PSD - Dr Eleni Matechou

1. Scalable and Adaptive Statistical Modelling for Environmental DNA Surveys of Biodiversity in China

Supervisor: Dr Eleni Matechou and Professor Silvia Liverani

Project description:

Understanding where species live and how their populations change is fundamental to conserving biodiversity. Yet gathering this information across large, remote regions—where most remaining biodiversity is found—is difficult using traditional surveys, which are labour-intensive and limited in scope. 

Recent advances in environmental DNA (eDNA) metabarcoding offer a transformative alternative: by analysing DNA in small samples of water, soil, or air, it is possible to detect hundreds of species quickly and non-invasively. This PhD project will focus on eDNA data from the Gaoligongshan mountains along the China–Myanmar border, a region of exceptional biodiversity. River and stream samples collected during a single month of fieldwork included DNA from hundreds of species, including 35 officially listed as threatened. These data represent a scalable and transferable method for determining species distributions in biodiverse regions, with the potential to inform conservation priorities nationally and globally [1]. 

The supervisory team has already made key contributions to statistical modelling of DNA-based survey data for single species [2, 3] and communities [4]. Building on this foundation, the PhD will develop new statistical approaches that address the unique challenges of eDNA data while remaining broadly applicable to other large-scale ecological datasets. 

A central aim is to create scalable models for high-dimensional eDNA data. Current approaches often struggle as the number of species and sampling points increases, but meaningful measurements of biodiversity across large areas require much larger datasets. The project will explore computationally efficient models, incorporating approximate inference and parallel computing where appropriate. These methods are also relevant to other fields where high-dimensional, noisy data are common, including epidemiology and environmental monitoring. 

Another innovation is adaptive and active learning for study design. By using models to guide where and when samples should be collected, the project will develop adaptive strategies that maximise information gain on species distributions from eDNA data. This will ensure that additional sampling in the Gaoligongshan mountains and beyond provides the most informative data, reducing costs while improving model accuracy. Such approaches are of broad methodological interest, bridging statistical theory, machine learning, and applied ecology. 

The project will also address variable selection and model interpretability for complex ecological data. Many factors influence species presence, including environmental covariates, spatial correlation, and inter-species interactions. The project will develop methods to identify key predictors while accounting for multi-species dependencies, enabling more interpretable and actionable ecological insights. 

By combining novel statistical methodology with eDNA data from China, this PhD will equip students with valuable skills in hierarchical modelling, computational statistics, and adaptive study design. Beyond applications in the Gaoligongshan region, the approaches developed will be relevant to biodiversity monitoring, ecological forecasting, and other fields requiring scalable inference from complex observational data. 

In summary, this PhD will advance statistical science while contributing directly to biodiversity research in China and beyond. It will provide a platform for students to develop expertise in methods that are both theoretically innovative and practically important for conservation and environmental monitoring. 

References:

[1] https://www.authorea.com/doi/full/10.22541/au.174412107.76832286 

[2] https://academic.oup.com/jrsssc/article/69/2/377/7058533?login=false 

[3] https://link.springer.com/article/10.1007/s42519-025-00477-9 

[4] https://www.tandfonline.com/doi/full/10.1080/01621459.2024.2412362 

Funding Notes:

Further information: 
How to apply 
Entry requirements 
Fees and funding

PhD Information Session 2026:
On Wednesday 14 January, we will be holding a short information session about PhD studies in Mathematics at QMUL. For full details about the event, please visit: https://www.qmul.ac.uk/maths/postgraduate/postgraduate-research/phd-information-session-2026/

As one of the UK’s most diverse universities, QMUL fosters an inclusive and supportive academic community.

The School of Mathematical Sciences is committed to the equality of opportunities and to advancing women’s careers. As holders of a Bronze Athena SWAN award, we offer family-friendly benefits and support part-time study. 

2. Integrating AI and Statistical Ecology: Bridging Automated Photo-Recognition and Inference of Population Demographics

Supervisor: Dr Eleni Matechou and Dr Kostas Papafitsoros

Project description:

Artificial Intelligence (AI) and computer vision have transformed biodiversity monitoring. AI-based re-identification (re-ID) (photo-recognition) offers a non-invasive, low-cost way to automatically and rapidly recognise individual animals from photographs exploiting natural markings, such as scales, stripes, spots or more subtle morphological characteristics. This technique expands the scale of ecological studies and citizen-science projects, with respect to estimating population size, survival, reproductive activity and recruitment, all central to conservation biology. 

