To refer to this article use this url:

Contribution to Zoology, 76 (2) – 2007

Sponge diversity and community composition in Irish bathyal coral reefs

Rob W.M. van Soest1, Daniel F.R. Cleary1,2, Mario J. de Kluijver1, Marc S.S. Lavaleye3, Connie Maier3, Fleur C. van Duyl3

1.  Zoological Museum of the University of Amsterdam, P.O. Box 94766, 1090 GT Amsterdam, the Netherlands

2.  Institute for Biodiversity and Ecosystem Dynamics, Faculty of Science, University of Amsterdam, P.O. Box 94766, 1090 GT, Amsterdam, the Netherlands

3.  Royal Netherlands Institute for Sea Research, P.O. Box 59, 1790 AB Den Burg, Texel, the Netherlands

Keywords: coldwater, multivariate analysis, North Atlantic, ordination, PCA, Porifera, RDA


Sponge diversity and community composition in bathyal cold water coral reefs (CWRs) were examined at 500-900 m depth on the southeastern slopes of Rockall Bank and the northwestern slope of Porcupine Bank, to the west of Ireland in 2004 and 2005 with boxcores. A total of 104 boxcore samples, supplemented with 10 trawl/dredge attempts, were analyzed for the presence and abundance of sponges, using microscopical examination of (sub)samples of collected coral branches, and semi-quantitative macroscopic examination. Approximate minimum size of identified and counted sponge individuals was 1 mm. Literature data were added to the Porcupine Bank results to compensate for a less intensive sampling program in that location. Species richness and abundance were determined at local (sample diversity, pooled-sample diversity, local reef diversity), between-reef (diversity of two reef areas at 15 km distance), and regional scales (diversity of three reef areas over a distance of 200 km). Abiotic and biotic parameters including depth, the presence and cover of live coral, dead coral and sand, local reef, and orientation towards the nearest reef mound summit, were included in a constrained ordination technique (RDA); a Monte Carlo forward selection procedure was used to obtain significant predictors of variation in composition. The results of this analysis were compared with unconstrained ordination (PCA) and cluster analysis. The presence of live coral, depth and the local reefs C1 and C3 proved to be significant predictors of variation in sponge composition. The PCA and cluster analysis confirmed these results. Sample species richness was consistently heterogeneous from zero species and individuals up to 57 species and 90 individuals per (boxcore) sample. Species richness of local reefs determined from pooled samples showed the three localities studied to have similar species richness, namely 105-122 species in each location. Species richness was highest in samples with relatively low live coral cover. As in the RDA, live coral presence and depth appeared to be responsible for most of the variation observed in the cluster results. Cluster analysis of Bray-Curtis dissimilarity values of the pooled samples of all three reef localities using presence / absence data of all available samples indicated that distance appeared to structure the composition of the sponge assemblages of the three reef mound areas, but much less so within and among local reefs. Bathyal reefs of the regions to the west of Ireland were found to have a combined sponge species richness of 191 species, exceeding the richness of individual reef mound areas by c. 38-45%. Sponge presence in CWRs is clearly structured and controlled by biotic and abiotic factors. In particular, live coral presence appears a significant predictor of CWR sponge composition and diversity.


Northwestern European continental slopes and oceanic banks/seamounts contain extensive coral reef formations at depths of 200-1200 m (Freiwald and Roberts, 2005). Reefs are predominantly built by two scleractinian coral species, Lophelia pertusa and Madrepora oculata (with Stylaster gemmascens as an additional structural species). Elsewhere in the world, in similar environments, similar reefs exist consisting of the same or a few other species. The paucity of reef-building species in these coldwater reefs (CWRs) contrasts with high biodiversity and great abundance of suspension feeders, grazers, scavengers and predators (Jensen and Frederiksen, 1992), even rivaling tropical shallow water reef communities in faunal diversity.

The size and frequency of occurrence of these reef communities along the European coasts have previously been reported (e.g. Broch, 1922; Dons, 1944), but were largely ignored, until Wilson (1979) ‘rediscovered’ them. Since then, mapping of CWRs has received unprecedented attention from geologists and biologists. It appears that size and frequency increase from south to north along the continental margins of Europe, with the most extensive reefs reported from areas to the south and west of Ireland (De Mol et al., 2002), off northwestern Scotland (Roberts et al., 2000), the Far Oer (Bruntse and Tendal, 2001) and along the continental slope of Norway (Broch, 1922; Dons, 1944; Fosså et al., 2005). Southwards, along the coasts of France, the Iberian peninsula, and in the Mediterranean coral reefs are apparently only present in smaller formations (e.g. Duineveld et al., 2004; Alvarez-Perez et al., 2005; Taviani et al., 2005), but there is ample evidence for the existence of extensive reef formations of Pleistocene and early Holocene age in these areas (De Mol et al., 2005). This is consistent with a scenario of Holocene northward extension of previously Mediterranean-Atlantic CWRs, with concurrent extinction and dying off of reefs in the south. Possibly, this scenario is a repetition of similar scenarios in pre-Pleistocene ages, as borings in locations of the northern reefs have yielded reef remnants as old as 10 million years (De Mol et al., 2005).

The reefs along the continental slopes of Europe depend on variable abiotic factors, which are currently under study in many different research groups. The precise environmental parameters, however, which make a flourishing reef possible, have not (yet) been identified. Mapping of CWRs has demonstrated the occurrence of island-like reef mounds in the midst of highly sedimented deep sea gullies and plains. (Fosså et al., 2005; Foubert et al., 2005; Wheeler et al., 2005). This island-like distribution of reefs invokes questions about connectivity and interactions of inhabitants of different reefs. It is generally thought that reef faunas - both shallow-water and deepwater reefs - require larval replenishment from proximate reef locations. However, current directions in the Eastern North Atlantic (Bersch, 1995) are a complicated interplay of (1) eastward directed surface currents, (2) northward directed intermediate and bottomwater currents, as well as - in the northern North Atlantic and more to the west - (3) southward directed currents, and (4) strong residual tidal currents (White et al., 2005). Also, larval dispersal capabilities of reef organisms may be limited to scales of 10-100 km, even in groups that developed long-distance (teleplanic) larvae, as has been established for shallow-reef fishes (Cowen et al., 2006).

The present study focuses on sponges, rather than on an integral community approach (Jensen and Frederiksen, 1992; Mortensen and Fosså, 2006). Often integral community approaches focus on well-known taxa and fail to adequately resolve lesser-known taxa. Mortensen and Fosså’s (2006) comprehensive study of the species diversity and spatial distribution of invertebrates in Norwegian CWRs is a case in point; they mention the occurrence of 16 sponge species of which 5 were only identified to higher taxon, whereas a conservative estimate predicts that there may be 5 to 10 times more species in these reefs. Statements about numerical richness and abundance of various invertebrate groups made by these authors cannot be taken as definitive in view of their disregard for the true sponge diversity of Norwegian reefs.

