Author's personal copy Continental Shelf Research 31 (2011) S93–S109 Contents lists available at ScienceDirect Continental Shelf Research journal homepage: www.elsevier.com/locate/csr Quantitative characterisation of seafloor substrate and bedforms using advanced processing of multibeam backscatter—Application to Cook Strait, New Zealand Geoffroy Lamarche a,n, Xavier Lurton b, Anne-Laure Verdier a, Jean-Marie Augustin b a b National Institute of Water and Atmospheric Research (NIWA) Ltd, Private Bag 14-901, Wellington 6241, New Zealand Institut Franc- ais de Recherche pour l’Exploitation de la Mer (IFREMER), BP70, 29280 Plouzané, France a r t i c l e in fo abstract Article history: Received 10 August 2009 Received in revised form 9 May 2010 Accepted 1 June 2010 Available online 11 June 2010 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 highresolution 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. Keywords: Backscatter Multibeam echo-sounder Sediment waves Habitat mapping 1. Introduction The increasing need for the sustainable management of the marine environment and resources requires obtaining objective and quantitative descriptors of seafloor habitats from remotesensed data. The potential for characterizing the nature of the seafloor using the acoustic reflectivity (or backscatter) collected routinely with multibeam echo-sounders (MBES) has been recognised for at least 20 years (e.g., de Moustier, 1986). However, quantitative analysis of backscatter data for seafloor characterisation 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 acoustic backscatter and geological substrate properties fall into three n Corresponding author. Tel.: +64 4 386 0465; fax: + 64 4 386 21 53. E-mail address: g.lamarche@niwa.co.nz (G. Lamarche). 0278-4343/$ - see front matter & 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.csr.2010.06.001 categories: (1) physical process modelling addressing links between physical processes (sound propagation in inhomogeneous media and acoustic wave scattering by rough interfaces and volume heterogeneity); (2) geoacoustical modelling relating the physical parameters of the sediments with their geological description; and (3) phenomenological models, based on heuristical descriptors and parameters, aimed at a global model of intricate physical phenomena perceived as too complex either for justifying a rigorous physical approach or for making possible inversion processes. Physical and geoacoustical models are best suited to interpreting fundamental phenomena observable on simplified or ideal features (Fonseca et al., 2002; Hughes Clarke et al., 1996; McRea et al., 1999; Mourad and Jackson, 1989; Novarini and Caruthers, 1998). Phenomenological models have been developed for experimental data exploitation (Augustin et al., 1996; Le Chenadec et al., 2007; Lurton, 2003a). Accurate physical modelling provides insight into the phenomena involved in sonar signal backscattering—in the sense that it makes explicit the intrinsic potential dependence of backscattered echoes upon geological substrate Author's personal copy G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 composition (Mulhearn, 2000; Pratson and Edwards, 1996), However this approach is unavoidably limited to ideally simplified configurations, making it possible to build a theoretical model of acceptable complexity; hence it is generally inappropriate for providing unsupervised (objective) inversion tools for accurately retrieving sediment characteristics directly from the backscatter data, except in tightly constrained contexts where the model applicability is known a priori (Hughes Clarke et al., 1997; Jackson et al., 1986). It therefore remains still very much questionable as to whether such numerical models may provide the expected link between seafloor remote sensing and habitat mapping. A difficult and often under-estimated problem is that the accurate retrieval 174°30'E Kapiti Is. of physical BS values has to go through a thorough compensation of the sensor characteristics (Ryan and Flood, 1996). Despite progress in the most recent MBES systems, the built-in compensation and basic processing do not allow an accurate enough determination of absolute BS and incidence angles required for any quantitative analysis of the backscatter signal (Hellequin et al., 1997; Hughes Clarke et al., 1996; Le Gonidec et al., 2003). This study aims at showing the potential provided by fully processed backscatter data to characterize the seafloor in a variety of geological and sedimentological environments, using an MBES dataset from Cook Strait, New Zealand (Fig. 1). The first part of the paper undertakes a qualitative interpretation of the normalised 175°'E 175°30'E 176°E Hi ku ran gi Ma rgin S94 AUS ar lb So oro un ug ds h 40°S M Mana Is 41°S PAC 175°E 180° Wa ir ara pa Fig.3 190° Wellington Fig.6 Narrows t aul eF in Alp Pall liser Bay NB NC WC 41°30'S f el Sh a p a irar Wa Cloudy Bay Ck l ta en in lf t n e Co Sh PC Ck PaC Campbell Bank B Pa PB OR Fig.13 OC 42°S h ug ro iT g ran ku Hi 0 10 20 km 40 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. Author's personal copy G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 backscatter imagery, with a particular focus on well-developed sediment waves, from which we discuss the potential benefit brought by backscatter data to quantify the geometrical and geological characteristics of sediment bedforms. We then apply an improved version of the heuristical model for angular BS proposed by Lurton (2003a) and Augustin and Lurton (2005) to various responses recorded in central Cook Strait. This approach enables us to propose a quantitative interpretation of the backscatter over a variety of environments and bedform, and work toward the classification and geological interpretation of BS using the model parameters. We attempt to retrieve the seafloor sediment types from the derived relationship between sediment mean grain size and BS angular response. We finally discuss the methodological difficulties encountered and propose practical solutions for future research aiming at a generalisation to this approach. Our purpose is not to propose new models in backscattering or geoacoustics. Rather, we seek to emphasise the value of quantitative backscatter data analysis for investigating the submarine environment at a regional scale from shallow continental shelf to deep ocean water depths, by using a pragmatic, rigorously controlled, approach to backscatter data analysis. S95 with rich, distinctive, chemosynthetic faunas, which stress the importance of marine environment management for the region. 3. Data and methods 3.1. Data acquisition MBES data were collected during six oceanographic surveys of R.V. Tangaroa between 2001 and 2005 in Cook Strait by the National Institute of Water and Atmosphere (NIWA) (Table 1). The total surveyed area is ca. 8500 km2 in water depths greater than ca. 150 m (Fig. 1). Tangaroa has a hull-mounted Kongsberg EM300 (32 kHz) MBES (Table 2) that fully compensates the vessel position and motion (heave, pitch, roll, and yaw). Sound velocity measurements were performed at regular intervals to account for hydrology effects. Bathymetry data were routinely processed onboard using C&C Technologies HydroMap and ArcInfo software. A large geophysical and geological data dataset including multichannel and high resolution seismic reflection data, sediment cores, seafloor grab samples, and photographs are available for the region (Carter, 1992; Lewis et al., 1994) and provided means to ground-truth the remote-sensed data analysis. 