Deep-Sea Research I 67 (2012) 98–110

Contents lists available at SciVerse ScienceDirect

Deep-Sea Research I

0967-06

http://d

n Corr

E-m1 Pr

Evans B

journal homepage: www.elsevier.com/locate/dsri

Response of megabenthic assemblages to different scales of habitatheterogeneity on the Mauritanian slope

Daniel O.B. Jones n, Michael E. Brewer 1

National Oceanography Centre, European Way, Southampton SO14 3ZH, United Kingdom

a r t i c l e i n f o

Article history:

Received 25 January 2012

Received in revised form

22 April 2012

Accepted 9 May 2012Available online 18 May 2012

Keywords:

Benthic

Megafauna

Biodiversity

Imaging

Seabed habitat

Spatial scales

37/$ - see front matter & 2012 Elsevier Ltd. A

x.doi.org/10.1016/j.dsr.2012.05.006

esponding author. Tel.:þ44 0 2380 596357.

ail address: dj1@noc.soton.ac.uk (D.O.B. Jones

esent address: National Institute of Water an

ay Parade, Greta Point, Wellington 6021, New

a b s t r a c t

The topographically complex deep seabed on the Mauritanian slope, from 990 to 1460 m water depth,

was imaged with video in an extensive quantitative survey of 17,199 m2 of seafloor using a Remote

Operated Vehicle (ROV). This study investigated the influence of habitat heterogeneity at two scales on

the megafaunal assemblages observed by ROV. Changes in megafaunal assemblages on the Mauritanian

slope were assessed at a broad scale, within depth zones, and at a finer scale, in response to changes in

local geomorphology associated with submarine landslides. Geomorphology was determined by

classification of habitat parameters (slope, aspect, bathymetric position, curvature, fractal dimension

and ruggedness) derived from an autonomous underwater vehicle-based multibeam bathymetry

survey. Habitat parameters were classified by Iterative Self Organizing Clustering into six major

geomorphological groups, four of which were assessed in the ROV video survey. A total of 29

megafaunal taxa were observed along the entire survey, with an overall average faunal density of

0.344 ind m�2. Megafaunal assemblage density, species richness and evenness varied significantly

across the depth range of the survey in the most common geomorphological zone (sedimentary plains

of low slope and complexity). Characteristic species inhabited the shallow areas (asteroid, ophiuroid,

anemone, small macrourid), intermediate areas (Benthothuria funabris, black cerianthid, squat lobster)

and deeper areas (the holothurians Enypniastes eximia and Elipidia echinata). Megafaunal density,

species richness and evenness were not significantly different between geomorphogical groups within

one depth zone (1300–1400 m). However, the steepest zone, on the edge of a major headwall feature,

had four unique taxa (Parapagurus pilosimanus, a comatulid crinoid, a gorgonian and its associated

ophiuroid).

& 2012 Elsevier Ltd. All rights reserved.

1. Introduction

Habitat heterogeneity at local scales is a pervasive feature ofbathyal continental margins and exerts a fundamental influence onthe diversity and structure of benthic communities (Levin andDayton, 2009). Broad-scale geological processes are important con-trolling mechanisms for benthic community structure (Carney,2005), for example, creating fluid flow to the seafloor. Less is knownabout the role of large mass movements in creating habitat of highstructural complexity in otherwise less complex soft-sedimentregions. Many studies show habitat related controls on the distribu-tions of benthic fauna, particularly in shallow waters (e.g., Kaiseret al., 2005). In the deep sea there is less information, although theassociation of megafauna with specific habitat types is revealedby very fine-scale patterns in sediment type (Auster et al., 1995),

ll rights reserved.

).

d Atmospheric Research, 301

Zealand.

mid-scale associations with specific habitats (Dolan et al., 2008) tothe broad-scale regional patterns (Bryan and Metaxas, 2006;Williams et al., 2010). At the broad scale, physical factors such asthose associated with depth (Jones et al., 2007a) and water massproperties (Mortensen and Buhl-Mortensen, 2004; Williams et al.,2010) tend to become more important in describing patterns ofbenthic fauna (Levin et al., 2001).

Habitat heterogeneity on the West African continental margin hasbeen increased by regional-scale submarine sliding (Wynn et al.,2000). Evidence of submarine slides off Mauritania is apparent alongmuch of the slope, the most well known feature being the MauritaniaSlide Complex which covers an area of 30,000 km2 between 800 and2000 m depth (Henrich et al., 2008; Krastel et al., 2006). This featureinvolved a multiphase submarine slide, which occurred �10.5 ka agoand was followed by retrogressive failure that created a series ofstepped headwalls upslope before being halted at around 700 m by alinear carbonate mound (Colman et al., 2005). The latest eventsare debris flow tongues dated at 10.5 ka (Forster et al., 2010).The Mauritanian Slide Complex has produced varied deep-watertopography and exposed several stratigraphic units along the sliding

D.O.B. Jones, M.E. Brewer / Deep-Sea Research I 67 (2012) 98–110 99

planes (Fig. 3 in Henrich et al., 2008), breaking the deep-sea slopehabitat into areas of low relief interspersed with high relief anddiffering geology. The changes in seabed topography brought aboutby the sliding processes may give rise to changes in benthiccommunities, as has been observed on breaks and escarpmentselsewhere in the Atlantic (Mortensen et al., 2001). The effects ofchanges in habitat heterogeneity associated with major submarinelandslides on deep-water biology are unknown, despite these fea-tures being ubiquitous along the world’s continental margins(Masson et al., 2010) and particularly extensive in west Africa(Krastel et al., 2011).

Deep-sea megabenthic ecology has traditionally been relianton semi-quantitative sampling with trawls and sledges (Thurstonet al., 1994). More recent advances have used imaging methods toobtain quantitative data (Jones et al., 2007b). However, it has notbeen possible, in most cases, to obtain accurate spatial positionalinformation of individual organisms or even communities on theseafloor (Dolan et al., 2008). Advances in detailed navigationtechnology and their application to deep-sea studies (Barry andBaxter, 1992), coupled with the increasing availability of high-resolution, spatially-accurate acoustic data on submarine topo-graphy and sediment properties, is now making it possible to linkfine-scale patterns in biology to the broad-scale patterns inhabitat type and subsea landscape. Given the sheer size anddifficulties of accessing the deep-sea environment, accurateextrapolation of the fine-scale observations using informationon the important physical controls for distribution is vital todescribe the important broad-scale patterns in benthic biology.

This paper is the first study that investigates the variation inbenthic megafaunal assemblages across a major submarine land-slide. Quantitative remotely operated vehicle (ROV) imaging isused to assess the changes in megabenthic assemblages inresponse to changes in habitat at two scales. First, we assessbroad-scale patterns in megafaunal assemblages within depthbands. We then explore patterns within a depth band at a fine-scale by assessing the response of megafaunal assemblages tosurficial geomorphology, classified from a range of habitat vari-ables. Megafaunal assemblages will be linked to specific geologi-cal units exposed by multiphase sliding events off Mauritania. Asthese huge slide events affected large sections of the Mauritanianslope, with similar units exposed at similar depths, thismay enable extrapolation of the information on local benthicmegafaunal assemblages collected here to a larger proportion ofthe Mauritanian slope.

2. Methods

2.1. Study site

This work was conducted as part of an environmental surveyof the TIOF field, Mauritania (171550 N 161530 W). It covers an areaof around 54 km2 of seabed at a depth of 900–1500 m. Oneexploration well had been drilled in the area (location 171 560

17.67700 N 161 520 3.07800 W). The ROV passed this site during thissurvey, but all video data from within 500 m of the previouslydrilled site were excluded from analysis. This was done to ensurethat any effects of seabed disturbance did not confound theanalysis.

The Mauritanian margin has high primary productivity, causedby continuous trade-wind-driven oceanic upwelling. The highproductivity is concentrated in surface waters above the outershelf and shelf edge (van Camp et al., 1991) but offshore advec-tion means that much of the biogenic detritus is deposited on theupper- and mid-slope at water depths of �1000–1500 m(Futterer, 1983; Henrich et al., 2010). The high productivity off

north-west Africa leads to substantially denser and more diversedeep-water benthic communities (Thurston et al., 1998). This isconsistent with the high diversity of fish caught in trawl samplesoff Mauritania (Merrett and Marshall, 1981).

The Mauritania Slide Complex was discovered by Seibold andHinz (1974) and subsequently mapped and described by Jacobi(1976), who estimated the area of seafloor affected by the massmovement to be in the order of 34,300 km2, and the total volumeof excavated material to be about 400 km3. Recent data confirmthese early estimates, indicating that the total area affected by theslide is about 30,00074000 km2 (Henrich et al., 2008). Seismicprofiles of the Mauritania Slide Complex, taken �70 km south ofthe study site, reveal the headwall area is characterised by threedistinct morphological steps each up to 75 m high between 800and 1300 m water depth (Henrich et al., 2008). The headwalls cutwell-stratified slope sediments. Relatively thin (o30 m) blockydeposits dominate the area immediately downslope of the head-wall area to �2000 m water depth. Here, a reduction in slopeangle to �11 corresponds to a major change in depositionalpattern, with stacked tongue-shaped acoustically transparentdebris flow deposits occurring below the break in slope(Henrich et al., 2008). The morphology to the north-east of thesidewall is highly complex with several canyons and large distinctblocks (Henrich et al., 2008).