Sponge diversity in tropical shallow-water coral reefs has been the subject of a considerable body of literature (see e.g. Hooper and Kennedy, 2002; Hooper et al., 2002; Cleary et al., 2005; Becking et al., 2006; De Voogd et al., 2006), in which estimates of local reef and regional diversity are presented, and attempts at explaining this diversity with abiotic and biotic factors are made (e.g. Wilkinson and Cheshire, 1989; Wilkinson and Evans, 1989; Schmahl, 1990; De Voogd et al., 2006). In contrast, sponge diversity in CWRs has been largely ignored, despite the widely noted extensive presence of sponges in these habitats (e.g. Bruntse and Tendal, 2001; Reitner and Hoffmann, 2003; Mortensen and Fosså, 2006). Sponges are thought to be vital components of the reefs as filter feeders capturing organic particles from prevailing currents, harboring microsymbiont communities with a plethora of unknown but likely important functions, and as excavating agents. Studies of sponge diversity and frequency of occurrence in CWRs are limited to two recent papers, van Soest and Lavaleye (2005) on a Rockall Bank CWR recorded 95 species and Longo et al., (2005) on a Mediterranean CWR recorded 39 species. A major setback for diversity studies of sponges in deeper water is the relatively complicated taxonomy of sponges, both in the sense of being technically demanding (high magnification microscopic examination is necessary) and the lack of easily accessible identification guides, such as are available for certain shallow-water environments (e.g. van Soest et al., 2000). This means, that generalist ecologists usually focus on megabenthic, colourful or otherwise characteristic species (e.g. Mortensen and Fosså, 2005), ignoring the smaller and cryptic sponges, which often are the most numerous.

In a previous publication, we presented, for the first time, quantitative data on sponge diversity and abundance in extensive CWRs on the southeastern slopes of the mid-oceanic Rockall Bank (van Soest and Lavaleye, 2005). This study revealed high sponge diversity (95 species), low sponge biomass, and an apparent lack of a clear association between coral cover and sponge diversity and abundance. In a follow-up study, presented below, we revisited the Rockall Bank CWR systems, as well as a CWR area on the northwestern Porcupine Bank on the Irish side of the Rockall Trench. In the present study we aim to assess the diversity and community composition of sponges across gradients of coral cover (both dead and alive), depth, local reef, and orientation towards local reef summits. We were in particular interested if the presence of live coral cover affected the composition of sponges. Our biodiversity study is intended to increase the understanding of the diversity of CWR biota along the coasts of Western Europe.

Materials and methods

Sponges were collected during two cruises on board the research vessel ‘Pelagia’, Moundforce 2004 (M2004; Mienis and de Haas, 2004; van Soest and Lavaleye, 2005) and BIOSYS/HERMES 2005 (B2005; Van Duyl and Duineveld, 2005),


Fig. 1. Maps of study areas. Left: overview of North Atlantic to the west of Ireland with southeastern Rockall Bank (squares) and N Porcupine Bank (P). Right: detail of southeastern Rockall Bank with CLAN (C) and HAAS (H) mound locations.


Fig. 2. HAAS mound areas H1, H2 and H3. A, topographic map with all sampling stations (triangles). B, schematized representation of HAAS mounds with detailed positions of exhaustively studied boxcore samples situated north of H1 summit (4/32, 4/32, 4/32A, 4, 10, 14, 48, 49), south of H1 summit (4/29, 4/30, 4/31, 4/34, 4/34A, 4/37, 105, 153), and W of H3 summit (4/33, 4/33A, 23, 28, 71, 92, 93, 96, 159, 161). Station numbers beginning with 4/ indicate samples taken in 2004, without 4/ samples taken in 2005. Scale bar is 0.8 nm (nautical miles).

organized by the Royal Netherlands Institute for Sea Research. Sponges were collected during both cruise programs, but mainly during the multidisciplinary research program BIOSYS, which had the study of live corals from CWRs as its main objective.

Study areas (Figs 1-4)

In the southeast Rockall Bank (Fig. 1; Moundforce and BIOSYS/HERMES target areas c. 15 km apart, collectively known as Lochachev Mound area), two sites (i.e., hereafter known as the ‘HAAS’ and ‘CLAN’ mounds; 55.4°N 15°W and 55.5°N 16°W respectively) were sampled. Both sites consisted of a complex of reef mounds separated by deeper gullies and plains (Figs 2 and 3). Average depths were 596.7 (range: 562-715 m) for HAAS and 750.5 m (range: 629-826 m) for CLAN mounds. Current directions at these sites were complex and variable over a 24 hour period. Nevertheless, preliminary data indicate that the prevailing current direction at least over the top of the mounds is southeasterly from


Fig. 3. CLAN mound areas C1, C2, C3, and C4. A. Topographic map with all sampling stations (triangles). B. Schematized representation of CLAN mounds with detailed positions of exhaustively studied boxcore samples situated S of C1 summit (4/40, 4/40A, 4/41, 4/41B, 4/41C, 4/44), W of (near the summit of) C2 (4/42 ,4/42A, 114, 115), and S of the summit of C3 (4/43, 4/43A, 37, 72, 73, 78, 168, 173). Station numbers beginning with 4/ indicate samples taken in 2004, without 4/ samples taken in 2005. Scale bar is 1.5 nm.


Fig. 4. Topographic map of N Porcupine Bank, showing the position of the stations. Scale bar is 1.0 nm.


Table 1. List of MOUNDFORCE, BIOSYS and HERMES stations taken in and around CWRs west of Ireland during 2004 and 2005 from which sponges were obtained, with data on their geographic position and depth, and (a)biotic content. (BX = Boxcore, DRE = Dredge, DTR = Trawl sample, H = HAAS Mounds, C = CLAN Mounds, Porc = Porcupine Bank, Orientation refers to nearest reef mound summit).

the shallower parts of Rockall Bank downwards (Van Duyl and Duineveld, 2005). Current strengths near the bottom varied over the investigated areas between 3 and 30 cm s-1 (Van Duyl and Duineveld, 2005). Temperature and salinity were also variable over a 24 hour period with an amplitude of 2°C (range: 8-10°C) and 1%o (34-35.5 %o). T and S values measured in an individual location, W-HAAS Mound (H1), on July 10 varied between 8 and 9°C and 35.22-35.36 %o, respectively.

At both sites 10 boxcore samples were taken in 2004 during the exploration of the Moundforce 2004 cruise, and in 2005 50 (HAAS) and 29 (CLAN) boxcore samples were taken during the BIOSYS 2005 cruise. In addition, 4 dredge and 2 trawl samples were taken in the sandy gullies surrounding the reef mounds (4 at HAAS mounds at 508-753 m and 2 at CLAN mounds at 750-863m). All sample locations are shown in Figs. 2 and 3. Samples from 2004 and 2005 were taken at least partially in approximately the same locations. Sample results from 2004 were used to attempt to find optimal live coral localities in 2005 in order to serve the main goals of the BIOSYS program, with limited success, however, since resampling exactly the same locations with the boxcores is extremely difficult.

Based on their positions on individual reef mounds and on their orientation North, West or South of the nearest reef summit, samples were also arranged in informal geographic groups dubbed H1-3 (Fig. 2) and C1-4 (Fig. 3) (see Table 1 for further details). In the case of exhaustively studied samples these groups were reduced to H1 and H3 and C1, C2 and C3 (Figs. 2 and 3); no exhaustively examined samples were available from H2 and C4.