2. Cook Strait geological setting 3.2. Backscatter data processing The diverse and complex geomorphology of Cook Strait is the result of dynamic climatic, tectonic, and oceanic forcings that impacted on the region since at least the last post-glacial period (Lewis et al., 1994). Cook Strait, between the North and South Islands of New Zealand (Fig. 1), is of strategic public and economic interest for New Zealand containing electrical power cables joining the two islands; major tectonic faults (Van Dissen and Berryman, 1996); and economically important fish stocks and biological resources. A series of canyons has been carved into the continental shelf of Cook Strait (Mountjoy et al., 2009), with walls up to 1100 m in height. Instability of the canyon walls is suggested by multiple semi-circular scars in their upper section with debris accumulating in the canyon floors. Beyond the continental slope, the canyons flow into the Hikurangi Trough infilled with thick turbidite sequences and host to the Hikurangi channel (Lewis et al., 1994). The shelf edge is defined at  150 m water depth beyond which the continental slope descends to more than 2500 m water depth in the Hikurangi Trough. During the Last Glacial Maximum, the continental shelf flanking 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 by a wedge of post-glacial mud. However, strong waves and tidal forcing have left some areas of the shelf bare of post-glacial sedimentation so that the erosional surface outcrops at the seafloor (Lewis et al., 1994; Mountjoy et al., 2009). Today powerful tidal currents, with peak flows reaching 1.5 m/s in the narrowest passage of Cook Strait, known as the Narrows, coupled with strong sediment delivery from North and South island’s rivers (Hicks and Shankar, 2003), have produced distinctive patterns of sedimentation. In particular, several sediment wave fields were broadly recognised in the Narrows Basin (Carter, 1992). In addition to the ecological diversity to be expected from such a complex environment, Cook Strait hosts a large diversity of faunal ecosystems associated with active and relict fluid seeps such as carbonate chimneys and concretions, and reflective plumes generated (Barnes et al., 2010; Lewis et al., 1994, Law et al., 2010). Seeps and cold-fluid vents are usually associated The aim of backscattered signal processing is to provide an interpretable image, and to estimate descriptive parameters relating the BS with geological properties. The data is inherently noisy and requires processing to remove artefacts attributable to the recording equipment, seafloor topography, and water column properties (Hellequin et al., 1997, 2003; Le Gonidec et al., 2003). The echo level (EL) at the receiver’s output is primarily associated with the seafloor backscattering strength (BS(y)), where y is defined as the ray-path incidence angle onto the seafloor obtained from the refraction estimated during bathymetry processing and from the locally computed digital terrain model (DTM). EL depends on transmission loss (TL), which depends on transmission range (R) and on water column Table 1 R.V. Tangaroa surveys in Cook Strait for EM300 acquisition (Figs. 2 and 5). Survey Year Region TAN0105 TAN0204 TAN0211 TAN0215 TAN0309 TAN0510 Apr 2001 Mar 2002 Aug 2002 Oct 2002 Jun 2003 Aug 2005 Narrows Basin South Narrows Basin Campbell Plateau, Cook Strait Canyon Cook Strait canyons Northern Marlborough, Hikurangi Trough Wairarapa Shelf—Hikurangi Trough Table 2 Technical characteristics of NIWA’s EM300 multibeam system. Central frequency Number of beams Beam width Angular swath range Overall swath coverage Water depth range Survey navigation systems Positioning and motion compensation Sound velocity calibration 32 kHz 135 11 n 21 (along n across track) 1401 5.5 n water depth 150–3500 m HydroPro and HydroMap TSS DMS TSS POS-MV 320 system Expendable bathythermographs (XBTs) Digibar DB-1100 AML SVplus sound velocity profiler Author's personal copy S96 G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 absorption coefficient (a); the sonar characteristics (transmitted source level SL, signal duration T, transmission and reception directivity patterns DT and DR; receiver gain GR; hydrophone sensitivity SH); and the insonified area A(R, y, T) (Lurton, 2003a). SL, DT, SH, DR, and GR are hardware dependent. Hydrology conditions influence absorption (a) and incidence angle (y), whereas the geometrical configuration (water depth, topography) controls both the range (R) and (y). These parameters are gathered into the conventional sonar equation (Lurton, 2003a, 2003b): EL ¼ SL þDT 2:TL þ 10log A þ BS þSH þDR þ GR ð1Þ Practically, there are two issues in processing BS data for enhanced application for geological interpretation: sonar calibration, which controls the physical signal levels provided by the sensor, and recorded data correction required for qualitative and quantitative backscatter estimation and exploitation. 3.2.1. Sonar calibration Two levels of amplitude calibration may be applied on any sonar data: (1) An elementary calibration to ensure that the sonar parameters (SL, T, DT, DR, GR) are correctly accounted for, in the recorded amplitude levels; this guarantees that the data are repeatable, and do not depend on the system settings; ideally this is done through built-in compensations applied in the receiver. (2) A full calibration procedure to provide exact referenced values for the transmitted level and the receiving sensitivity. This requires factory calibration of transducers and electronics, and implementation of control tools and procedures to ensure that these parameters are checkable and compensable onboard. Despite significant improvements, the built-in corrections applied by MBES (Kongsberg, 1997) still remain imperfect and may result in artefacts in the backscatter data, precluding accurate qualitative and quantitative data analysis. To improve data quality, built-in compensations applied at data acquisition must therefore be removed and replaced by more accurate ones. Unfortunately, some of the system characteristics are not directly accessible and have to be deduced from the recorded data. For instance the echo level modulation caused by the array directivity patterns is corrected by fitting a modelled pattern on data recorded over several assumed-homogeneous seafloor areas (Fig. 2) (Lurton et al., 1997). This technique is similar to that applied to satellite-borne radar reflectivity data for a posteriori compensation of directivity patterns (Maı̂tre, 2001). The amplitude calibration used in fishery acoustics, measuring the response of a reference spherical target (e.g. Foote, 1980), can be applied to high-frequency MBES in a controlled environment (Foote et al., 2005); however it is not practical for low-frequency MBES featuring long arrays already mounted on ship’s hull. In this study, we retained the manufacturer values for the transmitted levels and receiver sensitivity; the risks of this simplification being either an inaccurate qualification of the hardware by the manufacturer or a possible drift of the system with time. Manufacturer calibration gain function Sonarscope calibration gain function BS (dB) -20 Gain (dB) 0 -4 -80 0 Transmission angle 80 -36 -60 0 Transmission angle 60 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 recalculated 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. Author's personal copy G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 3.2.2. Data correction For qualitative analysis of the sonar images, an echo-level compensation needs to be applied on the data. This process removes artefacts inherent in the sensor and corrects the signal