Análisis de la expansión y transformación de la mancha urbana en Bogotá (Colombia) y St. Louis (EE. UU.) durante el periodo 2014–2024 mediante procesamiento digital de imágenes

Abstract

This study analyzes the urban expansion of Bogotá (Colombia) and St. Louis (United States) between 2014 and 2024 using remote sensing techniques. Landsat 8 OLI (2014-2015) and Sentinel-2 MSI (2016-2024) satellite images were used, which were preprocessed with radiometric corrections, spatial harmonization (bilinear resampling at 10 m), and temporal normalization (October-December period), ensuring compatibility between sensors. Land cover classification was developed using a hybrid approach: unsupervised classification (K-means) followed by supervised classification with machine learning algorithms (Random Forest and SVM), supported by spectral indices (NDVI, NDBI, BSI, MNDWI). This allowed for the identification of six land cover classes according to the LCCS2 scheme, differentiating between urban density (dense vs. dispersed) and vegetation cover (dense vs. sparse). Validation was performed using: (i) RPAS photogrammetry in pilot areas, obtaining very high-resolution orthomosaics, and (ii) official land cover products (CORINE Land Cover for Bogotá, National Land Cover Database for St. Louis). The results allow for the generation of multi-temporal maps (2014, 2018, 2021, 2024), the quantification of urban expansion rates, and the identification of distinct growth patterns between the two cities. This multi-scale, binational methodological approach offers a replicable model for urban dynamics studies.