In the northwestern Porcupine Bank (Fig. 4; HERMES target site, known as the Pelagia Mound area, a part of what is known as the Hovland Mound Province), two close-lying reef mounds at approximately 53°N 13°W were sampled by 6 boxcores (at 659-765 m) and 4 trawls/dredges (at 508-747 m). The number of observed species was low (see below), due to insufficient sampling, so additional sponge biodiversity data from the literature were added to our results derived from 37 dredge/trawl samples from nearby reef locations (51-54°N, 10-11°W, Stephens, 1915, 1921). Since these data were collected over a larger area and by methods different from the BIOSYS program, we only included results expressly stated to have come from Lophelia beds, and only for the purpose of determining regional diversity.

Boxcores consisted of cylinders that upon impact completely closed off both the under and upper sides, bringing up standard samples of the sediment, the epibenthic fauna and the near bottom ambient seawater. Boxcore samples were usually 50 cm in diameter (surface area approx. 2000 cm2), but occasionally 30 cm diameter boxcores were used (surface area 700 cm2). The content of each boxcore was photographed prior to any subsampling in order to record the cover of live and dead coral, other important biota and sand (Van Duyl and Duineveld, 2005: Appendix IIIa). In addition the samples were briefly characterized by dominant faunal components and dead remnants (Van Duyl and Duineveld, 2005: Appendix IIIb). Depth and geographic position were recorded for each sample (Table 1).

In each boxcore sample we identified and counted macroscopically recognizable sponges (see below). Of these sponges, only fragments or representative specimens were retained as voucher specimens. Approximately half the boxcore samples taken in 2005 were exhaustively studied. A subsample of the collected (dead) coral branches consisting of a minimum of 20 coral branches of 10 cm or more in length were taken out, and these were preserved in 96% alcohol for investigation of the microscopical sponges, both on board and in the lab in Amsterdam. All 2004 boxcore samples were similarly exhaustively studied (Van Soest and Lavaleye, 2005), including those with no or only a few coral branches. These exhaustively studied samples were used to relate variation in composition to environmental variables. The remaining samples were included only in the estimates of regional diversity and reef/non-reef diversity.

Boxcore photos were imported into the program ImageJ 1.28 (public domain software by W. Rasmand, National Institute of Health, USA, cf. for image analysis. Coral cover (both dead and live) and bare sediment cover as projection on the boxcore bottom surface were thus obtained (listed in Table 1), and abundance of the macroscopically recognizable sponges was determined.

Other samples were taken with a 3m Beam-trawl with a bottom time of 10 minutes and a 1 m triangular dredge with similar bottom time. Samples were treated the same as the boxcore samples. Trawl and dredge samples were only used for regional scale studies and for reef/non-reef comparisons.

Sponge identification

Species identification followed van Soest and Lavaleye, 2005 (thick sections were dried on microscopical slides and mounted in Canada balsam). Voucher material and preserved branches have been incorporated in the collections of the Zoological Museum of the University of Amsterdam. Sponge species were also classified in one of four trait categories (see van Soest and Lavaleye, 2005): (1) thin small hispid crusts, (2) hollow bladder-like forms, (3) massive, large highly silicified forms, and (4) other forms. All identified sponges are listed in Table 2. A table with species and their abundance per sample site is available upon request as supplementary material.


Species richness

Total species richness was assessed (using the program ESTIMATES! (Colwell, 2000) and ECOSIM 6 (Gotelli and Entsminger 2001) using only exhaustively studied samples (boxcores) per study site in order to examine how complete our sampling of the HAAS and


Table 2. List of sponge species, arranged in to systematic order, obtained from CWRs west of Ireland by MOUNDFORCE, BIOSYS and HERMES sample activities. Species are assigned to four growth form categories (crust, bladder, silica and other). Samples are pooled in informal clusters C1-4, H1-3 of proximate locations with similar orientation towards local reefmounds. See also maps of Figs 2-4. ‘x’ denotes occurrence of species over the clusters. Also, the total number of sponge individuals (N) of each species is given. (C1-4 = CLAN Mounds, H1-3 = HAAS Mounds, Porc = Porcupine Bank).

CLAN areas was. In the results we present sample and individual-based rarefaction estimates of cumulative species richness in addition to Chao2 and second order Jackknife nonparametric richness estimators (see Collwell and Coddington, 1994). The Chao2 estimator gave the least biased estimate of true species richness for small numbers of samples in a study of eight species richness estimators (Colwell and Coddington, 1994) and has been previously used to assess marine benthic diversity (Ellingsen, 2002). For estimates of individual-based rarefaction we used 1000 iterations and independent sampling of randomly chosen individuals from the total species pool in each area (HAAS or CLAN).

Ordination and Permanova

Community composition was assessed with Permanova (Anderson and Ter Braak 2003) using PRIMER 6 and PERMANOVA+ β3 (PRIMER-E Ltd), unconstrained ordination, i.e., Principal Components Analysis (PCA), and constrained ordination, i.e, Redundancy Analysis (RDA), using CANOCO for Windows 6.1 (Ter Braak and Smilauer, 1998). Input for the Permanova, PCA and RDA consisted of log10 (x+1) species abundance data that were first ‘transformed’ within the programme TRANSFORMATION (freely available at: Transformation consisted of modifying the species abundance data such that subsequent analyses, such as PCA and RDA, preserved the chosen distance among objects (transects). The species abundance data were transformed because of the inherent problems of the Euclidean-based distance metric (in standard PCA) and the Chi-square metric (in correspondence analysis, CA) for community data (see Legendre and Gallagher, 2001). In the present case, the Hellinger (Rao, 1995) distance was used, which gave very good results in a comparison of various distance metrics (Legendre and Gallagher, 2001). In particular, the Hellinger distance gave low weights to rare species, was monotonically related to the geographic distance along a model gradient, and reached an asymptote for sites with no species in common. It also produced little ‘horseshoe effect’ or inward folding of sites at opposite ends of the gradient, in ordinations (Legendre and Gallagher, 2001).

Using PERMANOVA we tested for an effect of live coral presence on community composition with a two-way Permanova; the factors were site (HAAS or CLAN) and live coral (present or absent). Live coral presence was included as a fixed factor in the analysis and area was included as a random factor. Constrained gradient analysis using RDA (within CANOCO for Windows v4; Ter Braak 1986; Ter Braak and Smilauer 1998) was used to assess environmental gradients in the species data matrix. RDA arranges sites and species in a multidimensional space whereby the axes are constrained to be linear combinations of environmental variables (Ohmann and Spies 1998). In RDA, the amount of species variation explained by the explanatory variables, or the total variation explained (TVE) is the sum of all constrained eigenvalues divided by the total variation (TV) in the species data. The en-vironmental dataset used in the RDA consisted of the following variables: (1) depth; (2) live coral present (yes: 1 or no: 0); (3) percentage live coral cover, (4) percentage dead coral cover; (5) percentage sand; (6) reef orientation (North, South or West coded as dummy variables); (7) reef (C1, C2, C3, H1, and H3 coded as dummy variables).

