Author's personal copy

Quantitative characterisation of seafloor substrate andbedforms using advanced processing of multibeambackscatter—Application to Cook Strait, New Zealand

Geoffroy Lamarche a,n, Xavier Lurton b, Anne-Laure Verdier a, Jean-Marie Augustin b

a National Institute of Water and Atmospheric Research (NIWA) Ltd, Private Bag 14-901, Wellington 6241, New Zealandb Institut Franc-ais de Recherche pour l’Exploitation de la Mer (IFREMER), BP70, 29280 Plouzane, France

a r t i c l e i n f o

Article history:

Received 10 August 2009

Received in revised form

9 May 2010

Accepted 1 June 2010Available online 11 June 2010

Keywords:

Backscatter

Multibeam echo-sounder

Sediment waves

Habitat mapping

a b s t r a c t

A comprehensive EM300 multibeam echo-sounder dataset acquired from Cook Strait, New Zealand, is

used to develop a regional-scale objective characterisation of the seafloor. Sediment samples and high-

resolution seismic data are used for ground-truthing. SonarScopes software is used to process the data,

including signal corrections from sensor bias, specular reflection compensation and speckle noise

filtering aiming at attenuating the effects of recording equipment, seafloor topography, and water

column. The processing is completed by correlating a quantitative description (the Generic Seafloor

Acoustic Backscatter—GSAB model) with the backscatter data. The calibrated Backscattering Strength

(BS) is used to provide information on the physical characteristics of the seafloor. The imagery obtained

from the BS statistical compensation is used for qualitative interpretation only; it helps characterizing

sediment facies variations as well as geological and topographic features such as sediment waves and

erosional bedforms, otherwise not recognised with the same level of detail using conventional

surveying. The physical BS angular response is a good indicator of the sediment grain size and provides

a first-order interpretation of the substrate composition. BS angular response for eight reference areas

in the Narrows Basin are selected and parameterised using the GSAB model, and BS angular profiles for

gravelly, sandy, and muddy seafloors are used as references for inferring the grain size in the reference

areas. We propose to use the calibrated BS at 451 incidence angle (BS45) and the Specular-To-Oblique

Contrast (STOC) as main global descriptors of the seafloor type. These two parameters enable global

backscatter studies by opposition to compensated imagery whose intensity is not comparable from one

zone to the other. The results obtained highlight the interest of BS measurements for seafloor remote

sensing in a context of habitat-mapping applications.

& 2010 Elsevier Ltd. All rights reserved.

1. Introduction

The increasing need for the sustainable management of themarine environment and resources requires obtaining objectiveand quantitative descriptors of seafloor habitats from remote-sensed data. The potential for characterizing the nature of theseafloor using the acoustic reflectivity (or backscatter) collectedroutinely with multibeam echo-sounders (MBES) has beenrecognised for at least 20 years (e.g., de Moustier, 1986). However,quantitative analysis of backscatter data for seafloor characterisa-tion has only resurfaced recently for these types of applications(Anderson et al., 2008; Ehrhold et al., 2006; Le Gonidec et al.,2003; Medialdea et al., 2008; Wright and Heyman, 2008).

Quantitative models of the relationship between the acousticbackscatter and geological substrate properties fall into three

categories: (1) physical process modelling addressing links betweenphysical processes (sound propagation in inhomogeneous mediaand acoustic wave scattering by rough interfaces and volumeheterogeneity); (2) geoacoustical modelling relating the physicalparameters of the sediments with their geological description; and(3) phenomenological models, based on heuristical descriptors andparameters, aimed at a global model of intricate physical phenom-ena perceived as too complex either for justifying a rigorousphysical approach or for making possible inversion processes.

Physical and geoacoustical models are best suited to inter-preting fundamental phenomena observable on simplified or idealfeatures (Fonseca et al., 2002; Hughes Clarke et al., 1996; McReaet al., 1999; Mourad and Jackson, 1989; Novarini and Caruthers,1998). Phenomenological models have been developed for experi-mental data exploitation (Augustin et al., 1996; Le Chenadec et al.,2007; Lurton, 2003a). Accurate physical modelling providesinsight into the phenomena involved in sonar signal backscatter-ing—in the sense that it makes explicit the intrinsic potentialdependence of backscattered echoes upon geological substrate

Contents lists available at ScienceDirect

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

Continental Shelf Research

0278-4343/$ - see front matter & 2010 Elsevier Ltd. All rights reserved.

doi:10.1016/j.csr.2010.06.001

n Corresponding author. Tel.: +64 4 386 0465; fax: +64 4 386 21 53.

E-mail address: g.lamarche@niwa.co.nz (G. Lamarche).

Continental Shelf Research 31 (2011) S93–S109

Author's personal copy

composition (Mulhearn, 2000; Pratson and Edwards, 1996),However this approach is unavoidably limited to ideally simpli-fied configurations, making it possible to build a theoretical modelof acceptable complexity; hence it is generally inappropriate forproviding unsupervised (objective) inversion tools for accuratelyretrieving sediment characteristics directly from the backscatterdata, except in tightly constrained contexts where the modelapplicability is known a priori (Hughes Clarke et al., 1997; Jacksonet al., 1986). It therefore remains still very much questionable asto whether such numerical models may provide the expected linkbetween seafloor remote sensing and habitat mapping. A difficultand often under-estimated problem is that the accurate retrieval

of physical BS values has to go through a thorough compensationof the sensor characteristics (Ryan and Flood, 1996). Despiteprogress in the most recent MBES systems, the built-in compen-sation and basic processing do not allow an accurate enoughdetermination of absolute BS and incidence angles required forany quantitative analysis of the backscatter signal (Hellequinet al., 1997; Hughes Clarke et al., 1996; Le Gonidec et al., 2003).

This study aims at showing the potential provided by fullyprocessed backscatter data to characterize the seafloor in a varietyof geological and sedimentological environments, using an MBESdataset from Cook Strait, New Zealand (Fig. 1). The first part of thepaper undertakes a qualitative interpretation of the normalised

OCOC

PBPBPCPC

OROR

NarrowsNarrows

NCNCWCWCNBNB

Kapiti Is.

Mana Is

Marlbo

rough

S

ound

s

Cloudy Bay

Pallliser Bay

Wair

arap

a

CkCk

CkCkCampbell Bank

PaBPaB PaCPaC

Hikurangi Trough

Hikurangi Trough

0 20 4010 km

174°30'E 175°'E 175°30'E 176°E

41°S

41°30'S

42°S

Wellington

Wairarapa Shelf

Fig.

3

Fig.

6

Fig.13

Continental

Continental

ShelfShelf

40°S

175°E 180°

Alpine Fault

Hiku

rang

i Mar

gin

190°

PAC

AUS

Fig. 1. Bathymetry of Cook Strait. The intense colour DTM is gridded at 10 m, and compiled from MBES surveys. The pastel colours are from New Zealand regional

bathymetry. NB: Nicholson Bank; NC: Nicholson Canyon; WC: Wairarapa Canyon; Ck: Cook Strait Canyon, PB: Palliser Bank; PC: Palliser Canyon; OR: Opouawe Ridge;

OC: Opouawe Canyon; PaB; Pahaua Bank; PaC; Pahaua Canyon. Insert: The Pacific–Australia (PAC–AUS) plate boundary (teeth line) and principal faults.

G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109S94

Author's personal copy

backscatter imagery, with a particular focus on well-developedsediment waves, from which we discuss the potential benefitbrought by backscatter data to quantify the geometrical andgeological characteristics of sediment bedforms. We then apply animproved version of the heuristical model for angular BS proposedby Lurton (2003a) and Augustin and Lurton (2005) to variousresponses recorded in central Cook Strait. This approach enablesus to propose a quantitative interpretation of the backscatter overa variety of environments and bedform, and work toward theclassification and geological interpretation of BS using the modelparameters. We attempt to retrieve the seafloor sediment typesfrom the derived relationship between sediment mean grain sizeand BS angular response. We finally discuss the methodologicaldifficulties encountered and propose practical solutions for futureresearch aiming at a generalisation to this approach.

Our purpose is not to propose new models in backscatteringor geoacoustics. Rather, we seek to emphasise the value ofquantitative backscatter data analysis for investigating thesubmarine environment at a regional scale from shallow con-tinental shelf to deep ocean water depths, by using a pragmatic,rigorously controlled, approach to backscatter data analysis.

2. Cook Strait geological setting

The diverse and complex geomorphology of Cook Strait is theresult of dynamic climatic, tectonic, and oceanic forcings thatimpacted on the region since at least the last post-glacial period(Lewis et al., 1994). Cook Strait, between the North and SouthIslands of New Zealand (Fig. 1), is of strategic public and economicinterest for New Zealand containing electrical power cablesjoining the two islands; major tectonic faults (Van Dissen andBerryman, 1996); and economically important fish stocks andbiological resources.

A series of canyons has been carved into the continental shelfof Cook Strait (Mountjoy et al., 2009), with walls up to 1100 m inheight. Instability of the canyon walls is suggested by multiplesemi-circular scars in their upper section with debris accumulat-ing in the canyon floors. Beyond the continental slope, thecanyons flow into the Hikurangi Trough infilled with thickturbidite sequences and host to the Hikurangi channel (Lewiset al., 1994). The shelf edge is defined at �150 m water depthbeyond which the continental slope descends to more than2500 m water depth in the Hikurangi Trough.

