How AI is transforming our understanding of the deep sea
The ocean covers over 70% of the world’s surface with 90% of that being the deep sea. When it comes to reaching areas of the deep sea, however, it is extremely difficult and poses many challenges. This is felt in many industries including the research sector where traditional deep-sea research methods such as dive teams, manned submersibles and robot vehicles are often met with challenges. The intense environmental conditions of the deep sea restrict opportunities for human exploration while its vast size produces overwhelmingly large data to be analysed. This is where the use of artificial intelligence (AI) becomes very useful.
Artificial intelligence is a rapidly growing field that has advanced oceanography over the past few years. Artificial intelligence is computer software that mimics human intelligence to perform tasks such as reasoning, learning and analysing. This tool is transformative for oceanography, enhancing our ability to explore, study and protect our vast, unknown ocean, particularly the deep sea. In this article, we will examine how AI will increase our understanding of deep-sea biodiversity, help us map and explore the deep-sea floor, and enhance oceanography research techniques.
Understanding Deep-sea Biodiversity
The deep sea is one of the most biodiverse ecosystems in the world. According to the NOAA, there are an estimated 1 million species in the ocean (excluding microorganisms) with roughly two-thirds of these species yet to be discovered, the majority of which are found in the deep ocean. The unknowns of deep-sea biodiversity are highlighted in the recent discovery of over 100 new deep-sea species in a research expedition exploring the underwater seamounts off the coast of Chile. Major discoveries such as this emphasise how much more is yet to be discovered about the biodiversity found in the deep ocean and its overall function within the ecosystem.
AI-driven technology is changing this. For example, AUVs (automated underwater vehicles) are being programmed with Machine Learning integrated tracking (ML tracking) algorithms to track and follow deep-sea species for long-duration observations. The AUVs use advanced 3D stereo trackers and multi-object detectors to detect what species they are recording and autonomously follow the animals for up to 5 hours. This imagery has allowed deep-sea scientists to study the behaviour of different species in more clarity than ever before. In addition to this, the extent and continuity of the data can uncover migration routes and breeding patterns to allow for a better prediction of future movements of deep-sea species.
The integration of AI into identifying deep-sea species and their behaviour supports the enforcement of better conservation and management of the deep sea. By understanding the feeding and breeding patterns of deep-sea species, conservation regulations such as Marine Protected Areas can be proposed and established to impose the sustainability of the ecosystem from fisheries and deep-sea mining. AI-driven models can also be used to simulate management scenarios to guide policymakers in making the most informed decisions. Balancing ecological sustainability with economic decisions regarding fisheries and habitat management will help implement the most effective policies for the protection of deep-sea species.
AI-Driven Innovation in Deep Sea Floor Mapping
In 2014, Malaysia Airlines Flight 370 disappeared after departing with 239 passengers aboard. Despite military and scientific efforts as well as huge financial incentives from multiple countries to explore the Indian Ocean, they were unable to find the Boeing 777. The failure of the search highlighted that we still understand very little of the topography of our ocean floor. However, mapping the deep sea is extremely challenging. The deep sea is a vast, rugged and unpredictable terrain. As a result, traditional remotely human-operated vehicles are impractical for scouring the depths of the deep sea.
The use of AI-enabled drones is helping change this. Advanced AI algorithms can provide underwater drones with autonomous navigation and the ability to recognise their surroundings with high precision. Each machine can make self-sufficient decisions to avoid obstacles and navigate through seamounts, ocean trenches and hydrothermal vents. The AI-powered drones also use advanced sensors and high-quality underwater cameras to capture images and data of the deep-sea terrain. This data is used to create detailed maps and can also help to identify essential marine ecosystems.
Furthermore, traditional research methods require automated underwater vehicles (AUVs) to re-surface so new instructions can be processed. This process is not only time-consuming but also disrupts data continuity. AI machines on the other hand, instantly respond to on-board data inputs and can accommodate changes in the research interest without the need for human intervention.
The use of these AI machines is revolutionary for not only mapping the deep sea but oceanography overall. Projects such as Seabed 2030 are on a mission to have mapped the entire ocean floor by 2030. The outcome of this mission would be profound in our understanding of the ocean. Knowing the seafloor’s shape is fundamental to our improvement in tsunami prediction, fishing resources, climate modelling, telecom services and underwater geo-hazards. However, as of August 2024, only 2.7% of depths between 5750-11,000 metres have been mapped by an 800 x 800 grid cell size. Thus, to achieve the Seabed 2030 mission, AI will be imperative.
The use of AI-powered machines will dramatically increase the efficiency of deep-sea exploration and provide a more detailed picture of the deep-sea than ever before. This can be used to better our ability to understand and protect the biodiversity and ecosystem of this unknown habitat.
The Impact of AI on Data Processing Efficiency and Accuracy
The field of deep-sea research is labour-intensive, time-consuming, and financially demanding. This is due, in part, to the ocean's enormous vastness and the copious amounts of data that follow. Advancements in deep-sea research techniques have provided scientists with big data. Analysing thousands of datasets is a laborious affair for scientists. For example, the Monterey Bay Aquarium Research Institute (MBARI) has recorded 28,000 hours of deep-sea video and gathered over 1 million deep-sea images over the past 35 years.
The development of AI algorithms has accelerated data processing significantly. AI has made it possible to automate the analysis of big data at a scale of efficiency that humans could never match. As a result, tasks that once took months can now be completed in weeks or sometimes days. MBARI has recently piloted its machine-learning model to label videos of the deep-sea species taken by remotely operated underwater vehicles. This process is estimated to have reduced human effort by 81 per cent and increased the research output tenfold. Thus, deep-sea scientists can focus their expertise and time on the interpretation and application of the results. In addition, the use of AI algorithms can also increase the accuracy of data analysis by stopping human error. This improves the quality of data analysis significantly.
AI algorithms have also been developed as predictive models. Scientists can now use AI models to forecast the health of our deep sea. Using complex and historical datasets, AI models can predict the likelihood of environmental stresses such as marine heat waves and climate change. These predictions and data are integral for the future implementation of conservation measures. Using predictive data is a cost-effective and time-saving form of evidence for the implementation of more suitable measures and policies to better protect the future of the deep sea.
Final Thoughts
With the earth on the brink of catastrophic climate warming and disaster, using AI to guide us to discover and better protect this unique environment will be imperative. AI-driven autonomous underwater vehicles can be used to explore deeper and with more precision than ever before. AI can also process large volumes of data and distinguish clear patterns more efficiently than humans ever could.
With the development of AI still in its infancy, the scope and scale at which it will advance deep-sea research further is exponential. The increasing use and development of AI-driven technology, deep learning models and AI algorithms will advance our ability to reveal the mysteries of the deep sea. These findings will be paramount to the implementation of effective conservation strategies and policies and the sustainability management of the deep sea for the benefit of future generations.