Forest Fire Risk Assessment and Mapping Using Remote Sensing and GIS Techniques: A Case Study in Ngh$${{\underset{\raise0.3em\hbox{$\smash{\scriptscriptstyle\cdot}$}}{\hat {e}} }}$$ An Province, Vietnam

Abstract—This paper presents the results of modeling the risk of forest fires in the west of Ngh $${{\underset{\raise0.3em\hbox{$\smash{\scriptscriptstyle\cdot}$}}{\hat {e}} }}$$ An Province (north-central Vietnam) using remote sensing and GIS data. We built models for the occurrence of forest fires using machine learning methods, including Random Forest (RF), Suppor Vector Machine (SVM), and Classification and Regression Trees (CART). The models took into account nine factors influencing the risk of forest fires, including vegetation cover (the normalized difference vegetation index (NDVI)), surface evapotranspiration, elevation, slope, aspect, wind speed, ground surface temperature, average monthly precipitation, and population density. Various parameters are tested in the RF, SVM, and CART algorithms to select the algorithm with the highest accuracy in forest fire risk prediction. The results show that the RF algorithm with the value of the “numberOfTrees” parameter equal to 100 has the highest accuracy in predicting the risk of forest fires in the study area.