Automated underwater image analysis reveals sediment patterns and megafauna distribution in the tropical Atlantic – Nature

 

Report on Automated Seafloor Analysis in the Tropical Atlantic and its Contribution to Sustainable Development Goals

Executive Summary

This report details the application of automated, Artificial Intelligence (A.I.)-driven workflows to analyze deep-sea seafloor imagery from the tropical North Atlantic. The primary objective was to characterize benthic habitats and megafaunal distribution patterns to provide critical data for marine ecosystem management. This work directly supports the United Nations Sustainable Development Goals (SDGs), particularly SDG 14 (Life Below Water), by enhancing scientific knowledge and developing research capacity for ocean conservation. Key findings reveal distinct habitat clusters driven by biogenic activity and significant regional differences in megafauna abundance, strongly linked to environmental drivers like depth, temperature, and food availability. The methodologies and findings presented here offer a scalable solution for monitoring marine biodiversity, contributing to SDG 9 (Industry, Innovation, and Infrastructure) through technological advancement and SDG 17 (Partnerships for the Goals) through collaborative scientific endeavor.

Introduction: Aligning with Sustainable Development Goals

The deep sea, Earth’s largest biome, remains largely under-explored, yet faces increasing pressure from anthropogenic and natural factors. The global decline in marine biodiversity necessitates urgent, globally coordinated efforts to monitor and protect these remote ecosystems. This research addresses this challenge by developing and deploying automated analysis tools, aligning with several key SDG targets:

• SDG 14 (Life Below Water): The core of this study is to “conserve and sustainably use the oceans, seas and marine resources.” By providing comprehensive baseline information on deep-sea habitats and species, this work directly informs efforts under Target 14.2 (sustainably manage and protect marine ecosystems) and Target 14.a (increase scientific knowledge and develop research capacity).
• SDG 9 (Industry, Innovation, and Infrastructure): The logistical and financial challenges of deep-sea exploration require innovative solutions. This report showcases the use of A.I. workflows as a technological innovation (Target 9.5) that enhances scientific research efficiency, transforming terabyte-scale image datasets into actionable insights for sustainable ocean management.
• SDG 13 (Climate Action): Understanding the influence of environmental drivers such as temperature and organic carbon flux on marine life provides crucial data on the impacts of climate change on deep-sea ecosystems, contributing to the knowledge base for climate action.

This report outlines a workflow that expedites the characterization of seafloor geology, biology, and ecology, providing a model for future large-scale marine ecosystem surveys essential for achieving the UN Decade of Ocean Science for Sustainable Development.

Methodology for Sustainable Ocean Monitoring

Study Area and Data Acquisition

The study was conducted in the tropical North Atlantic, offshore Mauritania, during the RV METEOR expedition M182. A total of 8,838 high-resolution still images were acquired across seven deployment stations using an Ocean Floor Observation System (XOFOS). The survey area was divided into a shallower Eastern region and a deeper Western region, allowing for comparative analysis of biodiversity patterns across different depths and geomorphological settings. This targeted data collection in a scientifically significant area contributes to filling knowledge gaps, a key component of SDG 14.a.

Technological Innovation for SDG 9 and SDG 14

To overcome the infeasibility of manual image annotation, a semi-automated, A.I.-driven process was implemented. This represents a significant transfer of marine technology (Target 14.a) and an enhancement of technological capabilities (Target 9.5). The workflow consisted of several key stages:

1. Image Visibility Improvement: Raw images were processed to correct for color distortion and uneven illumination, ensuring data quality and consistency for subsequent analysis.
2. Automated Seafloor Substrate Classification: An unsupervised machine learning approach was used. A pre-trained convolutional neural network extracted visual features from each image, which were then grouped using a K-means clustering algorithm. This automated sorting allowed for objective and repeatable habitat classification, essential for mapping ecosystems under SDG 14.2.
3. Automated Megafauna Detection: The pre-trained FaunD-Fast A.I. model was deployed to detect, localize, and provide initial classifications of megafaunal organisms. These “weak annotations” were then refined by human experts, drastically reducing manual effort while retaining scientific accuracy.
4. Data Standardization and Analysis: Absolute organism counts were converted to standardized abundances (individuals per square meter) to account for variable camera altitudes. This robust dataset was then analyzed using multivariate statistics to assess community composition and spatial autocorrelation analysis to identify biodiversity hotspots.

Key Findings and SDG Implications

Seafloor Habitat Characterization for SDG 14.2

The unsupervised classification successfully partitioned the seabed into seven distinct habitat clusters, with further sub-partitions indicating subtle local variability. A key finding was the identification of a clear gradient in sediment disturbance from biogenic activity. This gradient, visualized in feature space, distinguished between areas of low bioturbation and those with vigorous sediment reworking. This detailed habitat characterization provides an essential foundation for marine spatial planning and the identification of vulnerable marine ecosystems, directly supporting the sustainable management and protection goals of SDG 14.2.

