Downloaded By: [Wilson, Margaret] At: 09:27 8 May 2007
Marine Geodesy, 30: 3–35, 2007
Copyright © Taylor & Francis Group, LLC
ISSN: 0149-0419 print / 1521-060X online
DOI: 10.1080/01490410701295962
Multiscale Terrain Analysis of Multibeam
Bathymetry Data for Habitat Mapping on the
Continental Slope
MARGARET F. J. WILSON, BRIAN O’CONNELL, COLIN
BROWN, JANINE C. GUINAN, AND ANTHONY J. GREHAN
Department of Earth and Ocean Sciences, National University of Ireland, Galway
Multibeam surveys can provide detailed bathymetry data for the continental slope from
which quantitative descriptors of the seabed terrain (e.g., slope) may be obtained. We
illustrate the value of these descriptors for benthic habitat mapping, and highlight the
advantages of multiscale analysis. We examine the application of these descriptors as
predictor variables for species distribution models, which are particularly valuable in
the deep sea where opportunities to directly survey the benthic fauna remain limited. Our
initial models are encouraging and suggest that wider adoption of these methods may
assist the delivery of ecologically relevant information to marine resource managers.
Keywords Multibeam bathymetry, terrain analysis, scale, habitat suitability modelling,
continental slope
Introduction
Ecologists involved in terrestrial habitat and vegetation mapping studies have developed
their science against a backdrop of improving topographic information. Visualization
and analysis tools too have advanced and simple contour maps have now largely given
way to Digital Terrain Models (DTM), which may be readily analyzed in a geographic
information system (GIS) environment. Parameters derived from these DTMs have been
used in numerous studies to classify habitat and to predict soil, vegetation cover, and species
distributions at a variety of scales (Franklin 1995; Franklin et al. 2000; Hirzel 2001; Schmidt
and Hewitt 2004; Wu and Smeins 2000).
Until recently, no detailed terrain data were available for the marine environment.
Historically, bathymetry data were acquired primarily for navigation purposes and, while
certain coastal areas may be reasonably well charted, vast areas of the deep sea remain
largely unexplored. Even when studies such as Le Danois (1948) provided evidence that the
deep sea supported a variety of benthic fauna and began describing its regional distribution,
attempts at habitat mapping were limited without this baseline bathymetry information.
With only a basic knowledge of seafloor bathymetry coupled with imprecise position
information for biological samples, studies of the deep-sea benthic fauna tended to focus
Received 21 June 2006; accepted 4 November 2006.
Address correspondence to Margaret F. J. Wilson, Geological Survey of Norway, 7491
Trondheim, Norway. E-mail: margaret.wilson@ngu.no
3
Downloaded By: [Wilson, Margaret] At: 09:27 8 May 2007
4 M. F. J. Wilson et al.
on ecology (e.g., Billett 1991) biodiversity (e.g., Rex et al. 1997), or biogeography of certain
species (e.g., Riemann-Z
¨
urneck 1986) without any explicit link to the seabed terrain.
With the advent of multibeam technology (Kenny et al. 2003), marine benthic habitat
mapping has entered a new era. Ship borne multibeam bathymetry surveys, such as those
conducted under the Irish National Seabed Survey (INSS) (GOTECH 2002), have provided
spectacular detail of deep sea terrain revealing numerous previously unrecognized features.
Multibeam surveys provide the detailed bathymetry data necessary for the production of
submarine DTMs which, unlike traditional charting f ormats (contours, soundings), lend
themselves well to terrain analysis. Multibeam data have proven their value for habitat
mapping (Kostylev et al. 2001; Parnum et al. 2004; Van De Beuque et al. 1999) and studies
of the distribution of benthic fauna (Kostylev et al. 2003). Shipborne multibeam data are
also being complemented by Remotely Operated Vehicle (ROV)-based multibeam surveys,
which provide a means to obtain submeter level bathymetry i n deeper waters (e.g., Wilson
2006).