Within CANOCO a forward selection procedure using a Monte Carlo permutation test (999 permutations) and the full model option (Ter Braak and Smilauer 1998) was used to test environmental variables for significance (Ter Braak and Verdonschot 1995). In the results the conditional effects of environmental variables on composition (lambdaA ) are presented in addition to P values from the Monte Carlo test. The conditional effects (lambdaA ) represent the additional fit or increase in eigenvalue with each consecutively selected environmental variable. Only variables with P < 0.10 were included in the final model. The significance of the association between the species and environmental datasets was also assessed using Monte Carlo simulations (999 permutations) of constrained ordination scores against environmental variables. This analysis was performed only with samples from HAAS and CLAN mounds, only at the sample scale, only with exhaustively studied samples and only with samples containing 10 or more individuals.

Cluster analysis

As a complement to the PCA and RDA we performed a cluster analysis with the Bray-Curtis dissimilarity coefficient with subsequent UPGMA classification, using the MVSP package (Kovach, 1999; de Kluijver, 1997). Clusters were formally analyzed for the presence and dominance of species confined to each cluster using the process called ‘inverse analysis’ (full data matrix and inverse analysis results are available upon request). Obtained clusters were compared informally with cover of live coral, dead coral, sediment, depth and geographic position. This approach has been criticized for its neutral effects: data are clustered, but what the clusters mean is not formally analyzed. Nevertheless, empirical evidence (de Kluijver, 1997) suggests this approach is quite useful for detecting meaningful patterns.

Regional scale diversity was determined by cluster analysis of combined HAAS, CLAN and Porcupine samples, including trawl and dredge samples (the full data matrix is available upon request). Data from Stephens (1915, 1921) were included in the Porcupine data to complement the insufficient sampling effort during BIOSYS/HERMES 2005. Reef/non-reef samples were evaluated only for species contents, due to incomparable sampling devices and number of sampling attempts.

Finally, we used RLQ analysis to relate growth forms (excluding the growth form ‘other’) to environmental variables (Dolédec et al., 1996; Cleary et al., in the press). RLQ analysis is an extension of coinertia analysis; ‘R’ is a table of p environmental variables at m locations, ‘L’ a contingency table representing the abundance of k species at m locations, and ‘Q’ a table of k species with n biological traits. With RLQ analysis both traits and environmental variables affected by disturbance, as well as their interrelationships can be assessed (Dolédec et al., 1996). A detailed description of the analysis can be found in Dolédec et al., (1996), Cleary and Renema 2007 and De Voogd and Cleary 2007.


Coldwater reef sponges from the region W of Ireland

A total of 157 species (Table 2) and 2565 individuals (see supplementary material) were identified during the M2004 and B2005 cruises. Based on an overview comparison of our data with those listed in the World List of Porifera (van Soest et al., 2006 at, the composition of the CWR sponge fauna deviates from shallow-water sponge faunas in general by a relative predominance of Myxillina species (especially Hymedesmia), Calcineans, hexasterophoran Hexactinellida, and a poor representation of Dictyoceratida, Haplosclerina, and Calcaroneans.

Species richness

Boxcore samples contained between zero and 100 % live coral cover with a similar range in dead coral cover, and bare sand (see Table 1). Sponge species richness varied from zero to 57 species and sponge abundance from zero to 90 individuals (Table 1), demonstrating substantial small-scale spatial heterogeneity.

The full data matrix of species presence and abundance in individual samples is available upon request as supplementary material. Evenness (Pielou’s J) values of 41 exhaustively studied boxcores varied from 0.74-1.0 (average 0.914), which confirms and slightly exceeds values obtained by van Soest and Lavaleye (2005).

Pooled samples based on nearby similarly oriented locations taken at individual reef mounds (C1-5, H1-H3, see figs 2B and 3B and Table 2) yielded similar heterogeneity; average species richness and abundance varied from 11 to 86 species and 26-349 individuals in exhaustively studied sets of 2-8 samples (results not shown).

Of the three reef mound complexes we investigated, there appeared to be some difference between HAAS and CLAN with HAAS somewhat richer based on the number of samples taken and using non-parametric richness estimators (Chao2 and second order Jackknife). With a similar number of samples (18) we found 7 more species in HAAS than CLAN (109 versus 102) (Fig. 5). The Chao2 and Jackknife estimates showed even greater disparity (Chao2: 157.7 versus 142.6 and Jackknife: 170.4 versus 157.8) (Fig. 5). Note, however, that both nonparametric richness estimators showed no evidence of reaching an asymptote so that actual species richness in both areas should be considerably higher than observed species richness. The difference, however, between both areas will probably turn out to be less pronounced as can be seen in the individual-based rarefaction curve for both areas (Fig. 6), which shows very little difference. The value for CLAN (100.4 ± 2.4) is in fact well within the 95% confidence intervals of the value for HAAS (102.8 ± 6.3) at a similar number of individuals (n = 500 individuals).

Because Porcupine Bank was clearly understudied we pooled our data with Stephens’ (1915, 1921) sponge data obtained from 37 trawl/dredge samples on Lophelia banks to the east of our study site (51-54°N 10-11°W). The B2005 Porcupine Bank value (39 species) was obviously the result of insufficient sampling (with only 10 samples taken), but when pooled with Stephens’ data, a similar number of species were observed (105 species) as found at both Rockall localities (see Table 3). The aggregate observed species richness of all three investigated reef complexes, including Stephens’ data, is 191 species exceeding local reef diversity by 38-45%.


Table 3. Species richness of pooled M2004 and B2005 CLAN and HAAS Rockall Bank samples and pooled B2005 Porcupine Bank samples and Porcupine literature data (Stephens, 1915, 1921), showing aggregate species numbers in the same size range.


Fig. 5. Observed (Sobs) and estimated species richness of HAAS and CLAN mounds based on the number of samples taken and based on non-parametric richness estimators (Chao2 and second order Jackknife, Jackknife2).


Fig. 6. Individual-based rarefaction curve for species richness of HAAS and CLAN mounds.

Community composition

There was both a significant effect of live coral presence (Pseudo-F = 1.987, P < 0.001) and site (Pseudo-F = 1.437, P = 0.031) on the composition of sponges, although the contribution of live coral presence (components of variation: estimate: 0.099, square root: 0.315) was substantially larger than site (components of variation: estimate: 0.029, square root: 0.169).