Several biological studies focus on the deep-water demersaland benthic megafaunal assemblages off the north-west Africancontinental slope and abyss, including off Mauritania (Duineveldet al., 1993a; Duineveld et al., 1993b; Galeron et al., 2000;Henriques et al., 2002; Keller and Pasternak, 2002; Merrett andDomanski, 1985; Merrett and Marshall, 1981; Pfannkuche et al.,1983; Sibuet et al., 1993), the Canaries (Uiblein et al., 1996), theMeteor Seamounts (Fock et al., 2002; Piepenburg and Muller,2004; Uiblein et al., 1999) and Mid-Atlantic Ridge and Azores(Bergstad et al., 2008a; Bergstad et al., 2008b; Gebruk et al., 2010;Moss, 1992). These studies show a generally diverse and abun-dant deep-water fauna, but with high variability between sitesand depths studied.

2.2. Broad-scale habitat classification

Bathymetric data were obtained from a commercial survey of thearea (Woodside Petroleum) using the World Geodetic System ’84reference datum. The 200 kHz multibeam bathymetry data werecollected using a Kongsberg Maritime Hugin 3000 AutonomousUnderwater Vehicle (AUV) operated by C & C Technologies, Inc.AUV positioning was achieved by Kongsberg ‘‘High Precision Acous-tic Positioning System’’ ultra-short baseline navigation system, aninertial navigation system and Doppler velocity log integrated in aKalman filter. Data were provided for this study by WoodsidePetroleum, post-processed and gridded to 90 m spacing. The Maur-itanian slope was divided by depth into polygons representing100 m depth bands using ArcGIS.

2.3. Fine-scale geomorphological classification

Fine-scale classification of geomorphology was based on sixprimary derivatives of the bathymetry dataset (slope, aspect,bathymetric position, curvature, fractal dimension and rugged-ness). These were then combined using Iterative Self Organizing(ISO) clustering into a single geomorphological classification.

2.3.1. Bathymetry derivatives

Slope, or the measure of steepness (a first-order derivative),was derived using the ArcGIS Spatial analyst extension’s surfaceanalysis. Aspect, or orientation of benthic habitat, is relevant to

16°45'0"W16°50'0"W16°55'0"W

18°0'0"N

-120

0-140

0

-130

0

D.O.B. Jones, M.E. Brewer / Deep-Sea Research I 67 (2012) 98–110100

benthic faunal interaction with local and regional currents,potentially affecting a number of biological factors including foodsupply and colonisation (Gage and Tyler, 1991). Aspect calcula-tions were performed with a 3�3 cell (272�272 m) movingwindow in Landserf (Wood, 2005).

Bathymetric position index (BPI), a second-order derivative ofbathymetry, evaluates elevation differences between the each celland the mean elevation of the surrounding cells within a user-defined annulus. Here two BPI layers were calculated using theBenthic Terrain Modeller (BTM) software (Lundblad et al., 2006)within ArcGIS, a broad-scale BPI (circle of radius 1.008 km¼11cells) and a fine-scale BPI (circle of radius 181 m¼2 cells). Thesescales were chosen to represent broad-scale (e.g., slope collapsefeatures) and fine-scale (local topography e.g., mounds/blocks)features identified by examination of the bathymetry. Theselayers were standardised to a mean of 0 and standard deviationof 100 to enable effective comparison. Positive values represent acell that is higher than its neighbouring cells (ridge) and negativevalues represent the opposite (valley; Lundblad et al., 2006).

Curvature of seabed terrain has a number of measures(Schmidt et al., 2003), which have been shown to be importantin describing terrestrial habitats (Shary et al., 2002). Here, threemeasures were calculated: profile curvature, plan curvature andmean curvature. Profile curvature describes the curvature of thesurface in the steepest down-slope direction and may be useful tohighlight concave and convex slopes on the seabed (Wilson et al.,2007). Plan curvature is the curvature of a contour drawn throughthe central pixel and may be useful in describing ridges, valleysand slopes (Wilson et al., 2007). Both profile and plan curvatureare based on the steepest slope. Maximum and minimum curva-ture can be calculated, independent of slope, based solely onsurface geometry. Mean curvature is the average of the minimumand maximum calculated using the methods of Evans (1980).These calculations were implemented in Landserf (Wood, 2005)for a 3�3 cell analysis window.

The fractal dimension (D) is used to measure surface complex-ity (Mandelbrot, 1983). It varies between 2 (flat surface) and 3(space-filling rough surface) (Wilson et al., 2007). Here fractaldimension was calculated for a 3�3 cell moving window aroundeach pixel in the raster by the variogram method (Herzfeld andOverbeck, 1999; Mark and Aronson, 1984). These calculationswere performed using the FocalD calculator add-on for Landserf(Wood, 2005) with the minimum 9�9 cell analysis window.

Terrain variability (complexity) was measured from bathyme-try data using the Terrain Ruggedness Index (TRI), a measure ofthe local variation in seabed terrain about a central pixel (Rileyet al., 1999). TRI is calculated by comparing a central pixel with itsneighbours (within a 3�3 cell grid), taking the absolute value ofthe differences and averaging the result (Wilson et al., 2007). Thiswas done automatically in ArcInfo Workstation using a freelyavailable AML script (Evans, 2004) available through ArcScripts(http://arcscripts.esri.com/).

17°55'0"N

-110

0-1

000

-900

-800

-150

0

Fig. 1. Location of study site on Mauritanian margin. Depth contours are shown

every 50 m, 100 m isobaths have thicker lines and depth labels. Biological

sampling sites are marked as grey filled circles.

2.3.2. Data reduction and grouping

A total of nine layers of information were generated (slope,aspect, TRI, fractal dimension, broad-scale BPI, fine-scale BPI,profile curvature, plan curvature and mean curvature) and anunsupervised classification was used to reduce this volume ofdata. Prior to analysis, all bands were normalised to the samevalue range using ArcGIS. The ArcGIS Iterative Self Organizing(ISO) Cluster function in Spatial Analyst was used to generate acluster signature file (based on the methods of Ball and Hall,1965). Initially 11 classes were differentiated based on therecommendation to calculate double the desired classes (ERDAS,2005). Each pixel was classified using the Maximum Likelihood

Method based on the cluster signature file. Between-classdistance was calculated using the multivariate dendrogram toolin ArcGIS Spatial Analysis. Several classes were found to be verysimilar, these were combined (where between-class distanceswereo5) and the raster reclassed with a total of six groupings. Amajority filter was applied for a 3�3 cell neighbourhood toreduce noise in the data, classifying habitats at an extent ofapproximately 250 m�250 m.

ROV video data for sample stations was divided by spatialposition into each of the six generated clusters using the IntersectPoint tool in the Hawth’s Tools add-on to ArcGIS (Beyer, 2004).Any sample station within 50 m of the boundary betweenclassified areas was removed from analysis.

2.4. Biological methodology

2.4.1. Video survey

A video survey was designed to assess the effects of habitathetrogeneity at both scales (Fig. 1). Data were collected fromM.V. Boa Deep C (22–23 October 2004) using an industry-operatedwork-class Oceaneering Millennium Remotely Operated Vehicle(ROV) equipped with colour video camera (Remote Ocean Sys-tems) and digital still camera (Kongsberg OE14-208). Cameraswere mounted on a pan and tilt unit at the front of the ROV. Thevideo camera was zoomed out to maximum extent, panned tocentre and tilted to its most vertical angle (471), obtaining obliquevideo. Continuous video transects were conducted in a patterndesigned to cover as much as possible of each habitat. The ROVwas run in a straight line at a constant speed (o0.5 m s�1) andaltitude (�0.2 m) between waypoints. Transect width (0.91 m)was calculated from the camera acceptance angles and vehiclealtitude with an estimated precision of 70.1 m. Transect widthwas also verified by passing over objects of known size on theseafloor. In addition to transects, some detailed close up inspec-tion and still photography of examples of many of the megafaunawas carried out to aid identification. Ultra-short base-line naviga-tion was used, tied to shipboard differential global positioningsystem, giving seabed positioning to approximately 71 m accu-racy. Owing to the length of the transects, the vessel wasdynamically positioned to follow the ROV.

