The world is a change planet, and both terrestrial and marine ecosystems are in flux. In our research, we mostly focus on marine ecosystem. Understanding and preserving deep-sea ecosystems is essential for marine conservation efforts in a rapidly changing climate. These ecosystems, which are among the most biodiverse and least explored environments on the planet, play a critical role in regulating ocean health, carbon cycling, and global climate stability. However, assessing biodiversity and ecosystem resilience in these remote environments presents unique challenges due to the vast areas to cover and the difficulty of manual observation.

How can deep learning help us?

An example of our work is a research project focusing on automating the classification of deep-sea biota using advanced deep learning models trained on remotely operated vehicle (ROV) image datasets (Deo et al, 2024). By creating detailed habitat maps through automated object detection, we aim to improve biodiversity assessments and provide valuable data for evaluating the health and resilience of deep-sea ecosystems. In collaboration with marine research institutions, we have contributed to the development of a comprehensive image classification dataset featuring nearly 4,000 images of deep-sea species across 33 distinct classes. These images were manually labeled using a human-in-the-loop process, ensuring high-quality annotations for training machine learning models.

Examples of the classes of organisms in our Deep Dive dataset (Deo et al, 2024)

Our work leverages state-of-the-art deep learning models such as ResNet, DenseNet, Inception, and Inception-ResNet to address the challenges posed by class imbalance in the dataset, which is a common issue in biological image classification. Benchmarking these models on our dataset demonstrated that Inception-ResNet provides the most promising results, achieving classification accuracies of up to 65% and AUC scores exceeding 0.8 for each class. These results highlight the potential of AI to accelerate marine biodiversity assessments and aid in the conservation of deep-sea ecosystems by providing scalable, automated solutions to analyze vast quantities of underwater imagery.

By improving our ability to monitor deep-sea habitats, we are better equipped to understand the impacts of climate change and human activity on these fragile ecosystems, ensuring that conservation efforts are guided by data-driven insights.

The geological record provides invaluable insights into how past climate systems operated under different atmospheric and oceanic conditions. We leverage the wealth of the geologic record to learn to predict future environmental impact based on past climate change.

For example, our work on Middle Jurassic carbonate ramp deposits in the Musandam Peninsula (United Arab Emirates) highlights how high-energy shallow marine environments generate a high probability of lateral facies variation (Hönig and John, 2015). By combining sedimentological data with stable isotope analyses, we mapped ancient carbonate facies and identified patterns of diagenesis caused by fluctuating sea levels impacting wave base. These findings have important implications for understanding how past sea-level variations were driven by global climate cycles — a pattern that continues today, with modern coastlines vulnerable to rising seas.

In the UAE, Jurassic carbonates show high degree of lateral variability in facies demonstrated quantitatively usine Markov-chain transition in our work (Hönig and John, 2015).

Our research extends beyond stratigraphy to include multi-proxy isotopic analyses that reveal connections between regional climate changes and global climatic events. A study of the Hampshire Basin (southern England) during the Middle Eocene shows how sea-level changes and ocean gateways influenced regional water temperatures and salinity levels. During this period, the English Channel acted as a critical connection between the warmer Atlantic Ocean and the cooler North Sea, with periodic openings and closings driven by glacio-eustatic fluctuations. By analyzing oxygen isotope values and ostracod assemblages, we demonstrated how climate fluctuations, such as the Middle Eocene Climatic Optimum (MECO), impacted regional environments (Marchegiano and John, 2022). These past climate events help contextualize modern ocean circulation changes and regional climate impacts caused by current global warming.

We used ostracodes and clumped isotopes to show evidence of intermittent connection between the Artic Ocean and the Atlantic in the course of the Middle Eocene Climate Optimum (MECO). Work published in Marchegiano and John, 2022

Our isotopic work also extends to high-resolution climate reconstructions from the latest Miocene to the early Pliocene, a period of major Earth system changes. By studying benthic foraminiferal isotope records from the equatorial Pacific and comparing them to the North Atlantic, we identified orbital-scale climate variability and uncovered evidence for two distinct climate states during this time. The Messinian Salinity Crisis (MSC) and the Late Miocene Carbon Isotope Shift (LMCIS) are pivotal events that occurred during periods of reduced global ice volumes and moderate cryosphere sensitivity to orbital forcing. These findings are crucial for understanding the dynamics of Earth’s carbon cycle and how ice sheets respond to climatic changes, offering analogues for the cryosphere’s future behavior in a warming world (Drury et al, 2015).