In the RDA, the sum of all constrained (canonical) eigenvalues was 0.190 thus explaining 19% of the variation in the dataset (Figs 7 A and B). The species-environment correlations of the first four axes were high (range: 0.872 to 0.917) indicating a strong association between the species matrix and the environmental matrix. A Monte Carlo test showed the first axis and trace (sum of all canonical eigenvalues) to explain a highly significant amount of variation in community composition (p < 0.001 for both tests). Significant environmental variables selected with the Monte Carlo forward selection procedure included the presence of live coral (lambdaA = 0.08, P < 0.001), depth (lambdaA = 0.04, P = 0.010), C1 (lambdaA = 0.04, P = 0.010), and C3 (lambdaA = 0.03, P = 0.071). Ordinations of the RDA constrained using environmental variables are presented in Figs 7 A and B whereby


Fig. 7. Ordination based on redundancy analysis showing site scores for the first and second axes. Eigenvalue axis 1: 0.086. Eigen-value axis 2: 0.044. Arrows represent significant environmental variables superimposed onto the ordination; the length of the arrow indicates the correlation between the environmental variable and the ordination axis. Species codes in (a) represent: Ac-er: Acanthella erecta, Ap-be: Aphrocallistes beatrix, Ap-ye: Aplysilla yellow, Br-sc: Breitfussia schulzei, Ch-no: Chelonaplysilla noevus, Cl-as: Clathrina ascandroides, Cl-bi: Clathrina biscayae, Cl-ge: Cladorhiza gelida, Cl-re: Clathrina reticulum, Cy-sp: Cyamon spinispinosum, De-in: Desmacella aff. inornata, De-pe: Desmoxya pelagiae, Ep-sp: Epipolasis spissa, Er-nu: Erylus nummulifer, Ha-fl: Haliclona (Gellius) flagellifera, Ha-fo: Haliclona (Gellius) foraminosa, Ha-si: Haliclona (Gellius) sp. sigmas, He-de: Hexadella dedritifera, Hi-th: Higginsia thielei, Hy-he: Hymedesmia (Hymedesmia) helgae, Hy-ko: Hymedesmia (Hymedesmia) koehleri, Hy-mu: Hymedesmia (Hymedesmia) mucronata, Hy-sp: Hymeniacidon/Parasyringella sp., Hy-ve: Hymeraphia verticillata, Le-as: Leuclathrina asconoides, Li-di: Lissodendoryx diversichela, My-ma: Mycale (Raphidotheca) marshallhalli, My-tu: Mycale (Aegogropila) tunicata, Pl-am: Plocamionida ambigua, Ps-fi: Psammoclema finmarchicum, Ro-no: Rossella nodastrella, Sc-no: Sceptrella normani, To-gl: Topsentia glabra.


Fig. 8. Principal components ordination of the species abundance matrix showing (A) axes 1 and 2 and (B) axes 3 and 4. Eigenvalue axis 1: 0.116. Eigenvalue axis 2: 0.077. Samples with no live coral (No) and live coral (Yes) are indicated by symbols. Note that most samples with no live coral were sampled in 2004. Interestingly the only sample (C-77) with no live coral sampled in 2005 grouped with the other samples with live coral present. Alternatively samples collected in 2004 with live coral (C-41*, C-43* and H33*) grouped together with the other samples collected in 2005 with live coral.arrows represent significant environmental variables superimposed onto the ordination; the length of the arrow indicates the correlation between the environmental variable and the ordination axis.

Species strongly associated with live coral cover included Hexadella dedritifera , Lissodendoryx diversichela and Hymeraphia verticillata (see PCA Axis 1 in Fig. 8A as well as RDA Axis 1 in Fig. 7A) whereas species associated with reefs with no live coral include Cyamon spinispinosum, Desmoxya pelagiae and Acanthella erecta.

Species associated with shallower samples (axis 2 in Fig. 7B) included Rossella nodastrella, Clathrina biscayae and Haliclona (Gellius) flagellifera whereas species associated with deeper samples include Hymedesmia (Hymedesmia) helgae, Cladorhiza gelida and Higginsia thielei.

The significant separation of sites in C1 and C3 when compared to other sites is related to the occurrence of Cladorhiza gelida, Higginsia thielei and Hymedesmia (Hymedesmia) helgae in C3, and Sceptrella normani and Psammoclema finmarchicum in C1, supported by both the PCA and RDA (Figs. 7B and 8B); note that the orientation towards the nearest reef top is similar for both areas. In addition to local environmental conditions this effect may be related to the year in which samples were taken (C1 was predominantly sampled in 2004, C3 predominantly in 2005). Longterm and detailed studies, however, would be required to ascertain a temporal effect on sponge composition.

Cluster analysis: (Table 4; Fig. 9; supplementary material) yielded in addition to a number of unexplained single sample or two-sample clusters, all from 2004,


Fig. 9. Results of the cluster analysis of Bray-Curtis dissimilarity coefficient values obtained from 41 individual exhaustively studied boxcore samples with subsequent UPGMA classification.


Fig. 10. Results of the cluster analysis of Bray-Curtis dissimilarity coefficient values obtained from all available boxcore and trawl/dredge samples - pooled for each of the three locations CLAN-, HAAS-Mound and PORCUPINE (including literature data) - with subsequent UPGMA classification.

two major clusters I and II (Fig. 9) at approx. 0.8 dissimilarity, which appear to a high extent to conform to the presence (I) or absence (II) of live coral in the samples (see Table 4). However, as in the ordination results, the cluster representing zero live coral presence also consisted predominantly of samples from 2004, yielding a mixture of signals of the sample year and live coral presence. A further subdivision of the live coral presence cluster appears to be possible along a depth gradient: deeper, (average: 728 m; range: 644 - 777) (IB), vs shallower (average: 632 m; range: 568 - 778) (IA). The deeper cluster IB has less species per boxcore (average: 14.8; range: 10 - 28) than the shallower IA cluster (average: 25.7; range: 13 - 57) (Table 4; Fig. 9). As in the ordination, the samples from HAAS and CLAN appear to be mixed, therefore, the 15 km distance between both areas does not appear to explain much variation in the observed compositional patterns.

The inverse analysis (de Kluijver, 1987; not shown, but available upon request as supplementary material) identifies many species that support the clusters by only a low frequency of occurrence. Exceptions are Hexadella dedritifera and Hymeraphia verticillata (characteristic and dominant for cluster I), and Cyamon spinispinosum, Chelonaplysilla noevus and Haliclona (Gellius) flagellifera (characteristic and dominant for clusters I and II combined). No species appear to dominate the subclusters IA and IB.

Regional diversity. - Pooled presence / absence data of each of the three reef areas (boxcore and dredge/trawl samples combined) were subjected to cluster analysis (datamatrix available upon request). The result – HAAS and CLAN locations are less dissimilar than each is to Porcupine Bank (Fig. 10) - indicates that distance may be an important predictor of similarity at larger spatial scales.

Reef / non-reef diversity

Both boxcore and trawl/dredge samples of all three reef complexes were pooled and compared to all samples obtained from sedimented non-reef areas. Since the number of samples from the sedimented gullies and plains was much lower than that of the reef mounds, only a comparison of the species content was considered meaningful. Gullies and

plains yielded 44 species, of which 6 were unique to these habitats, the remaining 38 species dwelling on rocks and stones were almost in their entirety also found on the reefs.

Growth forms

There was no significant association between growth forms and environmental variables (Monte Carlo test: P = 0.138). Spcies with different growth forms appeared to occur in similar proportions in each of the habitats and reef mound areas (results not shown, but may be inferred from Table 2).


