Análise espacial da restauração no bioma Cerrado: automação e representação cartográfica na plataforma Araticum

• Ano de defesa: 2025
• Autor(a) principal: Rocha, Jairo Matos da
• Orientador(a): Ferreira, Manuel Eduardo
• Banca de defesa: Ferreira, Manuel Eduardo , Galbraith, David Roberto , Teixeira, Denilson
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Goiás
• Programa de Pós-Graduação: Programa de Pós-graduação em Ciências Ambientais (PRPG)
• Departamento: Pró-Reitoria de Pós-graduação (PRPG)
• País: Brasil
• Palavras-chave em Português: Política ambiental; Aprendizado de máquina; Mapeamento; Acesso à informação
• Palavras-chave em Inglês: Environmental policy; Machine learning; Mapping; Access to information
• Área do conhecimento CNPq: CIENCIAS AGRARIAS::RECURSOS FLORESTAIS E ENGENHARIA FLORESTAL::CONSERVACAO DA NATUREZA::RECUPERACAO DE AREAS DEGRADADAS
• Link de acesso: https://repositorio.bc.ufg.br/tede/handle/tede/14467

Monitoring Conservation Units in the Cerrado Brazilian biome involves challenges related to effective monitoring and verification of the conservation status of each area in situ. This work becomes more efficient with a comprehensive approach that combines remote sensing techniques, serving as a complement to machine learning and online processing resources. Therefore, the objective of this study was to build an automated high-spatial-resolution classification, and textual and visual description methodology for restoration areas in the Cerrado. Georeferenced data from restoration areas were used for land use classifications based on three auxiliary products with a 4 m resolution, and then the sampling points were delimited using the geohash method. Decision-making at the time of classification was carried out by the Random Forest machine learning algorithm. A total of 8,161 polygons of restoration areas were identified in 11 federative units, with the states that stood out most in terms of the amount of restored area being São Paulo (7,705 ha) and Piauí (3,714 ha). The classification of the areas, with 80% accuracy, covered 19 land use classes, ranging from forest cover, agriculture and pasture. With the program developed for generating automated reports, it was possible to synthesize the descriptive information of the restoration areas in text and image format, in addition to a final land use classification map