During the Last Glacial Maximum, the continental shelfflanking Cook Strait was an emergent coastal plain (Carter,1992; Lewis et al., 1994). As the sea level rose from ca. 150 m,a wave-based erosional surface was formed around the strait.Most of this transgressive surface was subsequently blanketed bya wedge of post-glacial mud. However, strong waves and tidalforcing have left some areas of the shelf bare of post-glacialsedimentation so that the erosional surface outcrops at theseafloor (Lewis et al., 1994; Mountjoy et al., 2009).

Today powerful tidal currents, with peak flows reaching1.5 m/s in the narrowest passage of Cook Strait, known as theNarrows, coupled with strong sediment delivery from North andSouth island’s rivers (Hicks and Shankar, 2003), have produceddistinctive patterns of sedimentation. In particular, severalsediment wave fields were broadly recognised in the NarrowsBasin (Carter, 1992).

In addition to the ecological diversity to be expected from sucha complex environment, Cook Strait hosts a large diversity offaunal ecosystems associated with active and relict fluid seepssuch as carbonate chimneys and concretions, and reflectiveplumes generated (Barnes et al., 2010; Lewis et al., 1994, Lawet al., 2010). Seeps and cold-fluid vents are usually associated

with rich, distinctive, chemosynthetic faunas, which stress theimportance of marine environment management for the region.

3. Data and methods

3.1. Data acquisition

MBES data were collected during six oceanographic surveys ofR.V. Tangaroa between 2001 and 2005 in Cook Strait by theNational Institute of Water and Atmosphere (NIWA) (Table 1).The total surveyed area is ca. 8500 km2 in water depths greaterthan ca. 150 m (Fig. 1). Tangaroa has a hull-mounted KongsbergEM300 (32 kHz) MBES (Table 2) that fully compensates the vesselposition and motion (heave, pitch, roll, and yaw). Sound velocitymeasurements were performed at regular intervals to account forhydrology effects. Bathymetry data were routinely processedonboard using C&C Technologies HydroMap and ArcInfo software.A large geophysical and geological data dataset includingmultichannel and high resolution seismic reflection data,sediment cores, seafloor grab samples, and photographs areavailable for the region (Carter, 1992; Lewis et al., 1994) andprovided means to ground-truth the remote-sensed data analysis.

3.2. Backscatter data processing

The aim of backscattered signal processing is to provide aninterpretable image, and to estimate descriptive parametersrelating the BS with geological properties. The data is inherentlynoisy and requires processing to remove artefacts attributableto the recording equipment, seafloor topography, and watercolumn properties (Hellequin et al., 1997, 2003; Le Gonidecet al., 2003).

The echo level (EL) at the receiver’s output is primarilyassociated with the seafloor backscattering strength (BS(y)),where y is defined as the ray-path incidence angle onto theseafloor obtained from the refraction estimated during bathyme-try processing and from the locally computed digital terrainmodel (DTM). EL depends on transmission loss (TL), whichdepends on transmission range (R) and on water column

Table 1R.V. Tangaroa surveys in Cook Strait for EM300 acquisition (Figs. 2 and 5).

Survey Year Region

TAN0105 Apr 2001 Narrows Basin

TAN0204 Mar 2002 South Narrows Basin

TAN0211 Aug 2002 Campbell Plateau, Cook Strait Canyon

TAN0215 Oct 2002 Cook Strait canyons

TAN0309 Jun 2003 Northern Marlborough, Hikurangi Trough

TAN0510 Aug 2005 Wairarapa Shelf—Hikurangi Trough

Table 2Technical characteristics of NIWA’s EM300 multibeam system.

Central frequency 32 kHz

Number of beams 135

Beam width 11 n21 (along n across track)

Angular swath range 1401

Overall swath coverage 5.5 nwater depth

Water depth range 150–3500 m

Survey navigation systems HydroPro and HydroMap

Positioning and motion

compensation

TSS DMS TSS POS-MV 320 system

Sound velocity calibration Expendable bathythermographs (XBTs)

Digibar

DB-1100 AML SVplus sound velocity

profiler

G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 S95

Author's personal copy

absorption coefficient (a); the sonar characteristics (transmittedsource level SL, signal duration T, transmission and receptiondirectivity patterns DT and DR; receiver gain GR; hydrophonesensitivity SH); and the insonified area A(R, y, T) (Lurton, 2003a).SL, DT, SH, DR, and GR are hardware dependent. Hydrologyconditions influence absorption (a) and incidence angle (y),whereas the geometrical configuration (water depth, topography)controls both the range (R) and (y). These parameters are gatheredinto the conventional sonar equation (Lurton, 2003a, 2003b):

EL¼ SLþDT�2:TLþ10log AþBSþSHþDRþGR ð1Þ

Practically, there are two issues in processing BS data forenhanced application for geological interpretation: sonar calibra-

tion, which controls the physical signal levels provided by thesensor, and recorded data correction required for qualitative andquantitative backscatter estimation and exploitation.

3.2.1. Sonar calibration

Two levels of amplitude calibration may be applied on anysonar data:

(1) An elementary calibration to ensure that the sonar parameters(SL, T, DT, DR, GR) are correctly accounted for, in the recordedamplitude levels; this guarantees that the data are repeatable,and do not depend on the system settings; ideally this is donethrough built-in compensations applied in the receiver.

(2) A full calibration procedure to provide exact referenced valuesfor the transmitted level and the receiving sensitivity. This

requires factory calibration of transducers and electronics, andimplementation of control tools and procedures to ensure thatthese parameters are checkable and compensable onboard.

Despite significant improvements, the built-in correctionsapplied by MBES (Kongsberg, 1997) still remain imperfect andmay result in artefacts in the backscatter data, precludingaccurate qualitative and quantitative data analysis. To improvedata quality, built-in compensations applied at data acquisitionmust therefore be removed and replaced by more accurate ones.Unfortunately, some of the system characteristics are not directlyaccessible and have to be deduced from the recorded data. Forinstance the echo level modulation caused by the array directivitypatterns is corrected by fitting a modelled pattern on datarecorded over several assumed-homogeneous seafloor areas(Fig. 2) (Lurton et al., 1997). This technique is similar to thatapplied to satellite-borne radar reflectivity data for a posteriori

compensation of directivity patterns (Maıtre, 2001).The amplitude calibration used in fishery acoustics, measuring

the response of a reference spherical target (e.g. Foote, 1980), canbe applied to high-frequency MBES in a controlled environment(Foote et al., 2005); however it is not practical for low-frequencyMBES featuring long arrays already mounted on ship’s hull. In thisstudy, we retained the manufacturer values for the transmittedlevels and receiver sensitivity; the risks of this simplificationbeing either an inaccurate qualification of the hardware by themanufacturer or a possible drift of the system with time.

0-36

-20

60-60Transmission angle

BS

(dB

)

Transmission angle

Gai

n (d

B)

0 80-80

Sonarscope calibration gain functionManufacturer calibration gain function

-4

0

Fig. 2. Calibration of the backscatter data for a 3-sector configuration sounder, undertaken using gain functions provided by the manufacturer (dotted line in A) and re-

calculated using the SonarScopes software (bold line in A). (B) Backscatter angular profiles generated with manufacturer (dotted line) and re-calculated (bold line) gain

functions. Datagrams with manufacturer (C) and re-calculated (D) calibrations. Arrows indicate sector transitions.

G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109S96

Author's personal copy

3.2.2. Data correction

For qualitative analysis of the sonar images, an echo-levelcompensation needs to be applied on the data. This process removesartefacts inherent in the sensor and corrects the signal angularresponse to produce a normalised image interpretable by end-users(Fig. 3). Largest errors are produced by the specular echo at nadir andthe reflectivity fall-off at grazing angles. Echo-level compensationrequires that the normalised level is taken at a significant andrelevant value, in particular respecting the contrast between the BS

levels associated with different sediment types. Finally, an additionalabsolute calibration of the BS data must be carried out for anysatisfactory quantitative analysis to be undertaken.

3.2.3. Methodology applied to the Cook Strait data

The Cook Strait dataset was processed using the SonarScopes

software developed by IFREMER, France (Augustin and Lurton,

2005) (http://www.ifremer.fr/fleet/acous_sism/sonarscope/index.html). In SonarScopes the physical parameters (bathymetry,backscatter level, incidence angle, motion, etc.) are displayed asdifferent layers available for simultaneous or combined proces-sing using either conventional or sonar-specific signal and imageprocessing modules. Advanced processing techniques includesignal calibration and compensation, speckle noise filtering,texture analysis, and image segmentation.

The compensation process was undertaken in three steps:(1) correction for transmission losses associated with absorptiongiven by changing hydrological conditions; (2) estimation of thearray directivity patterns, from which specific gains were appliedfor each angular sector of the sounder; and (3) correction for theinstantaneous signal footprint area, as a function of incident angle,directivity pattern aperture, and pulse duration.

