Underwater hyperspectral imaging as a tool for benthic habitat mapping

Since the 1950s, continuous technological advances have gradually improved our ability to map the seafloor and its associated benthic habitats. Historically, seafloor mapping has largely been carried out acoustically, using techniques such as echo sounding and sonar imaging. These techniques are able to cover large areas, and can provide valuable information related to bathymetry, sedimentology and the distribution of large biogenic structures. To provide insight into finerscale benthic properties (e.g., species distribution and community composition), other techniques must, however, be employed in addition. Traditionally, physical sampling (e.g., trawling, dredging and grab sampling) has widely been employed for these purposes, but in recent times, optical remote sensing has emerged as a viable non-intrusive alternative. Examples of optical remote sensing techniques that can be applied in relation to seafloor mapping include video recording, digital photography, and most recently, underwater hyperspectral imaging (UHI). The goal of this thesis is to evaluate the potential of the latter as a mapping tool for benthic habitats. As opposed to conventional digital cameras, which render colors using a red (R), a green (G) and a blue (B) waveband within the visible light spectrum, hyperspectral imagers quantify colors as contiguous spectra. This provides a substantially improved foundation for color-based identification and mapping of biogeochemical seafloor targets. In the presented work, it is demonstrated that a considerable biological color diversity can be found among benthic organisms. It is also shown that this diversity not necessarily is well represented in RGB imagery, and that particularly suitable targets for UHI surveys include a variety of echinoderms, arthropods, cnidarians, mollusks and sponges, as well as brown, green and red macroalgae. Through a series of examples and papers, it is further demonstrated that underwater hyperspectral imagers can be deployed on a range of sensor-carrying platforms, all of which have associated benefits and limitations. Moreover, the thesis covers several relevant data processing steps related to optical correction, georeferencing and classification of underwater hyperspectral imagery. Although the focus of the thesis is on marine biological applications, the presented work also features case studies of archaeological wreck sites. These examples illustrate that UHI may be used interdisciplinarily, with actors from multiple scientific fields involved. Ultimately, the potential role of UHI in future marine research is discussed in context with other currently employed seafloor mapping techniques. Overall, the presented findings suggest that UHI may serve as powerful tool for detailed studies of areas ≤1,000 m2, but that more user-friendly software solutions for hyperspectral data processing likely must be developed if the technique is to be adopted by end users without an extensive technological background.