The aim of the study was to explore different approaches of feature selection, extraction and reduction from backscatter angular response curves for a relatively complex seabed. The study area is located at Point Cloates along the coast of central Western Australia where water depths range from 6 to 200 m and is characterised by extensive sandy bedforms, flat sandy seabed and numerous reefs. A Simrad EM3002 300 kHz sonar system was used to collect multibeam data across an area of 281 km2 in 2008. A series of radiometric and geometric corrections were applied to the backscatter data. The angular response curves were derived separately for port and starboard by averaging 100 pings along the ship track. Seabed sediment texture was characterised from 90 samples that were analysed for grain size properties (gravel, sand, mud%) and classified into six sediment classes. Co-located towed-video transects from the survey were used to identify areas of rocky seabed. Four approaches of processing the angular response curves have been explored. The first approach used all effective beam angles (4o to 51o) with a manual feature selection method in the modelling process. The second approach used principal component analysis to condense the 48 variables into four (explained 99% data variance). The third approach extracted nine parameters from two domains of the angular response curves including slope, intercept, orthogonal distance and mean. The fourth approach derived continuum-removed angular response curves. Probability Neural Network was used as the classifier. The classification results show that the continuum removal approach performed the best with an overall accuracy of 73% when classifying the seven seabed classes (Figure 1).When merging the six sediment classes into four, which results in five seabed classes, the performance was improved for all approaches.

The Carnarvon Shelf Survey (SOL4769, GA survey #0308) was conducted on the R.V. Solander in collaboration with the Australian Institute of Marine Science between 12 August and 15 September 2008. The survey was operated as part of the Surrogates Program of the CERF Marine Biodiversity Hub. The survey was completed under a Memorandum of Understanding between GA and the AIMS and represents the first of three surveys planned under this agreement. The objective was to collect high-quality, accurately co-located data to enable the robust testing of a range of physical parameters as surrogates of patterns of benthic biodiversity. Underwater video footage and still images were collected from 122 stations from water depths of 13-125 m, although video quality varies among transects and some still images were not of suitable quality for analysis. Images from the still camera can be found in 'Image Library', and images from towed video screen captures can be found in 'Tow Video Stills'. Image files from screen captures are named according to area (1 = Mandu, 2 = Point Cloates, 3 = Gnarloo) followed by the station number and video identifier (TVA1). For example, 2_032TVA1 would represent a towed video transect from Station 32 at Point Cloates. See GA Record 2009/02 (Geocat #68525) for further details. Video footage was recorded to mini DV tapes, and copied to digital format. The original mini DV tapes are archived at AIMS-WA.

Geoscience Australia carried out marine surveys in Jervis Bay (NSW) in 2007, 2008 and 2009 (GA303, GA305, GA309, GA312) to map seabed bathymetry and characterise benthic environments through colocated sampling of surface sediments (for textural and biogeochemical analysis) and infauna, observation of benthic habitats using underwater towed video and stills photography, and measurement of ocean tides and wave generated currents. Data and samples were acquired using the Defence Science and Technology Organisation (DSTO) Research Vessel Kimbla. Bathymetric mapping, sampling and tide/wave measurement were concentrated in a 3x5 km survey grid (named Darling Road Grid, DRG) within the southern part of the Jervis Bay, incorporating the bay entrance. Additional sampling and stills photography plus bathymetric mapping along transits was undertaken at representative habitat types outside the DRG. Family per sample matrix generated by aggregating species level data in JBinfauna_species (25Oct10).xls using the information in JBinfauna_Taxa_info (25Oct10).xls.

The identification of marine habitats based on physical parameters is increasingly important for marine reserve design, allowing characterisation of habitat types over much wider areas than is possible from often patchy biological data. Marine management zones often contain a wide array of physical environments, which may not be captured in the biological sampling effort. The mismatch between biological and physical information leads to uncertainty in the application of bio-physical relationships at the broader management scale. In this study, a case study from northern Australia is used to demonstrate a methodology for defining uncertainties which result from the extrapolation of bio-physical associations across areas where detailed biological data is absent. In addition, uncertainties relating to the interpolation of physical data sets and that resulting from the cluster analysis applied to the physical data are calculated and mapped, providing marine managers with greater robustness in their analysis of habitat distributions.

