Proximal remote sensing of seafloor communities with underwater imaging techniques

Concerns regarding the effects of marine biodiversity loss on ecosystem services have brought into focus the importance of seafloor communities and their functional role in global oceanic processes. By area, the seafloor represents the largest ecosystem on earth. As such, it should be characterized in ways that are relevant to scientific, socio-economic, and conservation objectives. However, current approaches for studying seafloor communities, such as core sampling or underwater cameras, have shortfalls in their spatial, spectral, and/or temporal resolution that hinder the capacities of acquiring biodiversity or ecosystem process information, at the required spatio-temporal scales to track changes. This is a critical challenge for remote and difficult environments to work in, such as the Southern Ocean (SO) seafloor, resulting in a paucity of data from these regions. The study of the SO benthos is increasingly relying on underwater optical imaging techniques that increase scientific capacities to make observations, ranging from individual organisms to entire seascapes, across adaptable spatial and temporal scales. The adoption of Structure-from-Motion photogrammetry (SfM) and underwater hyperspectral imaging (UHI) have the capability of increasing SO benthic monitoring capacities, yet there are scarce examples, and data analysis bottlenecks are widespread. This Thesis harnesses the capacities of both SfM and UHI in efforts to increase the effectiveness of their implementation in the study of benthic biodiversity and ecosystem function. I focused my efforts on the acquisition of biological traits from two Antarctic benthic foundational species associated with reef-building and increasing biodiversity; dense aggregations of serpulid polychaetes, and crustose coralline algae.
Chapter 2 provides an extensive literature review of the challenges and research gaps of seafloor studies incorporating standard camera based Red-Green-Blue (RGB) sensors, with emphasis on the role of UHI and its potential to alleviate challenges in seafloor research. This review explains how the widely adopted RGB imaging systems used in biological surveys have now led to image annotation 'bottlenecks' in our human capability to extract meaningful biological information from the increasingly expanding seafloor imagery datasets. It highlights the use of underwater platform technologies and UHI to acquire seafloor imagery in remote and challenging environments, such as the deep or polar benthos, across a range of spatial scales (e.g., m to km). Novel applications of UHI can resolve image annotation bottlenecks and retrieve quantitative information on biodiversity and ecosystem function through the acquisition of hundreds of continuous narrow spectral bands (e.g., 2 nm). The review concludes by comparing how different studies harness the capacities of UHI, the types of methods used to validate observations, and the current challenges for accurate and replicable optical imaging surveys.
Chapter 3 demonstrates the use of small remotely operated vehicles and underwater SfM to measure biogenic structural complexity, a key benthic functional trait known to be associated with biodiversity, yet understudied in the SO. This study is the first to reconstruct in three?dimensions (3D) a unique serpulid polychaete reef in Ellis Fjord, Antarctica to retrieve structural complexity metrics. Results show that an increase in fractal dimension values, a widely used structural metric, are associated with higher densities of polychaetes that are providing biogenic structural complexity; fractal dimensions can be used to monitor and detect changes in seafloor 3D spatial patterns. This Chapter’s case study provides relevant biological insights, and the challenges associated with, integrating structural complexity metrics in Antarctic benthic surveys. In this Chapter I highlight the importance of reproducible workflows for acquiring biological and fine-scale structural metrics of benthic communities, aiming to facilitate data interoperability between Antarctic research campaigns. The tools and analysis pipelines developed in this Chapter offer clear recommendations to assist in the acquisition of structural and biodiversity metrics essential for marine protected area designation and seafloor management policies, as well as to improve our understanding of the role of biogenic structures in sustaining Antarctic biodiversity.
Chapter 4 investigates the use of close-range hyperspectral imaging (in-air) to study the spatial relationship between spectral indices and photosynthetic pigment content in Antarctic crustose coralline algae (CCA). These algae are a foundational group in marine ecosystems but are understudied in the Southern Hemisphere. Quantifying the content of photosynthetic pigments allows the detection of CCA responses to environmental stressors, such as changes in light regimes that are known to be modifying Antarctic seafloor communities. In this study, close?range hyperspectral imaging, DNA barcoding, and pigment extractions were used to analyse differences in the microspatial distribution of the light-harvesting pigment R-Phycoerythrin (R-PE) between specimens of lithogenic CCA Tethysphytum antarcticum (Hapalidiales), collected from two locations within the Ross Sea, Antarctica. Results reveal that two different spectral indices, namely the double derivative at 563 nm (R² = 0.7), and the area under the curve normalized by maximum band depth ranging from 543 to 583 nm (R² = 0.67), were the most sensitive to R-PE content. Distinct micro-spatial variability of R-PE spatial distribution was found and attributed to adaptive growth forms from the T. antarcticum phenotype at each location. I discuss the implications of considering CCA phenotypes in Antarctic functional biogeography and highlight the role of remotely operated vehicles assisting with sample collection in under-ice environments, helping to increase the extent of CCA studies around Antarctica. The findings of this Chapter demonstrate the effectiveness of hyperspectral imaging to quantify micro-spatial variability of R-PE content in Antarctic coralline algae in relation to their phenotypic plasticity. These findings demonstrate the capacity to non-invasively monitor CCA pigment content both in vivo laboratory studies, and potentially in situ surveys using UHI systems.
The findings from Chapter 4 were required an assessment in vivo to determine their applicability. In Chapter 5, a laboratory UHI system and data processing pipelines were developed to validate previous Chapter’s findings by assessing differences between dry and underwater spectral reflectance traits of CCA communities. The successful retrieval of CCA spectral traits both in vivo and using desiccated specimens corroborated Chapter 4’s results, suggesting its capacity to be scaled up in situ using underwater platforms and to be used as a proxy for CCA photosynthetic activity. Variable spectral signatures and microspatial R-PE patterns were found across CCA morphologies and different degrees of bleaching, presenting interesting research avenues where individual traits can be linked to a DNA identity (pending), to assess inter- and intra-specific variability. I explain how acquiring laboratory UHI measurements are essential to link benthic community measurements with lower or higher levels of biological organization in biodiversity monitoring. As a marine imaging technique, UHI still holds challenges to meet scientific data criteria, such as FAIR (Findable, Accessible, Interoperable, Reproducible). To address these challenges, I developed data processing pipelines that could facilitate the automation and reproducibility of the following UHI steps required in benthic research: (1) underwater reflectance conversion, (2) quantification of R-PE content and its microspatial variability, and (3) unsupervised acquisition of CCA community spectral signatures. The pipelines were built using Python and Jupyter notebooks aimed to increase reproducibility and automation of UHI datasets and could facilitate the training and adoption of UHI in seafloor monitoring campaigns.
Together, these research chapters provide novel biological insights from methodological advances of underwater optical imaging techniques for the study of marine ecosystems, particularly of understudied habitats such as Antarctic biogenic reefs and poorly understood organisms such as coralline algae. The use of photogrammetry and hyperspectral imaging techniques presented in this thesis can facilitate the study of seafloor communities, their functional traits, and their surrounding habitat. These metrics are essential in scientific studies, ecosystem management, and conservation policies. Furthermore, bio-optical imaging technologies have the potential to bridge observations across biological and spatial scales, helping to increase our understanding of the impacts of biodiversity loss and environmental change on seafloor ecosystem functioning.