Background

Species distribution models (SDMs) aim to constrain the response and distribution of organisms to climate change. However, these models are built on ecological monitoring records spanning decades rather than the hundreds of thousands of years and broader climatic landscapes that species have persisted in. This means any model based on these records is highly overfit and the predicted response is biased to the current climatic window. SDM predictions therefore have previously unrecognised and unaccounted for uncertainties.

Palaeontology and historical ecology can address this, using fossil records, museum collections and archival data to extend baselines far beyond living memory. Advances in digitisation and AI-driven data extraction are now unlocking vast amounts of previously inaccessible “dark data”, enabling richer, longer-term ecological datasets. These deeper records can feed into ecological models to capture and reduce uncertainty in SDM predictions.

PhD Project

Marine ecosystems face escalating pressure from climate change, yet our ability to forecast their future is fundamentally limited by short-term datasets. This introduces unaccounted-for uncertainties into ecological models and limits their usefulness for informing policy.

This PhD will tackle this challenge by applying data mining to ‘dark data’, generating long-term datasets for two important marine groups, arthropods and corals. This will quantify and address unaccounted-for uncertainties in present ecological modelling approaches.

AI-based methods (e.g. text and image mining) will be developed and applied to extract ecological data from diverse sources at scale. This will be used to build robust ecological baselines spanning the Quaternary (last 2.5 million years), giving a long-term view of how these groups have responded to past environmental changes. This record exists scattered across historical survey reports, museum collections, and fossil archives but it is largely inaccessible, undigitised, or unstructured.

Uncertainty will be quantified by comparing predictions for datasets with and without long-term records. Evaluation of differences in prediction interval, extrapolation error, and hindcast skill against known climate transitions will validate the method and provide a framework that can be applied to risk projections for marine biodiversity and the ecosystem services they underpin across a range of taxonomic groups.

These baselines will feed into the development of new species distribution models, incorporating both taxonomic and functional ecology data, to generate improved, more confident predictions of how arthropods and corals will respond to future climate scenarios.

This project will be co-supervised by Dr Lewis A. Jones at the Natural History Museum, London, an expert in corals and ecological modelling, and provide access to large collections of material including modern, historical and fossil.

Applicant Profile

This project would suit students with a strong quantitative and computational background in a field such as environmental/Earth science, ecology, biology, palaeontology, computer science, or a related discipline. The student should be comfortable working with data in R or Python and keen to develop these skills further in an interdisciplinary research setting.

Essential is a genuine interest in applying data science and machine learning approaches to ecological and palaeontological questions. Prior experience in data analysis, statistics, or coding is highly desirable, as much of the project involves developing AI-driven tools to extract and analyse large, complex datasets.

No prior experience with fossils is required, though an interest in using the fossil record and deep-time perspectives to understand present-day biodiversity change would be an advantage. Similarly, a background in ecology (marine or otherwise) is welcome but not essential. What matters most is enthusiasm for combining ecological thinking with modern computational methods to answer questions about how species and ecosystems respond to environmental change over long timescales.

Other Information

– Dr Stephen Pates’ personal website (Primary Supervisor): https://www.stephenpates.com
– Dr Lewis A. Jones’ personal website (Secondary Supervisor): https://www.lewisajones.com
– Invertebrate Palaeontology Research Group:
https://www.ucl.ac.uk/mathematical-physical-sciences/earth-sciences/research-earth-sciences/research-groups-and-affiliated-institutes/palaeontology/invertebrate-palaeontology