Ashley A. Rowden, NIWA, Private Bag 14-901, Wellington, New Zealand. E-mail: firstname.lastname@example.org
Seamounts have often been viewed as specialized habitats that support unique communities; this notion has given rise to several hypotheses about how seamount ecosystems are structured. One, the ‘seamount oasis hypothesis’, predicts that invertebrates are more abundant, speciose and attain higher standing stocks on seamounts compared to other deep-sea habitats. Because this hypothesis has remained untested for biomass, we ask two questions: (i) Do seamounts support a higher benthic biomass than nearby slopes at corresponding depths? (ii) If they do, which particular taxa and trophic groups drive observed difference in biomass? Analysis of more than 5000 sea-floor images reveals that the mean biomass of epibenthic megafauna on 20 southwest Pacific seamounts was nearly four times greater than on the adjacent continental slope at comparable depths. This difference is largely attributable to the scleractinian coral Solenosmilia variabilis, whose mean biomass was 29 times higher on seamounts. In terms of trophic guilds, filter-feeders and filter-feeders/predators made up a significantly greater proportion of biomass on seamounts, whereas deposit feeders and those with mixed feeding modes dominated at slope habitats. Notwithstanding support for the seamount oasis hypothesis provided by this study, the hypothesis needs to be critically tested for seamounts in less productive regions, for seamounts with a greater proportion of soft substratum, and in other parts of the oceans where scleractinian corals are not prevalent. In this context, testing of seamount paradigms should be embedded in a broader ecological context that includes other margin habitats (e.g. canyons) and community metrics (e.g. diversity and body size).
Seamounts are topographically distinct sea-floor features, up to a million of which may exist globally (Pitcher 2007). Yet, despite their vast number and widespread distribution, the biota of very few seamounts have been well sampled (Stocks 2009). Nonetheless, scientific study has generated a range of hypotheses and paradigms about the ecological structure and function of seamounts, mostly related to the potential for seamounts to have an insular or island character and/or a morphology that generates particular environmental conditions. One hypothesis is that seamounts are biologically highly productive. This hypothesis originated from observations of higher abundance and biomass of fish associated with seamounts compared to elsewhere in the ocean (see review by Morato & Clark 2007). However, what originated as a paradigm for seamount fish has since been extended to include invertebrates, although without any direct evidence (McClain 2007). With the resurgence in seamount research in the late 1990s and early 2000s, the question of whether benthic invertebrates are relatively more abundant on seamounts was examined more critically.
Samadi et al. (2006) speculated that, like vent and seep habitats, seamounts could support populations of organisms with high abundances. They reasoned that seamounts were ‘places where a high trophic input allows an abundance of species and high population density [because] interactions between prominent topographic features and water masses increases turbulence and mixing, and enhances local biomass production by moving up nutrients in the euphotic zone’. This hypothesis, related to ones posed previously (e.g.Boehlert & Genin 1987), and an extension of a hypothesis posed and examined for chemosynthetic ecosystems in the deep sea (see Carney 1994), was formalized by Samadi et al. (2006) as the ‘seamounts as oases of productivity’ hypothesis (hereafter ‘seamount oasis hypothesis’). To date, this hypothesis has only been tested for larger invertebrates, using species richness as a proxy for abundance. Samadi et al. (2006) examined species richness data for squat lobsters (Galatheoidea) from Norfolk Ridge seamounts and an area of adjacent slope off New Caledonia (Southwest Pacific), and reported that ‘each individual seamount appeared to be richer than the restricted explored area on the slope, although a similar area to that of one seamount was sampled’. In another study, O’Hara (2007) found that brittle star (Ophiuroidea) species richness on seamounts was not elevated when compared to non-seamount areas. This author found that whilst seamounts can ‘exhibit high overall species richness for low number of samples … this did not increase with additional sampling at the rates found in non-seamount areas’ (O’Hara 2007). However, as noted by McClain (2007), testing the seamount oasis hypothesis by examining species richness is not ideal because diversity on seamounts could be controlled by competing explanations. McClain (2007) therefore suggested that future examinations of the seamount oasis hypothesis be restricted to tests for increased biomass.