Presence of live coral and occurrence of sponge species

The presence of live corals appears to be an important predictor not only of sponge composition (Fig. 7), but also of diversity. The latter is also evident from a comparison of the species richness of samples with and without live coral present (Fig. 11). The largest number of sponge species were, however, found in the lower live coral cover range, which is in accordance with findings of Mortensen and Fosså (2006) for overall faunal diversity of Norwegian reefs. Certain abundant sponge species, e.g., Hexadella dedritifera, appear to be closely associated with the presence of live coral. If this association should be confirmed in future studies, then their interaction is possibly important for the dynamics of the CWRs. As with many sponges, species inhabiting CWRs encrust dead corals, but also share their preferred habitat with live corals, presumably optimal environments for capturing food and oxygen. We have not found specific evidence that they compete with live corals for space in this habitat, but it cannot be excluded that they influence the performance and longevity of live corals along with some other invertebrates. Coral-overgrowth capabilities of certain sponges have been empirically and experimentally (e.g. Aerts and van Soest, 1997) demonstrated in shallow-water reefs, and it is quite likely such sponges occur in CWRs as well, with Hexadella dedritifera a likely candidate. In the northeast Pacific the sponge, Desmacella austini (Lehnert et al., 2005) encrusted dead skeletons of reef-building hexactinellid sponges and occasionally overgrew live individuals of the reef-builder. In another study (Van Soest et al., submitted) we demonstrated that the occurrence of another common species Rossella nodastrella, is negatively related to the presence of live coral, possibly a result of competitive exclusion.

Species composition and habitat

In the redundancy analysis, depth and presence of live coral were identified as the most important predictors of spatial variation in sponge composition. Depth was also identified as a structuring factor in the Mediterranean CWR studied by Longo et al., 2005. It is interesting to note that depth appears to be important for both deep-water communities and


Fig. 11. Average sponge species richness of exhaustively studied boxcore samples without (left column, N=14) and with (right column, N=27) live corals present in the samples. The difference is significant (Mann-Whitney U-test, p < 0.05).

shallow-water sponge communities (cf. De Voogd et al., 1999, 2006; Alcolado et al., 2004), although the causal relationship in deep-water remains unexplained, whereas the reasons for depth zonation in shallow-water are obvious. Bare sediment devoid of dead or live corals is an obvious negative factor for most sponges. The comparison of species composition between reefs and the surrounding sedimented gullies and plains led to the conclusion that the latter contain an impoverished ‘reef’ fauna and do not form a separate habitat for sponges. Alternatively, the results could also indicate that the term ‘reef’ sponges is fictional and that sponges merely prefer the concentrated hard substrate in the form of dead coral branches. They may thus merely profit from the same favorable abiotic circumstances that favor coral growth, i.e. increased food supply. A vital role for sponges in the reef habitat and a more than casual relationship between reefs and sponges still needs to be established.


Table 4. Results of the cluster analysis of Bray-Curtis dissimilarity coefficient values obtained from 41 exhaustively studied boxcore samples with subsequent UPGMA classification (see Fig. 12 for a graphic representation of the results). (Shannon = Shannon’s H’, Evenness is Pielou’s J’, provided by the MVSP Package, Kovach, 1999).

Growth forms

The four growth forms distinguished in this study appeared to occur in similar proportions in each of the reef mound areas. They do not seem to represent adaptations to different environmental regimes as was found by Bell and Barnes (2000) for sponges (and for within-species growth forms by Bell et al., (2002) in shallow-water habitats subjected to varying current regimes. Possibly, the deep-water growth forms represent different evolutionary solutions to the deep sea environmental regimes not present in the shallow-water habitats studied by Bell and Barnes (l.c.). An important factor is likely to be the lack of a steady food supply. One can imagine that thin hispid crusts and hollow bladder-like forms both do not need a steady high input of organic matter, and even the large sponge forms with their high inorganic silica content may survive for long periods with little organic food. In any case, most of the sponges in the studied CWRs are elsewhere also known only to occur in the indicated growth forms and do not appear to represent underdeveloped recruits from wide-ranging species.

Species richness of shallow-water reefs and CWRs

De Voogd et al. (2006) studied Indonesian shallow-water sponge diversity at all the levels addressed here. Their transect (14-77 species) and total species richness values (151-169 species) appear quite similar to the values obtained from our CWRs (Table 3). Among the factors that were identified by them as important for community composition are depth, live coral cover and distance, all of which also appear to be important in CWRs. There is however a major factor preventing a comparison based on equal conditions, because species forming thin encrustations, by far the most common growth form among CWR sponges (Van Soest and Lavaleye, 2005), were not included in De Voogd et al.’s study. Australian coral reef sponge diversity studies (Hooper and Kennedy, 2002) yielded individual reef diversity values of 43-83 sponge species, based on 8 small-scale regional reef-complexes, with overall regional species richness of > 250 species, showing again values in the same range as observed by us. However, as with De Voogd et al., (2006), thin encrustations were not included.

Although less species rich than these Indo-Pacific regions (e.g. van Soest, 1994), Caribbean reef sponge studies include several in which thin encrustations were included (e.g. Kobluk and van Soest, 1989; Meesters et al., 1991). Thus, species richness values from these reefs appear more directly comparable to our CWR studies. Species richness values from Caribbean studies, added to inventories of local species including larger growth forms (e.g. van Soest, 1981; Lehnert and van Soest, 1998; Alcolado et al., 2004), obtain values of 150-250 species for local island reefs (Curaçao, Jamaica, Cuba). CWR sponge diversity at local and regional scale appears to be in the same range as that of shallow-water reef sponge diversity. The present study thus confirms earlier studies (Levin et al., 2001) that deep-sea habitats contain a well-structured and dynamic fauna.

Sampling gear

The boxcore used in this study, developed by the NIOZ and employed here as a routine gear for CWR biodiversity research, have a great advantage over other sampling gear because of precisely known sample size combined with the absence of depth restrictions. Living or recently dead coral beds built on a sedimented coral rubble foundation are easily sampled quantitatively, including living material, with a minimum of damage or interference. A distinct disadvantage is that rocky or stony bottoms have to be avoided, so a prior video survey of the bottom is advised (Van Duyl and Duineveld, 2005).

Sample size

Chao2 and Jacknife 2 non-parametric richness estimates indicate that with 18 boxcore samples the observed species richness is still far from the expected richness. Few marine biodiversity studies have yet to demonstrate that asymptotes are reached in rarefaction curves, so we are confident our analysis is credible and state of the art. Our species richness estimates also confirm earlier computations using Kaandorp’s (1986) minimal area analysis (Van Soest and Lavaleye, 2005). For future studies using species richness and composition as indicators for connectivity and estimates of reef health, local reefs need more than 18 boxcores to be characterized, which is a considerable effort in a deep-ocean habitat.


The material of this research was collected with grants from the EUROMARGIN Programme of the European Science Foundation (Moundforce 2004 Project, 813.03.006/855.01.040), the Netherlands Organisation for Scientific Research (BIOSYS project 814.01.005/835.20.024), and the EU HERMES Project (contract no. GOCE-CT-2005-511234). The following colleagues and shipmates provided assistance: Drs Gerard Duineveld (NIOZ), Drs Magda Bergman (NIOZ), Dr Henk de Haas (NIOZ), Mr Arthur Palacs (International University, Bremen, Germany), and the captain and the crew of RV’Pelagia’ (NIOZ).


