Habitat destruction and degradation are currently the greatest threats to global biodiversity (Tilman et al 2017 Nature) and major threats to the benefits that people derive from ecosystems (IPBES 2019). High resolution datasets exist to investigate the spatial distribution and extent of habitat destruction (Hansen et al 2013 Science) but have not yet been used to robustly measure habitat degradation, despite its importance as a threat (Boucher et al 2020). This PhD project will develop novel methods to analyse new satellite datasets to identify rapid changes in habitat quality across the UK and, by working closely with the Joint Nature Conservation Committee (JNCC), will inform UK and regional monitoring and reporting processes, allowing cutting edge findings and up-to-date information to be rapidly incorporated into decision making by a range of stakeholders.
Earth observation tools represent an excellent opportunity for understanding the causes and consequences of changes in habitat quality (degradation and recovery) over large areas but at a fine spatial resolution. This is particularly true with the increasing availability of multi-spectral and radar data being collected at very high spatial and temporal resolutions, supplemented by lower temporal resolution lidar data. However, actually detecting changes from remote sensing data is fraught with difficulties: accessing consistent datasets; selecting appropriate algorithms; dealing with intra- and inter-annual variability; and obtaining adequate validation data. As such, automatically detecting changes in habitat quality with a high degree of accuracy and reliability is a major challenge and has not yet been demonstrated.
Aims and objectives: This PhD will (1) develop open source automated tools for detecting changes in habitat quality for one or more habitats within the UK; (2) apply these tools across the UK to develop national and subnational datasets on how habitats are changing; and (3) link these patterns to changes in biodiversity metrics collated from a range of datasets. Throughout the studentship will be guided by UK and national-level policy priorities and questions—led by JNCC partners.
Methodology: This PhD will develop next generation automated tools to take advantage of recently available imagery for the UK from the Sentinel 1 (radar) and 2 (multi-spectral) satellites. CASE-partner JNCC has developed automatic pre-processing of these data to a set standard, available weekly at a 10 m resolution from 2014/2015, providing the unprecedented potential for identifying small-scale changes in almost-real-time. The tools developed will account for natural seasonal variation in vegetation and identify variation from those patterns as well as abrupt changes due to management or natural disasters.
To account for this complexity, the project will first identify a broad range of potential approaches to detecting changes in habitat quality, including artificial intelligence, Bayesian statistical approaches, dynamic vegetation model data assimilation and machine learning. The most promising potential approaches will then be used to develop the tools and test them against verified changes in habitat quality, based on data from on-the-ground habitat surveys from JNCC’s stakeholders, and visually checking using high-resolution imagery. Next, changes in habitat quality will be linked to location-specific data on biodiversity that feeds into the UK’s biodiversity monitoring programmes. This will allow satellite-derived changes to be directly linked to on-the-ground changes in biodiversity indices.
The student will apply the tools across the UK to identify national and sub-national changes in habitat condition and biodiversity. This project offers a rare combination of cutting-edge methodological development, using next generation machine learning and AI techniques, with the opportunity to feed directly into policy development and a national level.
Supervisory team: The student will be supervised by Dr David Williams (University of Leeds, expertise in conservation science; land-use change and land-use policy; biodiversity); Dr Steven Hancock (University of Edinburgh, expertise in remote sensing and land surface modelling); Dr Gwawr Jones (JNCC, Senior Earth Observation Evidence Specialist and creator of automated system for standardising data from Sentinel 1 and 2); Dr Paul Robinson (JNCC, Senior Natural Capital Evidence Specialist, expertise in providing evidence for policy decisions). The student will be registered at, and based at, the University of Leeds, but with strong inter-institutional support and the opportunity to visit both Edinburgh and JNCC.
Candidate description: A strong quantitative background and interest in remote sensing and environmental issues (essential); interests in AI, machine learning, and spatial data analyses (desirable); interest in, and willingness to work with, policy makers and decision-making processes.