2.4.2. Data analysis

The 36 km ROV transect was divided into 100 m lengths(covering 91 m2 of seabed), each of which represented a samplingunit. The size of the sampling unit was selected as an appropriatescale for comparison with the finest scale of geomorphological

D.O.B. Jones, M.E. Brewer / Deep-Sea Research I 67 (2012) 98–110 101

classification (90�90 m grain). Sampling units where vehiclealtitude or speed was too great were excluded from analysis.Video transects were replayed at half speed and all visibleorganisms were counted. Infaunal species, when seen, werecounted if enough of their body was visible for identification.Species with direct and apparently obligatory symbiosis, forexample the hermit crab Parapagurus pilosimanus and its epizoiczoanthid Epizoanthus paguriphilus, were counted as one speciesfor analysis. Where habitat-forming species, for example thegorgonians, were present, all megafaunal species associated withthat organism were counted. The resolution of the ROV videoallowed all organisms greater than 50 mm total length to beassessed. Abundances were then standardised to numbers m�2.

2.4.3. Assessment of patterns in megafaunal assemblages

With limited a priori knowledge of the fauna it was assumedthat depth would be the most important environmental variabledetermining broad-scale patterns and that, within depth bands,variations in geomorphology would be important in controllingcommunity organisation at the scale of available multibeam data(see e.g., Gage and Tyler, 1991). This led to the formulation of twonull hypotheses:

1)

Megafaunal assemblages will not vary between 100 m depthbands (within a specific geomorphological zone).

2)

Fig. 2. Overall classification of the available bathymetry for the study area on the

Mauritanian margin. The rectangle shows the study area.

Megafaunal assemblages will be indistinguishable betweendifferent geomorphological zones (within a specific 100 mdepth band).

Owing to limited data availability, it was not possible to testhypotheses for all combinations of depth and geomorphological zone.For hypothesis 1 only the megafaunal assemblages that occurredwithin the most common geomorphological zone (E) were comparedwithin different depth bands. For hypothesis 2 the megafaunalassemblages of different geomorphological zones were comparedonly within the 1300–1400 m depth band.

2.4.4. Hypothesis testing

Several biological parameters were selected for hypothesistesting. Univariate parameters were: total faunal density, ameasure of species richness (number of species encountered)and a measure of evenness (Pielou’s J0). Species richness was alsoestimated by individual-based species accumulation (Mao-Tao)calculated using EstimateS software (version 7.51). Multivariatemeasures (Bray–Curtis similarity) were also employed andqualitative comparisons were made to support quantitativehypothesis testing.

Hypotheses were tested using a one-way ANOVA model. In theunivariate case, one-way ANOVA was used to test for significantdifferences in biological parameters. If data did not satisfy theassumptions of normality a non-parametric Kruskal–Wallis testwas used. When the factors were found to have significantdifferences, pairwise a posteriori Holm–Sidak tests were used toexplore relationships between factors. In the non-parametric case,Dunn’s test was used for pairwise testing. Tests were implementedin SigmaPlot version 11 (Systat Software Inc.). In the multivariatecase, one-way ANOSIM was used to test differences in Bray–Curtissimilarities. Densities for each sample unit were 4th root trans-formed prior to multivariate analysis to buffer the influence ofdominant taxa (Field et al., 1982). Similarities were calculatedusing Bray–Curtis coefficients (Bray and Curtis, 1957). Multivariateassessment was implemented in PRIMER (Clarke and Warwick,2001).

As habitat was found to exert significant effects on multivariatebenthic community composition, further testing was carried out to

confirm that top-down habitat classification had not missed anyimportant patterns. Unconstrained analyses were used to assess therelationships between the bathymetric derivatives and the multi-variate megafaunal community composition. The BIO-ENV routineof PRIMER (Clarke and Warwick, 2001) was used to assess theenvironmental parameters that were most important in describingthe observed megafaunal distribution patterns.

3. Results

3.1. Geomophological habitat classification

The water depth in the study area increases from 990 to 1460 m(Fig. 1), although the slope is not smooth. It is interrupted by a majorheadwall feature, �50–300 m wide, sloping at an angle of up to�211 and running approximately north to south. This featurecreates a major curvature of the seafloor, although it is locallysmooth (low TRI and fractal dimension). Immediately to the west, atthe base of the headwall, is a flat area (slope �21). For 430 kmup-slope of the headwall the seabed is generally smooth, featurelessand sloping gradually (�21 slope at �2701 aspect). In some places,close to the headwall, there are areas (�27 km2) of partial collapsewith a rougher seafloor. There are also numerous areas where theheadwall has collapsed leaving a blocky debris flow westwards of thefailure. These areas are characterised by variable slope, high rough-ness and bounded by regions of high plan curvature. Downslope of

D.O.B. Jones, M.E. Brewer / Deep-Sea Research I 67 (2012) 98–110102

the headwall a more complex situation exists with heterogeneousfeatures indicative of the debris flows from slope collapse. Roughnessis generally high with variable slope, aspect, curvature and BPI. Insome areas, although not in the study site, there are canyon featuresthat incise the slope. The nearest, Chinguetti canyon, is around 15 kmnorth of the study site.

The ISOCluster technique differentiated 6 classes over the entirearea where bathymetry data were available (Fig. 2). In the studyarea five of these were present (Fig. 3). There was no single factorthat differentiated these classes (Table 1) and so it was difficult toassign fully descriptive names to each class. Class B only occurrednear the steep side-walls of the major canyons; it was associatedwith high mean curvature, slope and positive BPI. It was not presentin the study site. Class A represented the areas at the bottom ofslopes and was represented by negative BPI (at broad-scale) andnegative mean curvature. It had relatively low TRI and high fractaldimension indicating a smooth terrain. Class C occurred principallyat the base of the canyon systems. There were several small patchesin the study area at the base of east–west channels. Class C wascharacterised by negative BPI, variable but high profile curvature butlow mean and plan curvature. It had high TRI and low fractaldimension indicating a relatively rough seabed. Class D was wide-spread; it was characterised by low roughness (low TRI and highfractal dimension) and near-zero bathymetric position and curva-ture (mean, plan and profile). Slope was low on average. Class E wasthe most common class. It was characterised by low roughness (lowTRI and high fractal dimension), near-zero curvature (in all mea-sures) and BPI. In this class, slope was fairly low on average butvariable. Class F represented the upper parts of the wall features. Itwas characterised by positive BPI (at both scales), negative profile

Fig. 3. Seabed categories/types within the Mauritanian margin study area. Class¼ IS

bathymetric position index. F BPI¼fine scale bathymetric position index. See methods

curvature, positive plan and mean curvature and intermediateroughness.

3.2. Fine-scale habitat heterogeneity observed in ROV video

At a finer scale the patterns were less obvious. At a scale of tensof metres to centimetres, observed in ROV video, most of the seabedappeared homogeneous both in sediment type (sub-metre) and inmorphology (metre to decimetre). The only regions that it waspossible to differentiate were the blocky debris flows and thesloping headwall. In the blocky debris flow region sedimentsappeared homogeneous, but the block features led to depth changesof up to 10 m within 20 m of transect (this was more apparent thansuggested by the bathymetry where the cell size of 90 m wouldhave averaged many of these local variations). The headwall regionappeared to have finer-grained sediment than other areas. The slopewas obvious in video and slope estimates from video records agreedwith bathymetry interpretation (�201). Pilot observations sug-gested stronger local currents than normal in this region.

3.3. Patterns in the megafaunal assemblages

A total of 29 megafaunal taxa were observed in ROV videocovering 17,199 m2 of seabed, with an overall average faunaldensity of 0.344 ind m�2. Species observed (Fig. 4) were predo-minantly fish (nine species, 0.148 ind m�2), echinoderms (sixspecies, 0.233 ind m�2), cnidarians (six species, 1.044 ind m�2)and arthropods (five species, 0.259 ind m�2). Molluscs andsponges were also observed.

OCLUSTER classified seabed. TRI¼Terrain Ruggedness Index. B BPI¼broad scale

section for full descriptions of indices.

Table 1Characteristics of geomorphological classifications for the Mauritanian continental slope.

Class A B C D E F

Area (km2) 198.88 44.34 64.40 982.50 4134.62 228.99

Depth (m) Mean 1125.6 980.8 1083.9 1277.4 877.5 1045.7

Std 446.1 367.6 415.2 464.7 561.2 440.2

Slope (1) Mean 6.46 17.67 12.52 2.50 1.89 5.99

Std 5.35 9.04 9.26 1.68 1.34 5.36

Aspect (1) Mean 234.9 214.6 208.0 253.8 266.6 236.5

Std 94.2 108.5 106.6 73.9 48.8 89.2

Fractal dimension Mean 2.418 2.317 2.371 2.444 2.385 2.382

Std 0.130 0.077 0.115 0.138 0.114 0.159

TRI Mean 29.07 75.35 56.81 11.31 8.03 26.91

Std 22.60 39.20 39.74 6.80 5.72 24.71

B BPI Mean �18.18 42.22 �67.01 �4.35 1.29 20.48

Std 12.94 38.57 38.43 4.65 3.47 15.07

F BPI Mean �2.42 3.98 �5.54 �0.50 0.14 2.35

Std 5.51 8.92 8.97 2.05 1.42 6.20

Profile curvature Mean 2085 �5836 7107 390 �103 �2353

Std 2109 5141 5293 737 470 2287

Plan curvature Mean �1272 1843 �2598 �425 109 1279

Std 1772 3993 3961 677 421 1762

Mean curvature Mean �3357 7680 �9705 �815 212 3631

Std 2508 6262 6140 1031 695 2590

D.O.B. Jones, M.E. Brewer / Deep-Sea Research I 67 (2012) 98–110 103

3.3.1. Broad-scale patterns in megafaunal assemblages related to

depth

Many species exhibited bathymetric trends, including depth-related changes in population densities, with high densities inshallow areas (asteroid, ophiuroid, anemone, small macrourid),intermediate areas (Benthothuria funabris, Black Cerianthid, Muni-did squat lobster) or deeper areas (the holothurians Enypniastes

eximia and Elipidia echinata) (Table 2). Other species showed noclear patterns with depth (Phormosoma placenta, Actinoscyphia

aurelia, Paralomis africana, cephalopods and most fish). Only twospecies were confined to one depth zone, an actinarian in theshallowest zone (900–1000 m) and a pennatulid in the deepest(1400–1500 m). These were both rare species.

