Subscribe to our newsletter
Get the latest news, events and developments from iAtlantic, straight to your inbox.
The use of cameras to observe the seafloor environment is in itself not new, but advances in imaging technology now provide unprecedented insights into life in the deep sea. This session will feature some of the latest camera technology developed for use on board iAtlantic’s deep-sea research expeditions: cameras that can image the deepest and darkest areas of the ocean, which – despite being far from land – are not immune to the impacts of climate change and human activities.
New hyperspectral camera systems can capture views of deep-sea ecosystems in incredible detail, enabling sophisticated scientific analysis of ecosystem structure through highly accurate image mosaicking and 3D reconstructions. This level of analysis enables us to better understand the functioning of deep-sea ecosystems and species, their relationship to the physical environment, and their vulnerability to environmental change or disturbance. Baited camera landers, deployed on the seafloor, allow us to observe the diversity, abundance and behaviour of scavengers in a particular area through time-lapse footage, giving us important information on ecosystem structure and diversity.
We will also demonstrate how imaging technology can be made far more accessible by reducing the cost, size and complexity of such systems and greatly improving their ease of deployment – for example, from small vessels such as fishing boats. This technology has great potential to be used not only for investigating the ecology of unknown areas of the ocean, but also in ecosystem assessment, ecosystem restoration, conservation actions, and a whole host of marine monitoring activities. The ease of deployment of low-cost camera systems means that they are suitable for use in community-led projects and in developing countries, and can help foster collaborative cross-sectoral working, for example between scientists and local fishing fleets, or local authorities and conservation initiatives. In this sense, we hope this event will appeal to participants from a range of sectors and across all geographic regions.
The new image processing and interpretation techniques that we will introduce show how technology is enabling us to obtain much greater scientific value from the huge volume of imagery collected during seafloor surveys. Novel visualisation techniques allow us to examine these data in greater detail and in an integrated way, so that it is easier to compare the characteristics of different areas, determine the level of change between repeat surveys, understand better the relationship between ecosystem/habitat distribution and environmental characteristics such as current flow, nutrient and oxygen levels, water temperature, and so on. At a time when environmental change presents a serious threat to the health of the ocean, new tools that allow us to extract greater value and insight from the data we collect at sea are critical to finding solutions.
The baited camera lander is deployed at the seafloor for a fixed period of time. A plate with a known weight of bait, like fish or squid, hangs at the bottom of the lander, with a camera supported by flashing lights pointing straight at it. Scavengers in the deep sea are attracted to the smell of the bait, and the camera takes pictures at regular intervals, documenting all the animals that eat the bait. A current meter mounted on the side of the lander records the direction and speed of the water passing by the lander over the time of the experiment. The lander is deployed for about 24 hours, after which an acoustic release allows the floats to bring the lander back to the surface for recovery. Data collected by the lander is used to determine the diversity of scavengers in the area, calculate their scavenging rate, and estimate the abundance of scavengers in the area using a mathematical model with the current speed and direction, and time of first arrival of the organisms.
Left: The baited camera lander, ready for action. Image © D. De Jonge, HWU. Below: a video guide to the camera lander, plus some time-lapse footage of scavengers taken with the lander on the iMirabilis2 expedition offshore Cabo Verde. [coming soon!]
Until recently, hyperspectral imaging data has been acquired mostly using passive sensors installed either on aerial platforms or satellites. These sensors use the sun as the light source and have limited use in studying underwater areas where sunlight is totally absent. As part of the iAtlantic project, a hyperspectral underwater camera
(UHI) is being tested, using an ROV platform, to extend the use of this technology to deep marine areas. It comprises a push-broom scanner that continuously records the intensity of light reflected from the seafloor in the spectral range between 378 and 800 nm. The seafloor is illuminated using two LEDs. Our main objective is to test the potential of underwater hyperspectral technology for deep-sea benthic habitat identification and their ecological status assessment.
The Azor Drift-Cam is an affordable and easy-to-use underwater video system for a rapid appraisal of benthic habitats, designed by researchers at IMAR/Okeanos (University of Azores). It uses the drift of the deploying vessel to ‘fly’ over the seabed and record images from the deep seafloor, in water depths up to 1000 m. This system was developed to be cost-effective, cover large areas in short periods of time, perform well over rough seafloor terrain, be operational from small vessels and have high chances of escaping lost long-lines – the most common fishing gear in the Azores region.
Left: The Azor Drift-Cam. Image © IMAR/Okeanos-UAz, Azor drift-cam
Exploring the Azores deep sea using a custom-made low-cost imaging tool: how much have we achieved with the Azor drift-cam? Extract from the iAtlantic Newsletter, January 2022 (PDF. 0.4MB)
Dominguez-Carrio et al. (2021) A cost-effective video system for a rapid appraisal of deep-sea benthic habitats: The Azor drift-cam. DOI: 10.1111/2041-210X.13617
Artificial interlligence (AI) is a data science toolkit that is rapidly evolving around deep learning methods that allow computers to play Go, fold proteins and understand images. In the oceans, AI can predict data to large spatial scales, increase the resolution of sparse data points, or detect and classify fauna in images. This kind of information extraction from images requires robust image annotation data, methods for FAIR sharing of images and annotations, and of course a lot of computing power – which is rarely available in the deep sea.
Left: Applying AI to underwater images needs proper tuning and training data so that, for example, a group of divers is not misclassified as a “vase”. Image © Timm Schoening.
Photogrammetry is a technique that allows the reconstruction of 3D high-resolution models based on imagery collection with a camera. For the past decade, it has been growing field for performing ecological or geological investigation in the deep sea using underwater vehicles. Photogrammetry allows us to accurately map the biology and the microtopography of the seabed with centimetre-scale resolution. We have results from an innovative study using a time series of 3D models to investigate the temporal variability of benthic assemblage distribution over 5 years on the Eiffel Tower hydrothermal edifice at Lucky Strike on the Mid-Atlantic Ridge.
Left: 3D reconstruction of the Capolinhos edifice. Image © Ifremer/MoMARSAT2020
Predictive habitat mapping is a term that encompasses a number of different modelling techniques, variously referred to in the current literature (sometimes with different emphases and meanings) as species distribution models, habitat suitability models or ecological niche models. All of them aim to predict the distribution of target species or habitats based on the correlation between presence records at known locations and environmental variables (e.g. depth, sediment type, temperature, etc) available in a GIS format, usually a raster. The growing available information about the seabed (high-resolution multibeam data, oceanographic variables from satellite data, etc.) together with the need to implement an ecosystem approach has made these models a key tool in a wide range of management applications, such as delineation of marine protective areas, mapping listed species or habitats, and forecasting the impact of climate change.
Left: Example of a seafloor habitat map around a seamount. Image © José Manuel Gonzalez