community shells could have been dissolved after the whale implantation, and our picture purely resembles the whale-fall community. A second explanation could instead be linked to a sampling bias. Because we collected only one sediment sample at the whale-fall site, it is not possible to exclude that T. equalis was present in the sediments close to the rest of the carcass. The whale-fall community has a lower diversity with respect to the background community located north of it (samples B2 and B3). To the contrary, the background community located 18 m south (sample B1) is closer in species number and diversity to the whale-fall community. This suggests that currents might have transported whale organics (e.g. blubber) or decaying seaweeds from the carcass to this control site. The organic load related to the minke whale might have partially enriched the sediments there, probably not sufficiently to attract T. sarsi but enough to have a detrimental effect on other species.
Both the whale-fall and the background communities record the presence of coastal species, such as the mytilids Mytilus edulis and Musculus cf. discors and littorinid gastropods. They were probably transported down-slope by bottom currents and as a consequence were not alive when sampled. Most of the gastropods found at the whale-fall site, including the rissoids, are known to dwell on seaweeds, and were possibly moved together with drifting algae, as shown by algal debris around whale bones ( Figure 2). Because alive and dead specimens were not distinguished, it is not possible here to support the hypothesis that some rissoid species could be adapted to live also in reducing environments, as recently found in some North Atlantic vents (Schander et al. 2010b).

Environmental gradients on the Swedish shelf
At the species level, multivariate comparison of the Kosterfjord samples with samples collected from soft-bottom sediments across the west Swedish coast allowed for a better interpretation of environmental parameters controlling the faunal composition at the Kosterfjord whale-fall. In this study, the ordination of samples along the principal axis of the DCA is controlled by water depth. The faunal composition changes continuously along a depth gradient, with shallower samples on the left side of the diagram and deeper on the right side (Figure 7). This result is in accordance with the interpretation that in marine environments water depth is the single most important factor indirectly controlling the distribution of benthic organisms (see Gauch 1982), as it has been amply proven in other case studies (Scarponi & Kowaleski 2004;Dominici et al. 2008;Konar et al. 2008).
The significance of the DC2 ordination is generally more difficult to interpret, because variations in water depth sum up continuously changing values of other parameters that directly affect the distribution of benthic species, such as food availability, water energy, substrate texture, seasonality, oxygen content and salinity. Samples from reducing soft bottoms show low DC2 scores (Figure 7). Among them are samples from the Gullmar Fjord, where bottom waters are periodically affected by low oxygen conditions (Josefson 1987;Josefson & Widbom 1988;Dando & Spiro 1993). These three samples were collected between 1985 and 1987, at a time of re-colonization of the sediments by Thyasira sarsi after a period of oxygen depletion resulting in the death of the bottom fauna in the winter 1979Á1980 (Josefson & Widbom 1988). Dando & Spiro (1993) report high concentrations of total reduced sulphur in the Gullmar Fjord sediments in 1986, together with negative d 13 C values in the gills of collected specimens of T. sarsi, indicating a significant carbon input from autotrophic endosymbiotic bacteria. In addition, samples from the large North Sea pockmark are characterized by the presence of high total sulphide concentrations in the surface layers, as compared to surrounding areas (Dando et al. 1991). In particular, both samples R8 and S1, which have low DC2 values, host the bivalve T. sarsi with negative d 13 C values (Dando et al. 1991). Since macrobenthic communities associated with fjords, organic-enriched sediments and high sulphide methane seeps all occur in the lower part of the diagram and communities from open marine, oxygenated, environments in the upper part (Figure 7), the DC2 score may be a direct measure of the degree of sulphide concentration and an indirect measure of oxygen level in the soft-bottom sediments.
Moreover, the whale-fall and the Gullmar Fjord samples might have a similar faunal composition because at the time of sampling they were in a similar successional stage of faunal recovery. W1 was in fact collected 5 years after the impact of the carcass on the sea floor and GLF1Á3 were collected 5Á7 years Molluscs from a shallow-water whale-fall 13 after the hypoxic event that eliminated all the fauna from the Gullmar Fjord.

Whale-falls in shallow waters
The general picture that can be drawn from our study of north European shelf molluscs shows that the benthic community structure at the Kosterfjord whale-fall is similar to that of communities developed in other organic-rich, sulphide environments living at similar water depth. In comparison with whale-falls sunk just below the 200 m depth threshold in the northwest Pacific (Fujiwara et al. 2007), the Kosterfjord whale-fall lacks molluscs typical of deep-water reducing environments, such as mytilid mussels (Adipicola pacifica) and cocculinid limpets. Among the bathymodiolin mussels, Idas simpsoni has been described from trawled bones on the North Sea shelf (Marshall 1900;Warén 1991;Tebble 1996) and in oil-polluted areas (Hartley & Watson 1993;Southward 2008). However, during multiple ROV surveys and the careful examination of multiple recovered bones and sediment samples, I. simpsoni was never found at the Kosterfjord whale-fall.
One of the few other examples of the monitoring of a whale-fall in relatively shallow waters (385m depth, northeast Pacific) does not bear evidence of the development of a complex megafaunal community (Braby et al. 2007). Concerning benthic molluscs, during 6 surveys across a 13-month period, only rare buccinid gastropods were found in the sediments surrounding the carcass. The causes might be related to high scavenging rates and frequent disturbance from sediment flows at the whale-fall site (Braby et al. 2007).
These available case studies, which show different possible scenarios, suggest how the community structure and dynamics of whale-falls in shallow water merit substantial further study.

Conclusions
Our evidence suggests that the Kosterfjord whale-fall mollusc community is structured around species that exploit a variety of food sources on the continental shelf, including heterotrophs and chemoautotrophs. This result is similar to that recorded from a fossil analogue from the Pliocene of Italy (Dominici et al. 2009;Danise et al. 2010). However, it is in contrast with the data on the polychaete fauna. To date, a total of seven new species of polychaetes have been recorded from the Kosterfjord whale-fall, of which just three are present at other organic-rich settings such as fish farms (Glover et al. 2005;Wiklund et al. 2009a, b). This may well be partly because the polychaete fauna of organic-rich habitats is less well studied than the mollusc fauna, but the presence of specialists such as Osedax mucofloris at the Kosterfjord whale-fall site is in contrast with the pattern for Mollusca. From an ecological perspective, our data suggest that shelf-depth whale-falls are a natural analogue to areas of organic pollution, such as oil spillages and fish farms, and as such may well offer interesting insights into natural bioremediation at these habitats. From an evolutionary perspective, small carcasses at shelf-depths may provide an avenue for speciation in polychaetes, but not necessarily in molluscs.