Total density of megafauna varied significantly between 100 mdepth bands (Table 2; one-way ANOVA F¼8.77, df¼5,150,po0.01). Multiple pairwise comparisons (Holm–Sidak) revealtwo main groups, a lower density group (1000–1100 m, 1300–1400 m and 1400–1500 m) and a higher density group (900–1000 m, 1100–1200 m and 1200–1300 m). Within these groupssites were not significantly different, whereas between groupssites were significantly different to each other. These groupingsare also apparent in analysis of the continuous data, particularlythe reduction in density at the deeper sites.

Both the evenness (J0; Kruskal–Wallis H¼94.90, df¼5, po0.001)and the richness (S; ANOVA F¼11.60, df¼5,150, po0.001) compo-nents of diversity were significantly different between 100 m depthbands (Table 2). Evenness was significantly lower (Dunn’s multiplecomparisons) in the 1200–1300 m depth band than in other bands.There were no other significant differences between depth bands.Analysis of the continuous data suggested a parabolic relationshipbetween depth and evenness. Richness was highest in the 1200–1300 m zone. This zone had significantly higher richness than the1000–1100 m, 1100–1200 m, and 1400–1500 m zones (Holm–Sidak multiple comparisons), but was not significantly differentfrom the other zones. The 900–1000 m zone had the lowest speciesrichness. It was significantly lower than both the 1100–1200 m andthe 1400–1500 m zones, but not significantly different from theother zones.

Within the most common habitat (E) there were large andsignificant changes in community type with depth (Fig. 5A; ANOSIMR¼0.382, po0.001). Every combination of depth zones revealedsignificant differences in multivariate community composition

(po0.05). However, the two-dimensional multidimensional scalingordination plot (Fig. 5A) had high stress (0.27) and so should betreated with caution.

3.3.2. Fine-scale patterns in megafaunal assemblages related to

geomorphology

Geomorphology F was distinctive in terms of the speciespresence (Table 3). Two species, the sepiid and the sabellid, thatwere absent from this zone were found in all the other zones. TheF geomorphology also yielded four species that were unique tothis habitat: Parapagurus pilosimanus, the comatulid crinoid, thegorgonian and its associated ophiuroid. Moreover, with theexception of Parapagurus pilosimanus, these species were foundat only one site (centre position 17156’49’’ N 16151’53’’ W), whichhad particularly high slope (4201).

Geomorphological zone A and E were characterised by distinctivespecies that were generally rare when present, or by species thatwere absent uniquely. These rare species were the orange sponge(only occurring at geomorphology A), the octopus Benthoctopus sp.(only absent from geomorphology A) and the holothurian Enyp-

niastes eximia (only occurring at geomorphology E). Four of thespecies associated with geomorphology D occurred at densitiesconsiderably higher than at other sites: the ubiquitous cerianthid,the common ophiuroid, the palemonid prawn and the fish Dicrolene

intronigra.The total density of benthic megafauna was highest for

geomorphology D (mean¼36.7 animals per 100 m2711; notethese are presented as sample means, rather than standardisedtotals, as given in Table 3) and lowest for E (mean¼28.0 animalsper 100 m278.8). There was considerable variation between theindividual transects within a geomorphological zone (the highestwas habitat F: mean¼28.8 animals per 100 m2716.2). As a resultof fairly similar means and high standard deviations there wereno significant differences in total density between geomorphol-ogy types (one-way ANOVA F¼0.74, df¼3,27, p¼ns). None of thethree numerically dominant deposit feeding species (ophiuroid,Phormosoma placenta, Polycheles sp.) displayed significant differ-ences in density between geomorphology types.

The different geomorphological zones were associated withsimilar levels of megafaunal diversity (Table 3). No significantdifferences were seen in either species richness (measured by S;ANOVA F¼1.31, df¼3,27, p¼ns) or evenness (measured by J�;

Fig. 4. Megafaunal species encountered. Sizes are in brackets, stated as horizontal length of whole organism (as shown) in mm with an estimated precision of710 mm.

From top left: Row 1: Cnidarians: (A) Actiniarian (150), (B) cf Actinoscyphia aurelia (200), (C) Cerianthid (100), (D) Gorgonian (150), (E) Pennatulid (80), (F) Scyphozoan (30);

Row 2: Echinoderms: (G) Benthothuria funabris (400), (H) Elipidia echinata (30), (I) Enypniastes eximia (120), (J) Ophiuroid (250), (K) Gorgonian associated Ophiuroid (70);

Row 3: (L) Comatulid crinoid (90), (M) Phormosoma placenta (80), (N) Goniasterid (60); Molluscs: (O) Benthoctopus sp. (300), (P) Sepiid (100); Row 4: Arthropods:

(Q) Palemonid prawn (70), (R) Munidid, (50) (S) Parapagurus pilosimanus (and Epizoanthus paguriphilus) (80), (T) Polycheles sp. (90), (U) Paralomis africana (600); Row 5:

Fishes: (1) Dalatiid cf Centroscyllium fabricii (700), (2) Bathyraja sp. (800), (3) Lophius budegasa (900), (4) Dicrolene intronigra (600): Row 5: (5) Small Macrourid (300),

(6) Macrourid (350), (7) Corphaenoides sp. (500); Row 6: (8) Alepocephalus sp. (500), (9) Notocanthid (600).

D.O.B. Jones, M.E. Brewer / Deep-Sea Research I 67 (2012) 98–110104

ANOVA F¼0.25, df¼3,27, p¼ns) components of diversitybetween the zones.

However, they did exhibit differences in multivariate commu-nity composition (Fig. 5B; ANOSIM R¼0.104, po0.05). Pairwisecomparisons revealed that these faunal differences resulted fromdifferences between geomorphology A and the two most commongeomorphologies (D and E). No other pairwise comparisons weresignificant. SIMPER analysis showed that three species, two fish

(Macrouridae and Dicrolene intronigra) and the Palemonid prawn,contributed most to the significant dissimilarities.

3.3.3. Unconstrained assessment of megafaunal assemblage patterns

There was a low but significant overall correlation betweenenvironmental factors and the multivariate community in allsampling units when analysed by BIOENV analysis (Rho¼0.26).

Table 2Density (number m�2) and composition of megafaunal communities within the major habitat type identified (E) for the Mauritanian slope. Data are divided by 100 m

depth bands. Species number (S) is used as an estimate of species richness. Evenness is calculated by Pielou’s evenness index (J0). The Shannon-Wiener (H0) diversity index

is presented for comparison with other studies.

Densities (no m�2)