Anthropogenic threats to benthic habitats do not pose an equal risk, nor are they uniformly distributed over the broad depth range of marine habitats. Deep sea benthic environments have, by and large, not been heavily exploited and most are in relatively good condition. In contrast, shelf and coastal habitats, and deep ocean pelagic fisheries, have been exploited extensively and human impacts here are locally severe. A critical point is that anthropogenic threats do not act in isolation; rather, they are cumulative and the impacts are compounded for every affected habitat. In general, the impacts of humans on benthic habitats is poorly understood. Habitat mapping provides condition assessments and establishes baselines against which changes can be measured. GeoHab scientists ranked the impacts on benthic habitats from fishing as the greatest threat, followed by pollution and litter, aggregate mining, oil and gas, coastal development, tourism, cables, shipping, invasive species, climate change and construction of wind farms. The majority of authors (84%) reported that monitoring changes in habitat condition over time was a planned or likely outcome of the work carried out. In this chapter the main anthropogenic threats to benthic habitats are reviewed in relation to their potential impacts on benthic environments.

The term 'surrogacy' is used in habitat mapping with reference to the biophysical variables that can be mapped with a quantifiable correspondence to the occurrence of benthic species and communities. Surrogacy research can be defined as an empirical method of determining which easily measured characteristics best describe the species assemblage in a particular space and at a particular time. These characteristics act as predictors (with some known probability and uncertainty) for the occurrence of species assemblages in unexplored areas. Abiotic variables are, in general, more easily and less expensively obtained than biological observations, which is a key driver for surrogacy research. However, the suite of abiotic factors that exert control over the occurrence of species (its niche) is also a scientifically interesting aspect of ecology that provides important insights into a species evolution and biogeography. This chapter provides a review of surrogates used by case study authors and of the methods used to quantify relationships between variables.

A number of terms used in this book are derived from the fields of biogeography and benthic ecology and these are defined in the glossary; the reader is also referred to the works cited at the end of this chapter for further information. Many of the case studies presented in this book refer to habitat classification schemes that have been developed based on principles of biogeography and ecology. For these reasons a brief overview is provided here to explain the concepts of biodiversity, biogeography and benthic ecology that are most relevant to habitat mapping and classification. Of particular relevance is that these concepts underpin classification schemes employed by GeoHab scientists in mapping habitats and other bioregions. A selection of published schemes, from both deep and shallow water environments, are reviewed and their similarities and differences are examined.

The planktobenthos is an important area with unique environmental conditions and represents the immediate link between the benthos and the water column. Despite its close links with the seafloor, the planktobenthos has never been deliberately sampled concurrently with the benthos. We have developed a new method for planktobenthic sampling which allows concurrent collection of benthic and suprabenthic specimens. The Mounted Assembly for Planktobenthic Sampling (MAPS) uses a novel tri-layered net with a seafloor-triggered opening and closing mechanism attached to an epibenthic sled. The MAPS was deployed on the Carnarvon Shelf in Western Australia and was successful at separately sampling both benthic and planktobenthic fauna. A wide variety of epibenthic and infaunal animals were collected from the sled while planktobenthic animals such as mysids were identified from all three nets. The tri-layered net was particularly effective at collecting a broad range of planktobenthic organisms, including smaller fragile larvae and adults which may have otherwise been destroyed during collection in a single net. The number of species in planktobenthic and benthic samples was correlated, although the strength and significance of this relationship varied among taxonomic groups. Importantly, the MAPS can be modified for use on a wide variety of benthic sleds to target a range of organisms. The concurrent collection of planktobenthic and benthic biota will contribute to a range of research areas, including larval ecology, nutrient cycling, biogeography and surrogacy research.

This introductory chapter provides an overview of the book's contents and definitions of key concepts including benthic habitat, potential habitat and seafloor geomorphology. The chapter concludes with a summary of commonly used habitat mapping technologies. Benthic (seafloor) habitats are physically distinct areas of seabed that are associated with particular species, communities or assemblages that consistently occur together. Benthic habitat maps are spatial representations of physically distinct areas of seabed that are associated with particular groups of plants and animals. Habitat maps can illustrate the nature, distribution and extent of distinct physical environments present and importantly they can predict the distribution of the associated species and communities.

Autonomous Underwater Vehicles (AUVs) have only recently become available as a tool to investigate the biological and physical composition of the seabed utilizing a suite of image capture and high-resolution geophysical tools. In this study we trialled the application of an AUV, integrating AUV image capture with ship-based high resolution multibeam bathymetry, to map benthic habitats and biodiversity in coastal and offshore waters of SE Tasmania. The AUV successfully surveyed a plethora of marine habitats and organisms, including high-relief kelp-dominated rocky reefs to deep mid-shelf reef and sediments that are otherwise difficult to access. To determine the spatial extent of these habitats within a broader-scale context, the AUV surveys were integrated with larger scale multibeam mapping surveys. The data collected using the AUV significantly improved our understanding of the distribution of benthic habitats and marine organisms in this region, with direct application to the management and conservation of these environments. Integrating the AUV data with the largescale mapping data provided the opportunity to quantify the relationships between the biological and physical variables, and to use thise data to develop predictive models of biodiversity across the region.