angular response to produce a normalised image interpretable by end-users (Fig. 3). Largest errors are produced by the specular echo at nadir and the reflectivity fall-off at grazing angles. Echo-level compensation requires that the normalised level is taken at a significant and relevant value, in particular respecting the contrast between the BS levels associated with different sediment types. Finally, an additional absolute calibration of the BS data must be carried out for any satisfactory 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, 41°10' 41°14' -2 dB -40 -15 dB -40 0 km 5 S97 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 as different layers available for simultaneous or combined processing using either conventional or sonar-specific signal and image processing modules. Advanced processing techniques include signal 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 absorption given by changing hydrological conditions; (2) estimation of the array directivity patterns, from which specific gains were applied for each angular sector of the sounder; and (3) correction for the instantaneous signal footprint area, as a function of incident angle, directivity pattern aperture, and pulse duration. Attenuation of the strong, sometimes obliterating, echo associated with the specular reflection in backscatter images is a fundamental step prior to any qualitative analysis (Fig. 3). In the present case, the specular signal was filtered out by statistical compensation (Augustin and Lurton, 2005) consisting in subtracting from the physical angle-dependent sonar image (Fig. 3C) an average BS(y), hence flattening the angular dependence while preserving the average BS value. The resulting image is more ‘‘readable’’ since the contrast is enhanced because the corrected BS values are close to the mean value (Fig. 3B). However, this compensated image has lost most of the BS level information, since it was normalised to an arbitrarily chosen reference level (in this study we used the uncompensated value at 451), and it is hence exploitable mainly in terms of relative intensity. It is emphasised that a fundamental distinction has to be made between the compensated backscatter, normalised at a conventional reference level, defined for imagery purposes and usable for qualitative analysis, and the physically correct calibrated BS used for quantitative analysis. The two types of data are expressed in dB per m2 (simplified as dB) and both are incorporated in the geological-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 firstly developed within the framework of directivity pattern compensations for MBES data (Hellequin et al., 1997) and used since in various 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 applied to a large regional dataset. The GSAB model represents the BS angular response as a combination of a Gaussian law for the specular angles and Lambert-like law for grazing angles (Fig. 4). In its original form, it uses four descriptors (A–D), with no explicit relationship with the geological or acoustical attributes used classically in these disciplines (e.g., Jackson et al., 1986). The model is given by 41°10' BSðyÞ ¼ 10log½A expðy2 =2B2 Þ þC cosD y 41°14' ð2Þ where 174°30’ 174°36’ 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.  A quantifies the specular maximum amplitude. In the tangent-  plane approach (Brekhovskikh and Lysanov, 1982). A is related to the coherent reflection coefficient at the water–seabed interface, 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 Author's personal copy S98 G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 + BS(θ) = 10 log[ A.exp(-θ²/2B²) + C. cosDθ ++ A + E.exp(-θ²/2F²)]. + + + + + B + E + + + + + + + + + F + + + + Lambert ++ ++ + + ++ C + ++ ++ + +++ + + + ++ cosDθ ++ ++ + ++ ++ Transitory + ++ +++ ++ Specular ++ -10 BS (dB) -20 -30 -40 -80 -60 -40 -20 0 20 40 60 80 θ : Incidence Angle (degrees) 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)¼ 10log[A exp(  y2/2B2) + C cosD y + E exp(  y2/2F2)].   is therefore an interface roughness descriptor. Note that, in the following, angular extent denotes the half-width of the specular peak actually measured by MBES working over both sides of nadir. C quantifies the average BS level at oblique incidence. It is the offset associated with Lambert’s law describing BS at intermediate angles for rough interfaces, and includes the contribution 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 values between  15 and  40 dB are commonly observed. D is the backscatter angular decrement, commanding the falloff at grazing angles. It is high for soft and smooth sediment interfaces. According to the laws of Lommel-Silliger and Lambert, D is equal to 1 and 2, respectively (Lurton, 2003a). While this four-parameter model (A, B, C, D) is adequate to model the backscatter angular behaviour in many configurations, in some cases the model may require an intermediate term to account for the smooth transition between the specular and lateral modes. This transitory component is modelled with a second Gaussian law; Eq. (2) becomes BSðyÞ ¼ 10log½A expðy2 =2B2 Þ þ C cosD y þ E expðy2 =2F 2 Þ ð3Þ where E is the transitory maximum level (dB) and F its angular half-extent (in deg) (Fig. 4). There has been no study on the relationship of E and F to geological and physical seafloor properties. Their secondary role will be addressed in the discussion. This improved form (3) of GSAB will be used in the following. Since GSAB aims at providing angle dependence descriptors of a swath sonar dataset, the coefficients are only valid for a given Table 3 Simplified grain size classification from Blott and Pye (2000). j* d (mm) Class o  1.0  1.0–4.0 44 42.00 2.00–0.063 o 0.063 Gravel Sand Mud n j ¼  log 2d, d—grain diameter (mm). sonar frequency. Today’s current MBES uses one of seven nominal frequencies (12; 32; 70; 100; 200; 300; and 450 kHz) each corresponding to different practical applications. 3.4. Sample analysis The sample database consists of 260 seafloor samples collected since the late 1950s by the New Zealand Oceanographic Institute (NZOI) and subsequently by NIWA, using gravity corer, piston corer, and seafloor grab. Sedimentological analysis consisted of grain size analysis and classification as a function of the percentage of mud, gravel, and sand undertaken using standard procedure for all samples. Mean grain sizes were obtained using a laser granulometer and the Gradistat programme (Blott and Pye, 2000) for statistical analysis of unconsolidated sediments. The samples are classified in three groups based on log-normal distribution from their j (phi) values, where j ¼  log2 d, where d is the grain diameter in millimetres (Table 3). 