900–1000 1000–1100 1100–1200 1200–1300 1300–1400 1400–1500

FishDalatiid cf Centroscyllium fabricii 0 0 0.0007 0.0015 0 0.0011

Bathyraja sp. 0.0011 0.0015 0.0004 0.0002 0 0

Lophius budegasa 0 0 0.0004 0.0002 0 0

Corphaenoides sp. 0.0085 0.0081 0.0089 0.0077 0.0088 0.0099

Macrourid 0.0007 0.0011 0 0 0.0011 0

Small macrourid 0.0014 0.0029 0.0032 0.0027 0.0022 0

Notocanthid 0.0007 0.0029 0.0004 0.0027 0.0044 0.0088

Dicrolene intronigra 0.0071 0.0066 0.0121 0.0105 0.0121 0.0110

Alepocephalus sp. 0.0004 0.0015 0.0007 0.0030 0.0044 0

MolluscsSepiid 0.0007 0.0015 0.0007 0.0005 0.0022 0

Benthoctopus sp. 0.0021 0.0007 0.0014 0.0007 0.0011 0.0011

ArthropodsMunidid 0 0.0077 0.0315 0.0005 0 0

Palemonid prawn 0.0727 0.0502 0.0454 0.0305 0.0363 0.0099

Paralomis africana 0 0.0004 0.0007 0.0002 0 0

Polycheles sp. 0.0011 0.0037 0.0060 0.0027 0.0077 0.0055

Parapagurus pilosimanus (and Epizoanthus paguriphilus) 0 0 0 0 0 0.0011

EchinodermsGoniasterid 0.0053 0.0062 0.0060 0.0012 0 0.0022

Ophiuroid (cf Bathypectinura heros) 0.0663 0.0491 0.0298 0.0195 0.0099 0.0176

Phormosoma placenta 0.0128 0.0260 0.0202 0.0052 0.0077 0.0132

Benthothuria funabris 0 0.0004 0.0004 0.0005 0 0

Enypniastes eximia 0 0 0 0 0.0033 0.0396

Elipidia echinata 0 0 0 0.0022 0.0011 0.0176

AnnelidsSabellid 0.0160 0.0018 0.0082 0.0005 0.0033 0.0220

CnidariansCerianthid 0.1425 0.1070 0.1663 0.2887 0.1967 0.1000

Actiniarian 0.0007 0 0 0 0 0

cf Actinoscyphia Aurelia 0.0025 0 0.0014 0.0005 0.0022 0

Scyphozoan 0 0.0040 0.0053 0.0002 0.0011 0.0011

Pennatulid 0 0 0 0 0 0.0011

Poriferansindet. Orange sponge 0 0 0 0.0002 0 0.0055

Total 0.3424 0.2832 0.3499 0.3829 0.3055 0.2681

Species richness (S) 18 20 22 24 18 18

ES(150) 12.1 15.2 14.2 11.5 15.7 15.8

Evenness (J�) 0.59 0.65 0.61 0.34 0.51 0.75

Diversity (H�) 1.70 1.95 1.88 1.09 1.47 2.16

D.O.B. Jones, M.E. Brewer / Deep-Sea Research I 67 (2012) 98–110 105

The combination of two environmental variables that bestdescribes the data was depth and broad-scale bathymetric posi-tion index (Rho¼0.25). A similar, but slightly lower correlationwas found with depth and fine-scale BPI (Rho¼0.25). The combi-nation of environmental variables that best describes the multi-variate community is a three variable combination of depth, TRIand mean curvature (Rho¼0.26).

4. Discussion

4.1. Geomorphological units identified

This study adds significant local geomorphological detail to thebroad-scale studies of the Mauritanian margin (Krastel et al.,2006). The slumping head wall and blocky deposit featuresassociated with the Mauritania slide (Antobreh and Krastel,2007) are clearly apparent in the bathymetry and its derivatives.The depth and structure of the features found here agrees withthat found in Parasound profile sections approximately 50 km

further south (Antobreh and Krastel, 2007). Indeed, these featureslikely extend all along the4150 km of Mauritanian margin thathas been affected by the Mauritanian slide (Henrich et al., 2008).This study site was located approximately in the centre ofthe slide area and it appears, from geological studies, that thefeatures analysed are representative of a much wider area(Krastel et al., 2006).

The slope of the headwall regions is steep for deep-waterenvironments (up to around 201). Slopes suggested by parasoundprofile sections were similar (Antobreh and Krastel, 2007). Between600 and 1400 m gradients were typically up to �2.51 and theheadwall scars themselves were up to �81 (Antobreh and Krastel,2007). The parasound profiles were based on data with a compar-able spatial resolution to the bathymetry, with an approximately100 m footprint diameter (at 1400 m depth), compared to the 90 mbathymetry grid. The data here extend previous observations andshow that the headwall features are fairly consistent throughout thestudy area.

The blocky deposits identified (Henrich et al., 2008; Krastel et al.,2006) are very obvious on the bathymetric derivatives as areas of

Table 3Density (number m�2) and composition of megafaunal communities at 1300–1400 m depth the Mauritanian slope. Data are divided by geomorphological classifications.

Species number (S) is used as an estimate of species richness. Evenness is calculated by Pielou’s evenness index (J0). The Shannon-Wiener (H0) diversity index is presented

for comparison with other studies.

Densities (no m�2)

A D E F

FishDalatiid cf Centroscyllium fabricii 0.0022 0.0037 0 0

Corphaenoides sp. 0.0066 0.0055 0.0088 0.0055

Macrourid 0.0044 0.0018 0.0011 0.0099

Small macrourid 0.0022 0.0037 0.0022 0

Notocanthid 0.0088 0.0055 0.0044 0.0209

Dicrolene intronigra 0.0022 0.0495 0.0121 0.0055

Alepocephalus sp. 0.0022 0.0147 0.0044 0.0044

MolluscsSepiid 0.0022 0.0018 0.0022 0

Benthoctopus sp. 0 0.0018 0.0011 0.0011

ArthropodsPalemonid prawn 0.0352 0.0897 0.0363 0.0044

Polycheles sp. 0.0022 0.0092 0.0077 0.0022

Parapagurus pilosimanus (and Epizoanthus paguriphilus) 0 0 0 0.0011

EchinodermsOphiuroid (cf Bathypectinura heros) 0.0176 0.0293 0.0099 0.0132

Ophiuroid 2 (gorgonian associate) 0 0 0 0.0033

Phormosoma placenta 0.0132 0.0037 0.0077 0.0055

Enypniastes eximia 0 0 0.0033 0

Elipidia echinata 0.0022 0 0.0011 0

Comatulid crinoid 0 0 0 0.0011

AnnelidsSabellid 0.0132 0.0018 0.0033 0

CnidariansCerianthid 0.2264 0.4176 0.1967 0.2286

cf Actinoscyphia aurelia 0 0.0018 0.0022 0

Scyphozoan 0 0.0055 0.0011 0

Gorgonian 0 0 0 0.0099

PoriferansOrange sponge 0.0022 0 0 0

Total 0.3429 0.6465 0.3055 0.3165

Species richness (S) 16 17 18 15

ES(150) 15.7 11.7 15.7 12.2

Evenness (J�) 0.50 0.47 0.51 0.46

Diversity (H�) 1.40 1.33 1.47 1.24

-0.5 0.0 0.5

-0.5

0.0

0.5

Stress = 0.27 900-10001000-11001100-12001200-13001300-14001400-1500

-0.5 0.0 0.5 1.0

-0.5

0.0

0.5

1.0

Stress = 0.18ADEF

Fig. 5. Non-metric multi-dimensional scaling (MDS) ordination of similarities of square root transformed megafaunal assemblage data for the Mauritanian margin. A)

Broad-scale patterns with depth, in metres B) fine-scale patterns with geomorphological zone at 1300–1400 m depth.

D.O.B. Jones, M.E. Brewer / Deep-Sea Research I 67 (2012) 98–110106

medium to high TRI and high fractal dimension. Fine-scale videoobservations support the suggestion that these areas are composedof sediment blocks, and suggest that there is some additional

complexity at a finer scale. The block features were approximately50 m long and about 10 m high, corresponding with estimates usingacoustic methods (Henrich et al., 2008).

Megafaunal density, no m-2

0.0001 0.001 0.01 0.1 1 10

Dep

th, m

0

1000

2000

3000

4000

5000

6000

EAtlantic (0 - 45 o N)MauritaniaThis study

Fig. 6. Deep-water (4200 m) megafaunal densities (no. m�2), from the present

study and literature sources, compared with depth for the Eastern Atlantic. East

Atlantic literature data from Bay of Biscay (451N) south to the Equator (Feldt et al.,

1989; Galeron et al., 2000; Lebrato and Jones, 2009; Sibuet et al., 1989, 1984;

Sibuet and Segonzac, 1985; Thurston et al., 1994), Mauritania data from EUMELI

sites (Galeron et al., 2000). Note log axis for density.

D.O.B. Jones, M.E. Brewer / Deep-Sea Research I 67 (2012) 98–110 107

4.2. Biological patterns with habitat change

Depth was the most important factor for describing commu-nity change on the Mauritanian margin. This is the case, at abroad-scale, in most areas of the world investigated (see reviewsin Carney, 2005; Rex and Etter, 2010). Very few species appearedto be at the limits of their bathymetric range (i.e., present atone end of the depth scale and absent from the other). However,there are clear and significant changes in the abundances of taxa.This was expected as taxon-specific preferences for a varietyof factors, either directly related to hydrostatic pressure orcovarying with depth, narrow the bathymetric zone in which aspecies is abundant (Howell et al., 2002). Faunal density, althoughsignificantly different between depth bands, does not displayany clear trends with depth, except notable reductions in densityin the deepest areas investigated. Although benthic faunal den-sities typically decrease with depth as a result of the reducedflux of organic matter (Rex and Etter, 2010), many studies havereported high local variation, usually related to a wide suiteof temporally and spatially varying environmental parametersor changes in the biology of abundant organisms (e.g., size)(Carney, 2005). There is not sufficient high-resolution dataon depth-related changes in environmental parameters on theMauritanian margin to speculate on specific controls that mayexplain the fine-scale depth-related differences in densityobserved in this study.

