article · Journal of Sustainable Forestry
Predicting wildfire occurrence helps forest managers plan fire safety measures, fuel management, and emergency team deployments. Using historical fire occurrence records from 2001 to 2021, machine learning models were developed to map wildfire susceptibility across the 58,950-hectare Talassemtane National Park in Morocco. The models incorporated environmental factors such as climatic conditions, vegetation characteristics, topography, and distance from roads. Predictions across 257,499 mapped locations classified land into five susceptibility tiers, ranging from very low to very high. The Random Forest model achieved superior predictive performance with an area under the curve score of 98.47, outperforming the Support Vector Machine model which scored 96.36. The resulting mapping determined that more than seven percent of the park, encompassing roughly 4,000 hectares of forest, faces high or very high wildfire susceptibility.
Talassemtane National Park represents one third of Morocco's biodiversity, making proactive wildfire risk assessment essential for conservation. Pinpointing specific zones with heightened fire probabilities provides forest managers and municipal authorities with evidence needed to deploy firefighting teams, design targeted prevention measures, and optimise fuel management strategies, thereby safeguarding vulnerable forest ecosystems against future outbreaks.
This predictive modelling approach serves public forest authorities and emergency planning agencies seeking decision-support tools for risk mitigation and resource deployment. The models represent applied and tested research at park scale. Integrating this methodology into operational geographic information systems or continuous environmental monitoring software could facilitate broader institutional adoption, though commercial uptake would require packaging the analysis into an accessible platform for municipal or regional forestry services.
AI-generated from the published abstract. Always read the original work before citing.
Assessing the probability of forest fires by analyzing the main fire factors can provide forest managers with an essential basis for making crucial decisions on aspects such as prevention strategies, fuel management, fire safety measures, emergency plans and the deployment of firefighting teams. The main factors influencing fires, such as climatic conditions, vegetation characteristics, topographical factors and proximity to elements such as roads and residential areas, were considered in order to produce forest fire susceptibility maps. Machine learning algorithms have become a powerful tool for predicting the probability of forest fires. The aim of this study is to produce a forest fire susceptibility map using widely used machine learning models: Random Forest (RF) and Support Vector Machine (SVM). The study was conducted in Talassemtane National Park (TNP), located in the city of Chefchaouen in Morocco, covering an area of 58,950 ha and representing one third of the country’s biodiversity (Naoual, s.d.). The factors influencing fires that were taken into account were altitude, tree species, slope, exposure, LULC, NDVI, wind speed, soil moisture, temperature, precipitation and distance from roads. And we integrated forest fire occurrence data from 2001 to 2021 in the TNP into the training process. The accuracy of the fire susceptibility maps produced was assessed using the Area Under the Curve (AUC) value. After running the machine learning models, estimates were made for 257,499 points on the map, which were classified into five levels of fire susceptibility: very low, low, medium, high and very high. The results revealed that the fire susceptibility map generated by the RF model had better accuracy (AUC = 98.47) than that produced by the SVM model (AUC = 96.36). According to the probability maps, more than 7% (approximately 4,000 ha) of the forests in the study area were classified as having high or very high fire susceptibility levels.
This page summarises published work. The authoritative version sits with the publisher.
DOI: 10.1080/10549811.2026.2717429
Is something wrong with this record? Report it or request removal.
Discussion
Have you built on this work, tried to replicate it, or seen it applied in practice? Share what you know. Verified researchers and MARATTO™ domain experts can open a discussion, and any member can reply. Contributions are reviewed before they appear.
No discussion yet. Open the first thread.
New to MARATTO™? Create a free account.
