Mapping the risk of forest fires in Peru’s Amazon and Andean forest regions using the AdaBoost algorithm and Geographic Information Systems

Abstract

The use of fires represents one of the principal methods to convert tropical forest ecosystems into other land-uses. To better target forest fire management and regulation policies, a better understanding of the areas prone to forest fires is needed. This study focuses on forest fires in the Peruvian Amazon and tropical Andes regions. First, the areas where forest fires are most prevalent are identified using Kernel Density Analysis. Second, the risk on fire within two forest fire hotspot locations is assessed using the AdaBoost algorithm. The results of this study show that forest fire is most extensive in the departments of Ucayali/Huanuco and San Martin. Furthermore, forest fire risk maps show that most of the highly to very highly susceptible areas in Ucayali/Huanuco are clustered within a single region, while in San Martin, susceptible areas are more widely scattered throughout the department. The fire risk maps generated in this study could contribute to forest fire management and research efforts in the Peruvian Amazon and tropical Andes regions.