The benthic megafaunal assemblages seem to only respond tothe more extreme variations in geomorphology observed here.The assemblages of geomorphology F are the most variable intaxon composition, reflecting the high habitat heterogeneity. Atthe level of overall density or species diversity measures, thesedifferences are not detectable, presumably because the habitatvariation is too subtle to cause an overarching trend in all speciesencountered. Different species will respond in different ways toimportant habitat characteristics, or changes in other factorsrelated to habitat change, e.g., local current increases in moresteeply sloping areas (Palardy and Witman, 2011; Wildish andKristmanson, 1997). Multivariate techniques for communityassessment showed small but detectable change, reflecting theincreased sensitivity of this method (Clarke and Warwick, 2001).The low densities of organisms may reduce the power to detectchanges, particularly as many taxa are only represented by one ortwo individuals within a sampling unit. This may partially explainlack of observed pattern between zones and high stress values inmultivariate analysis. Changes in densities of individual speciesbetween geomorphological zones were sometimes major. Forexample, two likely suspension feeding taxa (gorgonians andassociated ophiuroids) were only present in one geomorphologi-cal zone. Many of the species found on areas with higher slopewere suspension feeders, perhaps taking advantage of localincreases in current flow (Leichter and Witman, 1997; Wildishand Kristmanson, 1997). The majority of the benthic megafaunaobserved here were deposit feeders, which generally did notexhibit any clear patterns with geomorphological zone. If therewere major changes in habitat between zones we would expect tosee changes in deep-sea deposit feeders, as they are known torespond rapidly and consistently to changes in habitat quality(Ruhl and Smith, 2004). The changes in habitat between geomor-phological zones were generally not extreme and did not appearfrom the video to affect the gross composition of the sedimentaryenvironment. It appears likely that small changes in geomorphol-ogy have negligible effect on benthic species composition on theMauritanian margin. The geomorphological classification meth-ods used here provide a highly sensitive approach to classificationof seabed types, which may have more discriminating power thanis meaningful for benthic megafauna.

4.3. Regional-scale assessment of Mauritanian megafaunal

assemblages

Megafaunal densities in this study are comparable to thosederived from other deep-water studies on the West Africanmargin (Fig. 6). The nearest sites for comparison, in terms ofgeographical location, were the eutrophic EUMELI site (eEs), alsoon the Mauritanian margin (Galeron et al., 2000). This was around300 km NNW of the location of this study and slightly deeper(1600–2100 m). The densities obtained here are considerablylower than those found at the eEs. Seasonal upwelling results inhigh surface production (500 g Cm�2 yr�1) at the eEs and asso-ciated increases in flux of organic material to the benthos. Thesites investigated here do receive elevated flux, but not as high asthe eEs (Helmke et al., 2005). Most of the megafaunal density ateEs was represented by cnidarians, which were an order ofmagnitude more abundant than found here. Other taxa at eEswere also significantly more abundant than in the present study(Galeron et al., 2000). Variations in methodology may explaindensity differences, although photographic surveys typically findhigher megafaunal densities, especially for soft-bodied organisms(Thurston et al., 1994). Few other data are available for the EastAtlantic nearby at equivalent depths. Our preliminary evaluationshows an approximate log linear reduction in megafaunaldensities with depth in the eastern Atlantic, as has been pre-viously noted (Rex et al., 1990). The densities for the sitesevaluated here do not deviate from this trend.

Most of the species identified during the present study arecommon to the West Africa region in general (see Supplementarymaterial for species-level comparisons with the literature) andmany have been encountered in other surveys of megafauna offMauritania (Galeron et al., 2000; Merrett and Marshall, 1981).Taxon richness is lowest at the centre of the bathymetric rangeinvestigated. This may be because this range does not encompassany faunal transition zones (Carney, 2005), which tend to includespecies from adjacent zones and therefore have higher speciesrichness (Rex and Etter, 2010). Further sampling at deep andshallow stations is required to confirm this hypothesis.

Megafaunal community structure assessed here was variablewithin and between strata. The deepest site was fairly similar inthe proportion of megafauna in each phylum to that found at eEs

D.O.B. Jones, M.E. Brewer / Deep-Sea Research I 67 (2012) 98–110108

(Galeron et al., 2000), with the most abundant faunal groupsbeing cnidarians and echinoderms. These patterns, however, areextremely difficult to interpret as there were huge changes inmegafaunal density and composition between years at the sameeEs site (Galeron et al., 2000). With high temporal variation it isperhaps not surprising that patterns between studies are notclear. Even within the narrow depth bands studied here, there arelarge changes in megafaunal composition. It is also likely thattemporal and broad-scale spatial variability is very high in thepresent study region, resulting from the variable surface produc-tion (Helmke et al., 2005).

4.4. Evaluation of the approach

The ISOCLUSTER technique effectively summarised the bathy-metric derivative layers into areas visually determined as havingdiffering geomorphology, for example creating distinct classes forthe headwall and canyon features. It was less effective indifferentiating the areas of blocky deposits, which are mostclearly visible in the TRI, mostly likely as the scale of thesefeatures were very close to the resolution of the data. TheISOCLUSTER technique is based on the natural breaks within eachof the layers (Ball and Hall, 1965). These breaks are likely to bebiologically meaningful; for example, areas of major transition inrugosity or slope are likely to represent faunal boundaries(Carney, 2005). As a result, the geomorphological zones aresuitable for determining whether fine-scale habitat-related pat-terns are important in determining the composition of biologicalassemblages. However, the ISOCLUSTER technique ignores therelative importance of each input variable and makes it difficult todetermine which input variables are responsible for the greatestvariation in the eventual classified geomorphological zones. ISO-CLUSTER is relatively robust to the effects of the potentialstatistical data redundancy that could occur as all layers used inclassification were derivatives of bathymetry (Ball and Hall,1965). Owing to the relative importance of the geomorphologicalclassification method in our interpretation of fine-scale biologicalpatterns it was important to use the BIOENV analysis as anindependent assessment. Based purely on the biological data,the variables that best describe faunal patterns were a combina-tion of broad-scale processes (associated with depth) and thefine-scale BPI. The fine-scale BPI highlights seabed areas (in thiscase 90�90 m) with large differences in elevation from thesurrounding area. As this index highlights areas of fine-scaleanomalies in topography it shows the greatest change along theheadwall areas of the study site. The variation in geomorphologyacross the headwall features is close to the limit of resolution ofthis study, but it appears to be the source of the greatestbiologically-meaningful habitat heterogeneity at the scale of theinvestigation.

Bathymetry is commonly gridded with pixel sizes from 20 to100 m in deep-water surveys. This work suggests that, althoughdifferences can be detected, there was substantial biologically-important habitat variation at a finer scale than can be recognisedfrom bathymetry at this scale. Video observations suggest that theheadwall features (and potentially blocky deposits) would be muchmore readily discriminated with pixels of 10 m or less. The samplesites where these features are clearly visible in video are oftenclassified in the same ‘habitat’, although the averaging effects ofthe large pixel sizes means that the fine-scale habitat variation ismasked and non-slope features are also included in generatedpolygons. Where obvious features are observed in video records,the associated faunal communities are clearly different. Theevidence suggests that the headwalls and their associated com-munity are discrete and cover a small area but could readily bediscriminated by a finer-scale analysis.

5. Conclusion

This study focuses in on a typical section of the Mauritanianmargin, with the slumping head-wall and blocky deposits asso-ciated with the Mauritania slide. Broad-scale variations in thephysical environment associated with depth resulted in highlysignificant differences in megafaunal assemblage density, speciesrichness and evenness. Assemblage composition was also vari-able, with shallow, intermediate and deep assemblages recogni-sable and significant differences between depth zones inmultivariate analysis. At a finer scale, variations in geomorphol-ogy have surprisingly little influence on the benthic megafauna,with only steeply sloping head-walls having a notably differentmegafaunal assemblage composition. This may be partially aresult of the scale (grain) of the multibeam bathymetry data,which may miss biologically-important habitat heterogeneity.This study greatly enhances available information on bathyalmegafaunal assemblages on the Mauritanian margin. The patternsrevealed likely reflect assemblages on a provincial scale. Theresults could be extrapolated within depth zones to the majority(495%) of the Mauritanian margin that lacks major geomorpho-logical features, such as canyons. As changes in seafloor habitatresulting from submarine mass movements are common globally,occurring on every passive continental margin (Masson et al.,2010), the responses of the megafaunal assemblages observedhere may also help to explain patterns at a global scale.

Acknowledgements

The authors thank Woodside Petroleum Ltd. for supportingthis work. We also thank Tullow Oil Plc. for providing access tothe data. We acknowledge Mike Forde for participating in theproject and Jeremy Colman for his help. This study would havebeen impossible without the help of the captain and crew of theBoa Deep C, including the Oceaneering ROV team. We thankR. Wynn, D. Masson, V. Huvenne, A. Gooday and three anonymousreviewers for constructive comments on this manuscript. Thisinvestigation was carried out as part of the SERPENT Projectwww.serpentproject.com. This work was funded by the UKNatural Environment Research Council as part of the MarineEnvironmental Mapping Programme (MAREMAP).

