Through land use change (including agricultural intensification) agriculture is a key driver of global biodiversity loss, yet agriculture relies upon the ecosystem services that healthy, diverse agri-ecosystems provide, including biological control, pollination, and soil fertility. In the last three decades there have been increasing efforts to mitigate agriculture’s negative impact on the environment – Europe alone has spent approximately 50 billion EUR on agri-environmental schemes since 1992 – yet the majority of farmland biodiversity indicators show a continued decline. Critically, the level of monitoring of all farmland biodiversity is negligible, in large part due to the surveying cost, so that even when agri-environment schemes are put in place, we rarely understand their impact (Kleijn and Sutherland 2003). The goal of this PhD is to generate large scale, low cost methods that effectively determine what biodiversity is present in farmland, and how it responds to different treatments.
The technical challenge is to generate information at the relevant scales for farmers and land managers on the diversity of their managed areas. The opportunity here is to take advantage of earth observation from drones and satellites, and use machine learning methods to link in situ and remotely sensed data. Remote sensing provides the means to monitor and report on biodiversity across landscapes. Image analysis allows metrics to be derived from these data and related to in situ diversity data, for upscaling. Imagery will largely describe the vegetation structure and its pattern in the landscape, with indicators of vegetation diversity from spectral patterns of reflectance. Diversity of other groups must be inferred from vegetation pattern and diversity, i.e. habitat.
The student will work with a range of wildlife groups (such as: RSPB, Buglife, Plantlife) who already have ground data for specific taxa. The focus will be on the UK initially, with scope to include other countries depending on data sources. Using case study areas, the student will collect both fine resolution images using a drone equipped with a multi-spectral imager and high resolution camera; and coarse resolution data from satellites (e.g. Sentinels). The student will use machine learning methods to derive pattern analyses from images linked to in situ diversity data. The student may need to collect additional biodiversity information from the field to ensure adequate baseline data, and where possible data will include natural enemy and/or pollinator data. Relevant industrial and policy partners will be engaged to ensure that the results are appropriate to land managers, industry and policy.
Kleijn, David, and William J. Sutherland. 2003. 'How effective are European agri-environment schemes in conserving and promoting biodiversity?', Journal of Applied Ecology, 40: 947-69.