Since 2006, threatened reptile species have been reintroduced back to islands to rebuild lost Mauritian communities wher
PhD in Statistics Modelling changes in abundance of species over space and time in island ecosystems Primary supervisor: Dr Rachel McCrea, Statistical Ecology @ Kent, School of Mathematics, Statistics and Actuarial Science, University of Kent Project Partners: Dr Lianne Concannon, Durrell Wildlife Conservation Trust Dr Nik Cole, Durrell Wildlife Conservation Trust and Mauritian Wildlife Foundation
Background Information The Mauritius Islands Restoration Programme is a collaborative project between the Durrell Wildlife Conservation Trust, Mauritian Wildlife Foundation and National Parks & Conservation Service to restore island ecosystems in the Republic of Mauritius. Since 2006, threatened reptile species have been reintroduced back to islands to rebuild lost Mauritian communities where major threats have been removed. These actions are enhancing the distribution and abundance of threatened species and have already prevented at least one extinction event. To ensure success field staff need to closely monitor these communities, to make appropriate management decisions for the ongoing restoration of the island ecosystems. Commonly used techniques for estimating abundance, such as distance and capture-recapture when detection rates are very low (as is often the case with severely threatened species), produce poor results where comparisons over space and time are not possible. Furthermore, these methods are effort-intensive or challenging to teach to field staff. The use of models based on count data have a number of advantages in terms of ease of field techniques and they are less demanding in terms of field effort required, but also have potential pitfalls. Overview of the Project The PhD student will construct novel stochastic models to develop a robust modelling approach that will enable changes in abundance of Mauritian reptiles and birds to be detected over time and space. The new approaches will inform optimal study design for a major overhaul of field methods and will provide guidance for conservation managers. Models for estimating abundance from count data alone are becoming increasingly popular due to their relative simplicity (Royle, 2004). Adaptations of the basic count data models are constantly developing, relaxing previously restrictive modelling assumptions, such as closure (see for example Dail and Madsen, 2011) or single-species analyses. However, fundamental estimation problems can arise in such models (Dennis et al, 2015) and revised survey protocols should not rely on count data alone. Integrated population models (Besbeas et al, 2002; McCrea and Morgan, 2014, chapter 12) provide a potential solution as multiple data types can be analysed simultaneously. Estimates from integrated models generally result in improved precision (McCrea et al, 2010) and can overcome the risks of being reliant on a single data source (Cole and McCrea, 2016). The stochastic models need to be
sufficiently flexible to adapt to available data types, with automated solutions to overcome missing data problems and the potential to incorporate additional covariate information at a temporal and/or individual level. The student will investigate both frequentist and Bayesian modelling approaches and will scrutinise the potential of automated model selection routines in frameworks to make the model fitting achievable by non-specialists. References Besbeas, Freeman, Morgan and Catchpole (2002) Integrated mark-recapture-recovery and census data to estimate animal abundance and demographic parameters. Biometrics. 58, 540-547. Cole and McCrea (2016) Parameter redundancy in discrete state-space and integrated models. Biometrical Journal. 58, 1071-1090. Dail and Madsen (2011) Models for estimating abundance from repeated counts of an open metapopulation. Biometrics. 67, 577-587. Dennis, Morgan and Ridout (2015) Computational aspects of N-mixture models. Biometrics. 71, 237246. McCrea and Morgan (2014) Analysis of capture-recapture data. Chapman and Hall/CRC Press. Royle (2004) N-mixture models for estimating population size from spatially replicated counts. Biometrics. 60, 108-115.
Aims and Objectives The aim of the project is to develop new statistical models to provide a revised survey protocol for the monitoring of species on a complex island ecosystem. The objectives the student will achieve will be: (i) evaluate existing models for a variety of data types using historical data, including distance, capture-recapture, count data; consider relative information contained in different data types; (ii) evaluate which data types might be analysed in conjunction with one another to utilise shared information and develop new stochastic models; investigate the statistical power of combined data sources, factoring in costs and field site complexity; develop an optimal survey strategy for the island ecosystem, and forecast statistical power expected from survey protocols; (iii) develop a statistical toolbox of R programs and handbook for conservation practitioners with limited quantitative/R background. Using Mauritius as a model, this project responds to a critical need to overhaul current survey strategies for long-term benefits of island ecosystems. The student and supervisory team will be uniquely placed to facilitate this, bringing statistical expertise to provide rigorous and cutting-edge methodology to solve a real-world problem. Student Experience and Benefits The primary supervisor is an expert in applied stochastic modelling and the student will benefit from the academics and students of the Statistical Ecology @ Kent research group (https://www.kent.ac.uk/smsas/statistics/research/seak.html). This project is interdisciplinary,
exposing the student not only to cutting-edge applied statistics research (which will result in topquality statistics journal articles), but also to the importance of translating academic research into accessible methodology for a wide-variety of end-users. This opportunity will be excellent experience for future employment in either academia or non-academic jobs. The project partners will play a key role in linking the theoretical developments with applications and translate this into management decisions. If desired, the PhD student will be able to travel to the field site in Mauritius to experience the challenges faced by fieldworkers, thus broadening their understanding of what challenges need to be accounted for in the proposed survey strategy resulting from the project. The student will develop collaborative skills during bi-monthly research visits between the supervisors at Durrell Wildlife Conservation Trust and the University of Kent, supplemented by fortnightly skype meetings. Towards the end of the project, the student will run an analytical training workshop with conservation practitioners, gaining key skills for future employment either in academia or industry.