Data-driven modelling of the Southern Ocean biological carbon sink

Applications are invited from prospective students for a PhD scholarship at Imperial College London. The successful applicant will produce a novel machine learning tool and maps of krill distribution and their associated carbon sinks based on environmental and in-situ variables (e.g., drifter-derived quantities, such as flux, disturbance, etc.). Mapping krill-carbon hotspots will allow for characterising how fishing and climate change may be impacting the carbon sink and the wider Southern Ocean ecosystem, of which krill are an essential part. The candidate will be based in the Department of Life Sciences, at the Silwood Park Campus, supervised by Dr Emma Cavan, with co-supervision from Dr Adam Sykulski and Dr André Ribeiro Amaral in the Department of Mathematics.

Project details:

The Southern Ocean is one of the most important oceanic carbon sinks in the global oceans, both physically and biologically. Its productive and rich ecosystems help sink carbon to the deep sea, where it can be stored for centuries. Antarctic krill are one of the most abundant species on our planet and a key food source to many charismatic megafauna residing in the Southern Ocean. Research has highlighted their crucial role in driving carbon sequestration in the region though their faecal pellets. As krill are fished, it is imperative we understand exactly where they reside in the Southern Ocean, so we can advise krill management and appropriate fishing limitations can be set to protect the krill faecal pellet carbon sink. At present, it is not possible to map the abundance and distribution of krill dynamically and remotely.

In this Ph.D. a data-driven and machine learning approach to fuse heterogeneous data sources across multiple scales will significantly improve predictions of the distribution of Antarctic krill. The statistical challenge is significant, as these data sources, while producing vast databases in isolation, are multiscale in nature and are complex to integrate in one prediction methodology due to their non-linear interactions. To overcome this, the Ph.D. project will implement state-of-the-art and recently developed methodologies in spatial machine learning (e.g., amortised inference, a neural network-based class of methods where parameter estimation is performed through fast feedforward operations), while also integrating ideas from classical spatio-temporal statistics.

The main outcome of the project will be a novel machine learning tool and maps of krill distribution based on environmental and in-situ (e.g., drifter-derived quantities, such as flux, disturbance, etc.) variables. The Ph.D. project will also use the results of the krill distribution work to map krill-carbon hotspots while correctly propagating the estimation uncertainty arising from the different modelling steps, and characterise how fishing and climate change may be impacting the carbon sink and the wider Southern Ocean ecosystem, of which krill are an essential part. The results will be relevant to krill fishery management and can help guide spatial and temporal planning when implementing temporary fishery closures.

Requirements:

• A First-Class Honours Degree in environmental science, mathematics, oceanography, or closely related disciplines.
• A Masters level degree qualification, e.g. MSc/MRes.
• Excellent communication skills, including strong writing abilities

How to apply:

Applicants are recommended to contact Dr Emma Cavan (e.cavan@imperial.ac.uk) for further details, informal discussions and information about the project, prior to the closing date 10^th January 2025. Please include a 2-page CV.

Funding:

If successful, the studentship will provide funding for 3.5 years from October 2025, including tuition fees at the level of Home students and a living stipend.