Quantitative Comparison of Benthic Habitat Maps Derived From Multibeam Echosounder Backscatter Data

Abstract

In the last decade, following the growing concern for the conservation of marine ecosystems, a wide range of approaches has been developed to achieve the identification, classification and mapping of seabed types and of benthic habitats. These approaches, commonly grouped under the denominations of Benthic Habitat Mapping or Acoustic Seabed Classification, exploit the latest scientific and engineering advancements for the exploration of the bottom of the ocean, particularly in underwater acoustics. Among all acoustic seabed-mapping systems available for this purpose, a growing interest has recently developed for Multibeam Echosounders (MBES). This interest is mainly the result of the multiplicity of these systems’ outputs (that is, bathymetry, backscatter mosaic, angular response and water-column data), which allows for multiple approaches to seabed or habitat classification and mapping. While this diversity of mapping approaches and this multiplicity of MBES data products contribute to an increasing quality of the charting of the marine environment, they also unfortunately delay the future standardization of mapping methods, which is required for their effective integration in marine environment management strategies. As a preliminary step towards such standardization, there is a need for generalized efforts of comparison of systems, data products, and mapping approaches, in order to assess the most effective ones given mapping objectives and environment conditions. The main goal of this thesis is to contribute to this effort through the development and implementation of tools and methods for the comparison of categorical seabed or habitat maps, with a specific focus on maps obtained from up-to-date methodologies of classification of MBES backscatter data. This goal is attained through the achievement of specific objectives treated sequentially. First, the need for comparison is justified through a review of the diversity characterizing the fields of Benthic Habitat Mapping and Acoustic Seabed Classification, and of their use of MBES data products. Then, a case study is presented that compare the data products from a Kongsberg EM3000 MBES to the output map of an Acoustic Ground Discrimination Software based on data from a Single-beam Echosounder and to a Sidescan Sonar mosaic, in order to illustrate how map comparison measures could contribute to the comparison of these systems. Next, a number of measures for map-to-map comparison, inspired from the literature in land remote sensing, are presented, along with methodologies for their implementation in comparison of maps described with different legends. The benefit of these measures and methodologies is demonstrated through their application to maps obtained from the acoustic datasets presented previously. Finally, a more typical implementation of these measures is presented as a case study in which the development of two up-to-date classification methodologies of MBES backscatter data is complemented by the quantitative comparison of their output maps. In the process of developing and illustrating the use of methods for the assessment of map-to-map similarity, this thesis also presents methodologies for the processing and classification of backscatter data from MBES. In particular, the potential of the combined use of the spatial and angular information of these data for seabed classification is explored through the development of an original segmentation methodology that sequentially divides and aggregates segments defined from a MBES backscatter mosaic on the basis of their angular response content.