Appendix A. Supporting information

Supplementary data associated with this article can be foundin the online version at http://dx.doi.org/10.1016/j.dsr.2012.05.006.

References

Antobreh, A.A., Krastel, S., 2007. Mauritania Slide Complex: morphology, seismiccharacterisation and processes of formation. Int. J. Earth Sci. 96 (3), 451–472.

Auster, P.J., Malatesta, R.J., Larosa, S.C., 1995. Patterns of Microhabitat Utilizationby mobile megafauna on the Southern New-England (USA) continental-shelfand slope. Mar. Ecol.Prog. Ser. 127 (1-3), 77–85.

Ball, G.H., Hall, D.J., 1965. A Novel Method of Data Analysis and PatternClassification. Stanford Research Institute, Menlo Park, California.

Barry, J.P., Baxter, C.H., 1992. Survey design considerations for deep-sea benthiccommunities using ROVs. Mar. Technol. Soc. J. 26 (4), 21–26.

Bergstad, O.A., Falkenhaug, T., Astthorsson, O.S., Byrkjedal, I., Gebruk, A.V.,Piatkoweski, U., Priede, I.G., Santos, R.S., Veccione, M., Lorance, P., Gordon,J.D.M., 2008a. Towards improved understanding of the diversity and abun-dance patterns of the mid-ocean ridge macro-and megafauna. Deep Sea Res. 2(55), 1–5.

Bergstad, O.A., Menezes, G., Høines, A.S., 2008b. Demersal fish on a mid-oceanridge: distribution patterns and structuring factors. Deep Sea Res. 2 (55),185–202.

D.O.B. Jones, M.E. Brewer / Deep-Sea Research I 67 (2012) 98–110 109

Beyer, H.L., 2004. Hawth’s Analysis Tools for ArcGIS. Available at /http://www.spatialecology.com/htoolsS.

Bray, J.R., Curtis, J.T., 1957. An ordination of the upland forest of southernWisconsin. Ecolo. Monogr. 27, 225–349.

Bryan, T.L., Metaxas, A., 2006. Distribution of deep-water corals along the NorthAmerican continental margins: relationships with environmental factors.Deep Sea Res. Part I 53 (12), 1865–1879.

Carney, R.S., 2005. Zonation of deep biota on continental margins. Oceanogr.Mar.Biol. Annu. Rev. 43, 211–278.

Clarke, K.R., Warwick, R.M., 2001. Changes in Marine Communities: An Approachto Statistical Analysis and Interpretation. Plymouth Marine Laboratory, U.K.

Colman, J.G., Gordon, D.M., Lane, A.P., Forde, M.J., Fitzpatrick, J.J., 2005. CarbonateMounds off Mauritania, north-west Africa: Status of Deep-Water Corals andImplications for Management of Fishing and Oil Exploration Activities. In:Freiwald, A., Roberts, J.M. (Eds.), Cold-Water Corals and Ecosystems. Springer-Verlag, Berlin, pp. 417–441.

Dolan, M.F.J., Grehan, A.J., Guinan, J.C., Brown, C., 2008. Modelling the localdistribution of cold-water corals in relation to bathymetric variables: addingspatial context to deep-sea video data. Deep Sea Res. Part I 55 (11),1564–1579.

Duineveld, G.C.A., Lavaleye, M.S.S., Noort, G.J.v., 1993a. The trawlfauna of theMauritanian shelf (Northwest Africa): density, species composition, andbiomass. Hydrobiologia 258 (1–3), 165–173.

Duineveld, G.C.A., Wilde, P., Berghuis, E.M., Kok, A., 1993b. The benthic infauna andbenthic respiration off the Banc d’Arguin (Mauritania, Northwest Africa).Hydrobiologia 258 (1–3), 107–117.

ERDAS, 2005. ERDAS Field Guide. Leica Geosystems, Atlanta, GA.Evans, I.S., 1980. An integrated system of terrain analysis and slope mapping.

Zeitschrift fur Geomorphologic Suppl-Bd 36, 274–295.Evans, J., 2004. Topographic Ruggedness Index. AML script for ArcInfo

Workstation.Feldt, W., Kanisch, G., Vobach, M., 1989. Deep-sea biota of the northeast Atlantic

and their radioactivity. In: Nyffeler, F. (Ed.), Interim Oceanographic Descrip-tion of the North-East Atlantic Site for the Disposal of Low-Level RadioactiveWaste, Vol. 3. Nuclear Energy Agency, OECD, Paris, pp. 178–204.

Field, J.G., Clarke, K.R., Warwick, R.M., 1982. A practical strategy for analysingmultispecies distribution patterns. Mar. Ecol. Prog. Ser. 8, 37–52.

Fock, H., Uiblein, F., Koster, F., von Westernhagen, H., 2002. Biodiversity andspecies-environment relationships of the demersal fish assemblage at theGreat Meteor Seamount (subtropical NE Atlantic), sampled by different trawls.Mar. Biol. 141 (1), 185–199.

Forster, A., Ellis, R.G., Henrich, R., Krastel, S., Kopf, A.J., 2010. Geotechnicalcharacterization and strain analyses of sediment in the Mauritania SlideComplex, NW-Africa. Mar. Pet. Geol. 27 (6), 1175–1189.

Futterer, D.K., 1983. The modern upwelling record off Northwest Africa. In: Thiede,J., Suess, E. (Eds.), Coastal Upwelling: Its Sediment Record, Part B. SedimentaryRecords of Ancient Coastal Upwelling. Plenum Press, London, pp. 105–121.

Gage, J.D., Tyler, P.A., 1991. Deep Sea Biology. Cambridge University Press,Cambridge.

Galeron, J., Sibuet, M., Mahaut, M.L., Dinet, A., 2000. Variation in structure andbiomass of the benthic communities at three contrasting sites in the tropicalNortheast Atlantic. Mar. Ecol.Prog. Ser. 197, 121–137.

Gebruk, A., Budaeva, N.E., King, N., 2010. Bathyal benthic fauna of the Mid-AtlanticRidge between the Azores and the Reykjanes Ridge. J. Mar. Biol. Assoc. theUnited Kingdom 90 (1), 1–14.

Helmke, P., Romero, O., Fischer, G., 2005. Northwest African upwelling and itseffect on offshore organic carbon export to the deep sea. Global Biogeochem.Cycles 19, 4.

Henrich, R., Cherubini, Y., Meggers, H., 2010. Climate and sea level inducedturbidite activity in a canyon system offshore the hyperarid Western Sahara(Mauritania): The Timiris Canyon. Mar. Geol. 275 (1–4), 178–198.

Henrich, R., Hanebuth, T.J.J., Krastel, S., Neubert, N., Wynn, R.B., 2008. Architectureand sediment dynamics of the Mauritania Slide Complex. Mar. Pet. Geol. 25(1), 17–33.

Henriques, C., Priede, I.G., Bagley, P.M., 2002. Baited camera observations of deep-sea demersal fishes of the northeast Atlantic Ocean at 15–28 degrees N offWest Africa. Mar. Biol. 141 (2), 307–314.

Herzfeld, U.C., Overbeck, C., 1999. Analysis and simulation of scale-dependantfractal surfaces with application to seafloor morphology. Comput. Geosci. 25(9), 979–1007.

Howell, K.L., Billett, D.S.M., Tyler, P.A., 2002. Depth-related distribution andabundance of seastars (Echinodermata: Asteroidea) in the Porcupine Seabightand Porcupine Abyssal Plain, N.E. Atlantic. Deep Sea Res. Part I 49, 1901–1920.

Jacobi, R.D., 1976. Sediment slides on the northwestern continental margin ofAfrica. Mar. Geol. 22, 157–173.

Jones, D.O.B., Bett, B.J., Tyler, P.A., 2007a. Depth-related changes in the arcticepibenthic megafaunal assemblages of Kangerdlugssuaq, East Greenland. Mar.Biol. Res. 3 (4), 191–204.

Jones, D.O.B., Bett, B.J., Tyler, P.A., 2007b. Megabenthic ecology of the Faroe-Shetland Channel: a photographic study. Deep Sea Res. Part I 54, 1111–1128.

Kaiser, M., Attril, M., Jennings, S., Thomas, D.N., Barnes, D., Brierley, A., Raffaelli,N.P.D., Williams, P.L.B., 2005. Marine Ecology: Processes, Systems, andImpacts. Oxford University Press, Oxford.

Keller, N.B., Pasternak, F.A., 2002. Madreporic and Gorgonian corals (Anthozoa:Scleractinia, Gorgonacea) in the benthos of the African shelf and slope in theCanary upwelling area. Oceanology 42 (4), 521–525.

Krastel, S., Wynn, R.B., Georgiopoulou, A., Geersen, J., Henrich, R., Meyer, M.,Schwenk, T., 2011. Large-Scale Mass Wasting on the Northwest AfricanContinental Margin: Some General Implications for Mass Wasting on PassiveContinental Margins. Submarine Mass Movements and their Consequences V.Springer, Dordrecht, pp. 189–199.