The generalization that invertebrates are more productive, more numerous or attain higher standing stocks on seamounts has persisted largely because of observational data showing dense concentrations of filter-feeding organisms, such as corals, on the peaks of seamounts (see references in review by Rogers et al. 2007). Although McClain et al. (2009) did demonstrate that the frequency of occurrence of some species on Davidson Seamount was higher than in nearby Monterey Canyon, seamount data are very sparse, and thus the oasis hypothesis, as related to biomass, remains quantitatively untested. Consequently, the two questions addressed here are: (i) Do seamounts support a higher benthic biomass than nearby slopes at corresponding depths? (ii) If so, which particular taxa and trophic groups drive the observed differences? We examine both questions using a large dataset of sea-floor images from seamounts and adjacent continental margins in the Southwest Pacific.
The two study regions are located in the southwest Pacific Ocean, one east of New Zealand on the Chatham Rise, and the other southeast of Australia off Tasmania (Fig. 1a). The Australian study area included the ‘Tasmanian seamounts’, where 12 features ranging in elevation from 125 to 565 m, and the adjacent slope off southern Tasmania were sampled (Fig. 1b). The New Zealand study region included the ‘Graveyard seamount complex’, where eight features ranging in elevation from 100 to 350 m and the adjacent slope on the Chatham Rise were sampled (Fig. 1c). The presence of natural fish aggregations on these seamounts (Koslow et al. 2001; Clark & O’Driscoll 2003) suggests that they likely provide an elevated food supply compared to adjacent slope areas.
Seamounts in the Australian and New Zealand study regions have been subjected to bottom trawling since the early 1980s and mid-1990s, respectively. Fishing on the Tasmanian seamounts ceased in 1999, and on three of the Graveyard seamounts in 2001, when these seamounts received protected status (Koslow et al. 2001; Brodie & Clark 2003). Further details about the seamounts and slope habitats, and fishing effort are provided by Althaus et al. (2009) and Clark & Rowden (2009).
Australian data were sourced from two CSIRO companion voyages in 2006 (SS200611) and 2007 (SS200702). A towed stereo camera system (Shortis et al. 2008) and an epibenthic sled (1.2 × 0.6 m mouth opening and 25-mm net mesh size; Lewis 1999) were used to sample the sea floor. Camera transects were distributed in regular radial pattern on the seamounts (running from summit to base), and randomly on the slope. Slope transect sites for the present study were selected to be unfished and adjacent, and at similar depths, to the seamounts. The mean length of camera transects was 3500 ± 1300 m (SE). Oblique images were taken automatically at 10–15-s intervals at a target height of 2–4 m above the seabed. Sled tows were co-located with camera transects on seamounts (22 stations), and the slope (13 stations) to the extent possible, or otherwise randomly located on closely adjacent areas on seamounts and slope.
New Zealand data were sourced from an NIWA Seamounts Programme survey of the Graveyard seamounts in 2006 (TAN0604) and a New Zealand Ocean Survey 20/20 survey of the Chatham Rise in 2007 (TAN0705). A towed camera system (the Deep-Towed Imaging System, DTIS, Hill 2009) and an epibenthic sled (1 × 0.4 m mouth opening and 30-mm stretched net mesh size) similar to that used to sample Tasmanian seamounts and slope were deployed on both voyages. Camera transects were distributed in a regular radial pattern on the seamounts (running from summit to base), and randomly on the slope, within strata designed for broad-scale habitat mapping purposes. Slope transect sites were selected to be unfished (one exception) and adjacent, and at similar depths, to the seamounts. The mean lengths of camera transects on seamounts and slopes were 824 ± 48 m (SE) and 1270 ± 102 m (SE), respectively. Vertical images were taken automatically at 20-s intervals on seamounts and at 15-s intervals on the slope at a target height of 2–3 m above the seabed. Sled stows were distributed randomly on seamounts (41 stations), and were run along the line of the camera transects at slope sites (six stations). Table 1 summarizes the seabed image sampling effort and distribution among seamount and slope habitat, and study regions (see also Fig. 1). Seabed images provided the primary data for the present study, whilst fauna recovered by benthic sleds were used to construct biomass conversion factors.