Aerts LAM, van Soest RWM. 1997. Quantification of sponge/coral interactions in a physically stressed reef-community, NE Colombia. Marine Ecology Progress Series 148: 125-134.

Alcolado PM, Grovas-Hernandez G, Marcos Z. 2004. General comments on species inventory, fisheries, culture and some community features of the Porifera in Cuba. Bollettino dei Musei e degli Instituti Biologici dell Universita di Genova 68: 175-186.

Alvarez-Perez G, Busquets P, de Mol B, Sandoval NG, Canals M, Casamor JL. 2005. Deep-water coral occurrences in the Strait of Gibraltar. In: Freiwald A, Roberts JM, eds, Cold-water corals and ecosystems. Springer, Berlin, Heidelberg, 207-221.

Anderson MJ, ter Braak CJF. 2003. Permutation tests for multi-factorial analysis of variance. Journal of Statistical Computation and Simulation 73: 85-113.

Becking LE, Cleary DFR, de Voogd NJ, Renema W, de Beer M, van Soest RWM, Hoeksema BW. 2005. Beta-diversity of tropical marine benthic assemblages in the Spermonde Archipelago, Indonesia. Marine Ecology 27: 76-88.

Bell J, Barnes DKA. 2000. The influences of bathymetry and flow regime upon the morphology of sublittoral sponge communities. Journal of the Marine Biological Association of the United Kingdom 80: 707-718.

Bell J, Barnes DKA, Turner JR. 2002. The importance of micro and macro morphological variation in the adaptation of a sublittoral demosponge to current extremes. Marine Biology 140: 75-81.

Bersch M. 1995. On the circulation of the northeastern North Atlantic. Deep-Sea Research (I) 42: 1583-160.

Broch H. 1922. Rifkorallen im Nordmeer einst und jetzt. Die Naturwissenschaften 37: 1-3.

Bruntse G, Tendal OS, eds. 2001. Marine biological investigations and assemblages of benthic invertebrates from the Faroe islands. Kaldbak Marine Biological Laboratory (Faroe): 1-42.

Cleary DFR, Becking LE, de Voogd NJ, Renema W, de Beer M, van Soest RWM, Hoeksema BW. 2005. Variation in the diversity and composition of benthic taxa as a function of distance offshore, depth and exposure in the Spermonde Archipelago, Indonesia. Estuarine Coastal Shelf Science 65: 557-570.

Cleary DFR, Renema W. 2007. Relating species traits of foramini-fera to environmental parameters in the Spermonde Archipelago, Indonesia. Marine Ecology Progress Series 334: 73-82.

Colwell RK. 2000. EstimateS version 6.0b1: Statistical estimation of species richness and shared species from samples. Freeware published at

Colwell RK, Coddington JA. 1994. Estimating terrestrial biodiversity through extrapolation. Philosophical Transactions of the Royal Society London, series B 345:101-118.

Cowen RK, Paris CB, Srinivasan A. 2006. Scaling of connectivity in marine populations. Science 311: 522-527.

De Mol B, van Rensbergen P, Pillen S, van Herreweghe K, van Rooij D, McDonnell A, Huvenne V, Ivanov M, Swennen R, Henriet JP. 2002. Large deep-water coral banks in the Porcupine Basin, southwest of Ireland. Marine Geology 188: 193-291.

De Mol B, Henriet JP, Canals M. 2005. Developments of coral banks in Porcupine Seabight: do they have Mediterranean ancestors? In: Freiwald A, Roberts JM, eds. Cold-water corals and ecosystems. Springer, Berlin, Heidelberg, 515-533.

De Voogd NJ, Cleary DFR. 2007. Relating species traits to environmental parameters in Indonesian coral reef sponge assemblages. Marine and Freshwater Research 58: 240-249.

De Voogd NJ, Cleary DFR, Hoeksema BW, Noor A, van Soest RWM. 2006. Sponge beta diversity in the Spermonde Archipelago, SW Sulawesi, Indonesia. Marine Ecology Progress Series 309: 131-142.

De Voogd NJ, van Soest RWM, Hoeksema BW. 1999. Cross-shelf distribution of Southwest Sulawesi reef-sponges. Memoirs of the Queensland Museum 44, 147-154.

Dolédec S, Chessel D, ter Braak CJF, Champely S. 1996. Matching species traits to environmental variables: a new three-table ordination method. Environmental and Ecological Statistics 3: 143-166.

Dons C. 1944. Norges korallrev. Kongelige Norske Videnskabers Selskabs Forhandlinger 16: 37–82.

Duineveld GCA, Lavaleye MSS and Berghuis EM. 2004. Particle flux and food supply to a seamount cold-water coral community (Galicia Bank, northwestern Spain). Marine Ecology Progress Series 277: 13-23.

Ellingsen KE. 2002. Soft-sediment benthic biodiversity on the continental shelf in relation to environmental variability. Marine Ecology Progress Series 232: 15-27.

Ellingsen KE, Gray JS 2002. Spatial patterns of benthic diversity: is there a lattudinal gradient along the Norwegian continental shelf. Journal of Animal Ecology 71: 373-389.

Fosså JH, Lindberg B, Christensen O, Lundälv T, Svellingen I, Mortensen PB, Alvsvåg J. 2005. Mapping of Lophelia reefs in Norway: experiences and survey methods. In: Freiwald A, Roberts JM, eds, Cold-water corals and ecosystems. Springer, Berlin, Heidelberg, 359-391.

Foubert A, Beck T, Wheeler AJ, Opderbecke J, Polarstern ARK-XIX/3a Shipboard Party. 2005. New view of the Belgica Mounds, Porcupine Sea Bight, NE Atlantic: preliminary results from the Polarstern ARK-XIX/3a ROV cruise. In: Freiwald A, Roberts JM, eds, Cold-water corals and ecosystems. Springer, Berlin, Heidelberg, 403-415.

Freiwald A, Roberts JM, eds. 2005. Cold-water corals and ecosystems.Springer (Berlin, Heidelberg), i-xxxi, 1243.

Gotelli NJ, Entsminger GL. 2001. EcoSim: Null models software for ecology. Version 6.0. Acquired Intelligence Inc. & Kesey-Bear,

Hooper JNA, Kennedy JA. 2002. Small-scale patterns of sponge biodiversity (Porifera) on Sunshine Coast reefs, eastern Australia. Invertebrate Systematics 16: 637-653.

Hooper JNA, Kennedy JA, Quinn RJ. 2002. Biodiversity ‘hotspots’, patterns of richness and endemism, and taxonomic affinities of tropical Australian sponges (Porifera). Biodiversity Conservation 11: 851-885.

Jensen A, Frederiksen R. 1992. The fauna associated with the bank-forming deep-water coral Lophelia pertusa (Scleractinia) on the Faroe shelf. Sarsia 77: 53-69.

Kobluk DR, van Soest RWM. 1989. Cavity-dwelling sponges in a souther Caribbean coral reef and their paleontological implications. Bulletin of Marine Science 44: 1207-1233.

Legendre P, Gallagher ED. 2001. Ecologically meaningful transformations for ordination of species data. Oecologia 129: 271-280.