4. Results 4.1. Bathymetry and backscatter imagery Although the morphological complexity of Cook Strait is well expressed in the bathymetry (Fig. 1), the compensated backscatter imagery provides complementary detailed information about seafloor features and substrates (Fig. 5). At the scale of the Cook Strait region, we define three basic imagery facies: (1) a homogeneous, weak-to-moderate reflectivity (dark grey) facies, (2) a homogeneous, highly reflective (light grey) facies, and (3) a highly heterogeneous reflective facies. Using these three facies as end-members of a continuum we define four regions: the Narrows, the continental shelf, the continental slope and canyons, and the Hikurangi Trough. The Narrows region corresponds to the 350 m-deep, N–S trending oval Narrows Basin and its steep, linear, eastern and western walls. The continental shelf on either sides of the basin is less than 6 km wide. It is smooth and homogeneous on the west and rough to the east, along the Wellington Peninsula. In terms of imagery, the Narrows Basin divides into a northern weak-tomoderate (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 strong imagery extends well southward of the Narrows to the head of the Cook Strait canyon (Fig. 6). The BS dynamic over the Narrows Basin from Oceanographic voyage Tan0105 alone is 18 dB above the 3% quantile (Table 4). The floor and eastern wall of the Narrows Basin is dominated by fields of sediment waves; these are characterised by a distinct E–W orientation, i.e. perpendicular to the long axis of the Narrows Basin, and asymmetric shapes with steeper NE flanks. The largest sediment waves are up to 10 m in height with a ca. 250 m wavelength, and the smallest waves are 2–4 m high with a wavelength less than 100 m (Fig. 7). Author's personal copy G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 S99 Tan0211 41°S 09 03 n Ta 6 T an 01 05 3 B Wellington 04 n02 Ta A 0 km 20 NC 41°30'S 0 51 n0 Ta 1 21 n0 Ta Ta n0 21 5 PaB 13 PB 0 km 2 OC gh 9 030 Tan u Tro gi n ura Hik Strong -18 Moderate dB Weak -36 PaB 175°E 175°30’E 42°S 176°E 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. The backscatter over the sediment wave field is characteristically very weak to weak (dark). The continental shelf is poorly covered by EM300 since this equipment is not well-adapted for water depths shallower than 100 m. At this relatively coarse regional scale, the shelf appears smooth and homogenous. The shelf break is well marked at 160 m below the sea level, except in areas where landsliding has prograded and eroded the top of the continental slope (Mountjoy et al., 2009). The continental shelf shows a weak-tomoderate (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 a very rough topography associated with canyon flanks and steep gullies. The steep gradients, blocky facies on the canyons floors, and multiple circular scarps at the top of the continental result in a highly heterogeneous imagery facies (Fig. 5), with a BS dynamic of 19 dB above the 3% quantile. The canyon floors along the continental slope show a strong to very strong reflectivity contrasting with the low values on the flanks. This is best exemplified over the Opouawe and Pahaua canyons. The Hikurangi Trough has a smooth topography associated with a moderate (light grey), very homogeneous imagery facies. The 16 dB dynamic of the BS above the 3% quantile is likely to be the result 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 provides an excellent means to identify bedforms. In the Narrows Basin, the multibeam bathymetry provides clear information on the geometry of sediment waves (amplitude, orientation, wavelength, asymmetry, and along-crest length, Figs. 7 and 8). Low backscatter level is consistently observed on the sediment wave crests, and strong levels in the troughs: a typical pattern already noted by e.g., Fonseca et al. (2009) and Goff et al. (2000). Over the Author's personal copy G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 Contours interval 250 m 70 35 0 300 200 S100 2 0 1 25 metres 0 30 30 0 390 300 4 20 0 41°10'S 3 15 0 25 0 7 6 0 35 35 5 41°15'S 15 0 100 15 0 250 20 0 0 8 174°20'E 174°25'E 0 174°30'E 2 km 174°35'E 4 1 2 4 41°10'S 7 3 6 8 5 -20 BS45 dB -30 41°15'S 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). Narrows Basin the BS varies with a very strong dynamic of 13 dB, for bathymetric amplitudes ranging from 3 to 5 m. The average crest-to-trough BS dynamics is  6 dB across any of the sediment waves, i.e. well within the equipment 1 dB resolution (Kongsberg, 1997). The variation in reflectivity across sediment waves is not related to an effect of the incidence angle. In some areas within the Narrows Basin the BS modulation contrast is still high with crest-to-trough BS dynamics from 3 to 7 dB producing sediment wave patterns on the BS imagery (Fig. 8A). However, these sediment waves cannot be detected in the bathymetry (Fig. 8B). The incidence angles across the sediment waves (Fig. 8C) range from 521 to 571, with a typical amplitude of 21. Assuming a Lambert dependence, a 21 incidence angle difference would generate 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, the average angular response profile (BS(y)) over the sediment wave field 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 a poorly defined specular shape, i.e. with a progressive transition from the specular to the transitory regime. At higher incidence Author's personal copy G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 S101 Table 4 Minimum 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 Alla Alla Alla Alla Min Max Amplitude Mean  65 12 77 3%d  34  16 18  257 5  55 18 73 3%d  38 9 29  25 77  47 16 63 3%d  23 1 22  12 7 6  47  11 46 3%d  36  18 18  26 75  54 9 45 3%d  42  22 20  317 6 a All angle of incidence (  701 to 701). 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) 7 1 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. b -305 m Bathymetry -315 Calibrated Reflectivity vvb -30 -40 db -30 -35 Processed Reflectivity Deg 55 45 Incidence angle 0 200 400 600 Distance (m) 800 1000 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. angle, BS decreases slowly to  33 dB at 501. The BS values orthogonal to the sediment wave show fluctuations up to 10 dB between troughs and crests at all angles. The two observations suggest that the reflectivity contrasts recorded across the sediment waves are controlled by sediment type variations rather than by incidence angle and that, in some situations, analysis of the BS may be more efficient for their detection than that of bathymetry. Although some angular dependence is certainly noticeable in the BS, its effect is clearly superseded by the sediment type. Consequently bedforms can be detected from backscatter data even 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 used from the Tan0105 voyage (Fig. 6). Tan0105 covers the entire Narrows Basin incorporating a variety of geological features and water depths. The GSAB parameters were generated for eight reference areas selected from the bathymetry and the BS maps (Fig. 6). The areas were chosen as homogeneous in terms of both morphology and average BS (Fig. 10). Each area typically includes more than 100 complete pings, thus spanning the entire BS angular response and providing a significant average. The average BS(y) generated for the eight reference areas have distinctly different shapes (Fig. 11A) generated by the unique seafloor characteristics (interface roughness, impedance, and volume scattering) within each region. The GSAB parameters (Table 5) provide a means for further classification of seafloor types from the measured BS. Author's personal copy S102 G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 20 dB Reflectivity (dB) 30 Bathymetry (m) m 270 280 20 deg 52 Incidence angle (deg) 30 54 56 400 600 800 1000 1200 Distance (km) 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 5 10 I.A.=0o BS 15 0 200 ping 400 20 I.A.=25o 25 BS 5 30 0 200 ping 400 10 I.A.=45o 30 34 0 200 Backscatter Strength (BS) 38 I.A.=60o 32 BS ping 400 36 BS 20 40 0 200 ping 400 25 30 35 40 80 60 40 20 0 20 40 60 80 Incidence Angle (I.A.) 