Dr James Brown, Dr Petra Bosilj (LoVE) and Dr Lan Qie (School of Natural Sciences) have been awarded £500,000 by the Natural Environment Research Council to develop new methods for habitat classification from ground-based imagery. Working in collaboration with the UK Centre of Ecology and Hydrology (UKCEH), the promises to develop technology that may be used to automatically assess habitat type and condition using lightweight machine learning models, deployed on mobile devices without internet connectivity.

Habitat recognition as a proxy for biodiversity is essential to the success of Biodiversity Net Gain (BNG) policies, which were enacted across England in January 2024. The UK Habitat Classification (UKHab), upon which BNG metrics will be based, provides a consistent spatial framework for habitat classification. Despite significant advances in the tools to recognise and monitor species, there are no equivalent tools for habitat recognition from ground-based imagery. Using tens of thousands of images from UKCEH’s long-term nationally representative Countryside Survey, this project will leverage the latest technological advances in artificial intelligence to produce a tool (AI-Hab) that can provide a user with habitat information in line with UKHab classifications from field-based imagery.

The resulting product will be integrated into a prototype version of an existing and widely used AI plant identification app developed by UKCEH – the E-Surveyor. The app will be optimised for use in new and existing land mapping applications for BNG assessment and for future uses, including potentially the Defra EBN (Environmental Benefits for Nature) tool. The product will also provide an excellent platform for the future integration of species/habitat data within species recognition tools and for improving the accuracy and resolution of habitat mapping from satellite data.