Table 1. Summary of image sampling effort and distribution on slope and seamount habitat off Australia and New Zealand (see Fig. 1 for geographic position of sampling locations).
mean depth (m)
min. depth (m)
max. depth (m)
Huon 1000 – stn 16
Huon 1000 – stn 45
Huon 1000 – stn 47
Huon 1000 – stn 53
Tasman 1000 – stn 73
Tasman 1200 – stn 72
Tasman 1200 – stn 45
Tasman canyon – stn 62
Sister 1 (south)
Sister 2 (north)
Fauna that were distinctly identifiable in the images (>1 cm, but mainly >3 cm) are hereafter termed ‘epibenthic megafauna’ and were identified to the lowest practicable taxonomic level and counted with the aid of image analysis software packages (PHOTOSHOP, IMAGEJ). Data for both unitary and colonial taxa were recorded as numbers of individuals per image frame. The only taxa for which this was potentially ambiguous were matrix-forming scleractinian corals (primarily Solenosmilia variabilis), which often cover large areas of seabed but in which live polyps are concentrated in localized patches (i.e. much of the skeletal matrix is not living). Because live polyps are distinctive in colour (orange-pink), it was possible to count patches of living coral in the images to obtain a count-based abundance estimate.
To combine the CSIRO and NIWA data, a common level of taxonomic grouping – based on commonalities in taxon occurrence and identification – was decided upon (Table 2). Data were then checked for consistency of identifications between regions and, where necessary, images were re-analyzed prior to data analysis. The main feeding mode of each taxonomic grouping (Table 2) was assessed based on existing knowledge of the biology and ecology of the organisms. The area of each seabed image was determined either by calculation from calibrated stereo image pairs (CSIRO data, see Althaus et al. 2009 for detail) or from parallel-scaling lasers projected in the image (NIWA data). In images where the two lasers were not visible (6.5%), scale was estimated by reference to fauna of known size (e.g. orange roughy Hoplostethus atlanticus; mean standard length 35 cm). The average area per image was 4.8 ± 0.07 m2 (SE) for the Australian data, and 6.9 ± 0.07 m2 (SE) for the New Zealand data. The area of each image was used to calculate the abundance of each taxon per unit area (m−2). For the purposes of the numerical analysis, the basic sampling units were transects, with data from individual images averaged by transect.
Table 2. Biomass of epibenthic megafauna encountered in deep-sea images on slope and seamount habitat off Australia and New Zealand.
mean individual weight (ww, g)
Tabulated values are wet weights (ww, g) derived mostly from on-board measurements of fresh sled tow samples. Because small (<1 cm) animals are not identified in images but form part of the physical catches, calculated weight statistics used the upper quartile (≥Q75) of weight determinations only. Q75 and Q100 therefore represent the range of values from which the mean individual weights were calculated.
aWeights from collections only.
bShip-based values replaced with collection weights because they were apparently fragments.
F = filter-feeder; O = omnivore; P = predator; D = deposit-feeder. Multiple codes denote mixed feeding modes within the taxonomic group or the trophic plasticity of the taxa).
Chrysogorgiidae gen. A sp. A
Crinoids (not stalked)
Biomass conversion factors
Fauna recovered with sleds were wet-weighed on board the vessels using motion-compensated balances (±1 g). These weights were compiled according to the taxonomic categories used for the image analysis. For a few taxa, weights from physical sampling were based on only small fragments or they were not captured in this study. In these cases, wet weights were obtained from specimens held at the CSIRO and NIWA invertebrate collections which had been sampled from the same study areas on previous voyages. Because small animals were not visible and therefore not counted in the seabed images, but were captured in the sled samples, weight statistics were calculated using only the upper quartile (≥Q75) of the measured weights (Table 2). The mean weight per individual of each taxonomic group was used to convert the abundance data generated by the image analysis into biomass data for the primary analysis. These data have been supplied to Seamounts Online (http://seamounts.sdsc.edu). Clearly, there are biases involved in deriving biomass values from images; however, in lieu of good direct biomass data it is, in our opinion, worth accepting these biases in order to attempt to test the seamount oasis hypothesis. Future examinations using biomass values derived from images might be stimulated to improve upon our methodology.