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. On the profiles generated from the measured BS (Fig. 11A, Table 4), the specular levels range from  23 to 1 dB. This is larger than the GSAB parameters, where the averaged specular levels (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 the contrast between the specular amplitude and the lateral value at 451, range from 4.5 dB (Area 4) to 21 dB (Area 3). The 4.5 dB relative specular amplitude for Area 4 denotes a weak, or lack of, coherent specular reflection, suggestive of a hard and rough Author's personal copy G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 0 BS (dB) -10 Area 5 -20 -30 Area 2 0 20 40 60 S103 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 seafloors show distinctive shapes (Fig. 12) and A–F parameters (Table 5). The muddy seafloor gives the weakest reflectivity on the whole angular range, with a modelled specular level A ¼  14 dB (Table 6). Gravelly and sandy seafloors give similar specular levels of A¼  9 dB. All the specular angular extents have similar values (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 (typically 6 dB) is well above the separability given by the relative 1 dB measurement capability of the EM300. Likewise, the modelled transitory levels (E) have distinctive reference level values for gravel (  18 dB), sand ( 21 dB), and mud (  25 dB), but the angular extents have comparable values (F¼ 161 and 181). The angular extents of the specular and transitory levels do not seem to help in deciphering between different grain sizes. Incidence Angle 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. 5. Discussion 5.1. Qualitative interpretation seafloor (Augustin et al., 1997; Hughes Clarke et al., 1997; Le Chenadec et al., 2007). The angular extent of the specular reflection is relatively stable around 51, suggesting medium to highly scattering interfaces associated with inhomogeneous material. No very narrow specular ( o21) was found comparable to those observed on soft, homogeneous seafloor in the deep-sea basins (Augustin et al., 1997). The modelled angular extents of the specular reflection are in the range 3.6–9.51. In any case, the angular full width of the specular peak cannot be smaller than the sounder beam width (21 for the EM300), which gives the angle measurement resolution. Over the eight reference areas, the modelled Lambert references (C in Eq. (3)) range from  31 to  17 dB (Table 5). This 14 dB range indicates important variations in seafloor roughness or sediment volume scattering across the reference areas. The transitory levels for the modelled BS (E) range from  21 to 15 dB, with an average angular extent (F) ranging from 10.11 to 16.61. These values differ from both the angular extent of the modelled specular reflection (3.6–9.51) and the angular width of a Lambert law (451), confirming the relevance of using a transitory level 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) is only perceptible in the Cook Strait data (Tables 5 and 6). In order to further assess the relationship between BS angular profiles and grain size, we selected a small number of samples with average j parameters characteristic of mud, sand, and gravel (Tables 3 and 6). We purposefully used an over-simplified model with three end-member characteristics corresponding to soft, medium, and rough seafloor. The number of classes is not specifically important as it can be statistically estimated at a later stage for segmentation purposes (Lucieer and Lucieer, 2009). The BS angular response models were established for the three sediment types following the methodology described in Section 3.3, for a small area assumed to be representative of a typical grain size as they were selected from homogeneous imagery regions around the samples (Fig. 12). The areas cover between 4 and 20 km2 around the sample locations, which is large compared with the sample size, but are required so that a wide angular range (ideally 0–701) is available to model the full BS angular response. This is mitigated by selecting several areas with The backscatter imagery provides particular fine details that cannot be obtained from the bathymetry alone; the imagery can provide additional information on the origin, nature, and structure of landforms. The combined use of backscatter and microtopography processing enhances the interpretation of fine scale geological and topographical features visible in EM300 multibeam data, such as post-glacial scouring, erosional landforms and carbonated platform concretions (175137’E–41138’S, PaB in Fig. 5), and on canyons floor (OC in Fig. 5). The strong reflectivity on the floor of the Opouawe Canyon suggests gravel deposits, indicating active sediment delivery. Relicts of the last glacial transgression are identified on the outer shelf with ripples marks, gravel lags, and sediment waves observed in the Campbell Bank (Fig. 13). This latter example shows that the backscatter imagery provides even finer details than from the microtopography. An E–W trending,  50 m-high scarp is well imaged in the bathymetry and corresponds to strong backscatter on the scarp, suggesting a rough seafloor, most likely associated with gravelly material. Immediately north of the fault scarp, Ntrending black and white lineaments emphasize seabed features related with contrasting seafloor deposits, but with no topographic expression. By contrast, small depressions visible in the bathymetric image in the NW quadrant of Fig. 13 do not exhibit any contrasting backscatter change, suggesting a gentle change in seafloor composition. A similar situation can be seen on Pahaua Bank, where a strongly reflective patch at the shelf break (175137’E 41138’S, Fig. 5) shows fine details of morphological and geological origin associated with carbonated concretion recognised by Law et al. (2010). 5.2. Use of GSAB for seafloor characterisation While theoretical models of BS angular response may prove useful for characterising the seafloor locally, their benefits for quantitative analysis of regional datasets are limited. There are two main reasons for this situation. The first reason is that, by definition, accurate physical models are formulated to account for extremely complex processes (e.g., Guillon and Lurton, 2001) requiring a high number of input parameters. This makes inversion results possibly ambiguous as different input parameters may provide similar effects. In the Author's personal copy S104 G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 -5 Calibrated Reflectivity -10 -15 BS (dB) -20 -25 -30 -35 -40 -45 -80 -60 -40 -20 0 20 40 60 80 60 80 -5 -10 Models -15 BS (dB) -20 -25 -30 1 North East 2 North 3 Sandwave field 4 West 5 South 6 Central South 7 West (Inhomogeneous) 8 South west -35 -40 -45 -80 -60 -40 -20 0 20 40 Incidence Angle 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. worst case, the inversion may be technically impossible. Moreover, the higher the complexity, the greater the specialisation of a model for a given seafloor type. The second reason is that the measured BS may be dominated by non-anticipated phenomena such as sediment layering, accidental highly random scatterers, shortscale 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 a restrained range of seafloor characteristics; a single physical modelling approach cannot be valid on both, e.g., fine-grain sediments, 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 Author's personal copy From the Gradistat sediment grain size classification software (Blott and Pye, 2000). Based on the visual interpretation of Fig. 5. c Specular-to-oblique contrast ¼A  C from Eq. (2). d Mean BS45 calculated between 431 and 471 incidence angles on both starboard and port sides. b a  7.2 4.51  20.3 2.2  19.2 12.81 13  23.3 7 2.8  3.9 9.51  16.6 2.1  15.4 11.21 13  19.77 2.3  12.9 4.81  17.4 2.5  16.6 16.61 4.5  21.4 7 2.3  9.8 5.31  30.8 2.2  20.6 15.01 21  34.3 7 3.0 Specular level Angular extent Lambert reference Lambert decrement Transitory level Angular extent STOCc BS45d  8.5 4.21  28.4 2.2  18.8 14.91 19.8  29.9 72.7  8.7 3.61  21.7 1.9  17.9 10.11 12.9  25.0 7 6.3 No. of samples Backscatter imageryb A B C D E F A–C BS45 43 Moderately weak Sediment charactera The six independent parameters (A–F) are generated from modelling of the angular response of the backscatter intensity. BS values are in dB re. 1 m2. A 71 dB absolute uncertainty is assumed on all BS values (see text). 