Attenuation of the strong, sometimes obliterating, echoassociated with the specular reflection in backscatter images isa fundamental step prior to any qualitative analysis (Fig. 3). In thepresent case, the specular signal was filtered out by statisticalcompensation (Augustin and Lurton, 2005) consisting in subtract-ing from the physical angle-dependent sonar image (Fig. 3C) anaverage BS(y), hence flattening the angular dependence whilepreserving the average BS value. The resulting image is more‘‘readable’’ since the contrast is enhanced because the correctedBS values are close to the mean value (Fig. 3B). However, thiscompensated image has lost most of the BS level information, sinceit was normalised to an arbitrarily chosen reference level (in thisstudy we used the uncompensated value at 451), and it is henceexploitable mainly in terms of relative intensity.

It is emphasised that a fundamental distinction has to be madebetween the compensated backscatter, normalised at a conven-tional reference level, defined for imagery purposes and usable forqualitative analysis, and the physically correct calibrated BS usedfor quantitative analysis. The two types of data are expressed indB per m2 (simplified as dB) and both are incorporated in thegeological-oriented analysis of reflectivity data.

3.3. Backscatter angular profile modelling

We used the Generic Seafloor Acoustic Backscatter (GSAB)model to represent the BS angular response. The GSAB was firstlydeveloped within the framework of directivity pattern compensa-tions for MBES data (Hellequin et al., 1997) and used since invarious contexts (Augustin and Lurton, 2005; Guillon and Lurton,2001; Le Chenadec and Boucher, 2005; Le Chenadec et al., 2007;Lurton, 2003b). However, it has never been systematically appliedto a large regional dataset. The GSAB model represents the BS

angular response as a combination of a Gaussian law for thespecular angles and Lambert-like law for grazing angles (Fig. 4).In its original form, it uses four descriptors (A–D), with no explicitrelationship with the geological or acoustical attributes usedclassically in these disciplines (e.g., Jackson et al., 1986). Themodel is given by

BSðyÞ ¼ 10log½A expð�y2=2B2ÞþC cosD y� ð2Þ

where

� A quantifies the specular maximum amplitude. In the tangent-plane approach (Brekhovskikh and Lysanov, 1982). A is relatedto the coherent reflection coefficient at the water–seabedinterface, and is therefore high for smooth sediment interfaces(at the acoustic wavelength’s scale), and for strong water–sediment impedance contrasts (Lurton 2003a).� B quantifies the angular extent of the specular regime. In the

tangent-plane model, it represents the facet slope variance and

174°30’ 174°36’

41°10'

41°14'

41°10'

41°14'

0 5km-2 -40dB

-15 -40dB

Fig. 3. (A) Calibrated reflectivity in the Narrows Basin. The backscattering strength

(BS) includes the full extent of the specular echo and the lateral decrease with

angle has been compensated by Lambert’s law. (B) Fully processed backscatter

reflectivity. The compensation of the backscatter was undertaken by applying a

gain function as shown in (C). An 11�11 pixel speckle filter is also applied to the

data. The compensated data have lost their absolute reference level (see text).

(C) The statistical compensation is performed by subtracting a bias value (thin

grey) to the measured calibrated backscatter (thick grey) and provides the

compensated angular response.

G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 S97

Author's personal copy

is therefore an interface roughness descriptor. Note that, in thefollowing, angular extent denotes the half-width of the specularpeak actually measured by MBES working over both sides ofnadir.� C quantifies the average BS level at oblique incidence. It is the

offset associated with Lambert’s law describing BS at inter-mediate angles for rough interfaces, and includes the con-tribution of the volume inhomogeneity backscatter. C

increases with frequency, seafloor roughness and impedance,and heterogeneities present inside the sediment volume.Typically 10log C ranges from �20 to �30 dB, but valuesbetween �15 and �40 dB are commonly observed.� D is the backscatter angular decrement, commanding the fall-

off at grazing angles. It is high for soft and smooth sedimentinterfaces. According to the laws of Lommel-Silliger andLambert, D is equal to 1 and 2, respectively (Lurton, 2003a).

While this four-parameter model (A, B, C, D) is adequate tomodel the backscatter angular behaviour in many configurations,in some cases the model may require an intermediate term toaccount for the smooth transition between the specular andlateral modes. This transitory component is modelled with asecond Gaussian law; Eq. (2) becomes

BSðyÞ ¼ 10log½A expð�y2=2B2ÞþC cosD yþE expð�y2=2F2Þ� ð3Þ

where E is the transitory maximum level (dB) and F its angularhalf-extent (in deg) (Fig. 4). There has been no study on therelationship of E and F to geological and physical seafloorproperties. Their secondary role will be addressed in thediscussion. This improved form (3) of GSAB will be used in thefollowing.

Since GSAB aims at providing angle dependence descriptors ofa swath sonar dataset, the coefficients are only valid for a given

sonar frequency. Today’s current MBES uses one of seven nominalfrequencies (12; 32; 70; 100; 200; 300; and 450 kHz) eachcorresponding to different practical applications.

3.4. Sample analysis

The sample database consists of 260 seafloor samples collectedsince the late 1950s by the New Zealand Oceanographic Institute(NZOI) and subsequently by NIWA, using gravity corer, pistoncorer, and seafloor grab. Sedimentological analysis consisted ofgrain size analysis and classification as a function of thepercentage of mud, gravel, and sand undertaken using standardprocedure for all samples. Mean grain sizes were obtained using alaser granulometer and the Gradistat programme (Blott and Pye,2000) for statistical analysis of unconsolidated sediments. Thesamples are classified in three groups based on log-normaldistribution from their j (phi) values, where j¼� log2 d, whered is the grain diameter in millimetres (Table 3).

4. Results

4.1. Bathymetry and backscatter imagery

Although the morphological complexity of Cook Strait is wellexpressed in the bathymetry (Fig. 1), the compensated backscatterimagery provides complementary detailed information aboutseafloor features and substrates (Fig. 5). At the scale of the CookStrait region, we define three basic imagery facies: (1) ahomogeneous, weak-to-moderate reflectivity (dark grey) facies,(2) a homogeneous, highly reflective (light grey) facies, and (3) ahighly heterogeneous reflective facies. Using these three faciesas end-members of a continuum we define four regions: theNarrows, the continental shelf, the continental slope and canyons,and the Hikurangi Trough.

The Narrows region corresponds to the 350 m-deep, N–Strending oval Narrows Basin and its steep, linear, eastern andwestern walls. The continental shelf on either sides of the basin isless than 6 km wide. It is smooth and homogeneous on the westand rough to the east, along the Wellington Peninsula. In terms ofimagery, the Narrows Basin divides into a northern weak-to-moderate (dark) and a southern moderate-to-strong (light) facies.The floor of the Narrows Basin has a weak backscatter facies,whereas it is strong on its flanks. A region of strong to very strongimagery extends well southward of the Narrows to the head of theCook Strait canyon (Fig. 6). The BS dynamic over the NarrowsBasin from Oceanographic voyage Tan0105 alone is 18 dB abovethe 3% quantile (Table 4). The floor and eastern wall of theNarrows Basin is dominated by fields of sediment waves; theseare characterised by a distinct E–W orientation, i.e. perpendicularto the long axis of the Narrows Basin, and asymmetric shapeswith steeper NE flanks. The largest sediment waves are up to 10 min height with a ca. 250 m wavelength, and the smallest wavesare 2–4 m high with a wavelength less than 100 m (Fig. 7).

0 20-20 40-40

-40

-30

-20

BS

(dB

)

-10

60-60-80 80

θ : Incidence Angle (degrees)

+++

++++++

++ +

++

+

++

++

+

++++++++++++++++++++++++++

+

++++++++++++++++

A

EB

F

C

cosDθ

Transitory

Lambert

BS(θ) = 10 log[ A.exp(-θ²/2B²) + C. cosDθ + E.exp(-θ²/2F²)].

Specular

Fig. 4. Modelling of the BS angular response using the Generic Seafloor Acoustic

Backscatter (GSAB) model. The parameterised BS angular function model (thick

grey) fits the averaged measured BS values for a selected number of pings in a

given area (crosses) and is the sum of two Gaussian functions for the specular

reflection (short black dash) and transitory level (long dark grey dash) and

a Lambert law (dark grey) for high incidence angles. The three curves are

defined using the 6 parameters A–F in the equation BS(y)¼10log[A exp(�y2/2B2)+

C cosD y+E exp(�y2/2F2)].

Table 3Simplified grain size classification from Blott and Pye (2000).

j* d (mm) Class

o�1.0 42.00 Gravel

�1.0–4.0 2.00–0.063 Sand

44 o0.063 Mud

n j¼� log 2d, d—grain diameter (mm).

G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109S98

Author's personal copy

The backscatter over the sediment wave field is characteristicallyvery weak to weak (dark).

The continental shelf is poorly covered by EM300 since thisequipment is not well-adapted for water depths shallower than�100 m. At this relatively coarse regional scale, the shelf appearssmooth and homogenous. The shelf break is well marked at�160 m below the sea level, except in areas where landslidinghas prograded and eroded the top of the continental slope(Mountjoy et al., 2009). The continental shelf shows a weak-to-moderate (medium grey) and homogeneous backscatter. The BS

dynamic over the continental shelf is ca. 10 dB above the 3%quantile.

