Projects
AI for Collective Intelligence (AI4CI) hub
We are part of the £12.8M UKRI AI4CI hub, which utilises smart collectives to improve decision making. The Environmental Intelligence theme is helping develop new technologies, such as chatbots and smart agents, to enable greater access to and utilisation of environmental and climate data by organisations, policy makers and the general public.
Academic Lead: Professor Hywel Williams
Partners: Met Office, Natural Resources Wales, Ordnance Survey
Renewing biodiversity through a people-in-nature approach (RENEW)
Partnership with the National Trust, landowners, businesses, and communities, utilising EI driven decision-support tools, to reshape understanding and action on biodiversity and restore woodlands, wetlands and farmland across the UK.
Environmental Intelligence theme Lead: Professor Peter Challenor
Partner: National Trust
Centre for Climate Communication and Data Science (C3DS)
We conduct research at the intersection of AI, social science, and climate communications to support climate action
The Centre for Climate Communication & Data Science’s (C3DS) research aims to improve strategic communications to deliver meaningful action on climate change. We work with a diverse community of stakeholders — from civil society, media, government, and academia — to monitor, evaluate, and analyse the ways in which the modern information environment promotes or hinders comprehensive climate action.
Academic Leads: Professor Saffron O’Neill, Dr Travis Coan
AI-elevated Decision-support via Digital Twins for Restoring and Enhancing Ecosystem Services (ADD-TREES)
A pioneering AI research project to develop user-bespoke digital twins and co-created AI-enhanced decision support tools to aid crucial decisions about land use changes, focussing on creation of new woodlands and forests.
Academic Lead: Professor Danny Williamson
Partners: Defra, MoD, National Trust, Network Rail, Forestry England, National Forest Company and the Woodland Trust
GreenSight
This UKRI funded project uses AI to twin satellite imagery of Green Belt land in the UK with climate models in the Met Office to better predict the instance and spread of wildfires.
Academic Leads: Professor Hywel Williams, Dr Milto Miltiadou, Dr Remy Vandaele
Partner: Met Office
Machine Learning for UK Rivers and Waterways
Research project utilising Computer Vision and CCTV to automatically identify when sluice gates in local rivers have become blocked with rubbish and to classify the nature of such blockages. This approach supports decision-making around the deployment of maintenance teams, saving time and improving river quality, whilst also supporting flood prediction and prevention.
Academic Leads: Professor Hywel Williams, Dr Remy Vandaele
Partner: Staffordshire County Council
© Copyright Richard Sutcliffe and licensed for reuse under this Creative Commons Licence.
Local Data: Global Plankton and Ocean Carbon
Teaming up with Plymouth Marine Laboratory and IBM, scientists at the Centre for Environmental Intelligence are gathering and utilising local data on phtyo plankton and carbon in the ocean to improve train and improve geospatial foundation models. This work will create a key capability for more accurate carbon and climate modelling.
Academic Leads: Professor Chunbo Luo, Professor Hywel Williams, Dr Remy Vandaele
Partners: Plymouth Marine Laboratory, IBM
Habitat ID
Natural England monitor and classify habitats, for instance as ‘Coniferous Woodland’, ‘Heathland’ or ‘Lowland Meadow’, as part of their national scale LivingEngland habitat map. Scientists at the Centre for Environmental Intelligence are working with them to develop an application that will use site photos, from field ecologists and potentially citizen scientists, and a checklist to automatically identify and record these, increasing the range of observations and data that can be utilised for habitat mapping.
Academic Lead: Dr Sareh Rowlands
Partner: Natural England
UK Climate Projections Chatbot
We have developed a Chatbot interfacing with the UK Climate Projects (UKCP) data and reports produced by the Met Office, to provide accurate evidence-based answers about climate change to scientists, the general public and journalists. This capability ensures the Met Office can provide 24/7 factual responses drawing on a curated archive of peer reviewed papers. This tool counters misinformation and by drawing on a verified source of information does not struggle with the inaccuracies many untrained chat bots such as Chat GPT are associated with. The Chatbot provides reliable answers in plain English and is easily used by the wider general public including schools.
Academic Lead: Professor Hywel Williams
Partner: Met Office
Earth Rover Program
The Earth Rover Program is a non-profit initiative focused on transforming soil assessment by applying seismology—a mature, low-cost technology—for agricultural and scientific benefit. The program uses geophones and seismic wave techniques to “see” into soil structure, moisture, volume, and potentially carbon and nutrient levels—without disturbing the ground.
The program’s goal is to empower farmers—regardless of scale or income—to monitor soil health in real time, boosting crop yields while reducing environmental impact . By generating spatially explicit soil health data and building operational hubs globally, the program aims to support both precision agriculture and soil science, tackling widespread soil degradation and helping secure sustainable food production for a growing global population.
Academic Lead: Professor Tarje Nissen-Meyer
Partner: Bezos Earth Fund
Centre for Net Positive Health and Climate Solutions
The Centre is a five-year £10M initiative funded by UKRI conducting research on climate change and its impacts on health.