2 Moderately strong  11.3 3.61  18.3 2.2  15.5 12.81 7  22.2 7 2.3 11 Moderately weak to weak  5.5 3.91  18.8 2.8  15.8 10.31 13.3  23.8 72.7 SE facing slope, smooth Muddy gravel, med-fine sand, very fine sand Morphology 26 Weak 15 Very weak 45 Strong Gravel, very coarse sand concretion, muddy sand 24 Moderately weak Gravel, sand, pebbles, shelly muddy sand Basin wall central Basin wall south Smooth deep basin floor Coarse to very coarse including pebbles and boulders 43 Very strong Very shallow rough Sandy to gravelly mud, some pebbles Sediment wave field, basin floor Mud, med-fine sand Flat, shallow, smooth Firm grey mud, med-fine sand, shingle Smooth shallow shelf Muddy sand, very fine gravel 7 6 5 4 3 2 1 Area Table 5 Geological, geomorphological and backscatter characteristics of the eight test areas identified in the Narrows Basin. 8 G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 S105 seafloor model over a wide area. Therefore the correlation of a set of physical descriptor values with a regional dataset becomes less valuable. This line of reasoning argues in favour of using a more empirical modelling approach focused on a simplified description of the seafloor characteristics, restricted to the identification of major trends but able to cope with strong local variations. In using GSAB, we are attempting to describe physical observations through a restricted set of parameters. The approach is sufficiently versatile to accommodate the diversity of geological environments, while remaining accurate enough for a goodquality correlation to experimental data. Our purpose is not to retrieve quantitative values of seafloor physical or geological properties, but rather to determine descriptor values making it possible to classify the seafloor inside a limited set of classes that may be ground-truthed. Obviously, the accuracy of these descriptors cannot compete with the refinement of the input parameters 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 applying the BS signal to enhance geological interpretation. The gravelly, sandy, and muddy GSAB patterns can be used as basic references to provide a first-order interpretation on the average grain size across a large seafloor area. We apply this approach to the Narrows Basin using the GSAB patterns generated in the eight reference areas (Figs. 11 and 14). At low-incidence angles (0–151) a close correspondence (e.g. in the lobe angular width) between the mud, sand, and gravel GSAB profiles and those of the eight areas is difficult to assess because of the closeness of all profiles. The correspondence is much clearer around 451 incidence angle, e.g., for Area 8 where the BS angular profile compares well with the gravel BS profile (Fig. 14). This is shown numerically by the parameters A¼  11 dB and C ¼  18 dB modelled for Area 8 comparing well with A ¼  9 dB and C ¼  18 dB for gravel (Tables 5 and 6). These observations suggest that the continental shelf on the west side of the Narrows Basin likely consists of coarse material. The strongest BS facies in the Narrows Basin is observed in the south basin floor (Area 5) with a modelled specular A ¼  4 dB and Lambert reference C¼  17 dB. The specular is higher than for the gravel profile (A¼  9 dB), suggesting very coarse material in Area 5. A smooth interface between water and soft sediment would have also generated a strong specular, but with low B and C values, which is not observed in Area 5. Overall the GSAB profiles for areas 4, 5, and 8 show the highest average BS around the 451 incidence angle, and hence can be interpreted as representing areas with the coarser material. Areas 6 and 7, along the western flank of the Narrows Basin, show very similar characteristics (A ¼  7 and 6 dB, C ¼  20 and  19 dB) and are intermediate between the sand and gravel profiles at intermediate angles. The specular levels for the gravel and sand are identical (A¼  9 dB) and hence not discriminating. However, the three BS level parameters (A, C, and E) are higher for Area 6 than for Area 7, suggesting coarser deposits in the south than in the central basin wall. Interestingly, Area 3 is largely dominated by sediment waves and presents the weakest BS response, with Lambert reference C¼  31 dB and specular A¼  10 dB. In details, the BS varies across dunes and sediment waves, whereas on a regional scale the homogenised reflectivity suggests on the average fine sandy–mud material on the floor of the Narrows Basin. The angular extent of the specular reflection (B) is an indicator of the interface roughness. The maximum and minimum values for B are observed for areas 5 (9.51) and 1 (3.61), respectively (Table 5). These are consistent with coarse to very coarse material including pebbles and boulders retrieved from Area 5, whereas Author's personal copy S106 G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 Table 6 Six parameters (A–F) generated from BS angular responses models for characteristic mud, sand, and gravelly samples collected throughout Cook Strait. Samples Gravel (%) Sand (%) Mud (%) Specular level Specular angular extent Lambert reference Lambert decrement Transitory level Transitory angular extent n A* B C* D E* F Gravel Sand Mud Mud with underlying sediments C118 C125 C126 C129 100 0 0  9.1 4.91  17.9 2  17.5 16.21 C61 C62 C90 C229 10.65 83.25 6.1  9.1 5.91  23.9 2  20.9 17.91 U630 Q222B 0 1.5 98.5  13.6 5.51  29.1 2  24.5 16.21 0 6.9 93.1  6.1 5.11  27.3 2  20.7 12.41 In dB re. 1 m2. A 71 dB absolute uncertainty is assumed. −10 Mud + underlying Sed Mud Sand: 1.5% Mud: 98.5% Gravel: 10.6% Sand: 83.3% Mud: 6.1% Gravel Sand BS (dB) −20 −30 Mud Gravel: 100% Sand: 0 % Mud: 0 % −40 −50 0 50 −50 0 Incidence Angle 50 −50 0 50 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. fine sand was retrieved from Area 1. Areas 4 and 5 show contrasting values for the specular (  13 and  4 dB, respectively), and a marked difference in the angular extent of the specular (4.81 and 9.51), which suggests that the floor of the Narrows Basin (Area 5) has a substantially higher water–sediment impedance contrast and interface roughness than the small basin on the continental shelf (Area 4). The similar Lambert reference (C¼  17 dB) for areas 4 and 5 indicates similar sediment characteristics in terms of volume homogeneity, likely to be coarse 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 beam width to generate an accurate enough descriptor. The data also indicate that the transitory level (E) may be characteristic of large-scale seafloor roughness, as when E is high and contrasting with A and C as in areas 3 (E¼  21 dB) the rugosity is a particular case as it includes sediment waves. We infer that the transitory (E and F) components are indicators of a scale rugosity in the order of magnitude (  100–500 m) of the sediment wave wavelength, rather that at the scale of local slope, in relation with the sounder signal footprint to be considered in the specular behaviour modelled by the facet theory (Brekhovskikh and Lysanov, 1982). 