With the advent of ROV-based video surveys of the benthic fauna, we can now also
obtain precisely georeferenced visual data on the distribution of at least the larger animals
colonizing the seafloor – the benthic megafauna. These are generally visible in video data
and, thanks to precise position information from accurate underwater positioning systems,
we know exactly the location for each observation. Now that biological information can be
directly georeferenced to the underlying seabed terrain, we can begin to characterize the
distribution of the fauna in relation to this terrain and develop predictive habitat and species
distribution models. Renewed commercial interest in deep sea resources, coupled with the
need for sustainable management, demands habitat information and predictive models may
be the only practical way to address this demand in the deep sea.
While the distribution of benthic fauna may be controlled by a combination of
environmental and biological factors, it is generally recognized that many animals show
a particular affinity for certain types of terrain (e.g., D
ˇ
zeroski and Drumm 2003; Roberts
et al. 2003; Wilbur 2000), which provide the physical habitat or structure that is directly
or indirectly suited to their mode of living. Characterization of the seabed in terms of
terrain parameters s uch as slope, aspect, or curvature may therefore offer a valuable tool for
delineating regions of the continental slope that are likely to support particular fauna and
thereby provide a distinct habitat. Recent work in shallower water has indicated the potential
for these types of techniques (Bekkby et al. 2002; Dartnell and Gardner 2004; Lundblad
et al. 2006), but there has been little work in deeper waters beyond the continental shelf.
It is perhaps in the relatively inaccessible deep sea, where the expense of direct surveying
makes ground-truth observations scarce, that techniques such as terrain modeling based on
multibeam data can make a significant contribution to the prediction of benthic habitat.
One important issue that should be addressed by any attempt at habitat mapping,
observed or predicted, is that of spatial scale. With seabed habitats and benthic fauna
spanning a continuum of scales this is not an easy challenge to meet. Scale is especially
important in relation to terrain analysis, since both the initial DTM resolution and the
analysis scale will influence the results. For habitat mapping it is important to try
and match the data and analysis scales to those relevant to habitat size and the fauna
themselves. Previous investigations of deep-sea habitat using video (e.g., Klages et al. 2004)
have indicated that distribution of fauna exhibits patterns of variability on spatial scales
smaller than the resolution of ship-borne multibeam at continental slope depths. However,
these investigations have also revealed that certain fauna (e.g., cold water corals) exhibit
tendencies to associate with larger scale features of the terrain such as carbonate mounds
(De Mol et al. 2002). This suggests that larger-scale features may indeed be important
Downloaded By: [Wilson, Margaret] At: 09:27 8 May 2007
Multiscale Terrain Analysis of Multibeam Bathymetry Data 5
contributors to the distribution of some fauna. In practice, for animals exhibiting a preference
for particular types of terrain it is likely that a combination of large- and small-scale features
contribute to their chosen habitat. We investigate the interaction of spatial scales with
reference to a case study area in this paper. Several technical approaches that inherently
address issues of scale can be recognized, including multiscale GIS visualization and
analysis (Wood 1999), fractal dimension (Wallace et al. 2004), and wavelet analysis (Csillag
and Kabos 2002), each of which will be discussed further in this paper.
We begin with a review of methods for terrain analysis using deep-sea multibeam
data. We then provide examples in a case study on the upper continental slope to the
southwest of Ireland. Results of analyses at multiple scales are compared with each other
and with video observations of benthic fauna with the aim of addressing the following
questions:
Ĺ What terrain information can be derived from multibeam data?
Ĺ What are the benefits of analyzing data at multiple spatial scales?
Ĺ What spatial scales are most relevant to observed seabed habitat?
Ĺ Can the derived terrain variables be used to predict the distribution of benthic
fauna?
Bathymetric Terrain Analysis
The terrain analyses focus on techniques that have most potential relevance to benthic
habitat. Some of these methods are common to terrestrial applications; others have been
modified for the marine environment. Terrain analysis methods can be grouped into four
types of information (Figure 1):
Ĺ Slope,
Ĺ Orientation (aspect)
Ĺ Curvature and relative position of features
Ĺ Terrain variability
Each of these parameters potentially gives important information for the delineation
and characterization of habitats and may be valuable inputs to predictive habitat modeling.
In the following sections, we provide an overview of terrain analysis with notes on any
Figure 1. Major classes of terrain parameters that may be derived from bathymetry data.
Downloaded By: [Wilson, Margaret] At: 09:27 8 May 2007
6 M. F. J. Wilson et al.
special considerations for deep-sea data and the relevance of each parameter to seabed
habitat.
