On 17–18 March 2025, the InterAcademy Partnership (IAP), in collaboration with the Royal Spanish Academy of Sciences and with support from the U.S. National Academies of Sciences, Engineering and Medicine, convened an international workshop on Wildfire Modelling and Artificial Intelligence.
The workshop brought together global experts in modelling and tool development to assess state-of-the-art wildfire prediction models, explore strategies to mitigate wildfire impacts on built environments and identify key challenges, gaps and opportunities for international collaboration.
The workshop aimed to advance wildfire risk mitigation by focusing on three key objectives. First, it sought to evaluate existing wildfire models, their accuracy, and computational requirements, when possible, as well as their application to capture wildfire damage and risk. Second, it aimed to close the gap between technical fire simulations and their practical use in shaping policies. Finally, the workshop addressed challenges in scaling models and standardising data across national boundaries to support coordinated and effective fire management strategies.
The workshop focused primarily on the impact of wildfires on the built environment, especially within the wildland–urban interface (WUI). The workshop brought together 26 experts from ten countries, including Australia, Chile, France, Italy, New Zealand, Portugal, Spain and the United States.
The workshop's sessions explored a spectrum of approaches—physics-based, semi-physics-based, and empirical/AI-driven modelling—each with distinct strengths and limitations.
Physics-based models such as the Fire Dynamics Simulator (FDS), developed by the U.S. National Institute of Standards and Technology (NIST), and Nek5000 spectral element, a highly scalable computational fluid dynamics code, offer high accuracy and detailed analysis of fire behaviour mechanisms, making them essential for policy development and structural planning; however, their computational intensity, expertise requirements, and limited accessibility restrict broader operational use.
Semi-physics-based models, including AGNI-NAR and urban hybrid models, balance physical realism with efficiency by integrating simplified physics into probabilistic frameworks, offering faster run times and adaptability for planning and training, though they face challenges in generalisability, adoption, and stakeholder awareness.
Empirical, statistical, and AI-driven models, such as Cell2Fire and vulnerability mapping tools, leverage historical and environmental data to deliver scalable, accessible wildfire risk assessments ideal for strategic planning and public engagement, though they lack real-time predictive capacity and may underperform under shifting climate conditions.
Together, these model types form a complementary toolkit for advancing wildfire understanding and preparedness, with workshop participants emphasising the need for international collaboration, training, and targeted investment to optimise their use in diverse operational contexts.
The workshop did not explore the uncertainties associated with these models, and not all presenters provided numerical information on the prediction and accuracy of the models, highlighting potential areas for further discussion.
Workshop participants explored the following concrete opportunities for advancing wildfire modelling efforts:
This report summarises the presentations and discussions that took place during the workshop. It is authored by Dr. Hussam Mahmoud of Colorado State University and now at Vanderbilt University, who served as the workshop chair and rapporteur. The report captures key insights from the sessions, outlines proposed priorities for advancing wildfire modelling capabilities and applications, and identifies proposed next steps for research and collaboration.
