Estimate forest fire risk using machine learning: An example of Taşköprü Forestry Directorate

This study was conducted at the Taşköprü Forest Management Directorate of the Kastamonu Regional Directorate of Forestry. The aim was to predict forest fire risk using machine learning. Data on fires that occurred between 2017 and 2022 in the Taşköprü Forest Management Directorate were used along with data on areas where no fires had occurred. Four main criteria (stand characteristics, topography, environmental factors, and weather conditions) and 12 sub-criteria (tree species, age class, canopy closure, slope, aspect, elevation, distance to settlements, distance to agricultural areas, distance to roads, distance to water sources, temperature, and wind speed) were identified for predicting fire risk using machine learning. Multilayer Perceptron, Random Forest, and Support Vector Machine models were used to predict fire risk through machine learning. Experimental studies using machine learning achieved an accuracy rate of at least 94%. This finding suggests that machine learning methods could effectively predict forest fire risk. In conclusion, this study highlights the importance of using Machine Learning models to predict forest fire risk. The significance of this study lies in its contribution to preventing potential forest fires by predicting fire risks and enabling early intervention measures before fires occur.