GIS-Based Analyses
By converting the bathymetric DTM to a raster grid, terrain analysis can be performed
conveniently in a GIS environment. The raster resolution (also referred to by pixel or cell
size) will be determined by the original multibeam data and the purpose for which the
data are being prepared. One important point concerning multibeam data is that resolution
and data density decrease with depth as a consequence of the beam geometry and lower
multibeam frequencies used (Wilson 2006). Regardless of raster resolution, terrain analysis
may be performed. However, it is important to note that the cell size, the analysis scale, and
the choice of algorithm for the calculation of terrain parameters will all influence the results
obtained (Albani et al. 2004). Although we have gridded our data at an optimal resolution
for the study area, the majority of scientists and managers are restricted to working with
agency-produced DTMs at a particular resolution. Unless DTMs at the highest meaningful
resolution are available within the study area, the choice of algorithms and analysis scale
will be restricted and so it is particularly important to understand the impact these will have
on the analysis.
While certain terrain analyses may be performed on the basis of the raster grid and
algebraic interactions between pixels, a more elegant and robust solution is to use some
continuous representation of the DTM as a double-differentiable surface (e.g., Wood 1996).
As we will demonstrate, this approach offers great flexibility in the choice of algorithms
and the scales at which these analyses may be performed.
Using Evans’ (1980) method, the DTM surface is approximated by a bivariate quadratic
equation
Z = aX
2
+ bY
2
+ cX Y + dX + eY + f (1)
where Z is the height of the DTM surface and X and Y are the horizontal coordinates.
The coefficients in Eq. (1) can be solved within a window using simple combinations of
neighboring cells, the basis for terrain analysis in most commercial GIS, whether they
use grid-based methods or a mathematical representation of the DTM. To perform terrain
analysis across a variety of spatial scales, Wood (1996) solves this equation for an n by
n matrix with a local coordinate system (x, y, z) defined with the origin at the central
pixel (Figure 2). These matrix algebra methods are implemented in Landserf software
(Wood 2005) where the user may specify any odd number (n) for the size of the square
analysis window defining the portion of the raster DTM to be analyzed in relation to each
central pixel in turn. Calculation of the various terrain parameters (e.g., slope, aspect) is
achieved on the basis of a particular algorithm evaluated through solution of Eq. (1). Other
surface representations are possible (e.g., Horn 1981; Travis et al. 1975; Zevenbergen and
Thorne 1987). However, Evans’ (1980) method is one of the most precise methods at least
for first-order derivatives (Shary et al. 2002). While it might not be the best method for
all applications, it performs well in the presence of elevation errors (Albani et al. 2004;
Florinsky 1998).
Albani et al. (2004) present an analysis of the effect of scale on derived parameters for
terrestrial data. They note how small-scale analyses (e.g., using a standard 3 by 3 window)
are influenced by errors in the elevation surface or in the case of i nterpolated data prone to
exhibit properties of the interpolation process. They also note how at large window sizes
Downloaded By: [Wilson, Margaret] At: 09:27 8 May 2007
Multiscale Terrain Analysis of Multibeam Bathymetry Data 7
Figure 2. Raster grid, showing numbering system for cells in analysis window where Z is the value
of the raster. The central cell is the origin of the local coordinate system (x, y) and the positions
relative to this are denoted by subscripts. To simplify notation we use N = (n − 1)/2foranyn ×
n analysis window where n may be any odd integer smaller than the number of cells shortest side of
the raster. These are shown in full for a 3 × 3 window. Larger values of n mean that more cells (larger
area) are considered in the analysis.
the quadratic equation will be less likely to be a good descriptor of the terrain surface in
the neighborhood of the central cell suggesting a potential upper limit in the useful range
of analysis windows. Here we assess the choice of window sizes and their influence on
derived terrain variables visually and with reference to habitat suitability models.
Slope
Slope is thought to be an important factor in determining benthic habitat and colonization
in the deep sea at a variety of scales. Flat areas tend to exhibit different seabed facies and
support communities that are different from those on steeply sloping areas (Dartnell and
Gardner 2004; Iampietro et al. 2004; Lundblad et al. 2006). Slope may also contribute to
current flow amplification (Mohn and Beckmann 2002), which has consequences for the