The principal tests of interest was for differences in biomass between slopes and seamounts. The analytical framework was a multiple regression model followed by tests for differences between habitat types. However, because deep-sea megafaunal biomass decreases with increasing depth (Rex et al. 2006), and because bottom-trawling can influence the assemblage composition of epibenthic megafauna on seamounts (Koslow et al. 2001; Clark & Rowden 2009, Althaus et al. 2009), we explicitly controlled for the possible influence of these two factors on the main contrast between habitat types (i.e. slopes versus seamounts) by including them as co-variates in the model. Our aim was to maximize the inclusion of data from seamounts from each study region, and in so doing it was necessary to include data from fished seamounts. We did not want to specifically address the influence of fishing on seamount biomass – considering that a topic for a separate analysis. Any variation due to regional differences (i.e. independent of habitat effect) was accounted for by partitioning region out of the analysis as a randomized block effect.
As biomass values were not normally distributed (and could not be normalized by transformation), differences between slopes and seamounts were tested for using a non-parametric equivalent of analysis of covariance (Stokes et al. 2001). The co-variates (average depth of transect and number of commercial bottom trawl tows) and the response variable (i.e. biomass) were converted to ranks. We then fitted a multiple linear regression model to the ranks. The means of the residuals from the regression model were then compared using the Cochran–Mantel–Haenszel test (Mantel & Haenszel 1959).
Seamounts support a significantly higher density of epibenthic megafaunal biomass compared to slopes in the same region. On average, epibenthic megafauna biomass on the unfished seamounts was 3.8 times greater than on unfished slopes at comparable depths (Table 3A, Fig. 2). This difference is largely attributable to the scleractinian coral Solenosmilia variabilis, the biomass of which was 29 times higher on seamounts than at comparable slope sites (seamounts: = 347.17 ± 23.11; slopes: = 11.94 ± 233.02 g·ww·m−2). Because S. variabilis accounts for 73% of whole assemblage biomass on seamounts, it largely drives the differences in total biomass observed between seamounts and slopes (Table 3A, Fig. 3a,b). Brisingid asteroids also displayed significantly higher biomass on seamounts. Other taxa for which biomass was notably higher (but spatially variable and patchy) on seamounts compared to slopes included stalked crinoids, octocoral species of the genera Narella and Corallium, and the scleractinian coral Enallopsamia (Table 3A).
Table 3. Contrast of epibenthic megafauna biomass (g·ww·m−2) between unfished slope and seamount habitat off Australia and New Zealand, for (A) taxa and (B) feeding mode groups identified in deep-sea images.
A – Taxon
slope (n = 1072)
seamounts (n = 1053)
(seamount versus slope)
Chrysogorgiidae gen. A sp. A
Crinoids (not stalked)
slope (n = 975)
Seamounts (n = 967)
B – feeding mode
(Seamount versus Slope)
Bold entries are significantly (P < 0.05) higher on seamounts. Entries marked with (*) have significantly higher mean biomass on slopes. P values refer to contrasts between habitat types using the Cochran–Mantel–Haenszel test (see Methods for explanation of regression model and test).
In contrast to the significantly higher biomass of only two taxa on seamounts, 14 taxa showed significantly higher biomass on the adjacent slopes (Table 3A). However, because these taxa accounted for only very small amounts of the total benthic biomass (i.e. only anemones, holothurians and hydroids contributed more than 2%), they did not greatly influence the overall pattern of higher total biomass on seamounts.
