Mapeamento pretérito e prognóstico da expansão urbana de Montes Claros/MG usando machine learning

• Ano de defesa: 2024
• Autor(a) principal: Laurentino, Carla Milena de Moura
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Uberlândia; Brasil; Programa de Pós-graduação em Geografia
• Palavras-chave em Português: Crescimento urbano; Montes Claros; MOLUSCE; Autômatos celulares; urban growth; cellular automata; CNPQ::CIENCIAS HUMANAS::GEOGRAFIA; Geografia; Crescimento urbano - Montes Claros (MG); Aprendizado do computador; Espaços urbanos - Montes Claros (MG); ODS::ODS 11. Cidades e comunidades sustentáveis - Tornar as cidades e os assentamentos humanos inclusivos, seguros, resilientes e sustentáveis.
• Link de acesso: https://repositorio.ufu.br/handle/123456789/41419; http://doi.org/10.14393/ufu.di.2024.159

Cities have shown steady growth in recent decades, and this growth is expected to continue in the coming years. Therefore, there is a greater need to monitor changes in land use and coverage, caused by the expansion of urban areas. It is these changes that cause impacts on the quality of urban life, especially depending on the social condition and the region of the city where one lives. The regions with greater environmental amenities are occupied by a population with greater purchasing power, on the other hand, the poor population resides in the most distant areas and with fewer urban resources. Therefore, this work aims to analyze changes in land use and coverage between the years 1985 and 2020, with emphasis on the urban area, and to develop the urban growth simulation of Montes Claros for the year 2050. To this end, Landsat 5, Landsat 8 and SRTM satellite images were used, which were manipulated and used to generate variables in the Qgis 2.18 software. To generate land use, the samples were inserted into the Random Forest (RF) classifier in the RStudio software. Then, the urban growth simulation was generated, using the cellular automata model (CA) and Artificial Neural Network (ANN) in the MOLUSCE plugin of the Qgis software. The results showed constant growth in the city of Montes Claros, between 1985 and 2020, there was an increase of 39.42 km² in the urban class. And, in the simulation for the year 2050, the city continues to grow, around 30.49 km². When generating land use, the model identified the variables MDE, Infrared, NDVI and NDWI as the most important for the good performance of the classification. Thus, when analyzing every year, that is, from 1985 to 2050, significant growth is identified in urban regions located in the east and south directions, where there is the presence of a population with lower purchasing power and popular subdivisions. In the simulated period, an urban expansion was identified towards Serra do Mel, an environmental protection area, however, this urban advance does not violate legal limits. Therefore, this area is expected to see an increase in gated communities and high-end homes, which are already being developed there.