The continental slope, in the east and SE of the study area has avery rough topography associated with canyon flanks and steepgullies. The steep gradients, blocky facies on the canyons floors,and multiple circular scarps at the top of the continental result ina highly heterogeneous imagery facies (Fig. 5), with a BS dynamicof 19 dB above the 3% quantile. The canyon floors along thecontinental slope show a strong to very strong reflectivity

contrasting with the low values on the flanks. This is bestexemplified over the Opouawe and Pahaua canyons.

The Hikurangi Trough has a smooth topography associated witha moderate (light grey), very homogeneous imagery facies. The16 dB dynamic of the BS above the 3% quantile is likely to be theresult of the particular contribution of the specular reflection,which remains high in all cases.

4.2. Characterisation of sediment bedforms

The backscatter dependence with the microtopography pro-vides an excellent means to identify bedforms. In the NarrowsBasin, the multibeam bathymetry provides clear information onthe geometry of sediment waves (amplitude, orientation, wave-length, asymmetry, and along-crest length, Figs. 7 and 8). Lowbackscatter level is consistently observed on the sediment wavecrests, and strong levels in the troughs: a typical pattern alreadynoted by e.g., Fonseca et al. (2009) and Goff et al. (2000). Over the

Tan0211

Tan0211

Tan0215

Tan0309

Tan0510

Tan0204

Tan0309

41°30'S

42°S

175°E 175°30’E 176°E

41°S

Wellington

0 km 20

OCOC

PBPB

63

NCNC

1313PaBPaB

Hikurangi Trough

dB

-18Strong

Moderate

Weak -36

A

B

0 km 2

PaB

Tan0

105

Fig. 5. Fully processed (calibrated and compensated) backscatter imagery of Cook Strait (A). Note that strong intensity shows in white. Thick dash lines indicate the

boundaries of individual datasets for which the survey number is indicated in bold. NC: Nicholson Canyon; OC: Opouawe Canyon; PB: Palliser Bank; PaB: Pahaua Bank.

See Fig. 1 for the full caption. (B) 3-D oblique view of the sediment waves in the Narrows Basin; looking to the NE, (PaB): calibrated and compensated backscatter

imagery of carbonated concretions on the Pahaua Bank. Main figure and insets generated using the Sonarscopes software. Frames indicate the locations of subsequent

figures.

G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 S99

Author's personal copy

Narrows Basin the BS varies with a very strong dynamic of 13 dB,for bathymetric amplitudes ranging from 3 to 5 m. The averagecrest-to-trough BS dynamics is �6 dB across any of the sedimentwaves, i.e. well within the equipment 1 dB resolution (Kongsberg,1997).

The variation in reflectivity across sediment waves is notrelated to an effect of the incidence angle. In some areas withinthe Narrows Basin the BS modulation contrast is still high withcrest-to-trough BS dynamics from 3 to 7 dB producing sedimentwave patterns on the BS imagery (Fig. 8A). However, thesesediment waves cannot be detected in the bathymetry (Fig. 8B).

The incidence angles across the sediment waves (Fig. 8C) rangefrom 521 to 571, with a typical amplitude of 21. Assuming aLambert dependence, a 21 incidence angle difference wouldgenerate a BS contrast of about 20log(cos 561/cos 541)¼0.4 dB,which is one order of magnitude smaller than the observed BS

variations associated with the sediment waves. Furthermore, theaverage angular response profile (BS(y)) over the sediment wavefield in the Narrows Basin (Fig. 9) shows a level of �5 dB at nadir,decreasing rapidly to �25 dB at 301 incidence angle, but with apoorly defined specular shape, i.e. with a progressive transitionfrom the specular to the transitory regime. At higher incidence

12

3

5

4

7

68

41°15'S

41°10'S

0 2 4km

200

250

300

350

300

150

350

250

30015

0

150

200

100

350

200

300

250

174°35'E174°30'E174°25'E174°20'E

41°15'S

41°10'S

Contours interval 250 m70

390

metres

12

3

5

4

7

6

8

-30

dB

-20BS45

Fig. 6. Seafloor morphology (top) and fully processed backscatter imagery (bottom) in the Narrows Basin. The backscatter imagery is compensated using the BS intensity at

451 incidence angle (BS45). Eight reference areas (numbered) were identified from the bathymetry and backscatter homogeneity for generating average BS angular profiles

(Fig. 11, Table 5).

G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109S100

Author's personal copy

angle, BS decreases slowly to �33 dB at 501. The BS valuesorthogonal to the sediment wave show fluctuations up to 10 dBbetween troughs and crests at all angles. The two observationssuggest that the reflectivity contrasts recorded across the sedimentwaves are controlled by sediment type variations rather than byincidence angle and that, in some situations, analysis of the BS maybe more efficient for their detection than that of bathymetry.Although some angular dependence is certainly noticeable in theBS, its effect is clearly superseded by the sediment type.Consequently bedforms can be detected from backscatter dataeven when their topographic expression is not measurable.

4.3. Backscattering strength angular profiles

The GSAB model was used to generate sets of descriptor values(Section 3.3) at key locations in the study area, to explore the

relationship between BS and incidence angle on the seafloor.To reduce inter-survey calibration errors, data was only usedfrom the Tan0105 voyage (Fig. 6). Tan0105 covers the entireNarrows Basin incorporating a variety of geological featuresand water depths. The GSAB parameters were generated foreight reference areas selected from the bathymetry and the BS

maps (Fig. 6). The areas were chosen as homogeneous interms of both morphology and average BS (Fig. 10). Eacharea typically includes more than 100 complete pings, thusspanning the entire BS angular response and providing a signi-ficant average.

The average BS(y) generated for the eight reference areas havedistinctly different shapes (Fig. 11A) generated by the uniqueseafloor characteristics (interface roughness, impedance, andvolume scattering) within each region. The GSAB parameters(Table 5) provide a means for further classification of seafloortypes from the measured BS.

Table 4Minimum and maximum processed backscatter values for Tan0105 survey.

Backscatter processing Compensatedb (dB) Calibratedc (dB re. 1 m2)

Incidence anglese�701 to 701 �701 to 701 (all swath) �101 to 101 (specular) 431 to 471 (BS45)f 681 to721 (grazing)

Alla 3%d Alla 3%d Alla 3%d Alla 3%d Alla 3%d

Min �65 �34 �55 �38 �47 �23 �47 �36 �54 �42

Max 12 �16 18 �9 16 �1 �11 �18 �9 �22

Amplitude 77 18 73 29 63 22 46 18 45 20

Mean �2575 �2577 �1276 �2675 �3176

a All angle of incidence (�701 to 701).b Statistical compensation undertaken by applying a gain function (Fig. 3C) 71 dB absolute uncertainty.c Calibration undertaken by fitting a modelled pattern of array directivity on data recorded over homogeneous seafloor areas (see text) 71 dB absolute uncertainty.d Minimum and maximum values inside the 3% quantile.e Positive incidence angles only as angular profiles are assumed symmetrical.f BS45 calculated between 431 and 471 incidence angles on both starboard and port sides.

Deg

Distance (m)0 200 400 600 800 1000

45

55

Incidence angle

db -35

-30

Processed Reflectivity

vvb

Calibrated Reflectivity

-40

-30

m

Bathymetry

-315

-305

Fig. 7. Bathymetric, calibrated BS, compensated backscatter, and incidence angle profiles (from top to bottom) across a sediment wave field in central Cook Strait (location

on small backscatter map). Profiles generated using a 5-m grid.

G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 S101

Author's personal copy

On the profiles generated from the measured BS (Fig. 11A,Table 4), the specular levels range from �23 to �1 dB. This islarger than the GSAB parameters, where the averaged specularlevels (A in Eq. (3)) range from �13 to �4 dB (Fig. 11B, Table 5).The oblique BS levels, at 451 incidence angle, range from �36 to

�18 dB. The relative specular amplitudes, represented by thecontrast between the specular amplitude and the lateral value at451, range from 4.5 dB (Area 4) to 21 dB (Area 3). The 4.5 dBrelative specular amplitude for Area 4 denotes a weak, or lack of,coherent specular reflection, suggestive of a hard and rough

400 600 800 1000 1200

56

54

52

Distance (km)

deg

Incidence angle (deg)

280

270

m

30

20dB

Reflectivity (dB)

Bathymetry (m)

30

20

Fig. 8. Compensated backscatter (A), bathymetric (B), and incidence angles (C) profiles across a sediment wave field in central Cook Strait. The sediment waves are only

detected using reflectivity, whereas the bathymetry does not show any morphological sediment wave pattern. The incidence angles vary little, which demonstrates that the

variation in BS is likely due to the geology and microtopography.

0 200 40038

34

30

pingpingBS

0 200 40040

36

32

0 200 40030

25

200 200 40015

10

5

0

80 60 40 20 0 20 40 60 80

40

35

30

25

20

10

5

Incidence Angle (I.A.)

Bac

ksca

tter S

treng

th (B

S) BS

ping

I.A.=45o

BS

ping

I.A.=25o

BS

ping

I.A.=0o

I.A.=60o

Fig. 9. Average angular backscatter strength (BS) profiles (thick grey) over a sand wave field. The standard deviation (dotted lines) indicates the backscatter distribution

variation with the incidence angle. Insets show the BS profiles at four beam incident angles (01, 251, 451, 601) over �600 pings.

