Next-Generation Biodiversity Assessments through Remote Sensing and Deep Learning University of Reading in United Kingdom

*Please note that this PhD will be hosted at Royal Botanic Gardens Kew*

BACKGROUND AND OBJECTIVE

There is an urgent need to develop rapid, cost-effective and robust ways to measure biodiversity in order to guide biological conservation and assess the progress of restoration efforts. However, most approaches rely on the identification of individual species, either through the sampling of physical specimens or their photographic documentation – both of which require considerable time with the handling and processing of specimens, taxonomic expertise, adequate reference collections and databases. These challenges have led to a dearth of high-quality biodiversity data at fine spatial scales, further characterised by strong taxonomic and temporal biases. To tackle this problem, this PhD project will evaluate the power of recently developed sensing technologies and artificial intelligence for predicting various biodiversity metrics without the handling and identification of individual species.

METHODOLOGY

Fieldwork will be conducted at two contrasting sites: the Thames Valley Science Park and neighbouring landscapes in the UK (a site of mutual interest to the Natural History Museum, Royal Botanic Gardens, Kew and the University of Reading for the development of collections infrastructure), and a nature reserve and surrounding region in Brazil’s Atlantic Forest (a mega-diverse region in urgent need for protection and restoration but where the identification of non-charismatic species is a major challenge for biodiversity assessments).

At each of the two sites, detailed biodiversity assessments will be carried out at a relatively high (c 100-150) number of randomly selected quadrants (10x10 m) where: i) plant, animal and fungal diversity will surveyed through appropriate sampling and documentation methodologies; and ii) various sensing technologies will be used to characterise the site. The project will explore and validate a range of technologies, including airborne and stationary lidar (Light Detection and Ranging, i.e. a 3D laser scan), multi- and hyperspectral sensors; audio recorders for air and soil bioacoustics; camera traps for large fauna and night-flying insects; cyclone samplers for airborne pollen and spores; and environmental DNA from air, insect traps, soil and water samples.

Supervised deep learning models will be used to predict Essential Biodiversity Variables from remote sensed data. The models will be trained based on the data collected on the ground within this project and other biodiversity monitoring data, available from networks such as s-plot and data held by our own organisations (e.g., the Nature Unlocked programme at RBG Kew’s second site Wakehurst in Sussex).

OUTCOMES AND IMPACT

The project is expected to deliver several publications in peer-reviewed international journals. While c. 3-4 should be led by the PhD student during the course of the project, the supervisor team will strive to include them in other work whenever appropriate and relevant to their professional development.

Given the wide public and scientific interest in the topic of measuring biodiversity / biodiversity metrics – a crucial step for delivering the Kunming-Montreal Global Biodiversity Framework – some of these publications could be competitive for even some of the most prestigious scientific journals, and may help set international standards in the field.

If successful, the approach developed and validated here could directly benefit on-the-ground efforts in biological conservation (e.g., to help identify the most important areas for protection; see https://doi.org/10.1038/s41893-022-00851-6 ), ecological restoration (e.g., measuring how different sites improve in their biodiversity conditions over time), and support the application of recently developed biodiversity credits as a means to fund conservation and restoration (see https://doi.org/10.1038/d41586-024-03475-2 ).

Training Opportunities:

A comprehensive training programme will be provided, comprising training both in applied AI and biodiversity, and transferable professional and research skills. The project includes a placement with an AI-INTERVENE project partner of between 3-18 months in duration. The student will present at national and international conferences, placing the student at the forefront of the discipline, leading to excellent future employment opportunities.

Student profile:

This project would be suitable for students with a degree in biology or a related field in environmental or physical sciences, with expertise in computational biology / data science and ideally also remote sensing. There is certain flexibility as to whether the thesis will be mainly field-based (e.g. through the collection of biological and sensing data) or desktop-based (e.g. through methods development and analyses), even if both elements will ideally be present and complement one another.

How to Apply:

To apply please use the Good Grants system at ai-intervene-dfa.grantplatform.com