Thus, the pattern of a significantly greater total biomass on seamounts caused by only one or two taxa contrasts with a more even distribution of biomass amongst taxa on the slopes. This pattern was also seen in the distribution of feeding modes between slope and seamount habitats (Table 3B). Because feeding by corals encompasses both filter-feeding and predation on zooplankton, the marked contrast in coral abundance between habitats, especially for S. variabilis, results in a significantly higher mean biomass of filter-feeders/predators on seamounts (Fig. 3c). Conversely, the biomass of predators, omnivores/predators and deposit-/filter-feeders was significantly higher on the slopes (Fig. 3d–f).
Within each habitat, the make-up of the epibenthic megafauna assemblages in terms of feeding modes differed markedly (Table 4). Filter-feeders made up a significantly greater proportion of biomass on seamounts ( = 42 ± 1.3%) than on slopes ( = 28 ± 1.1%) and the same pattern was evident for filter-feeders/predators (seamounts: 42 ± 1.3%, slopes: 30 ± 1.1%). Conversely, deposit feeders and those with mixed feeding modes comprised significantly more of the benthic biomass in slope habitats (Table 4).
Table 4. Comparisons of the distribution of biomass among feeding mode groups of epibenthic megafauna of unfished slope and seamount habitats off Australia and New Zealand.
slope (n = 1072)
seamounts (n = 1053)
(seamount versus slope)
Bold entries have significantly (P < 0.05) higher proportional biomass on seamounts, whereas those marked with (*) contribute significantly more to whole assemblage biomass on slopes. P values refer to contrasts between habitat types using the Cochran– Mantel–Haenszel test.
Our results support the central prediction of the seamount oasis hypothesis: there is elevated standing stock of benthic biomass on seamounts compared to adjacent habitats. The processes which cause such spatial contrasts in seamount biota have been linked to enhanced food supplies to organisms on seamounts (e.g.Genin et al. 1986). If seamounts are locations of enhanced trophic input, then benthic biomass will be greater than at comparable slope habitat, and the taxa largely responsible for this heightened biomass will be those that can most readily exploit the available food resource. Corals on seamounts are efficient at capturing and metabolizing food such as phytoplankton detritus and zooplankton (Duineveld et al. 2004). To support the high biomass of corals, an enhanced flux of particulate matter and zooplankton presumably occurs at the study seamounts, and such increased food supply may in fact be the principal reason for the higher biomass observed (Genin & Dower 2007).
Whereas the seamount biomass was dominated by a few taxa (most notably scleractinian corals), biomass on the slopes was more evenly distributed. For example, 14 taxa had higher average biomass on the slope, but none dominated to the extent of corals on the seamounts. The diversity of the fauna that made up the difference in biomass between the two habitats was reflected in the significant difference in the elevated biomass of omnivores/predators, deposit-/filter-feeders and predators on the slopes versus seamounts. Megafauna on slopes with similar mixed feeding modes constituted significantly greater proportions of the total epibenthic biomass than those of the same feeding modes on seamount habitat. In particular, the proportional biomass of omnivores/predators was greater on slopes than on seamounts (∼17%versus 2%).
A lower biomass on slopes, but one which is also more equally distributed among feeding modes, suggests that the benthos on the continental margins receives less food input, and that what food is available is exploited by a greater variety of taxa. However, the feeding mode pattern (particularly the prevalence of deposit-feeders) may reflect, to some degree, the greater proportion of soft substratum on the slopes versus the seamounts. Thus, the observed differences in epibenthic megafaunal biomass between slope and seamount habitats may result from more hard substratum on the study seamounts, which provides attachment sites for scleractinian corals such as Solenosmilia variabilis. The large influence of S. variabilis on seamount–slope differences raises the question of whether the same patterns would be observed at seamount and slope sites where matrix-forming corals did not exist, for reasons other than the absence of suitable substratum. Many seamounts do not occur in areas of high primary productivity or generate hydrological conditions that promote enhanced trophic input (Genin & Dower 2007; White et al. 2007). Thus, conditions favourable for corals on seamounts may not be ubiquitous. Indeed, the majority of seamounts exist in areas or at water depths of the ocean where conditions are predicted to be unsuitable for scleractinian corals (Tittensor et al. 2009). On these seamounts it would be reasonable to hypothesise that the benthic biomass is similar to adjacent slopes at similar depths.
