Identificação de florestas destinadas à produção de bioenergia no Estado do Tocantins utilizando imagens de satélite e mineração de dados

Abstract

Planted forests have attracted a lot of attention because of possibility of use in bioenergy applications and due to the global trend of prioritizing energy sources that provide greater environmental sustainability, more quality and security. In Brazil, the shifts in the geography of current agroforestry production chain towards the agricultural frontier areas (Midwest and North) are creating challenges to the adequacy of technical and scientific knowledge already established in other regions. So, the aim of this work is to assess the accuracy of the identification and classification of areas cultivated with plantation forests for energy, inside TM Landsat 5 images. Using statistical techniques for data mining, this study also evaluated the use of a broad set of attributes to identify improvements in the classification results. The research focused on samples of planted areas in the state of Tocantins, Northern Brazil. The data mining techniques used were effective in identifying of planted forests in Landsat 5 satellite images, both the classification performance, such as by reducing the amount of information needed to solve this kind of problem. Thus, the techniques employed in this study enable the development of robust classification models to aid in the planning and decision making on forest plantations in Brazil.