Megafaunal Abundance, Diversity, and Links to SDG 14

The analysis revealed significant regional differences in marine life, providing critical insights into ecosystem function and health.

• Higher Abundance in the East: Megafaunal abundances were 14 times higher in the shallower Eastern region compared to the deeper Western region. This disparity is likely driven by a combination of factors including higher Particulate Organic Carbon (POC) flux, warmer water temperatures, and proximity to nutrient-rich upwelling zones.
• Dominant Taxa: The most abundant taxa identified were Foraminifera, Echinodermata, and Lebensspuren (traces of life), providing a quantitative baseline of the community structure.
• Implications for Conservation: Understanding these drivers of abundance is fundamental to predicting how ecosystems might respond to environmental changes, such as those related to climate change (SDG 13). This knowledge is vital for developing effective conservation strategies under SDG 14.

Spatial Distribution and Environmental Drivers (SDG 13 & 14)

The study confirmed that megafauna are not randomly distributed but cluster in specific areas, highlighting the importance of geomorphology in structuring deep-sea communities.

• Biodiversity Hotspots: Statistically significant hotspots of megafauna were concentrated in topographically complex features, such as the slopes of submarine canyons and the tops of seamounts. These structures create diverse microhabitats and unique niches that support richer biological communities.
• Influence of Environmental Drivers: Ordination analysis demonstrated that depth and temperature were primary drivers separating the Eastern and Western communities. In the deeper, more stable Western region, subtle changes in bathymetric derivatives (e.g., slope, ruggedness) had a pronounced influence on community structure.
• Contribution to Climate Science: By linking biodiversity patterns to climatic variables like temperature, this research contributes to our understanding of the deep sea’s vulnerability to climate change (SDG 13), reinforcing the need to protect these ecosystems (SDG 14).

Conclusion: Accelerating Progress on Ocean Science for Sustainable Development

This research successfully demonstrates that A.I.-powered workflows can dramatically accelerate the analysis of marine imagery, providing timely and robust data for deep-sea ecosystem assessment. The integration of these innovative tools into marine science is critical for making progress on multiple Sustainable Development Goals.

• For SDG 14 (Life Below Water), this approach provides the scalable, high-resolution data needed to monitor biodiversity, map habitats, and identify areas for protection, thereby strengthening the scientific basis for conservation and sustainable use.
• For SDG 9 (Industry, Innovation, and Infrastructure), the development and application of these automated workflows represent a significant technological advancement that enhances research capacity and efficiency.
• For SDG 17 (Partnerships for the Goals), this work, conducted through international collaboration and with data shared on open platforms, exemplifies the cooperative approach needed to address global ocean challenges.

Automated image analysis is an indispensable tool for the UN Decade of Ocean Science for Sustainable Development. By efficiently transforming vast datasets into actionable knowledge, these technologies empower scientists and policymakers to better understand, manage, and protect our planet’s vital marine ecosystems for generations to come.

Analysis of Sustainable Development Goals in the Article

SDGs Addressed in the Article

• SDG 14: Life Below Water – The article directly addresses the conservation and sustainable use of oceans by studying deep-sea biodiversity, habitats, and ecosystem health.
• SDG 9: Industry, Innovation, and Infrastructure – The research focuses on developing and applying innovative AI technologies and automated workflows to enhance scientific research capabilities for marine data analysis.
• SDG 17: Partnerships for the Goals – The study is an example of international scientific collaboration, knowledge sharing, and technology transfer for environmental science.

Specific SDG Targets Identified

SDG 14: Life Below Water

1. Target 14.2: By 2020, sustainably manage and protect marine and coastal ecosystems to avoid significant adverse impacts, including by strengthening their resilience, and take action for their restoration in order to achieve healthy and productive oceans.
   • Explanation: The article contributes to this target by providing foundational data for the management and protection of deep-sea ecosystems. The research characterizes seafloor habitats, assesses megafaunal distribution, and investigates environmental drivers and the impact of biogenic activities (“sediment disturbance”). This information is crucial for understanding ecosystem health and identifying areas that may require protection. The abstract notes that such characterization “provides crucial insights into the health and resilience of our oceans.”
2. Target 14.a: Increase scientific knowledge, develop research capacity and transfer marine technology… in order to improve ocean health and to enhance the contribution of marine biodiversity.
   • Explanation: The study is a direct application of this target. It uses advanced imaging and AI to increase scientific knowledge about “largely under-sampled” deep-sea biodiversity in the tropical North Atlantic. It explicitly discusses the development and transfer of marine technology by adapting an AI workflow from the Pacific to the Atlantic, stating one objective was “to investigate the generalizability of the two workflows when presented with dataset from a completely different area and geological setting.” The data is also made publicly available to enhance global research capacity.

