Climate models are producing predictions too uncertain to guide decisions on melting ice sheets, air pollution, and land use. This project brings together statisticians, computer scientists, and environmental scientists to fix that. The core problem is that environmental models simulate individual processes—like ice flow or atmospheric chemistry—in sophisticated detail, but then stitch those components together crudely and calibrate them poorly against real-world data. The result: high uncertainty and weak policy guidance. As new satellite and sensor data flood in, and climate pressures intensify, this gap between model complexity and predictive reliability is widening. The team will develop a modular suite of machine learning and statistical tools designed to plug directly into existing scientific process models, improving how components integrate and how predictions are tested against observations. They will co-design the work with public and private stakeholders—environment agencies, land managers, air quality regulators—and train end-users to close the current skills gap. If successful, the research could sharpen forecasts that underpin infrastructure planning, pollution controls, and land-use decisions, making them more reliable as climate change accelerates.
View original technical description
We will develop a data science of the natural environment, deploying modern machine learning and statistical techniques to enable better-informed decision-making as our climate changes. While an explosion in data science research has fuelled enormous advances in areas as diverse as eCommerce and marketing, smart cities, logistics and transport, health and wellbeing, these tools have yet to be fully deployed in one of the most pressing problems facing humanity, that of mitigating and adapting to climate change. This project brings together world-leading statisticians, computer scientists and environmental scientists alongside an extensive array of key public and private stakeholder organisations to effect a step change in data culture in the environmental sciences. The project will develop a new approach to data science of the natural environment driven by three representative grand challenges of environmental science: predicting ice sheet melt, modelling and mitigating poor air quality, and managing land use for maximal societal benefit. In each motivational challenge, there is already an extensive scientific expertise, with intricate models of processes at multiple scales. However this sophisticated modelling of system components is usually let down by naive integration of these components together, and inadequate calibration to observed data. The consequence is poor predictions with a high level of uncertainty and hence poorly-informed policy making. As new forms of environmental data become available, and the pressures on our natural environment from climate change increase, this gap is becoming a pressing concern, and we bring an impressive team to bear on the problem. A key theme of the project is integration, developing a suite of novel data science tools which work together in a modular fashion, and with existing scientifically-informed process models. By building a team that spans the inter-disciplinary divisions between data and environmental scientists we can ensure the necessary interoperability of methods that is currently lacking. Working with the full range of stakeholder environmental organisations will enable continual co-design of the programme and training of end-user scientists to ensure a reduction of the skills gap in this area. The resultant culture shift in the data literacy of the environmental sciences will enable better decision-making as climate change places ever greater strains on our society.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know