Balancing the needs of a growing human population with safeguarding the world’s biodiversity is one of the most important challenges of the 21st Century. One of the most urgent, important, and a particularly difficult example of this challenge, is the conservation of many of the world’s remaining large mammals in Southern Africa: Africa has retained a larger proportion of its fauna compared to other tropical regions, but the impact of climate change and changing environment as well as direct impacts of human activity and increasing population means that the next few decades could see disastrous declines in many species (e.g. Tilman et al 2017 Nature; Williams et al 2021 Nat. Sust.)
To address this challenge requires an improved understanding of the state and functioning of the region’s ecosystems. However, even basic information is lacking on many crucial issues such as the abundance of wildlife species and how animals interact with their environments. Traditionally data have been collected on wildlife population using, for example, GPS tracking data, field and/or aerial surveys (King 2016 Annu. Rev. Stat. Appl.). However, new technologies provide exciting opportunities for collecting data at different geographical and temporal scales in order to try to improve our understanding. In particular, recent earth observation (EO) satellite data provide non-invasive opportunities at larger geographical scales to record environmental conditions and even observe large terrestrial animal populations via very high resolution data.
The aim of this project is to investigate the potential of data fusion for combining EO data at different scales with different forms of commonly data collected “on-the-ground” data (or “in-the-air” for aerial survey data) to provide new insights for improved conservation outcomes. The project will focus on data relating to elephants located within Southern Africa and will be in collaboration with Dr Niall McCann (National Park Rescue), Dr Mike Chase (Elephants without Borders) and Dr Murray Collins (co-supervisor; Space Intelligence) including a 3-month internship with Space Intelligence. This project will aim to provide novel insights into elephant populations and their associated habitat-use that will allow conservation organisations such as Elephants Without Borders to plan and promote conservation strategies and management policies.
The project will involve applying advanced statistical and machine learning techniques to different resolution multi-scale EO data (high resolution data 5m-30m (Sentinel 1 and 2); and very high resolution data <1m (eg Worldview 2/3)). A range of cutting-edge statistical techniques (e.g. generalised regression models, hidden Markov models, distance sampling) will be applied to combined high-resolution EO data and observational data to build a detailed understanding of how and where elephants use different habitats and land uses. Further, applying advanced machine learning object detectors (e.g. region CNN models) to very high resolution data offers an incredible opportunity to automate the processing of EO data to identify and count individual elephants from space. Finally, using data fusion techniques, the outputs from these models will be further combined with environmental data from larger geographical regions which will allow the team to generate regional population estimates (with associated measures of uncertainty).
This project provides an exciting opportunity to work at the forefront of modern interdisciplinary research. The successful student undertaking this project will gain knowledge and experience of working with modern EO data and its potential; develop cutting-edge data analytical skills; and also an in depth understanding of the associated ecological/conservation issues that these techniques can be applied to.