5.3. Quantitative BS for geological mapping From the eight reference areas in Cook Strait, we show that the six GSAB parameters can be used to quantitatively characterise BS for a variety of environments. However, these data are still too complex to develop a reflectivity map at a regional scale. We propose using the BS at 451 incidence angle (noted BS45) as a parameter representing the overall BS behaviour. In conventional MBES surveys, a swath width (aperture of ca. 601) covers approximately four times the water depth, and the 451 incidence angle represents the median value on both sides of the specular. For instance, the minimum and maximum calibrated BS values acquired (all angles put together) during the Tan0105 survey are  55 and 18 dB, with a mean of 25 dB (Table 4). The abnormally high BS values at specular level are due to the presence of a coherent reflection process superseding the backscattered return and are unreliable. At grazing angles (68–721), the BS is irregular Author's personal copy G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 S107 qualitative images comparable globally, thus enabling comparison from different regions. BS45 maps, however, still have limited classifying power other than relative contrast, since the information inherent to the angle response has disappeared. This is exacerbated if the sounder absolute calibration is not available; then only the contrast between various responses can be exploited. We propose using the relative specular level with regard to the oblique regime average level, given by A–C or Specular-To-Oblique Contrast (STOC), as an indicator of both the impedance contrast and the interface roughness. STOC, as B, D, F, and (E–C), have the significant advantage of being independent of the sounder absolutelevel 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 of 16.5 dB. A scatter plot of STOC vs. C for the eight reference areas (Fig. 15) suggests that these parameters may provide some classifying potential as they show a good inter-class separation. This is promising as class separation is a requisite for segmentation and classification algorithms to run effectively (e.g., Lucieer and Lucieer, 2009). Enabling robust class separation from BS measurements increases the rationale of using BS in habitat mapping. 6. Conclusion 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. due to strong angle dependence and poor signal-to-noise ratio and is not always validated. BS45 is therefore the most suitable to represent the BS in one synthetic value. The backscatter map, normalised at the BS45 value for survey Tan0105 (Fig. 6B), shows clearly separated regions: sufficiently homogeneous when considered individually (BS45 dynamics is 6 dB over the sediment waves). The contrast between the reference areas is sufficient for a fairly accurate visual classification (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 application to habitat mapping. At a regional scale in Cook Strait BS45 can discriminate distinct environments, which have different geological or micro-topographic properties. It therefore provides a tool for geological interpretation. The use of BS45 as a reference value is of course conditioned by the availability of the sounder calibration in absolute level. This will have to be validated from further studies, but we suggest that BS45 be used systematically for building compensated sonar images where one reflectivity value has to stand for one homogeneous seafloor facies. Such an approach would have the significant benefit of providing The development of a quantitative automated seafloor mapping tool has been a priority for geologists since acoustics was first used to map the ocean floor. In this paper we have refined the BS signal interpretation and isolated some of its complex relationships with the substrate. It is well established that BS is controlled by the sediment grain size, the surficial heterogeneity and the small-scale topography, and therefore relates to substrate composition and roughness. Absolute amplitude calibration is required for quantitative analysis whereas statistical compensation of the BS angular response is a prerequisite for qualitative interpretation. Both the calibrated and statistically compensated data are required for a complete quantitative and qualitative interpretation of the seafloor composition and interface characteristics. The BS angular response also provides insights into the seafloor primary sedimentary (grain size) and geomorphological (seafloor roughness) properties at both local and regional scales. Qualitatively, the backscatter data in the Cook Strait region are characterised by weak, homogeneous reflectivity patterns over the continental shelf and deep Hikurangi Trough indicating fine sediment, likely mud or sandy mud, and strong reflectivity on the canyons floors indicating gravel deposits and active sediment delivery. Strongly reflective spots on the continental shelf highlight carbonate concretions and erosional landforms. Backscatter imagery and bathymetry data together provide a tool for enhanced interpretation of fine scale structures that may not necessarily have geomorphic relief but have distinctive surface textures or roughness. In the Narrows Basin, the backscatter pattern of strong reflectivity on sediment wave’s troughs and weak reflectivity on the crests indicate that the BS variations across the sediment waves are associated with gravelly deposits at their base and fine sandy deposits at their top. In some areas, the backscatter data display a quasi-periodical structure emphasising fine sediment waves otherwise undetected in multibeam bathymetry. Modelling of the backscatter signal as a function of the incidence angle enables us to provide quantitative information on the seafloor. A Generic Seafloor Acoustic Backscatter (GSAB) improved model is used to define six physically significant parameters (A–F) for a number of homogeneous and distinctive areas in Cook Strait. The six quantitative parameters (A–F) for Author's personal copy S108 G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 -5 -10 -15 Gravel d BS (dB) Mu -20 -25 Sand -30 1 North East 2 North 3 Sediment wave field 4 West 5 South 6 Central South 7 Inhomogeneous West 8 South west -35 -40 -45 0 20 40 Incidence Angle 60 80 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). 25 3 A-C (STOC) 20 2 15 1 7 6 5 10 8 5 4 0 -40 -35 -30 -25 -20 -15 BS45 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. muddy, sandy, and gravelly seafloors are used as reference to provide information on the seafloor composition. We propose using two parameters to simplify the management of the multi-parameter models, and to define a classification of the average seafloor characteristic adapted to the regional scale relevant for many habitat mapping studies: the BS value acquired at 451 (BS45) incidence angle as a stable indicator of the interface sediment type, usable either as an absolute descriptor or as a comparison tool, depending on the sounder calibration status; and the specular-to-oblique (STOC) contrast, which provides an indication of the behaviour of interface roughness through its impact on the specular reflection. STOC has the added advantage of being independent of absolute data level calibration. Acknowledgments This project was funded by the New Zealand Foundation for Research Science and Technology (FRST) through their pro- gramme Consequence of Earth Ocean-Change (C01X0702). Thanks go to Hugues de Longevialle, Ifremer’s Direction of International Relations, for supporting this NIWA-IFREMER collaboration. Sediment analyses were performed at NIWA by Lisa Northcote and Melanie Herrmann. Joshu Mountjoy (NIWA) commented on an early version of the manuscript, and Richard Pickrill (Geological Survey Canada) commented on