Supervisors: This is a joint project between the University of Kent and Butterfly Conservation, and the PhD student will be supervised by a team with expertise in Statistics, Computing and Citizen Science.
University of Kent: Dr Eleni Matechou, Dr Marek Grzes, Dr Diana Cole, Prof Byron Morgan
Butterfly Conservation: Dr Emily Dennis, Dr Richard Fox
Training
The student will develop a strong, highly transferable skillset in data modelling and analysis using deep learning and other computational techniques applied to large unstructured data sets, with spatial and temporal replications. There will be ample opportunity for independent development, and the student will gain a transferable knowledge of modern data science that will be generally applicable in academia or industry.
Scientific background
At a time of biodiversity loss, including widely reported insect declines, monitoring changes in species’ populations and distributions is vital. To that effect, there is an ongoing growth of unstructured citizen science data, where species can be recorded by anyone, at any time and anywhere. The full potential of such sources of increasingly ‘big data’ for biodiversity monitoring has not yet been fully realized.
Analysing citizen science data of this nature presents unique challenges relating to their vast quantity and also associated sampling biases. Using cutting edge predictive modelling based on deep neural networks, this project will maximise the potential of these valuable datasets. Focusing on butterfly and moth data, we will enhance our understanding of these species’ phenology (flight periods), distribution and range dynamics, with the aim to help inform future conservation delivery and policy and better understand the drivers of species’ change. On the technical side, the project poses new research challenges that will require dedicated models based on deep neural networks and other methods for data analysis.
Research methodology
The student will develop new deep learning methods and associated efficient algorithms and apply them to large spatio-temporal citizen science data sets. The research will involve
- developing the first deep learning framework specifically tailored to citizen science data
- incorporating large numbers of environmental and remote sensing covariates (variables) to better describe drivers of species’ range and distribution change
- building methods for modelling phenology on varying spatio-temporal scales
- assessment approaches to determine reliability of inference for rare and/or under-recorded species from available data.
Research excellence
The student will be supervised by leading researchers in statistics and computing and will join the thriving Durrell Institute of Conservation and Ecology (DICE), recognised world-wide for its research and impact. DICE has recently been awarded substantial funding (£10m) to further support its growth. The School of Computing was recognised in the most recent Research Excellence Framework (REF) in 2021 with 100% of Computer Science and Informatics research classified as either 'world-leading' or 'internationally excellent' for impact. The PhD student will also be a member of the UK-wide National Centre for Statistical Ecology (NCSE), they will attend London Taught Course Centre training, and specialist seminars at the University of Kent, and they will present research results at a range of appropriate national and international conferences. There will be ample opportunity for independent development, with gaining transferable knowledge of modern data science and statistics.
Person specification
We seek a candidate with a strong quantitative or programming background, such as a degree in Mathematics, Statistics, Computer Science, Engineering, Quantitative Ecology, or closely related field.
Funding
The studentship is fully-funded for 3.5 years, covering any associated fees, (note only UK students are eligible). Stipend for 2024/25 is £19,237.00.
Applicants should follow the University of Kent’s online application process.
Please create an account and add your personal details as requested. Subsequently, you need to select your starting date (September 2024) and write your personal statement (see below). Choose "Other" for source of funding and "Definite" for funding. Add details of your qualifications, and then in the Research Information Tab, write "Dr Eleni Matechou" under supervisor and the title of the project ("Deep learning methods for large citizen science data sets") as the research topic. You do not need to add a research proposal.
As part of the process, you need to provide the following:
o details of your qualifications;
o two academic references;
o a personal statement
The statement must be maximum 500 words detailing (1) your reason for applying for a doctoral studentship (i.e, why do you want to pursue doctoral studies) and (2) your fit with the proposed project (how your educational/professional/personal background has prepared you well to undertake research in this topic).
Please email Dr Eleni Matechou ([Email Address Removed]) if you are interested in applying for the project or have any questions about the project or the application process.
Deadline for applications: 2nd of April
Interviews will take place on the week of the 15th of April via Microsoft Teams. Shortlisted candidates will be notified shortly after closing date.
Indicative relevant literature
Diana, A., Dennis, E. B., Matechou, E., & Morgan, B. J. T. (2023). Fast Bayesian inference for large occupancy datasets. Biometrics, 79(3), 2503-2515.
Johnston, A., Matechou, E., & Dennis, E. B. (2023). Outstanding challenges and future directions for biodiversity monitoring using citizen science data. Methods in Ecology and Evolution, 14(1), 103-116.
Wikle, C. K., & Zammit-Mangion, A. (2023). Statistical deep learning for spatial and spatiotemporal data. Annual Review of Statistics and Its Application, 10, 247-270.
Platanios, E. A., Al-Shedivat, M., Xing, E., & Mitchell, T. (2020). Learning from imperfect annotations. arXiv preprint arXiv:2004.03473.
Rhinehart, T. A., Turek, D., & Kitzes, J. (2022). A continuous‐score occupancy model that incorporates uncertain machine learning output from autonomous biodiversity surveys. Methods in Ecology and Evolution, 13(8), 1778-1789.
Continue with Facebook