Le Goff-Vitry MC, Pybus OG, Rogers AD. 2004. Genetic structure of the deep-sea coral Lophelia pertusa in the northeast Atlantic revealed by microsatellites and internal transcribed spacer sequences. Molecular Ecology 13: 537-549.

Lehnert H, Conway KW, Vaughn Barrie J, Krautter M. 2005. Desmacella austini sp. nov. from sponge reefs off the Pacific coast of Canada. Contributions to Zoology 74: 265-270.

Lehnert H, van Soest RWM. 1998. Shallow-water sponges of Jamaica. Beaufortia 48: 71-103.

Levin LA, Etter RJ, Rex MA, Gooday AJ, Smith CR, Pineda J, Stuart CT, Hessler RR, Pawson D. 2001. Environmental influences on regional deep-sea species diversity. Annual Review of Ecology and Systematics 32: 51-93.

Longo C, Mastrototaro F, Corriero G. 2005. Sponge fauna associated with a Mediterranean deep-sea coral bank Journal of the Marine Biological Association of the United Kingdom 85: 1341-1352.

Meesters E, Knijn R, Willemsen P, Pennartz R, Roebers G, van Soest RWM. 1991. Sub-rubble communities of Curaçao and Bonaire coral reefs. Coral Reefs 10: 189-198.

Mienis F, de Haas H. 2004. Report of the cruise ‘Moundforce 2004’ with Royal R.V. ‘Pelagia’, cruise 64PE229, Cadiz-Galway, 15 August – 9 September. Royal NIOZ, Texel: 1-99.

Mortensen PB, Fosså JH. 2006. Species diversity and spatial distribution of invertebrates on deep-water Lophelia reefs in Norway. Proceedings of the 10th International Coral reef Symposium (Okinawa): 1849-1868.

Ohmann JL, Spies TA. 1998. Regional gradient analysis and spatial patter of woody plant communities of Oregon forests. Ecological Monographs 68: 151-182.

Rao CR. 1995. A review of canonical coordinates and an alternative to correspondence analysis using Hellinger distance. Qüestii 19: 23-63.

Reitner J, Hoffmann F. 2003. Porifera-Zonierungen in Kaltwasser-Korallenriffen (Sula-Rücken, Norwegen). In: Gradstein SR, Willmann R, Zizka G, eds, Biodiversitätsforschung: Die Entsclüsselung der Artenvielvalt in Raum und Zeit. Kleine Senckenberg-Reihe 45. Schweizerbart’sche Verlagsbuchhandlung (Stuttgart): 75-87.

Reveillaud, J. 2005. What sponges can tell us about cold-water coral reefs. On-line HERMES News Update.

Roberts JM, Harvey SM, Lamont PA, Gage JA. 2000. Seabed photography, environmental assessment and evidence for deep-water trawling on the continental margin west of the Hebrides. Hydrobiologia 44: 173-183.

Schmahl G. 1990. Community structure and ecology of sponges associated with four South Florida coral reefs. In: Rützler K, ed, New perspectives in sponge biology. Smithsonian Institution Press (Washington), 376-383.

Stephens J. 1915. Sponges of the west coast of Ireland. I. The Triaxonia and part of the Tetraxonida. Fisheries Ireland Scientific Investigations 1914 (4): 1-43.

Stephens J. 1921. Sponges of the coasts of Ireland. II. The Tetraxonida (concluded). Scientific Investigations of the Fisheries Branch. Department of Agriculture for Ireland 1920: 1-75.

Taviani M, Freiwald A, Zibrowius H. 2005. Deep coral growth in the Mediterranean Sea: an overview. In: Freiwald A, Roberts JM, eds, Cold-water corals and ecosystems. Springer, Berlin, Heidelberg: 137-156.

Ter Braak CJF. 1986. Canonical correspondence analysis: a new eigenvector technique for multivariate direct gradient analysis. Ecology 67: 1167-1179.

Ter Braak CJF, Smilauer P. 1998. Canoco 4. Microcomputer Power. Ithaca, NY, USA.

Ter Braak CJF, Verdonschot PFM. 1995. Canonical correspondence-analysis and related multivariate methods in aquatic ecology. Aquatic Science 57: 255-289.

Van Duyl FC, Duineveld GCA. 2005. BIOSYS-HERMES 2005 cruise report with R.V. Pelagia. Cruise 64PE 238, Galway-Texel, 21 June-21 July 2005. Biodiversity, ecosystem functioning and food web complexity of deep water coral reefs in the NE Atlantic (Rockall Bank and Porcupine Bank). Royal NIOZ, The Netherlands, 1-32, i-xxii.

Van Soest RWM. 1981. A checklist of the Curaçao sponges (Porifera Demospongiae) including a pictorial key to the more common reef-forms. Verslagen en Technische Gegevens van het Instituut van Taxonomische Zoölogie, Universiteit van Amsterdam 31: 1-39.

Van Soest RWM. 1994. Demosponge distribution patterns. In: van Soest RWM, van Kempen TMG, Braekman JC, eds. Sponges in time and space.Balkema, Rotterdam: 213-223.

Van Soest RWM, Lavaleye MSS. 2005. Diversity and abundance of sponges in bathyal coral reefs of Rockall Bank, NE Atlantic, from boxcore samples. Marine Biology Research 1: 338-349.

Van Soest RWM, van Duyl FC, Maier C, Lavaleye MSS, Beglinger E, Tabachnick KR. 2007 (in press). Mass occurrence of the hexactinellid Rossella nodastrella Topsent in bathyal coral reefs of Rockall Bank, W of Ireland. In: Custódio MR, Lôbo-Hajdu G, Hajdu E, Muricy G, eds. Trends in Porifera Research-Biodiversity, Innovation, Sustainability Universidade de Sao Paulo, Brazil. Proceedings 7th International Sponge Conference, Buzios, 6-13 May 2006.

Wheeler AJ, Beck T, Thiede J, Klages M, Grehan A, Monteys FX, Polarstern ARK-XIX/3a Shipboard Party. 2005. Deep-water coral coral mounds on the Porcupine Bank, Irish Margin: preliminary results from the Polarstern ARK-XIX/3a ROV cruise. In: Freiwald A, Roberts JM, eds, Cold-water corals and ecosystems. Springer, Berlin, Heidelberg: 393-402.

White M, Mohn C, de Stigter, Mottram G. 2005. Deep-water coral development as a function of hydrodynamics and surface productivity around the submarine banks of the Rockall Trough. In: Freiwald A, Roberts JM, eds, Cold-water corals and ecosystems. Springer, Berlin, Heidelberg: 503-514.

Wilkinson CR, Cheshire AC. 1989. Patterns in the distribution of sponge populations across the central Great Barrier Reef. Coral Reefs 8: 127-134.

Wilkinson CR, Evans EA. 1989. Sponge distribution across Davies Reef, Great Barrier Reef, relative to location, depth and water movement. Coral Reefs 8: 1-7.

Wilson JB. 1979. ‘Patch’ development of the deep-water coral Lophelia pertusa (L.) on Rockall Bank. Journal of the Marine Biological Association of the United Kingdom 59: 165-177.