the reviewed version. Figures were generated using the Sonarscopes and GMT software. References Anderson, J.T., Van Holliday, D., Kloser, R., Reid, D.G., Simard, Y., 2008. Acoustic seabed classification: current practice and future directions. ICES Journal of Marine Science 65 (6), 1004–1011. doi:10.1093/icesjms/fsn061. Augustin, J.M., Dugelay, S., Lurton, X., Voisset, M., 1997. Applications of an image 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- Author's personal copy G. Lamarche et al. / Continental Shelf Research 31 (2011) S93–S109 bottom: examples of SOPACMAPS 3 and ZoNéCo 1 cruises. Marine Geophysical Researches 18 (2–4), 459–486. Augustin, J.M., Lurton, X., 2005. Image amplitude calibration and processing for 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 framework for 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 package for the analysis of unconsolidated sediments. Earth Surface Processes and Landforms 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 203755320 81pp. Buckingham, M.J., 2000. Wave propagation, stress relaxation, and grain-to-grain shearing in saturated unconsolidated sediments. Journal of the Acoustical Society of America 108 (6), 2796–2815. Carter, L., 1992. Acoustical characterisation of seafloor sediments and its relationship 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 from the deep seafloor with Sea Beam. Journal of the Acoustical Society of America 79 (2), 316–331. Ehrhold, A., Hamon, D., Guillaumont, B., 2006. The REBENT monitoring network, a spatially integrated, acoustic approach to surveying nearshore macrobenthic habitats: 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 range analysis of acoustic themes from Stanton Banks Ireland: a link between visual interpretation and multibeam echosounder angular signatures. Applied Acoustics 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-frequency backscattering angular response of gassy sediments: model/data comparison from the Eel River margin. California Journal of the Acoustical Society of America 111 (6), 2621–2631. Foote, K.G., 1980. Importance of the swimbladder in acoustic scattering by fish: a comparison of gadoid and mackerel target strengths. Journal of the Acoustical Society 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 the Acoustical Society of America 117 (4), 2013–2027. Goff, J.A., Olson, H.C., Duncan, C.S., 2000. Correlation of side-scan backscatter intensity 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: the equivalent input backscattering strength model. Journal of the Acoustical Society of America 109 (1), 122–132. Hellequin, L., Boucher, J.M., Lurton, X., 2003. Processing of high-frequency multibeam echo sounder data for seafloor characterization. IEEE Journal of Oceanic Engineering 28 (1), 78–89. doi:10.1109/JOE.2002.808205. Hellequin, L., Lurton, X. and Augustin, J.M., 1997. Postprocessing and signal corrections for multibeam echosounder images. Oceans Conference Record (IEEE), pp. 23–26. Hicks, D.M., Shankar, U., 2003. Sediment yield from New Zealand rivers. NIWA chart, Miscellaneous series No. 79. National Institute of Water and Atmospheric Research, Wellington, NZ. Hughes Clarke, J.E., Danforth, B.W., Valentine, P., 1997. Areal seabed classification using backscatter angular response at 95 kHz. In: High Frequency Acoustics in Shallow Water. NATO SACLANT Undersea Research Centre, Lerici, Italy. Hughes Clarke, J.E., Mayer, L.A., Wells, D.E., 1996. Shallow-water imaging multibeam sonars: a new tool for investigating seafloor processes in the coastal zone and on the continental shelf. Marine Geophysical Researches 18 (6), 607–629. S109 Jackson, D.R., Briggs, K.B., 1992. High-frequency bottom backscattering: roughness versus sediment volume scattering. Journal of the Acoustical Society of America 92 (2 I), 962–977. Jackson, D.R., Ishimaru, A., Winebrenner, D.P., 1986. Application of the composite roughness model to high frequency bottom backscattering. Journal of the Acoustical Society of America 79 (5), 1410–1422. Kongsberg, 1997. EM300 operator manual, N.160719/C. Kongsberg Simrad AS Horten, 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 biogeochemical variability of a New Zealand deep-water methane cold seep during an integrated 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 angular dependence of backscattering distributions. In: Proceedings of the IEEE Oceans—Europe 2005. Le Chenadec, G., Boucher, J.-M., Lurton, X., 2007. Angular dependence of K-distributed sonar data. IEEE Transactions on Geoscience and Remote Sensing 45 (5), 1224–1235. doi:10.1109/TGRS.2006.888454. Le Gonidec, Y., Lamarche, G., Wright, I.C., 2003. Inhomogeneous substrate analysis using EM300 backscatter imagery. Marine Geophysical Researches 24, 311–327. doi:10.1007/s11001-004-1945-9. Lewis, K.B., Carter, L., Davey, F.J., 1994. The opening of Cook Strait: interglacial tidal scour and aligning basins at a subduction to transform plate edge. Marine Geology 116 (3/4), 293–312. Lucieer, V., Lucieer, A., 2009. Fuzzy clustering for seafloor classification. Marine Geology 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 for swath bathymetric sonar. International Hydrographic Review 4 (2), 24–37. Lurton, X., Augustin, J.M., Dugelay, S., Hellequin, L., Voisset, M., 1997. Shallowwater seafloor characterization for high-frequency multibeam echosounder: image segmentation using angular backscatter. In: Pace, N. (Ed.), Proceedings of SACLANTCEN Conference CP-45, High Frequency Shallow Water Acoustics. Maı̂tre, H.,, 2001, Traitement des images de radar a synthe se d’ouverture. Paris: Hermes Sciences Europe Ltd, 351pp. McRea Jr., J.E., Greene, H.G., O’Connell, V.M., Wakefield, W.W., 1999. Mapping marine habitats with high resolution sidescan sonar. Oceanologica Acta 22 (6), 679–686. Medialdea, T., Somoza, L., León, R., Farrán, M., Ercilla, G., Maestro, A., Casas, D., Llave, E., Hernández-Molina, F.J., Fernández-Puga, M.C., Alonso, B., 2008. Multibeam backscatter as a tool for sea-floor characterization and identification 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 of submarine canyons across an active margin: Cook Strait sector of the Hikurangi 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 for bottom backscatter and forward loss. In: Proceedings of the IEEE OCEANS ’89 Proceedings, pp. 1168–1175. Mulhearn, P.J., 2000. Modelling acoustic backscatter from near-normal incidence echosounders—sensitivity analysis of the Jackson Model, DSTO-TN-0304. DSTO Aeronautical and Maritime Research Laboratory, Melbourne Victoria 3001 Australia, 36 pp. Novarini, J.C., Caruthers, J.W., 1998. A simplified approach to backscattering from a rough seafloor with sediment inhomogeneities. IEEE Journal of Oceanic Engineering 23 (3), 157–166. Pratson, L.F., Edwards, M.H., 1996. Introduction to advances in seafloor mapping using sidescan sonar and multibeam bathymetry data. Marine Geophysical Researches 18, 601–605. Ryan, W.B.F., Flood, R.D., 1996. Side-looking sonar backscatter response at dual frequencies. Marine Geophysical Researches 18 (6), 689–705. Stoll, R.D., 1985. Marine sediment acoustics. Journal of the Acoustical Society of America 77 (5), 1789–1799. Van Dissen, R.J., Berryman, K.R., 1996. Surface-rupture earthquakes over the last ca, 1000 years in the Wellington region, New Zealand, and implications for ground shaking hazard. Journal of Geophysical Research 101 (B3), 5999–6019. Wright, D.J., Heyman, W.D., 2008. Introduction to the special issue: marine and coastal GIS for geomorphology, habitat mapping, and marine reserves. Marine Geodesy 31 (4), 223–230. doi:10.1080/01490410802466306.