Unravelling the environmental interactions of subsea communication cables using Autonomous Robotic Survey Data and Deep Learning

Prof Blair Thornton, Dr Mike Clare, Dr Justin Dix, Dr Brian Bett, Prof Susan Gourvenec
Rationale: 

The economy relies on uninterrupted use of a network of seafloor telecommunications cables that are typically no wider than a garden hose. >99% of all digital data traffic worldwide is transferred via >400 fibre-optic systems that span >1.8 million km across the seafloor; underpinning the Internet and £trillions/day in financial transactions. To support growing demand, we need more cables. Understanding how submarine cables interact with sensitive deep seafloor environments is critical in order to plan diverse and viable, environmentally-sensitive routes and responsibly decommission out-of-service cables.

 

However, the deep-seabed is poorly explored due to the large costs involved and this presents a monitoring challenge.  Studies so far have been rare, opportunistic and have used very different (non-standardised) methods. While these have found telecoms cables only have minor impact on deep-sea habitats, encrustation of recovered cables has contributed to our knowledge of deep-sea marine biota growth rates and observations have highlighted complex sheltering behaviours of mobile organisms1,2.

 

This project will demonstrate how next-generation technology can transform marine environmental assessments of difficult-to-detect interactions of seafloor infrastructure, using ultra-high-resolution seafloor and photographic data acquired from Autonomous Underwater Vehicles (AUVs), and state-of-the-art Deep Learning to assess the impacts of human activities on deep-sea ecosystems3

Methodology: 

This project benefits from rare, extensive and high-resolution AUV photographic datasets gathered at the Southern Hydrates Ridges seafloor cable observatory, located off Oregon, USA (780m depth; surveyed in 2018, 4 years post-installation), and at Haig Fras (HF) UK Marine Protected Area (MPA) (100m depth; surveyed decades post-installation). Each dataset consist of >20,000 seafloor images that have millimetric resolution and show diverse biota adjacent to seafloor cables and in far-field areas (hundreds of metres away) over multi-hectare regions with several kms of cable. These provide a unique opportunity to develop/test automated detection of seafloor features and organisms, perform extensive habitat mapping, and quantify statistical relationships between species distribution/diversity with cables, substrates and other species.

The project will: [1] Manually identify objects (organisms, substrates, cables) to train and validate automated identification algorithms. [2] Determine distribution, diversity and abundance of visible biota, substrates and cable infrastructure using the automated methods. [3] Analyse spatial relationships between different species and natural substrates and cable-related infrastructure to ascertain if there are statistically-significant differences in the levels of interactions. [4] Test the efficacy when translating these approaches using data from other sites and/or commercial cable route survey data provided by project-partner International Cable Protection Committee (ICPC).

 

Location: 
University of Southampton
Training: 

The INSPIRE DTP programme provides comprehensive personal and professional development training alongside extensive opportunities for students to expand their multi-disciplinary outlook through interactions with a wide network of academic, research and industrial/policy partners. The student will be registered and hosted at the UoS with access to the NOC. Specific training includes: Processing of AUV-acquired seafloor and photographic data (Thornton/Bett), Deep Learning and automated analysis (Thornton, Gourvenec/Dix/Bett), ecology/substrate/infrastructure characterisation (Clare, Dix, Gourvenec/Bett), statistical analysis (Thornton/Gourvenec/Bett), environmental aspects of subsea cables (Clare/Dix/Bett/CASE Partner ICPC), python/C++ programming (Thornton), annotation software Biigle2.0 and Squidle+ (Thornton/Bett) and GIS (Dix).

The project will develop methods that can be used in diverse environmental and subsea infrastructure assessments beyond seafloor cables. The analytical skills gained will be equally valuable for academic and industry careers e.g. environmental assessment, telecommunications, renewables, decommissioning. Opportunities to join offshore research campaigns are expected; hence offshore survival training will be provided.

Eligibility & Funding Details: 

Please see https://inspire-dtp.ac.uk/how-apply for details.

Background Reading: 

1. Ospar Commission, 2009. Assessment of the environmental impacts of cables https://qsr2010.ospar.org/media/assessments/p00437_Cables.pdf.

 

2. Carter, L., 2010. Submarine cables and the oceans: connecting the world (No. 31). UNEP/Earthprint. https://www.unep-wcmc.org/system/dataset_file_fields/files/000/000/118/original/ICPC_UNEP_Cables.pdf?1398680911

 

3. Yamada, T., Prugel-Bennet, A., Williams, SB., Pizarro, O., Thornton, B., 2021, GeoCLR: Georeference Contrastive Learning for Efficient Seafloor Image Interpretation, arXiv preprint arXiv:2108.06421