Net-positive solutions aim to reduce the negative health impacts of climate mitigation and adaptation, whilst also contributing to positive outcomes – such as ecosystem recovery and improved human wellbeing - whilst ensuring new measures don’t worsen existing inequalities among those communities most impacted by climate change.
Individual projects are focusing on the design of urban environments, green and blue spaces, and food systems, as well as analysing associations between mental health and climate, and specific mechanisms linking thunderstorms and asthma.
Lead Academic: Professor Tim Taylor
Partners: UK Health Security Agency, National Trust, Forest Research, Met Office
UNRISK Centre for Doctoral Training
Funded by NERC, the UNRISK Centre for Doctoral Training will train students with the multidisciplinary knowledge and skills across climate science, data science and decision science to tackle the pressing challenge of reducing the risks associated with rapid climate change
Focussing on improving climate change predictions, understanding its impacts, and providing the best possible information to improve decision-making, many of the Centre’s research project will also involve industry partners including climate prediction centres, consultancies, regulatory bodies and AI start-ups.
Lead UoE Academic: Dr James Salter
Upskilling the UK’s Climate Scientists and Meteorologists
Through the Centre for Environmental Intelligence the University of Exeter is delivering Foundation and Advanced (Deep Learning) training in Machine Learning to over 150 Met Office staff. These programmes deliver research led education both in online and in person formats to ensure the Met Office remains at the cutting edge of climate research and meteorology.
Lead Academic: Dr Stefan Siegert
Partner: Met Office
AI enabled early warning system for earthquake intensity
Development of a physics-informed AI-based earthquake early warning (EEW) system that predicts ground shaking intensity within seconds of an earthquake's onset, enabling life-saving action.
Research paper: “An Interpretable and Domain-Informed Real-Time Hybrid Earthquake Early Warning for Ground Shaking Intensity Prediction”
Academic Lead – Jawad Fayaz
Partners – Prof. Rodrigo Astroza (University of the Andes), Prof. Sergio Ruiz (University of Chile)
Wildfire risk minimization using geospatial AI models
Through co-production and integration of Geospatial Foundation Models with the Met Office's land-use management supporting digital twin, this project aims to overcome current decision-support tool limitations by enabling the detection and subsequent management of small-scale fire events—a critical capability currently lacking in the UK.
Lead Academics: Professor Chunbo Luo, Professor Hywel Williams
Partners: – IBM, Plymouth Marine Laboratory, Met Office
AI Librarian for Ecological Archives
Many planning-relevant facts are buried in environmental statements, management plans, survey reports, and consultancy outputs.
This project turns those documents into structured, citable records (like species, habitats, interventions, locations, outcomes) to support faster and more systematic evidence use in environmental planning and decision-making. A core requirement is verifiability: answers must be grounded in source documents and traceable to specific pages and passages.
Technically, the system ingests a secure corpus of PDFs and metadata from Natural Resources Wales. A retrieval augmented generation (RAG) backend indexes this information and, when queried, retrieves the most relevant information, acting as a digital assistant by answering only when it can point to supporting text in the source documents to provide direct evidence for its response.
Lead Academics: Mahdis Tourian, Professor Hywel Williams
Partners: Natural Resources Wales
Chatbot for Living Wales
The Living Wales Chatbot project is part of the wider Living Wales initiative.
The project is developing a web-based chatbot that allows people to ask questions about the environment in Wales using plain language. Users can explore information on land cover, habitats, burnt areas, and flood mapping using satellite and other environmental data collected through the Living Wales programme and stored in the Welsh Data Cube (WDC). The goal is to make complex environmental data easier to access and use for researchers, policymakers, and the wider public.
The main challenge the project addresses is that earth observation and environmental data are difficult to access and interpret without programming skills or specialist software. The approach is to connect a large language model (LLM)-based chatbot directly to Living Wales datasets, enabling users to query the data via simple questions rather than through technical tools. The expected outcome is improved access to environmental information, better support for environmental decision-making, and increased use of Living Wales data in policy and public engagement.
Academic Leads – Oscar Hountondji, Professor Hywel Williams
Partners – Plymouth Marine Laboratory, Aberystwyth University
Developing a chatbot for accessing complex geospatial information
© Crown copyright and/or database rights 2026 OS (Research Licence).
Project to develop an AI-powered chatbot to facilitate retrieval of specific information from the Ordnance Survey (OS) National Geographic Database, a task that currently requires specialist expertise.
To make this information accessible to non-experts, the project uses "agents" (specialised AI programs) that work together to understand a user's question and find the right data. Introduction of a more autonomous multi-agent system, with each expert agent a specialist in a specific category, such as building or addresses, allows the chatbot to handle natural language questions (plain English) and navigate the database's complexity automatically, providing accurate geographical answers without the user needing technical training
Academic Leads – Professor Hywel Williams, Dr Hailun Xie, Arjun Biswas
Partners – Ordnance Survey