Krastel, S., Wynn, R.B., Hanebuth, T.J.J., Holz, C., Georgiopoulou, A., Henrich, R.,Schulz, H.D., 2006. Seafloor mapping of submarine geohazards offshoreMauritania, Northwest Africa. Norw. J. Geol. 86, 163–176.

Lebrato, M., Jones, D.O.B., 2009. Mass deposition event of Pyrosoma atlanticumcarcasses off Ivory Coast (West Africa). Limnol. Oceanogr. 54 (4), 1197–1209.

Leichter, J.J., Witman, J.D., 1997. Water flow over subtidal rock walls: relation todistributions and growth rates of sessile suspension feeders in the Gulf ofMaine—Water flow and growth rates. J. Exp. Mar. Biol. Ecol. 209 (1–2), 293–307.

Levin, L.A., Dayton, P.K., 2009. Ecological theory and continental margins: whereshallow meets deep. Trends Ecol. Evol. 24 (11), 606–617.

Levin, L.A., Etter, R.J., Rex, M.A., Gooday, A.J., Smith, C.R., Pineda, J., Stuart, C.T.,Hessler, R.R., Pawson, D., 2001. Environmental influences on regional deep-seaspecies diversity. Annu. Rev. Ecol. Syst. 32, 51–93.

Lundblad, E.R., Wright, D.J., Miller, J., Larkin, E.M., Rinehart, R., Naar, D.F., Donahue,B.T., Anderson, S.M., Battista, T., 2006. A benthic terrain classification scheme forAmerican Samoa. Mar. Geod. 29 (2), 89–111.

Mandelbrot, B.B., 1983. The Fractal Geometry of Nature. W. H. Freeman andCompany, New York.

Mark, D.M., Aronson, P.B., 1984. Scale-dependent fractal dimensions of topo-graphic surfaces: an empirical investigation with applications in geomorphol-ogy and computer mapping. Math. Geol. 16 (7), 671–683.

Masson, D.G., Wynn, R.B., Talling, P.J., 2010. Large Landslides on Passive ContinentalMargins: Processes, Hypotheses and Outstanding Questions. Submarine MassMovements and their Consequences IV. Springer, Dordrecht, pp. 153–166.

Merrett, N.R., Domanski, P.A., 1985. Observations on the ecology of deep-seabottom-living fishes collected off northwest Africa: II. The Moroccan Slope(27-34N), with special reference to Synaphobranchus kaupi. Biol. Oceanogr. 3(4), 349–399.

Merrett, N.R., Marshall, N.B., 1981. Observations on the ecology of deep-seabottom-living fishes collected off northwest Africa (08-27N). Prog. Oceanogr.9 (4), 185–244.

Mortensen, P.B., Buhl-Mortensen, L., 2004. Distribution of deep-water gorgoniancorals in relation to benthic habitat features in the Northeast Channel (AtlanticCanada). Mar. Biol. 144 (6), 1223–1238.

Mortensen, P.B., Hovland, M.T., Fossa, J.H., Furevik, D.M., 2001. Distribution,abundance and size of Lophelia pertusa coral reefs in mid-Norway in relationto seabed characteristics. J. Mar. Biol. Assoc. the United Kingdom 81 (4),581–597.

Moss, D.S., 1992. A summary of the Porifera collected during ‘‘Expedition Azores1989’’. Arquipelago, Life Earth Sci. 10, 43–53.

Palardy, J.E., Witman, J.D., 2011. Water flow drives biodiversity by mediating rarityin marine benthic communities. Ecolo. Lett. 14 (1), 63–68.

Pfannkuche, O., Theeg, R., Thiel, H., 1983. Benthos activity, abundance and biomassunder an area of low upwelling off Morocco, northwest Africa. MeteorForschungsergeb. Reihe D 36, 85–96.

Piepenburg, D., Muller, B., 2004. Distribution of epibenthic communities on theGreat Meteor Seamount (North-east Atlantic) mirrors pelagic processes. Arch.Fish. Mar. Res. 51 (1–3), 55–70.

Rex, M.A., Etter, R.J., 2010. Deep-Sea Biodiversity: Pattern and Scale. HarvardUniversity Press, Harvard.

Rex, M.A., Etter, R.J., Nimeskern Jr., P.W., 1990. Density estimates for deep-seagastropod assemblages. Deep Sea Res. Part I 37 (4A), 555–569.

Riley, S.J., DeGloria, S.D., Elliot, R., 1999. A terrain ruggedness index that quantifiestopographic heterogeneity. Intermountain J. Sci. 51 (1–4), 23–27.

Ruhl, H.A., Smith, K.L., 2004. Shifts in deep-sea community structure linked toclimate and food supply. Science 305 (5683), 513–515.

Schmidt, J., Evans, I.S., Brinkmann, J., 2003. Comparison of polynomial methods forland surface curavture calculation. Int. J. Geog. Inf. Sci. 17 (8), 797–814.

Seibold, E., Hinz, K., 1974. Continental slope construction and destruction, WestAfrica. In: Burk, C.A., Drake, C.L. (Eds.), The Geology of Continental Margins.Springer, New York, pp. 179–196.

Shary, P.A., Sharaya, L.S., Mitusov, A.V., 2002. Fundamental quantitative methodsof land surface analysis. Geoderma 107, 1–32.

Sibuet, M., Albert, P., Charmasson, S.S., Deming, J., Dinet, A., Galeron, J., Guidi-Guilvard,L., Mahaut, M.-L., 1993. The benthic ecosystem in three EUMELI sites in thenortheast tropical Atlantic: general perspectives and initial results on biologicalabundance and activities. T. Ann. Inst. Oceanogr. 69, 21–33.

Sibuet, M., Lambert, C.E., Chesselet, R., Laubier, L., 1989. Density of the major sizegroups of benthic fauna and trophic input in deep basins of the Atlantic Ocean.J. Mar. Res. 47 (4), 851–867.

Sibuet, M., Monniot, C., Desbruyeres, D., Dinet, A., Khripounoff, A., Rowe, G.T.,Segonzac, M., 1984. Peuplements benthiques et caracteristiques trophiques aumilieu dans la plaine abyssale de Demerara dans l’ocean Atlantique. Oceano-log. Acta 7, 345–358.

Sibuet, M., Segonzac, M., 1985. Abondance et repartition de l’epifaune mega-benthique. In: Laubier, L., Monniot, C. (Eds.), Peuplements Profonds du Golfede Gascogne. Institut Francaise de Recherche pour l’Exploration de la Mer,Brest, pp. 143–156.

Thurston, M.H., Bett, B.J., Rice, A.L., Jackson, P.A.B., 1994. Variations in theinvertebrate abyssal megafauna in the North Atlantic Ocean. Deep Sea Res.Part I 41 (9), 1321–1348.

D.O.B. Jones, M.E. Brewer / Deep-Sea Research I 67 (2012) 98–110110

Thurston, M.H., Rice, A.L., Bett, B.J., 1998. Latitudinal variation in invertebratemegafaunal abundance and biomass in the North Atlantic Ocean Abyss. DeepSea Res. Part II 45 (1–3), 203–224.

Uiblein, F., Bordes, F., Castillo, R., 1996. Diversity, abundance and depth distribu-tion of demersal deep-water fishes off Lanzarote and Fuerteventura, CanaryIslands. J. Fish Biol. 49, 75–90.

Uiblein, F., Geldmacher, A., Koster, F., Nellen, W., Krauz, G., 1999. Speciescomposition and depth distribution of fish species collected in the area ofthe Great Meteor Seamount, eastern central Atlantic, during cruise MM 42/3,with seventeen new records. Informes Tecnicos Instituto Canario de CienciasMarinas 5, 47–85.

van Camp, L., Nykjaer, L., Mittelstaedt, E., Schlittenhardt, P., 1991. Upwelling andboundary circulation off Northwest Africa as depicted by infrared and visiblesatellite observations. Prog. Oceanogr. 26, 357–402.

Wildish, D., Kristmanson, D., 1997. Benthic Suspension Feeders and Flow. CambridgeUniversity Press, Cambridge.

Williams, A., Althaus, F., Dunstan, P.K., Poore, G.C.B., Bax, N.J., Kloser, R.J.,

McEnnulty, F.R., 2010. Scales of habitat heterogeneity and megabenthosbiodiversity on an extensive Australian continental margin (100–1100 m

depths). Mar. Ecol. 31 (1), 222–236.Wilson, M.F.J., O’Connell, B., Brown, C., Guinan, J.C., Grehan, A.J., 2007. Multiscale

Terrain analysis of multibeam bathymetry data for habitat mapping on thecontinental slope. Mar. Geod. 30 (1), 3–35.

Wood, J., 2005. Landserf Version 2.2 (/www.landserf.orgS).Wynn, R.B., Masson, D.G., Stow, D.A.V., Weaver, P.P.E., 2000. The Northwest African

slope apron: a modern analogue for deep-water systems with complex

seafloor topography. Mar. Pet. Geol. 17, 253–265.