Universidad Pontificia Comillas. Madrid (España)
07 de julio de 2026
Resumen:
Wildfires are increasingly severe, frequent, and impactful, posing growing risks to human life, ecosystems, infrastructure, and economic activity, particularly within Wildland–Human Interfaces (WHI). Climate change, land-use change, and socio-economic dynamics are reshaping fire regimes and challenging traditional prevention and suppression strategies, creating the need for cost-effective, scalable, and data-driven systems to anticipate wildfire behaviour and reduce damage.
This thesis develops an integrated modelling and optimisation framework for landscape-scale wildfire prevention and mitigation, with emphasis on early detection using Wireless Sensor Networks (WSNs). The approach combines open-access environmental data, dynamic wildfire simulations, and multi-objective optimisation to autonomously design prevention strategies under uncertainty. A reproducible fire-modelling pipeline converts national-scale forest and environmental repositories into high-resolution fire behaviour predictors and introduces a Cellular Automaton (CA) to simulate wildfire spread. Its implementation integrates uncertainty in vegetation structure and environmental drivers, enabling predictions in terms of confidence intervals. The system is validated against a documented Mediterranean Wildland–Urban Interface (WUI) wildfire.
Building on this foundation, the design of Early Wildfire Detection Systems (EWDS) based on WSNs is formulated as a Multi-Objective Optimisation Problem (MOOP) that identifies Pareto-optimal sensor layouts balancing detection performance, operational constraints, and deployment cost across varying ignition likelihoods and fire-weather conditions. Optimised WSNs outperform uniform deployments of comparable size, enabling earlier detection across plausible high-risk scenarios. This application is extended to the Wildland–Industrial Interface (WII) by integrating WII geospatial data, enabling the optimal deployment to balance detection in wildland areas with the protection of industrial environments under different priority levels. In this way, it contributes to improving prevention in the context of the emerging WII risk, given its particular potential to trigger wildfire-induced Natech disasters. Overall, the thesis shows that simulation-driven optimisation can support low-cost, scalable, and adaptable prevention systems by coupling uncertainty-aware fire modelling with autonomous sensor network design for improved early detection and risk-informed decision-making in vulnerable communities.
Resumen divulgativo:
Esta tesis propone una optimización guiada por simulación que, a partir de bases de datos abiertas, deriva de forma autónoma y bajo demanda estrategias de prevención en paisajes vulnerables a incendios forestales, y aplica el método al despliegue de redes de sensores para detección temprana.
Descriptores: Programación no Lineal, Planificación, Simulación, Ciencia Forestal
Palabras clave: Wildfire prevention Simulation-driven optimisation Wireless Sensor Networks Fire-behaviour mapping Wildland-Industrial Interface
Cita:
J.L. Gómez, "Simulation-Driven Optimisation of Wildfire Prevention Systems", Tesis Doctoral, Universidad Pontificia Comillas, Madrid, España, 2026.