It must therefore be stressed that whilst our study provides some support for the seamount oasis hypothesis, the principal finding (i.e. higher biomass on seamounts versus adjacent slopes at comparable depths) may not apply equally to seamounts in less productive regions, seamounts with a greater proportion of soft substratum, in parts of the oceans where scleractinian corals are not prevalent, or seamounts distant from slope habitats. Thus, further study of biomass data from seamounts and slopes are required for a more robust examination of the seamount oasis hypothesis.
That the scleractinian coral S. variabilis was largely responsible for the difference in biomass between seamounts and slopes is important, because removal of this species is the main consequence of bottom trawling on seamounts in the region (Koslow et al. 2001; Althaus et al. 2009; Clark & Rowden 2009). Thus, trawling can reduce benthic biomass on seamounts to levels similar to or lower than on slopes. The partial or almost total removal of the considerable biomass of living scleractinian corals from seamount habitats is therefore likely to influence profoundly the trophic structure of the benthos. However, until detailed studies examine the trophic architecture of seamount ecosystems, the full extent of the impact of deep-water trawling will not be understood.
Future studies examining differences between seamounts and other habitats on the continental margins also need to consider the links between topography and productivity. For example, submarine canyons, which are significant features of some slopes, can be sites of enhanced trophic input based on detritus transported from the shelf (Vetter 1995). Studies have shown that benthic assemblages in canyons can be diverse and abundant (e.g.Schlacher et al. 2007, 2010), and are frequently dominated by filter-feeding/predatory taxa such as corals and sponges with high biomass where there is suitable substrate for attachment (e.g.Hargrave et al. 2004). Comparisons based predominantly on macro-infauna have shown that, like seamounts, canyons may also have relatively high benthic biomass compared to general slope habitat at the same depth. However, very few strictly comparative studies have been conducted and these are mostly limited to single-site comparisons (e.g.Houston & Haedrich 1984; Vetter & Dayton 1998). A wider comparison of benthic biomass (and assemblage composition) among non-canyon slope habitats, canyons and seamounts is needed to provide a more comprehensive test of the oasis hypothesis. Future studies should also determine the linkages between benthic biomass, trophic architecture and ecosystem function in the context of their sensitivity to human activities such as fishing and mining, and climate change effects (e.g.Danovaro et al. 2008). Such integrated studies are already underway (Weaver et al. 2009) and are now planned for the deep sea around Australia and New Zealand.
We thank the officers, crew and scientific staff of the CSIRO and NIWA voyages from which data were generated for the analysis reported here. In addition, we thank Felicity McEnnulty (CSIRO), Kareen Schnabel and Sadie Mills (NIWA) for providing specimen weights from the respective invertebrate collections. We would also like to thank Bruce Barker, Mark Green and Pamela Brodie (CSIRO) for data acquisition and databasing, and Karen Gowlett-Holmes (CSIRO) and taxonomists at NIWA for their expertise in helping to identify invertebrates from seabed images. A.W. and F.A. were supported through the Commonwealth Environment Facilities (CERF) Marine Biodiversity Hub, with funding for the Australian surveys on Marine National Facility vessel provided by CSIRO Wealth from Oceans Flagship, and the Department of Water, Environment, Heritage and the Arts. A.R., D.B., M.R.C., M.C. and R.S. were supported through NIWA’s Seamounts: their importance to fisheries and marine ecosystems project [funded by the New Zealand Foundation for Research, Science & Technology (CO1X0508), and the Ministry of Fisheries (ENV200516)] and Ocean Survey 20/20 (funded by Land Information New Zealand, Ministry of Fisheries, Department of Conservation, and NIWA). Lastly, we gratefully acknowledge the support of the Census of Marine Life field programme CenSeam (a global census of marine life on seamounts) and its Data Analysis Working Group, which funded and co-ordinated data analysis workshops that stimulated or contributed directly to the study reported here.