MODELING THE URBAN-RURAL CLASSIFICATION OF LAND USE AND LAND COVER USING LOGISTIC REGRESSION: THE CASE OF THE FEDERAL DISTRICT IN THE 2022 BRAZILIAN CENSUS

Abstract

Until the 2010 census, Brazil's urban-rural classification was based on administrative boundaries and local legislation. For the 2022 census, the Brazilian Institute of Geography and Statistics (IBGE) used high-resolution satellite imagery for the preliminary classification. Researchers then refined this classification in the field by integrating empirical knowledge with remote sensing. This study aims to model the urban-rural classification using land use and land cover categories derived from remote sensing products. Focusing on the Federal District, we used logistic regression, treating the urban-rural classification as the dependent variable and the proportions of vegetated, agricultural, and non-vegetated areas from the MapBiomas dataset as the independent variables. The model achieved an accuracy of 97.3%, with the proportion of non-vegetated areas showing the strongest association with urban classification. These findings can be used to estimate urban-rural status between censuses and support the planning of sample-based surveys in regions undergoing land use and land cover changes. This approach relies solely on data already available from MapBiomas. Additionally, the paper discusses the feasibility of using environmental characteristics for prediction instead of traditional approaches that primarily rely on social criteria to define urban and rural areas.