G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109S102

Author's personal copy

seafloor (Augustin et al., 1997; Hughes Clarke et al., 1997; LeChenadec et al., 2007). The angular extent of the specularreflection is relatively stable around 51, suggesting medium tohighly scattering interfaces associated with inhomogeneousmaterial. No very narrow specular (o21) was found comparableto those observed on soft, homogeneous seafloor in the deep-seabasins (Augustin et al., 1997). The modelled angular extents of thespecular reflection are in the range 3.6–9.51. In any case, theangular full width of the specular peak cannot be smaller than thesounder beam width (21 for the EM300), which gives the anglemeasurement resolution.

Over the eight reference areas, the modelled Lambert refer-ences (C in Eq. (3)) range from �31 to �17 dB (Table 5). This14 dB range indicates important variations in seafloor roughnessor sediment volume scattering across the reference areas. Thetransitory levels for the modelled BS (E) range from �21 to�15 dB, with an average angular extent (F) ranging from 10.11 to16.61. These values differ from both the angular extent of themodelled specular reflection (3.6–9.51) and the angular width of aLambert law (451), confirming the relevance of using a transitorylevel for accurately modelling the BS angular response.

The increase in BS with grain size noted in previous studies(Briggs et al., 2002; Goff et al., 2000; Jackson and Briggs, 1992) isonly perceptible in the Cook Strait data (Tables 5 and 6). In orderto further assess the relationship between BS angular profiles andgrain size, we selected a small number of samples with average jparameters characteristic of mud, sand, and gravel (Tables 3and 6). We purposefully used an over-simplified model withthree end-member characteristics corresponding to soft, medium,and rough seafloor. The number of classes is not specificallyimportant as it can be statistically estimated at a later stagefor segmentation purposes (Lucieer and Lucieer, 2009). TheBS angular response models were established for the threesediment types following the methodology described inSection 3.3, for a small area assumed to be representative of atypical grain size as they were selected from homogeneousimagery regions around the samples (Fig. 12). The areas coverbetween 4 and 20 km2 around the sample locations, which is largecompared with the sample size, but are required so that a wideangular range (ideally 0–701) is available to model the full BS

angular response. This is mitigated by selecting several areas with

characteristic samples for gravelly and sandy seafloor (Table 6).The areas for muddy seafloor only include one sample.

The BS(y) profiles for gravelly, sandy, and muddy seafloorsshow distinctive shapes (Fig. 12) and A–F parameters (Table 5).The muddy seafloor gives the weakest reflectivity on the wholeangular range, with a modelled specular level A¼�14 dB(Table 6). Gravelly and sandy seafloors give similar specularlevels of A¼�9 dB. All the specular angular extents have similarvalues (B¼5–61), the narrowest being for the gravelly seafloor.The Lambert references (C) are �18, �24, and �29 dB for gravel,sand, and mud, respectively. The inter-class spacing (typically6 dB) is well above the separability given by the relative 1 dBmeasurement capability of the EM300. Likewise, the modelledtransitory levels (E) have distinctive reference level values forgravel (�18 dB), sand (�21 dB), and mud (�25 dB), but theangular extents have comparable values (F¼161 and 181). Theangular extents of the specular and transitory levels do not seemto help in deciphering between different grain sizes.

5. Discussion

5.1. Qualitative interpretation

The backscatter imagery provides particular fine details thatcannot be obtained from the bathymetry alone; the imagery canprovide additional information on the origin, nature, andstructure of landforms. The combined use of backscatter andmicrotopography processing enhances the interpretation of finescale geological and topographical features visible in EM300multibeam data, such as post-glacial scouring, erosional land-forms and carbonated platform concretions (175137’E–41138’S,PaB in Fig. 5), and on canyons floor (OC in Fig. 5). The strongreflectivity on the floor of the Opouawe Canyon suggests graveldeposits, indicating active sediment delivery. Relicts of the lastglacial transgression are identified on the outer shelf with ripplesmarks, gravel lags, and sediment waves observed in the CampbellBank (Fig. 13). This latter example shows that the backscatterimagery provides even finer details than from themicrotopography. An E–W trending, �50 m-high scarp is wellimaged in the bathymetry and corresponds to strong backscatteron the scarp, suggesting a rough seafloor, most likely associatedwith gravelly material. Immediately north of the fault scarp, N-trending black and white lineaments emphasize seabed featuresrelated with contrasting seafloor deposits, but with notopographic expression. By contrast, small depressions visible inthe bathymetric image in the NW quadrant of Fig. 13 do notexhibit any contrasting backscatter change, suggesting a gentlechange in seafloor composition. A similar situation can be seen onPahaua Bank, where a strongly reflective patch at the shelf break(175137’E 41138’S, Fig. 5) shows fine details of morphological andgeological origin associated with carbonated concretionrecognised by Law et al. (2010).

5.2. Use of GSAB for seafloor characterisation

While theoretical models of BS angular response may proveuseful for characterising the seafloor locally, their benefits forquantitative analysis of regional datasets are limited. There aretwo main reasons for this situation.

The first reason is that, by definition, accurate physical modelsare formulated to account for extremely complex processes (e.g.,Guillon and Lurton, 2001) requiring a high number of inputparameters. This makes inversion results possibly ambiguous asdifferent input parameters may provide similar effects. In the

Area 2

Area 5

0 20 40 60Incidence Angle

-30

-20

-10

0

BS

(dB

)

Fig. 10. Average BS angular responses for reference areas 2 and 5 (Fig. 6) with

error bars indicating standard deviation computed for 51 angle bins.

G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 S103

Author's personal copy

worst case, the inversion may be technically impossible. Moreover,the higher the complexity, the greater the specialisation of a modelfor a given seafloor type. The second reason is that the measuredBS may be dominated by non-anticipated phenomena such assediment layering, accidental highly random scatterers, short-scale heterogeneities in the sediment bulk properties or roughness(e.g., Fonseca et al., 2002; Novarini and Caruthers, 1998), or short-

term changes of sediment facies. This makes irrelevant the a priori

application of a single theoretical model valid only locally on arestrained range of seafloor characteristics; a single physicalmodelling approach cannot be valid on both, e.g., fine-grainsediments, detritic material, rock structures, and vegetal cover

These shortcomings are exacerbated at a regional scale,making it improbable for one dataset to be applicable as a

-45

-40

-35

-30

-20

-25

-15

-10

-5

BS

(d

B)

-80 -60 -40 -20 0 4020 60 80

Incidence Angle

Models

0

-40

-35

-30

-25

-20

-15

-10

-5

BS

(d

B)

-80 -60 -40 -20 20 40 60 80-45

CalibratedReflectivity

1 North East2 North3 Sandwave field4 West5 South6 Central South7 West (Inhomogeneous)8 South west

Fig. 11. BS angular profiles of the calibrated backscatter (A) and generated using the Generic Seafloor Acoustic Backscatter (GSAB) model (B) as developed in this study

(see text) for the eight reference areas in central Cook Strait (Fig. 6). Parameters A–F generated by the models in Table 5.

G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109S104

Author's personal copy

seafloor model over a wide area. Therefore the correlation of a setof physical descriptor values with a regional dataset becomes lessvaluable. This line of reasoning argues in favour of using a moreempirical modelling approach focused on a simplified descriptionof the seafloor characteristics, restricted to the identification ofmajor trends but able to cope with strong local variations. In usingGSAB, we are attempting to describe physical observationsthrough a restricted set of parameters. The approach is suffi-ciently versatile to accommodate the diversity of geologicalenvironments, while remaining accurate enough for a good-quality correlation to experimental data. Our purpose is not toretrieve quantitative values of seafloor physical or geologicalproperties, but rather to determine descriptor values making itpossible to classify the seafloor inside a limited set of classes thatmay be ground-truthed. Obviously, the accuracy of thesedescriptors cannot compete with the refinement of the inputparameters used in some geoacoustic models (Buchanan, 2005;Buckingham, 2000; Fonseca et al., 2009; Jackson et al., 1986; Stoll,1985). Rather, the approach is a pragmatic solution to applyingthe BS signal to enhance geological interpretation.

The gravelly, sandy, and muddy GSAB patterns can be used asbasic references to provide a first-order interpretation on theaverage grain size across a large seafloor area. We apply thisapproach to the Narrows Basin using the GSAB patterns generatedin the eight reference areas (Figs. 11 and 14).

At low-incidence angles (0–151) a close correspondence (e.g. inthe lobe angular width) between the mud, sand, and gravel GSAB

profiles and those of the eight areas is difficult to assess becauseof the closeness of all profiles.

The correspondence is much clearer around 451 incidenceangle, e.g., for Area 8 where the BS angular profile compares wellwith the gravel BS profile (Fig. 14). This is shown numerically bythe parameters A¼�11 dB and C¼�18 dB modelled for Area 8comparing well with A¼�9 dB and C¼�18 dB for gravel(Tables 5 and 6). These observations suggest that the continentalshelf on the west side of the Narrows Basin likely consists ofcoarse material. The strongest BS facies in the Narrows Basin isobserved in the south basin floor (Area 5) with a modelledspecular A¼�4 dB and Lambert reference C¼�17 dB. Thespecular is higher than for the gravel profile (A¼�9 dB),suggesting very coarse material in Area 5. A smooth interfacebetween water and soft sediment would have also generated astrong specular, but with low B and C values, which is notobserved in Area 5.