SDG 9: Industry, Innovation, and Infrastructure

1. Target 9.5: Enhance scientific research, upgrade the technological capabilities of industrial sectors in all countries… including… encouraging innovation.
   • Explanation: The core of the article is about enhancing scientific research through technological innovation. It addresses the challenge that “modern seafloor imaging surveys typically generate thousands of images that are infeasible to manual annotation.” The solution presented is the development and deployment of “automated workflows” using AI and machine learning to “expedite the processing and annotation of images,” thereby upgrading the technological capability of marine science research.

SDG 17: Partnerships for the Goals

1. Target 17.6: Enhance North-South, South-South and triangular regional and international cooperation on and access to science, technology and innovation and enhance knowledge sharing on mutually agreed terms.
   • Explanation: The research exemplifies international cooperation. The study was conducted by German institutions (GEOMAR, Kiel University) during an expedition in the tropical North Atlantic, offshore Mauritania. The “Acknowledgements” section mentions the REEBUS project funded by the German BMBF. Furthermore, the commitment to knowledge sharing is demonstrated by the open-access publication status and the sharing of datasets on the BIIGLE portal, as mentioned in the “Data availability” section.

Indicators for Measuring Progress

Indicators for Target 14.2

1. Megafaunal abundance and diversity: The article provides quantitative data on the number and types of species.
   • Explanation: The study detected “10,189 megafaunal organisms belonging to 13 taxa groups” and calculated abundances, noting they were “14 times higher in the shallower Eastern region.” This data serves as a direct indicator of biodiversity levels in the surveyed area.
2. Spatial distribution of megafauna and habitat maps: The research produced maps showing where organisms and different habitats are located.
   • Explanation: The article uses choropleth maps (Fig. 6) to visualize the “spatial distribution of megafauna” and identifies “hotspots, coldspots and outliers” (Fig. 7). It also classifies the seafloor into “seven clearly distinct clusters” (habitats), which are essential metrics for ecosystem-based management.

Indicators for Target 14.a

1. Development and transfer of marine technology: The article details the creation and application of specific AI models.
   • Explanation: The development of the “FaunD-Fast” model and the “semi-automated classification of seafloor substrate” workflow are tangible outputs. The successful application of these tools, originally developed for the Pacific, to a new dataset from the Atlantic serves as an indicator of technology transfer and generalizability.
2. Publicly accessible marine datasets: The study makes its data available to the global scientific community.
   • Explanation: The “Data availability” section explicitly states that the datasets “can be found online in BIIGLE here https://annotate.geomar.de/projects/65.” This action is a direct indicator of increasing access to marine scientific data.

Indicators for Target 9.5

1. Development of innovative workflows for data analysis: The article’s main outcome is a novel, automated process for analyzing image data.
   • Explanation: The entire methodology, from image enhancement to AI-based classification and detection, represents an innovative technological process. The article concludes that “automated image analysis workflows have the capacity to efficiently extract actionable insights from terabyte-scale seafloor imagery.”

Indicators for Target 17.6

1. International scientific collaborations and publications: The project involves multiple institutions and results in a shared publication.
   • Explanation: The project was a collaboration conducted during the “RV METEOR cruise M182” and involved researchers from GEOMAR and Kiel University, funded by a German grant. The resulting open-access scientific paper is a clear indicator of knowledge sharing.

Summary Table of SDGs, Targets, and Indicators

SDGs	Targets	Indicators Identified in the Article
SDG 14: Life Below Water	14.2: Sustainably manage and protect marine and coastal ecosystems.	Quantification of megafaunal abundance and diversity (10,189 organisms, 13 taxa). Mapping of spatial distribution, including hotspots and coldspots of biodiversity. Classification of seafloor into distinct habitat types (seven clusters identified).
	14.a: Increase scientific knowledge and transfer marine technology.	Development and application of AI models (FaunD-Fast) for marine research. Testing the generalizability of technology from the Pacific to the Atlantic. Publication of datasets (8,838 images) on the open BIIGLE platform.
SDG 9: Industry, Innovation, and Infrastructure	9.5: Enhance scientific research and encourage innovation.	Creation of automated workflows to analyze terabyte-scale image datasets. Demonstrated efficiency gains over infeasible manual annotation methods.
SDG 17: Partnerships for the Goals	17.6: Enhance international cooperation on and access to science, technology and innovation.	International research expedition (RV METEOR cruise M182) involving German institutions. Publication of findings as an open-access article under a Creative Commons license.

Source: nature.com