Seabed biotope characterisation based on acoustic sensing

Abstract

The background to this thesis is Australia’s Oceans Policy, which aims to develop an integrated and ecosystem-based approach to planning and management. An important part of this approach is the identification of natural regions in regional marine planning, for example by establishing marine protected areas for biodiversity conservation. These natural regions will need to be identified on a range of scales for different planning and management actions. The scale of the investigation reported in this thesis is applicable to spatial management at 1 km to 10 km scale and monitoring impacts at the 10s of m to 1 km biotope scale. Seabed biotopes represent a combination of seabed physical attributes and related organisms. To map seabed biotopes in deep water, remote sensing using a combination of acoustic, optical and physical sensors is investigated. The hypothesis tested in this thesis is that acoustic bathymetry and backscatter data from a Simrad EM1002 multi-beam sonar (MBS) can be used to infer (act as a surrogate of) seabed biotopes. To establish a link between the acoustic data and seabed biotopes the acoustic metrics are compared to the physical attributes of the seabed in terms of its substrate and geomorphology at the 10s m to 1 km scale using optical and physical sensors. At this scale the relationship between the dominant faunal functional groups and both the physical attributes of the seabed and the acoustic data is also tested. These tests use data collected from 14 regions and 2 biomes to the south of Australia during a voyage in 2000. Based on 62 reference sites of acoustic, video and physical samples, a significant relationship between ecological seabed terrain types and acoustic backscatter and bathymetry was observed.These ecological terrain types of soft-smooth, soft-rough, hard-smooth and hard-rough were chosen as they were the most relevant to the biota in their ability to attach on or burrow into the seabed. A seabed scattering model supported this empirical relationship and the overall shape of backscatter to incidence angle relationship for soft and hard seabed types. The correlation between acoustic data (backscatter mean and standard deviation) and the visual and physical samples was most consistent between soft-smooth and hard-rough terrain types for a large range of incidence angles (16o to 70o). Using phenomenological backscatter features segmented into 10 common incidence angle bins from -70o to 70o the length resolution of the data decreased to 0.55 times depth. The decreased resolution was offset by improved near normal incidence (0o to 30o) seabed type discrimination with cross validation error reducing from 32% to 4%. A significant relationship was also established between the acoustic data and the dominant functional groups of fauna. Faunal functional groups were based on the ecological function, feeding mode and substrate preference, with 8 out of the 10 groups predicted with 70% correctness by the four acoustically derived ecological terrain types. Restricting the terrain classification to simple soft and hard using the acoustic backscatter data improved the prediction of three faunal functional groups to greater than 80%. Combining the acoustic bathymetry and backscatter data an example region, Everard Canyon, was interpreted at a range of spatial scales and the ability to predict the preferred habitat of a stalked crinoid demonstrated.Seabed terrain of soft and hard was predicted from the acoustic backscatter data referenced to a common seabed incidence angle of 40o. This method of analysis was selected due to its combined properties of high spatial resolution, consistent between terrain discrimination at the widest range of incidence angles and consistent data quality checking at varying ranges. Based in part on the research reported in this thesis a mid-depth Simrad EM300 multibeam sonar was purchased for use in Australian waters. A sampling strategy is outlined to map all offshore waters with priority within the 100 m to 1500 m depths.