AI-WUIFIRE is a decision-support solution designed to anticipate wildfire risk in Wildland-Urban Interface (WUI) areas and to provide operational tools for emergency responders when a wildfire happens. The platform combines predictive AI models with a GIS environment to support both prevention and response phases. Specifically, the solution addresses specific challenges of public administrations when dealing with this kind of problem: Slow uptake of innovative technologies. Public administrations face increasing wildfire risk in WUI areas, yet their capacity to adopt advanced technologies such as AI remains limited. AI-WUIFIRE addresses this barrier by delivering a ready-to-use AI-based system that enables risk-informed decision-making: -The predictive algorithm generates spatially explicit outputs estimating the probability of wildfire occurrence and the expected severity in WUI areas, producing actionable risk maps. -A multi-stage validation framework, including internal validation using a hold-out dataset and external validation against independent field data collected by the University of León for recent fire events. Limited situational awareness. Effective wildfire management in WUI areas requires the integration of multiple heterogeneous data sources, which are often distributed across separate applications, slowing down decision-making. AI-WUIFIRE covers the full emergency cycle within a single GIS environment: -Predictive wildfire risk and severity maps for WUI areas. -Operational tools such as wildfire spread simulation, real-time hotspot detection, and fire perimeter monitoring based on satellite data. -Real-time positioning networks of emergency responders and calculation of optimal arrival routes. -Geospatial datasets provided by cities through OGC-compliant services (e.g. fire defence infrastructure, cadastral parcels, protected monuments) or extracted from OpenStreetMap. -Non-spatial operational data, such as personnel shift schedules.
Contact
The AI-WUIFIRE proposal is fully aligned with European strategies promoting technological autonomy, innovation, and resilience in the face of environmental risks. It directly supports objectives outlined in the Special Report 16/2025: EU funding to tackle forest fires, which highlights the need for updated risk assessments, improved prevention and preparedness measures. It also contributes to the Open Strategic Autonomy and Resilient Europe 2030 Initiative, strengthening Europe’s technological capacities in AI and satellite data, reducing dependency on third countries in sensitive domains such as Emergency Management. In addition, it supports the EU Forest Strategy for 2030, which emphasizes forest resilience, adaptation to shifting vegetation zones, and the modernization of forest inventories to accurately assess wildfire risk. Beyond policy alignment, the platform delivers tangible benefits to the bioeconomy and emergency management sector across Europe. It enhances the tools used daily by specialists in control and communication centers, providing advanced situational awareness, that go beyond existing solutions. By combining AI-driven risk maps, satellite-derived data, and simulation tools, the platform introduces both technological and usage innovations, enabling authorities and operators to anticipate fire behavior, prioritize preventive actions, and optimize resource allocation. By delivering a clear, actionable overview of wildfire risk across European regions, AI-WUIFIRE empowers authorities to identify high-risk areas, implement preventive measures, and reduce socio-economic and environmental impacts. In doing so, the project strengthens Europe’s resilience, promotes innovation in emergency management, and positions the EU as a leader in autonomous, data-driven forest fire prevention and risk mitigation solutions.
The integration of ML algorithms with remote sensing observations is key to generating products that support decision-making. Open-access optical data from the Copernicus programme hold strong potential for wildfire management (Huot et al., 2022). These datasets have been used to classify fuel types and predict attributes such as structure and moisture content. However, open data alone often fail to meet city managers’ needs, as exploiting remote sensing requires programming skills, and significant computational capacity (Pettorelli, 2019). Currently, no application combines spatially explicit, large-scale fire severity predictions in WUI areas through an integrated modelling framework. No existing solution brings together: AI algorithms capturing complex fire-behaviour drivers; validated fuel type and condition databases; physically-based remote sensing models generalizable across ecosystems; and automated near real-time data ingestion to update WUI fire severity predictions. The AI-WUIFIRE algorithm consists of a comprehensive ML pipeline to predict fire severity in the wildland-urban interface. It combines physical vegetation models with Earth observation data: land cover classification (wooded vs. shrubby areas), Sentinel-2 multispectral bands (B2–B8a, B11, B12), and historical fire severity data from Iberian fires used as ground truth. The architecture follows a multi-stage approach. First, a Radiative Transfer Model simulates key biophysical vegetation parameters. Second, RTM outputs and Sentinel-2 bands are processed through a ML model (Random Forest/XGBoost), isolating four key variables: LAI, CWC, CCC, and LFMC. Finally, these variables are merged with GEDI LiDAR structural data and historical severity data to feed a second predictive model. The key innovation lies in data visualization: integrating all these outputs within a GIS environment to present complex satellite-derived geospatial information in an intuitive way.
Wildfires increasingly affecting cities represent one of the most pressing challenges for civil protection. Although the total number of wildfires worldwide decreased by 10% between 2005 and 2020, the proportion occurring in wildland-urban interface (WUI) areas increased by 23%, driven by urban expansion into forested land, declining land management, and worsening climate conditions such as prolonged droughts and rising temperatures (Tang et al., 2024). This trend shows that wildfire risk is no longer confined to traditionally fire-prone Mediterranean regions. It is becoming increasingly relevant in Northern Europe, including cities such as Trondheim (Norway), where urban areas directly interface with dense forests and wooden architectural heritage. In these contexts, fire can easily spread from vegetation to built structures through wind-driven embers, demonstrating the need for predictive and preventive tools beyond southern ecosystems (Sjöström et al. 70 Years of observational weather data show increasing fire danger for boreal Europe and reveal bias of ERA5 reanalysed data, 2025). At the same time, large-scale forest fires remain a major concern in purely natural environments.According to Greenpeace, eight of the ten largest wildfires in Spain this century occurred in 2025, with average burned areas per major fire rising to over 6,000 hectares, alongside fatalities and mass evacuations. Countries such as Portugal, Greece, Italy, and Romania consistently rank among those with the highest annual burned areas according to EFFIS data. AI-WUIFIRE addresses these challenges by shifting the focus toward prevention. Through AI-generated risk maps and wildfire simulation tools, authorities can detect high-risk areas before the summer season and prioritize vegetation clearance or prescribed burns. The fire simulator enables emergency services to test scenarios and assess potential fire spread, strengthening strategic planning.
Vexiza brings together a multidisciplinary team of experts in software development, GIS systems, and data science, as well as specialists in meteorology, remote sensing, and forestry. The team has extensive experience delivering projects for emergency management organizations in Spain, including the Spanish Emergency Military Unit (UME) and regional and local governments in some of the areas most affected by wildfires year after year in the Iberian Peninsula, such as Galicia and Castilla y León
Vexiza brings together a multidisciplinary team of experts in software development, GIS systems, and data science, as well as specialists in meteorology, remote sensing, and forestry. The team has extensive experience delivering projects for emergency management organizations in Spain, including the Spanish Emergency Military Unit (UME) and regional and local governments in some of the areas most affected by wildfires year after year in the Iberian Peninsula, such as Galicia and Castilla y León