Collectively, our research demonstrates that climate and environmental changes in Earth’s past were often driven by complex interactions between the carbon cycle, cryosphere, and ocean systems. These interactions are still at play today, but they are now compounded by human activities that are accelerating greenhouse warming and sea-level rise. By studying ancient climate transitions, we aim to provide long-term perspectives that can guide modern climate policies and help predict future climate thresholds. In this way, we align with the geological axiom that “the past may hold the key to our future” — past climate events reveal Earth system responses that can inform our understanding of ongoing changes and help society adapt to the challenges ahead.

Our research is not limited to fieldwork: we also use marine geology. We currently have funding from the National Environmental Research Council (NERC) for a research project focused on acquiring new seismic data to investigate the ancient reefs and atolls of the Mid Pacific Mountains and Emperor Seamount Chain in the mid-Pacific Ocean. This project aims to unlock crucial insights into how carbonate-producing organisms, such as corals, responded to significant oceanic and climate changes over the past 100 million years. By studying drowned atolls, known as guyots, we can reconstruct the history of reef development, global sea-level fluctuations, and ocean chemistry transitions — all of which are highly relevant to understanding the potential impacts of modern climate change on marine ecosystems. The idea for the project stems from work done by a former PhD student within our group (Elyamani et al, 2022).

The project centers on two key research questions. First, we aim to investigate how carbonate production adapted to changes in ocean chemistry during the transition from a ‘calcitic sea’ to an ‘aragonitic sea’, which occurred between the Early Cretaceous and the Oligocene. This shift, driven by major climatic changes, transformed the mineralogy of marine carbonates from predominantly calcite to aragonite, fundamentally altering how reefs grew and developed over time. Understanding how ancient reefs adapted to such dramatic environmental changes can provide critical insights into how modern reef systems might respond to ongoing ocean acidification and rising CO₂ levels caused by anthropogenic climate change. For instance, one of the questions we aim to answer is whether future changes in ocean chemistry will force carbonate producers, such as corals, to grow at different depths or change their reef-building behaviors, which could significantly impact the geometry and stability of modern atolls.

Proposed locations for the CARAPACE seismic survey

The second objective of this project is to explore global sea-level changes across the Cretaceous to Eocene period. The geological record preserved within these ancient reefs offers a unique opportunity to quantify past sea-level fluctuations, which is critical for predicting the rate and magnitude of future sea-level rise. Today’s accelerating ice sheet melt is leading to rapid sea-level rise, and by examining how sea-levels changed during past global warming events, we can improve climate models and sea-level projections. The carbonate reefs we are studying grew close to sea level, making them natural sea-level markers. By reconstructing the vertical positions of these reefs over millions of years, we can track past sea-level changes and better understand how warming climates impacted ice sheets and ocean levels in Earth’s history.

Our research involves conducting seismic reflection surveys on six different guyots in the mid-Pacific. This method allows us to image the internal architecture of ancient reefs by sending seismic waves to the seafloor and analyzing the reflected signals. These data will help us reconstruct the geometry and growth patterns of the reefs, revealing how they evolved during periods of oceanic and climatic transitions. This work will set the foundation for future deep-sea drilling campaigns, where we plan to recover rock samples from selected atolls to gain a more detailed understanding of paleoenvironmental conditions, reef growth dynamics, and sea-level changes.

In the long term, this research will contribute to our understanding of how marine ecosystems respond to changing climates and how sea-level rise has shaped Earth’s surface through time. By linking past environmental changes to present-day challenges, this project aims to provide valuable insights for modern climate adaptation efforts and global sea-level management strategies. As the planet faces unprecedented greenhouse warming, the study of ancient drowned reefs holds the potential to predict the future resilience of coral reef systems and better inform climate action policies.

Ultimately, the goal of CARAPACE will be to pave the way for future ocean drilling projects – here, the JOIDES Resolution photographed in the Gulf of Mexico (C. John, 2005).