As climate change accelerates the frequency and intensity of extreme wildland fires across Southern Europe, emergency responders face a critical challenge: predicting fire trajectories accurately and rapidly enough to protect lives and ecosystems. Bridging high-performance computing (HPC), artificial intelligence, atmospheric coupling, and real-time visualisation is at the heart of modern wildfire resilience.
As the leader of the Wildfires pilot in the EuroHPC JU HiDALGO2 project, MeteoGRID, together with key pilot partner ICCS (Institute of Communication and Computer Systems / NTUA), will present four research papers at the 10th International Conference on Forest Fire Research (ICFFR) in Coimbra, Portugal. In addition, MeteoGRID’s David Caballero Valero serves as an esteemed member of the conference Scientific Committee.
The Four Pillar Research Papers
MeteoGRID and ICCS span the complete operational lifecycle—from regional weather coupling and AI instant retrieval to detailed canopy physics and photorealistic 3D rendering.
1. Coupled Fire-Atmosphere Dynamics over the Iberian Peninsula
– Paper Title: High-resolution numerical simulation over the Iberian Peninsula with the coupled WRF-SFIRE fire-atmosphere model
– Lead Author & Speaker: Leydi Laura Salazar Domínguez (MeteoGRID)
Read the abstract
2. Overcoming the “Real-Time Barrier” with AI Retrieval
– Paper Title: A retrieval-based approach for real-time wildfire analysis using LSTM-driven similarity matching
– Lead Author & Speaker: Angeliki Dimitriou (ICCS – NTUA / MeteoGRID)
– Co-authors: Iliana Malliou, George Filandrianos, Angeliki Dimitriou, David Caballero Valero, Luis Torres Michelena, & Giorgos Stamou
Operational Impact:
In the initial minutes of an active wildfire, running full 3D physics-based simulations on HPC clusters can take too long for immediate tactical decisions.
To solve this, Angeliki Dimitriou and the joint pilot team built a data-driven retrieval engine:
– LSTM Spatiotemporal Embeddings: Recurrent Neural Networks encode evolving fire perimeters, terrain, and wind dynamics into compact vector representations.
– Instant Similarity Search: Matches active fire patterns against a vast database of pre-simulated and historical scenarios in seconds.
– Early Decision Support: Provides incident commanders with immediate, physics-grounded proxy trajectories while full-scale simulations initialise.
3. Fractal Canopy Fuel Initialisation for Physics Simulations
– Paper Title: From 3D vegetation models to physics-based wildfire simulation: a fractal barycentric method for Lagrangian particle initialisation in FDS/WFDS
– Lead Author & Speaker: David Caballero Valero (MeteoGRID)
Operational Impact:
Accurately modelling combustion inside dense forest canopies within physics engines like FDS/WFDS requires converting raw spatial scans into physical particles.
David Caballero Valero introduces a fractal barycentric initialisation method that maps dense 3D point cloud scans into discrete Lagrangian fuel particles. By preserving the spatial fractal structure of tree branches and foliage, this approach provides superior boundary conditions, heat release calculations, and flame-spread accuracy.
4. Interactive Real-Time Rendering in Unreal Engine 5
– Paper Title: Real-Time Physics-Based Visualisation of FDS/WFDS Simulations in Unreal Engine 5 Using Sparse Volume Textures
– Lead Author & Speaker: David Caballero Valero (MeteoGRID)
Operational Impact:
High-fidelity numerical simulations produce multi-gigabyte volumetric fluid and temperature fields that are difficult for field commanders to interpret quickly.
Using Sparse Volume Textures (SVTs) inside Unreal Engine 5, David establishes a real-time rendering pipeline capable of streaming dense volumetric fire, smoke, and temperature dynamics. Incident commanders gain an intuitive, interactive 3D interface to inspect physics simulation outcomes from any angle as conditions evolve.
Driving Resilience Through EuroHPC HiDALGO2
As lead partner for the HiDALGO2 Wildfires pilot, MeteoGRID, together with ICCS and European partners, continues to push the frontier of HPC and AI integration for emergency technologies funded by the EuroHPC Joint Undertaking (EuroHPC JU).
Conference Program & Details: 10th ICFFR Official Portal
Explore the Pilot: HiDALGO2 Wildfires Pilot