Overall the GSAB profiles for areas 4, 5, and 8 show the highestaverage BS around the 451 incidence angle, and hence can beinterpreted as representing areas with the coarser material. Areas6 and 7, along the western flank of the Narrows Basin, show verysimilar characteristics (A¼�7 and �6 dB, C¼�20 and �19 dB)and are intermediate between the sand and gravel profiles atintermediate angles. The specular levels for the gravel and sandare identical (A¼�9 dB) and hence not discriminating. However,the three BS level parameters (A, C, and E) are higher for Area 6than for Area 7, suggesting coarser deposits in the south than inthe central basin wall. Interestingly, Area 3 is largely dominatedby sediment waves and presents the weakest BS response, withLambert reference C¼�31 dB and specular A¼�10 dB. In details,the BS varies across dunes and sediment waves, whereas on aregional scale the homogenised reflectivity suggests on theaverage fine sandy–mud material on the floor of the NarrowsBasin.

The angular extent of the specular reflection (B) is an indicatorof the interface roughness. The maximum and minimum valuesfor B are observed for areas 5 (9.51) and 1 (3.61), respectively(Table 5). These are consistent with coarse to very coarse materialincluding pebbles and boulders retrieved from Area 5, whereasT

ab

le5

Ge

olo

gic

al,

ge

om

orp

ho

log

ica

la

nd

ba

cksc

att

er

cha

ract

eri

stic

so

fth

ee

igh

tte

sta

rea

sid

en

tifi

ed

inth

eN

arr

ow

sB

asi

n.

Are

a1

23

45

67

8

Mo

rph

olo

gy

SE

faci

ng

slo

pe

,

smo

oth

Fla

t,sh

all

ow

,

smo

oth

Se

dim

en

tw

av

e

fie

ld,

ba

sin

flo

or

Ve

rysh

all

ow

rou

gh

Sm

oo

thd

ee

p

ba

sin

flo

or

Ba

sin

wa

llso

uth

Ba

sin

wa

llce

ntr

al

Sm

oo

thsh

all

ow

she

lf

Se

dim

en

tch

ara

cte

raM

ud

dy

gra

ve

l,

me

d-fi

ne

san

d,

ve

ryfi

ne

san

d

Firm

gre

ym

ud

,

me

d-fi

ne

san

d,

shin

gle

Mu

d,

me

d-fi

ne

san

d

Sa

nd

yto

gra

ve

lly

mu

d,

som

e

pe

bb

les

Co

ars

eto

ve

ry

coa

rse

incl

ud

ing

pe

bb

les

an

d

bo

uld

ers

Gra

ve

l,v

ery

coa

rse

san

d

con

cre

tio

n,

mu

dd

ysa

nd

Gra

ve

l,sa

nd

,

pe

bb

les,

she

lly

mu

dd

ysa

nd

Mu

dd

ysa

nd

,

ve

ryfi

ne

gra

ve

l

No

.o

fsa

mp

les

43

26

15

45

43

24

11

2

Ba

cksc

att

er

ima

ge

ryb

Mo

de

rate

lyw

ea

kW

ea

kV

ery

we

ak

Str

on

gV

ery

stro

ng

Mo

de

rate

lyw

ea

kM

od

era

tely

we

ak

tow

ea

k

Mo

de

rate

ly

stro

ng

Sp

ecu

lar

lev

el

A�

8.7

�8

.5�

9.8

�1

2.9

�3

.9�

7.2

�5

.5�

11

.3

An

gu

lar

ex

ten

tB

3.61

4.21

5.31

4.81

9.51

4.51

3.91

3.61

Lam

be

rtre

fere

nce

C�

21

.7�

28

.4�

30

.8�

17

.4�

16

.6�

20

.3�

18

.8�

18

.3

Lam

be

rtd

ecr

em

en

tD

1.9

2.2

2.2

2.5

2.1

2.2

2.8

2.2

Tra

nsi

tory

lev

el

E�

17

.9�

18

.8�

20

.6�

16

.6�

15

.4�

19

.2�

15

.8�

15

.5

An

gu

lar

ex

ten

tF

10

.11

14

.91

15

.01

16

.61

11

.21

12

.81

10

.31

12

.81

ST

OC

cA

–C

12

.91

9.8

21

4.5

13

13

13

.37

BS 4

5d

BS 4

5�

25

.07

6.3

�2

9.97

2.7

�3

4.37

3.0

�2

1.47

2.3

�1

9.77

2.3

�2

3.37

2.8

�2

3.87

2.7

�2

2.27

2.3

Th

esi

xin

de

pe

nd

en

tp

ara

me

ters

(A–

F)a

reg

en

era

ted

fro

mm

od

ell

ing

of

the

an

gu

lar

resp

on

seo

fth

eb

ack

sca

tte

rin

ten

sity

.B

Sv

alu

es

are

ind

Bre

.1

m2.

A7

1d

Ba

bso

lute

un

cert

ain

tyis

ass

um

ed

on

all

BS

va

lue

s(s

ee

tex

t).

aFr

om

the

Gra

dis

tat

sed

ime

nt

gra

insi

zecl

ass

ifica

tio

nso

ftw

are

(Blo

tta

nd

Py

e,

20

00

).b

Ba

sed

on

the

vis

ua

lin

terp

reta

tio

no

fFi

g.

5.

cS

pe

cula

r-to

-ob

liq

ue

con

tra

st¼

A�

Cfr

om

Eq

.(2

).d

Me

an

BS 4

5ca

lcu

late

db

etw

ee

n4

31

an

d4

71

inci

de

nce

an

gle

so

nb

oth

sta

rbo

ard

an

dp

ort

sid

es.

G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 S105

Author's personal copy

fine sand was retrieved from Area 1. Areas 4 and 5 showcontrasting values for the specular (�13 and �4 dB, respec-tively), and a marked difference in the angular extent of thespecular (4.81 and 9.51), which suggests that the floor of theNarrows Basin (Area 5) has a substantially higher water–sedimentimpedance contrast and interface roughness than the small basinon the continental shelf (Area 4). The similar Lambert reference(C¼�17 dB) for areas 4 and 5 indicates similar sedimentcharacteristics in terms of volume homogeneity, likely to becoarse sand or gravelly sand. The 43 samples in Area 5 (Table 5)confirm the presence of coarser material than in Area 4. Overall,however, B may be too strongly linked with the vertical beamwidth to generate an accurate enough descriptor.

The data also indicate that the transitory level (E) may becharacteristic of large-scale seafloor roughness, as when E is highand contrasting with A and C as in areas 3 (E¼�21 dB) therugosity is a particular case as it includes sediment waves. Weinfer that the transitory (E and F) components are indicatorsof a scale rugosity in the order of magnitude (�100–500 m)of the sediment wave wavelength, rather that at the scale of localslope, in relation with the sounder signal footprint to be

considered in the specular behaviour modelled by the facettheory (Brekhovskikh and Lysanov, 1982).

5.3. Quantitative BS for geological mapping

From the eight reference areas in Cook Strait, we show that thesix GSAB parameters can be used to quantitatively characterise BS

for a variety of environments. However, these data are still toocomplex to develop a reflectivity map at a regional scale. Wepropose using the BS at 451 incidence angle (noted BS45) as aparameter representing the overall BS behaviour. In conventionalMBES surveys, a swath width (aperture of ca. 601) coversapproximately four times the water depth, and the 451 incidenceangle represents the median value on both sides of the specular.For instance, the minimum and maximum calibrated BS valuesacquired (all angles put together) during the Tan0105 survey are�55 and 18 dB, with a mean of �25 dB (Table 4). The abnormallyhigh BS values at specular level are due to the presence of acoherent reflection process superseding the backscattered returnand are unreliable. At grazing angles (68–721), the BS is irregular

Table 6Six parameters (A–F) generated from BS angular responses models for characteristic mud, sand, and gravelly samples collected throughout Cook Strait.

Gravel Sand Mud Mud with underlying

sediments

Samples C118 C61 U630 Q222B

C125 C62

C126 C90

C129 C229

Gravel (%) 100 10.65 0 0

Sand (%) 0 83.25 1.5 6.9

Mud (%) 0 6.1 98.5 93.1

Specular level A*�9.1 �9.1 �13.6 �6.1

Specular angular extent B 4.91 5.91 5.51 5.11

Lambert reference C*�17.9 �23.9 �29.1 �27.3

Lambert decrement D 2 2 2 2

Transitory level E*�17.5 �20.9 �24.5 �20.7

Transitory angular extent F 16.21 17.91 16.21 12.41

n In dB re. 1 m2. A 71 dB absolute uncertainty is assumed.

−40

−30

−20

−10

−50 0 50

Mud

BS

(dB

)

Sand: 1.5%Mud: 98.5%

Mud

−50 0 50

Gravel: 100% Sand: 0 %Mud: 0 %

Gravel

−50 0 50Incidence Angle

Sand

Gravel: 10.6%Sand: 83.3%Mud: 6.1%

Mud +underlying Sed

Fig. 12. Generic Seafloor Acoustic Backscatter (GSAB) models for characteristic samples of gravelly, sandy, and muddy seafloors defined from grain size analysis (Table 3).

