Prioritization of Areas for Forest Fire Combat and Prevention in the Amazon Biome for Optimized Response

• Grant number: 24/07354-7
• Support Opportunities: Scholarships in Brazil - Master
• Start date: October 01, 2024
• End date: September 30, 2025
• Field of knowledge: Agronomical Sciences - Forestry Resources and Forestry Engineering - Nature Conservation
• Principal Investigator: Liana Oighenstein Anderson
• Grantee: Ignácio Martins Pinho
• Host Institution: Instituto Nacional de Pesquisas Espaciais (INPE). Ministério da Ciência, Tecnologia e Inovação (Brasil). São José dos Campos , SP, Brazil
• Company: Universidade de São Paulo (USP). Escola Politécnica (EP)
• Associated research grant: 20/15230-5 - Research Centre for Greenhouse Gas Innovation - RCG2I, AP.PCPE

Abstract

The occurrence of fires in the Amazon rainforest threatens the integrity of this complex system that harbors great biological and human diversity and plays an important role in climate regulation and provision of ecosystem services. Fires in the Amazon occur through a combination of factors: anthropogenic ignition sources mainly from deforestation and the use of fire for agricultural and livestock management, climatic conditions characteristic of the dry season, and factors related to forest structure. There are several operational systems that forecast fire risk or occurrence to assist in fire prevention and control in the Amazon. However, there are limitations in these systems such as: (1) not considering all fire occurrence vectors; (2) not distinguishing between fire types; (3) delineation of excessively large priority areas, and (4) not using weather prediction models, focusing only on past events. Thus, the objective of this work is to develop a sub-seasonal fire occurrence prediction model, between 15 to 30 days, to delineate priority areas for fire prevention and control in the Amazon, based on remote sensing products and weather prediction models. The model will consider the various fire occurrence vectors in the Brazilian Amazon. The Random Forest algorithm is a simple, parsimonious machine learning method that exhibits good accuracy and also allows evaluating which variables are most important. It is expected to develop a model that assists managers and decision-makers in planning and implementing fire prevention and control policies.