Wildfire risk weighting and behaviour prediction using open geospatial data and ontologies

Abstract

This article presents a novel approach to wildfire risk assessment and behaviour prediction by leveraging open geospatial data and ontologies. The proposed methodology includes a spatially weighted index model and multicriteria analysis to represent the risk of forest fires in the affected area. It bridges gaps in theory and practice, offering a comprehensive solution for evaluating potential forest fire risk in near real time, predicting fire behaviour and elucidating the semantics of fire management. During dry and hot conditions, forest fires tend to escalate. Hence, we propose an algorithm that combines experts’ empirical criteria and open-source data to identify dangerous fires in near real time, aiding authorities in directing attention to the riskiest areas. The objective is to predict forest fire behaviour, a complex and nonlinear system influenced by dynamic factors such as weather conditions, topography and land use. Our methodology enables real-time assessment of potential forest fire risks, complemented by predictive fire behaviour scenarios and a descriptive ontology of fire management semantics. We examine existing fire-related ontologies and propose a comprehensive one encompassing incident descriptions, firefighting resources, actor interrelations and knowledge for effective action. By classifying fire sources, our algorithm enables strategic decision-making to prevent uncontrolled fires. This solution significantly enhances data using semantic and spatial relationships among wildfire resources. Furthermore, we demonstrate how ontologies improve data integration and interoperability among diverse systems and organisations involved in forest fire risk management, fostering better coordination and faster responses to critical situations. To facilitate decision-making, we create decision-making scenarios linked to analysed hot spots, drawing from open hot spot data such as National Aeronautics and Space Administration (NASA) Fire Information for Resource Management System (FIRMS), OpenStreetMap (OSM), OpenWeatherMap (OWM) and OpenTopoData (OTD). We propose an ordinal and linguistic classification system (F1–F5) denoting risk levels as low, moderate, high, very high and extreme. These values are obtained through factor aggregation and fuzzy logic. A publicly accessible, interactive web map displays the results derived from this model. Overall, our contributions to wildfire risk management provide authorities with a valuable tool to make informed decisions and mitigate the damaging effects of wildfires.