Thick grey: BS angular function model of the averaged measured values (dots). Dash grey lines: parameterised profile used for the GABS (Fig. 4; Section 3.3). A model for

mud with underlying compacted sediments is superimposed on the model for muddy seafloor. See Fig. 5 for sample locations. Samples average compositions (Table 6) are

indicated as percentage of gravel, sand, and mud.

G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109S106

Author's personal copy

due to strong angle dependence and poor signal-to-noise ratioand is not always validated. BS45 is therefore the most suitable torepresent the BS in one synthetic value.

The backscatter map, normalised at the BS45 value for surveyTan0105 (Fig. 6B), shows clearly separated regions: sufficientlyhomogeneous when considered individually (BS45 dynamics is6 dB over the sediment waves). The contrast between thereference areas is sufficient for a fairly accurate visual classifica-tion (BS45 typical values of �34 dB and �20 dB in areas 3 and 5,respectively Table 5).

BS45 is proving to be a robust single parameter for applicationto habitat mapping. At a regional scale in Cook Strait BS45 candiscriminate distinct environments, which have different geolo-gical or micro-topographic properties. It therefore provides a toolfor geological interpretation. The use of BS45 as a reference valueis of course conditioned by the availability of the soundercalibration in absolute level. This will have to be validated fromfurther studies, but we suggest that BS45 be used systematicallyfor building compensated sonar images where one reflectivityvalue has to stand for one homogeneous seafloor facies. Such anapproach would have the significant benefit of providing

qualitative images comparable globally, thus enabling compar-ison from different regions. BS45 maps, however, still have limitedclassifying power other than relative contrast, since the informa-tion inherent to the angle response has disappeared. This isexacerbated if the sounder absolute calibration is not available;then only the contrast between various responses can beexploited.

We propose using the relative specular level with regard to theoblique regime average level, given by A–C or Specular-To-ObliqueContrast (STOC), as an indicator of both the impedance contrast andthe interface roughness. STOC, as B, D, F, and (E–C), have thesignificant advantage of being independent of the sounder absolute-level calibration, and can therefore be used globally. The STOC

parameter for the eight representative areas in the Narrows Basin(Table 5, Fig. 15) ranges from 4.5 to 21 dB, i.e. a total dynamic of16.5 dB. A scatter plot of STOC vs. C for the eight reference areas(Fig. 15) suggests that these parameters may provide someclassifying potential as they show a good inter-class separation.This is promising as class separation is a requisite for segmentationand classification algorithms to run effectively (e.g., Lucieer andLucieer, 2009). Enabling robust class separation from BS

measurements increases the rationale of using BS in habitat mapping.

6. Conclusion

The development of a quantitative automated seafloor map-ping tool has been a priority for geologists since acoustics wasfirst used to map the ocean floor. In this paper we have refined theBS signal interpretation and isolated some of its complexrelationships with the substrate.

It is well established that BS is controlled by the sedimentgrain size, the surficial heterogeneity and the small-scaletopography, and therefore relates to substrate composition androughness. Absolute amplitude calibration is required for quanti-tative analysis whereas statistical compensation of the BS angularresponse is a prerequisite for qualitative interpretation. Both thecalibrated and statistically compensated data are required for acomplete quantitative and qualitative interpretation of the seafloorcomposition and interface characteristics. The BS angular re-sponse also provides insights into the seafloor primary sedimen-tary (grain size) and geomorphological (seafloor roughness)properties at both local and regional scales.

Qualitatively, the backscatter data in the Cook Strait region arecharacterised by weak, homogeneous reflectivity patterns overthe continental shelf and deep Hikurangi Trough indicating finesediment, likely mud or sandy mud, and strong reflectivity on thecanyons floors indicating gravel deposits and active sedimentdelivery. Strongly reflective spots on the continental shelf high-light carbonate concretions and erosional landforms.

Backscatter imagery and bathymetry data together provide atool for enhanced interpretation of fine scale structures that maynot necessarily have geomorphic relief but have distinctive surfacetextures or roughness. In the Narrows Basin, the backscatterpattern of strong reflectivity on sediment wave’s troughs and weakreflectivity on the crests indicate that the BS variations across thesediment waves are associated with gravelly deposits at their baseand fine sandy deposits at their top. In some areas, the backscatterdata display a quasi-periodical structure emphasising fine sedi-ment waves otherwise undetected in multibeam bathymetry.

Modelling of the backscatter signal as a function of theincidence angle enables us to provide quantitative informationon the seafloor. A Generic Seafloor Acoustic Backscatter (GSAB)improved model is used to define six physically significantparameters (A–F) for a number of homogeneous and distinctiveareas in Cook Strait. The six quantitative parameters (A–F) for

Fig. 13. Morphological scarp associated with the active Booboo Fault and seafloor

erosional landforms over the Campbell Bank. (A) Multibeam bathymetry,

generated from the log of the bathymetry with north illumination and (B)

compensated backscatter reflectivity. Location on Fig. 5.

G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 S107

Author's personal copy

muddy, sandy, and gravelly seafloors are used as reference toprovide information on the seafloor composition.

We propose using two parameters to simplify the manage-ment of the multi-parameter models, and to define a classificationof the average seafloor characteristic adapted to the regional scalerelevant for many habitat mapping studies: the BS value acquiredat 451 (BS45) incidence angle as a stable indicator of the interfacesediment type, usable either as an absolute descriptor or as acomparison tool, depending on the sounder calibration status;and the specular-to-oblique (STOC) contrast, which provides anindication of the behaviour of interface roughness through itsimpact on the specular reflection. STOC has the added advantageof being independent of absolute data level calibration.

Acknowledgments

This project was funded by the New Zealand Foundation forResearch Science and Technology (FRST) through their pro-

gramme Consequence of Earth Ocean-Change (C01X0702). Thanksgo to Hugues de Longevialle, Ifremer’s Direction of InternationalRelations, for supporting this NIWA-IFREMER collaboration.Sediment analyses were performed at NIWA by Lisa Northcoteand Melanie Herrmann. Joshu Mountjoy (NIWA) commented onan early version of the manuscript, and Richard Pickrill (Geolo-gical Survey Canada) commented on the reviewed version. Figureswere generated using the Sonarscopes and GMT software.

References

Anderson, J.T., Van Holliday, D., Kloser, R., Reid, D.G., Simard, Y., 2008. Acousticseabed classification: current practice and future directions. ICES Journal ofMarine Science 65 (6), 1004–1011. doi:10.1093/icesjms/fsn061.

Augustin, J.M., Dugelay, S., Lurton, X., Voisset, M., 1997. Applications of animage segmentation technique to multibeam echo-sounder data. In:OCEANS apos’97. MTS/IEEE Conference Proceedings, Halifax, NS, Canada,pp. 1365–1369.

Augustin, J.M., Le Suave, R., Lurton, X., Voisset, M., Dugelay, S., Satra, C., 1996.Contribution of the multibeam acoustic imagery to the exploration of the sea-

0

5

10

15

20

25

-40 -35 -30 -25 -20 -15BS45

A-C

(STO

C)

1

23

4

567

8

Fig. 15. Scatter plot of A–C vs. BS45. Error bars are indicated based on standard deviation on BS measurements as the GSAB model does not provide errors.

-45

-40

-35

-30

-20

-25

-15

-10

-5

BS

(dB

)

0 20 40 60 80Incidence Angle

Mud

Sand

Gravel

1 North East2 North3 Sediment wave field4 West5 South6 Central South7 Inhomogeneous West8 South west

Fig. 14. Modelled BS angular profiles for the eight representative areas (Fig. 6) in the Narrows Basin superimposed on the gravel, sand, and mud modelled BS angular

profiles. Only the positive incidence angles are represented on the figure as the models are symmetrical (Table 5).

G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109S108

Author's personal copy

bottom: examples of SOPACMAPS 3 and ZoNeCo 1 cruises. Marine GeophysicalResearches 18 (2–4), 459–486.

Augustin, J.M., Lurton, X., 2005. Image amplitude calibration and processingfor seafloor mapping sonars. In: Proceedings of the Oceans 2005—Europe,pp. 698–701.

Barnes, P.M., Lamarche, G., Bialas, J., Henrys, S., Pecher, I., Netzeband, G., Greinert, J.,Mountjoy, JJ., Pedley, K., Crutchley, G., 2010. Tectonic and geological frameworkfor gas hydrates and cold seeps on the Hikurangi subduction margin, New Zealand.Marine Geology 272 (1–4), 26–48, doi:10.1016/j.margeo.2009.03.012.

Blott, S., Pye, K., 2000. GRADISTAT: a grain size distribution and statistics packagefor the analysis of unconsolidated sediments. Earth Surface Processes andLandforms 26 (11), 1237–1248. doi:10.1002/esp.261.

Brekhovskikh, L., Lysanov, Y., 1982. Fundamentals of ocean acoustics. In: Felsen, L(Ed.), Springer Series in Electrophysics, vol. 8. Springer-Verlag, Berlin,Heidelberg, New York, pp. 249.

Briggs, K.B., Williams, K.L., Jackson, D.R., Jones, C.D., Ivakin, A.N., Orsi, T.H., 2002.Fine-scale sedimentary structure: implications for acoustic remote sensing.Marine Geology 182 (1–2), 141–159.

Buchanan, J.L., 2005. An assessment of the Biot-Stoll Model of a poroelastic seabed.NRL/MR/7140–05-8885, Naval Research Laboratory, Washington, DC 20375-5320 81pp.