However, AI outputs are never error-free, as they can introduce false matches (misidentifications) and missed matches (failure to recognise the same individual). These errors, even at low rates, can bias the demographic estimates that underpin conservation decisions. Yet, traditional statistical models for long-term data of this type assume perfect identification. As a result, demographic estimates derived from AI-processed data risk being biased and unreliable if identification uncertainty is ignored. 

As AI becomes standard in conservation, developing probabilistic approaches that quantify and propagate uncertainty is essential to ensure reliable inference from new ecological data streams. This project will develop probabilistic AI methods that integrate all stages of the ecological monitoring pipeline—from individual identification to demographic modelling—within a unified statistical framework. These methods will explicitly account for the uncertainty and bias introduced by automated recognition systems, ensuring that information about AI performance is propagated through the entire modelling process. By embedding AI outputs within statistical ecology models, the project will enable demographic parameters such as survival, population size, and recruitment to be estimated in a way that reflects both ecological variability and uncertainty in identification. 

The approach will combine innovations in probabilistic AI, which quantify uncertainty in machine-learning predictions, with scalable statistical computation to handle large datasets. This integration will allow the student to assess how different AI architectures, training data, and decision thresholds influence downstream demographic inference. The resulting framework will produce robust population-level estimates while providing practical guidance for balancing automation, accuracy, and computational efficiency in ecological monitoring. In doing so, the project will place probabilistic AI at the centre of ecological modelling—laying the foundation for how data from emerging technologies are analysed and interpreted in the future. 

The research will make use of two long-term re-ID datasets, which also provide opportunities for real-world impact. The first is a 25-year photo-database of loggerhead sea turtles (Caretta caretta) from Zakynthos Island, Greece. It was assembled from diverse sources, combining standardised in-water surveys, citizen-science observations, and systematically mined social-media imagery. Consequently, the dataset exhibits a wide range of photo qualities, with each collection method subject to different levels of automated re-ID error, making it ideal for developing and testing probabilistic AI models that explicitly characterise uncertainty. 

The second dataset concerns great crested newts (Triturus cristatus), a protected species with distinctive belly pattern markings that make individuals identifiable from photographs. This is the longest-running dataset of great crested newts in the UK and has never been analysed before using AI methods. It provides an exceptional opportunity to apply the proposed probabilistic framework to a species of high conservation importance, illustrating how uncertainty-aware AI can improve inference on survival, recruitment, and population trends in long-term monitoring programmes. 

Together, these applications will produce a quantitative framework for understanding and mitigating bias in AI-generated ecological data, along with open-source tools for integrating individual identification uncertainty into demographic models. The methods will apply broadly across automated monitoring systems, including camera traps, acoustic recognition, citizen science, drones, and remote sensing, and will underpin the next generation of biodiversity monitoring networks that combine these technologies with AI. By advancing probabilistic AI for ecological data, the project will contribute directly to the development of transparent, reliable, and future-ready models for biodiversity monitoring in an era of rapid technological change. 

References:

SeaTurtleID2022: A long-span dataset for reliable sea turtle re-identification,  
L. Adam, V. Cermak, K. Papafitsoros, L. Picek, 2024 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), (2024) 

WildlifeDatasets: An open source toolkit for animal re-identification, 
V. Cermak, L. Picek, L. Adam, K. Papafitsoros, 2024 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), (2024) 

Outstanding challenges and future directions for biodiversity monitoring using citizen science data,
Johnston, A., Matechou, E., & Dennis, E. B. (2023). Methods in Ecology and Evolution, 14(1), 103-116.

Funding Notes:

Further information: 
How to apply 
Entry requirements 
Fees and funding

PhD Information Session 2026:
On Wednesday 14 January, we will be holding a short information session about PhD studies in Mathematics at QMUL. For full details about the event, please visit: https://www.qmul.ac.uk/maths/postgraduate/postgraduate-research/phd-information-session-2026/

As one of the UK’s most diverse universities, QMUL fosters an inclusive and supportive academic community.

The School of Mathematical Sciences is committed to the equality of opportunities and to advancing women’s careers. As holders of a Bronze Athena SWAN award, we offer family-friendly benefits and support part-time study.