Wildfires that grow into the largest, fastest, or most intense events—megafires—are the ones that kill people, destroy homes, and release vast stores of carbon, yet scientists still cannot reliably predict when or where they will erupt. Current fire danger models are built for typical fires, not the extreme ones that do the most damage. This project fills that gap by combining a global database of millions of individual wildfires with machine learning to isolate the specific climatic, landscape, and human factors that push a routine fire into a megafire. If successful, the research will produce a global forecasting and warning system usable by disaster management centres and firefighting units days to decades ahead. It will also quantify how changing megafire risk threatens carbon offset projects that underpin national Net Zero plans, and identify land management interventions that reduce the chance of extreme fires. An interactive dashboard will make all outputs available to policymakers, the reinsurance sector, and carbon market actors. The project does not stop at prediction—it also feeds its AI-derived insights into dynamic vegetation models to project how future climate and land-use scenarios will alter megafire potential through this century.
View original technical description
The escalating frequency and intensity of wildfires globally pose significant challenges to society and the environment through loss of life, damage to homes and infrastructure, and disturbance to ecosystems and ecosystem services, including carbon storage. There is an urgent need to better target our research on the largest, fastest, or most intense wildfires—known as megafires—which are typically the most disastrous and challenging to control. Our project will harness the power of artificial intelligence (AI), including machine and deep learning (M/DL), to isolate the climatic and human factors that cause regionally 'typical' wildfires to evolve into megafires. We will predict megafire potential on timescales ranging from days to decades with global coverage. We previously established the Global Fire Atlas (GFA), a novel system for tracking millions of individual wildfires globally each year. The GFA presents unprecedented scope to disentangle the climatic and landscape factors that promote or inhibit megafires. Combining the GFA with cutting-edge AI, we will pioneer new methods in forecasting megafire potential and assess how changing megafire likelihood influences the vulnerability of at-risk communities and carbon stores. Our project will not only provide forecasts and warnings to support firefighting and disaster management, but also provide actionable information for optimising land use policies and firefighting resource allocation in the future. Our project will further inform the delivery of national Net Zero plans by quantifying the threat of extreme wildfires to carbon offset projects. First, we will build a global ‘data cube’ that enhances fire records from the Global Fire Atlas by integrating variables linked to megafire development, including meteorological conditions, fire danger indices, land cover, land use, and landscape. Second, we will use this data cube to identify regional fires with extreme properties, create the first global megafire database, and analyse regional trends in megafire frequency. Third, we will use M/DL models to determine the climatic, landscape, and management factors that promote or inhibit megafires regionally, also highlighting effective land management interventions. Fourth, we will use our models to predict megafire development beyond ‘typical’ fires, underpinning a global megafire forecasting and warning system useful to disaster management centres and firefighting professionals. In subsequent project phases, we will integrate key features from our AI-based models into a global dynamic vegetation model and use this to assess the impact of future climate and land use change scenarios on megafire potential this century. We will quantify trends in the exposure of at-risk communities and carbon offset projects to megafires over observational and future periods, allowing mitigation and adaptation capacities to be built where they are most required. All of our project outputs will be made available to stakeholders and the wider public via an interactive dashboard to maximise information available for cross-sectoral decision-making. Key beneficiaries include disaster management centres, firefighting units, policymakers (e.g., deciding resource allocation, developing land management policy), the reinsurance sector, and actors involved in Nature-based Solutions to climate change via the international and voluntary carbon markets. The project is not just about advancing our understanding of wildfires; it is about taking proactive steps to safeguard our communities, ecosystems, and vital carbon stores. Through groundbreaking research, capacity building, and the development of innovative tools, we will make a significant contribution to the global effort to address one of the most pressing environmental impacts of climate change.
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