Buckingham, M.J., 2000. Wave propagation, stress relaxation, and grain-to-grainshearing in saturated unconsolidated sediments. Journal of the AcousticalSociety of America 108 (6), 2796–2815.

Carter, L., 1992. Acoustical characterisation of seafloor sediments and itsrelationship to active sedimentary processes in Cook Strait, New Zealand.New Zealand Journal of Geology and Geophysics 35, 289–300.

de Moustier, C., 1986. Beyond bathymetry: mapping acoustic backscattering fromthe deep seafloor with Sea Beam. Journal of the Acoustical Society of America79 (2), 316–331.

Ehrhold, A., Hamon, D., Guillaumont, B., 2006. The REBENT monitoring network, aspatially integrated, acoustic approach to surveying nearshore macrobenthichabitats: application to the Bay of Concarneau (South Brittany, France).ICES Journal of Marine Science 63 (9), 1604–1615. doi:10.1016/j.icesjms.2006.06.010.

Fonseca, L., Brown, C., Calder, B., Mayer, L., Rzhanov, Y., 2009. Angular rangeanalysis of acoustic themes from Stanton Banks Ireland: a link between visualinterpretation and multibeam echosounder angular signatures. AppliedAcoustics 70 (10), 1298–1304. doi:10.1016/j.apacoust.2008.09.008.

Fonseca, L., Mayer, L.A., Orange, D., Driscoll, N.W., 2002. The high-frequencybackscattering angular response of gassy sediments: model/data comparisonfrom the Eel River margin. California Journal of the Acoustical Society ofAmerica 111 (6), 2621–2631.

Foote, K.G., 1980. Importance of the swimbladder in acoustic scattering by fish: acomparison of gadoid and mackerel target strengths. Journal of the AcousticalSociety of America 67 (6), 2084–2089.

Foote, K.G., Chu, D., Hammmar, T.R., Baldwin, K.C., Mayer, L.A., Hufnagle Jr, L.C,Jech, J.M., 2005. Protocols for calibrating multibeam sonar. Journal of theAcoustical Society of America 117 (4), 2013–2027.

Goff, J.A., Olson, H.C., Duncan, C.S., 2000. Correlation of side-scan backscatterintensity with grain-size distribution of shelf sediments, New Jersey margin.Geo-Marine Letters 20 (1), 43–49.

Guillon, L., Lurton, X., 2001. Backscattering from buried sediment layers: theequivalent input backscattering strength model. Journal of the AcousticalSociety of America 109 (1), 122–132.

Hellequin, L., Boucher, J.M., Lurton, X., 2003. Processing of high-frequencymultibeam echo sounder data for seafloor characterization. IEEE Journal ofOceanic Engineering 28 (1), 78–89. doi:10.1109/JOE.2002.808205.

Hellequin, L., Lurton, X. and Augustin, J.M., 1997. Postprocessing and signalcorrections for multibeam echosounder images. Oceans Conference Record(IEEE), pp. 23–26.

Hicks, D.M., Shankar, U., 2003. Sediment yield from New Zealand rivers. NIWAchart, Miscellaneous series No. 79. National Institute of Water and Atmo-spheric Research, Wellington, NZ.

Hughes Clarke, J.E., Danforth, B.W., Valentine, P., 1997. Areal seabed classificationusing backscatter angular response at 95 kHz. In: High Frequency Acoustics inShallow Water. NATO SACLANT Undersea Research Centre, Lerici, Italy.

Hughes Clarke, J.E., Mayer, L.A., Wells, D.E., 1996. Shallow-water imagingmultibeam sonars: a new tool for investigating seafloor processes in thecoastal zone and on the continental shelf. Marine Geophysical Researches 18(6), 607–629.

Jackson, D.R., Briggs, K.B., 1992. High-frequency bottom backscattering: roughnessversus sediment volume scattering. Journal of the Acoustical Society ofAmerica 92 (2 I), 962–977.

Jackson, D.R., Ishimaru, A., Winebrenner, D.P., 1986. Application of the compositeroughness model to high frequency bottom backscattering. Journal of theAcoustical Society of America 79 (5), 1410–1422.

Kongsberg, 1997. EM300 operator manual, N.160719/C. Kongsberg Simrad ASHorten, Norway, 398 pp.

Law, C.S., Nodder, S.D., Mountjoy, J., Marriner, A., Orpin, A., Pilditch, C.A., Franz, P.and Thompson, K., 2010. Geological, hydrodynamic and biogeochemicalvariability of a New Zealand deep-water methane cold seep during anintegrated three year time-series study. Marine Geology 272 (1–4), 189–208,doi:10.1016/j.margeo.2009.06.018.

Le Chenadec, G., Boucher, J.-M., 2005. Sonar image segmentation using the angulardependence of backscattering distributions. In: Proceedings of the IEEEOceans—Europe 2005.

Le Chenadec, G., Boucher, J.-M., Lurton, X., 2007. Angular dependence ofK-distributed sonar data. IEEE Transactions on Geoscience and Remote Sensing45 (5), 1224–1235. doi:10.1109/TGRS.2006.888454.

Le Gonidec, Y., Lamarche, G., Wright, I.C., 2003. Inhomogeneous substrateanalysis using EM300 backscatter imagery. Marine Geophysical Researches24, 311–327. doi:10.1007/s11001-004-1945-9.

Lewis, K.B., Carter, L., Davey, F.J., 1994. The opening of Cook Strait: interglacial tidalscour and aligning basins at a subduction to transform plate edge. MarineGeology 116 (3/4), 293–312.

Lucieer, V., Lucieer, A., 2009. Fuzzy clustering for seafloor classification. MarineGeology 264 (3–4), 230–241. doi:10.1016/j.margeo.2009.06.006.

Lurton, X., 2003a. Underwater Acoustics. An Introduction. Geophysical Sciences.Jointly published by Springer Praxis Books & Praxis Publishing, UK. 356pp.

Lurton, X., 2003b. Theoretical modelling of acoustical measurement accuracy forswath bathymetric sonar. International Hydrographic Review 4 (2), 24–37.

Lurton, X., Augustin, J.M., Dugelay, S., Hellequin, L., Voisset, M., 1997. Shallow-water seafloor characterization for high-frequency multibeam echosounder:image segmentation using angular backscatter. In: Pace, N. (Ed.), Proceedingsof SACLANTCEN Conference CP-45, High Frequency Shallow Water Acoustics.

Maıtre, H.,, 2001, Traitement des images de radar �a synth�ese d’ouverture. Paris:Hermes Sciences Europe Ltd, 351pp.

McRea Jr., J.E., Greene, H.G., O’Connell, V.M., Wakefield, W.W., 1999. Mapping marinehabitats with high resolution sidescan sonar. Oceanologica Acta 22 (6), 679–686.

Medialdea, T., Somoza, L., Leon, R., Farran, M., Ercilla, G., Maestro, A., Casas, D.,Llave, E., Hernandez-Molina, F.J., Fernandez-Puga, M.C., Alonso, B., 2008.Multibeam backscatter as a tool for sea-floor characterization and identifica-tion of oil spills in the Galicia Bank. Marine Geology 249 (1–2), 93–107.doi:10.1016/j.margeo.2007.09.007.

Mountjoy, J.J., Barnes, P.M., Pettinga, J.R., 2009. Morphostructure and evolution ofsubmarine canyons across an active margin: Cook Strait sector of theHikurangi Margin, New Zealand. Marine Geology 260 (1–4), 45–68.doi:10.1016/j.margeo.2009.01.006.

Mourad, P.D., Jackson, D.R., 1989. High frequency sonar equation models forbottom backscatter and forward loss. In: Proceedings of the IEEE OCEANS ’89Proceedings, pp. 1168–1175.

Mulhearn, P.J., 2000. Modelling acoustic backscatter from near-normal incidenceechosounders—sensitivity analysis of the Jackson Model, DSTO-TN-0304.DSTO Aeronautical and Maritime Research Laboratory, Melbourne Victoria3001 Australia, 36 pp.

Novarini, J.C., Caruthers, J.W., 1998. A simplified approach to backscattering from arough seafloor with sediment inhomogeneities. IEEE Journal of OceanicEngineering 23 (3), 157–166.

Pratson, L.F., Edwards, M.H., 1996. Introduction to advances in seafloor mappingusing sidescan sonar and multibeam bathymetry data. Marine GeophysicalResearches 18, 601–605.

Ryan, W.B.F., Flood, R.D., 1996. Side-looking sonar backscatter response at dualfrequencies. Marine Geophysical Researches 18 (6), 689–705.

Stoll, R.D., 1985. Marine sediment acoustics. Journal of the Acoustical Society ofAmerica 77 (5), 1789–1799.

Van Dissen, R.J., Berryman, K.R., 1996. Surface-rupture earthquakes over the lastca, 1000 years in the Wellington region, New Zealand, and implications forground shaking hazard. Journal of Geophysical Research 101 (B3), 5999–6019.

Wright, D.J., Heyman, W.D., 2008. Introduction to the special issue: marine andcoastal GIS for geomorphology, habitat mapping, and marine reserves. MarineGeodesy 31 (4), 223–230. doi:10.1080/01490410802466306.

G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 S109