Reconstructing 36 Years of Spatiotemporal Dynamics of Slums in Brazil by Integrating EO and Census Data

Officially, by 2019, more than five million households in 734 Brazilian municipalities were in slums, accounting for 7.8% of the total households. Due to the country’s continental dimensions, there is a diversity of forms and names for these settlements, officially called by the Brazilian Institute of Geography and Statistics (IBGE 2020a) substandard agglomerations (SAs), a proxy for slums, mapped by IBGE from official census data and other ancillary information and comprises areas containing at least 51 households on public or private land whose residents do not have legal land tenure, face substandard road structures and lot sizes, and have no adequate sanitation services. The SA data is available as georeferenced vector files, last updated in 2020, containing approximately 13,000 polygons. Although the delimitation of SA considers the entire country’s territory, it is only done every 10 years with the demographic census and considers population and household counts, not urbanized area expansion from a remote sensing perspective. On the other hand, the MapBiomas initiative produces annual collections of land use and land cover maps, between 1985 and 2020 in the case of collection 6, using Landsat time series (spatial resolution of 30 m) and applying artificial intelligence techniques such as random forest and U-net in a cloud computing environment (Google Earth Engine), with a total of 23 different classes of land use and land cover, including urbanized areas. By combining the two data collections (SA and MapBiomas), this research uses the SA boundaries (fixed in 2020) overlaid on the patches of urbanized areas in Brazil (annually between 1985 and 2020) and, from this, calculates the annual increments of urbanized area for each SA, subdividing the calculations into two spatial subsets: inside and outside the SA. This is done to answer the following question: What are the spatial–temporal dynamics of the SA (slums) urbanization in Brazil? Additionally, the predominant types of land cover transitions to urbanized areas in these spatial subsets are calculated. Finally, we aggregate the resulting statistics at regional (states) and local (municipalities) scales for all 5573 Brazilian cities, analyzing the most predominant patterns. The data processing is based on cloud processing techniques using the Google Earth Engine (GEE) platform. The results pointed out that, in 36 years throughout Brazil, there was an expansion of the occupation of SA areas of 84.000 ha, equivalent to 3.8 times the size of the city of Amsterdam, in the Netherlands. In the two main cities of the Amazon area, Manaus and Belém, the growth of the urban sprawl inside SAs corresponded to more than 50% of the total urbanized area growth in the time series, suggesting that slum growth is not an exception but almost a rule in the region. Compared to São Paulo and Rio de Janeiro, the two more consolidated Brazilian capitals where the slum growth corresponded to 24% and 10%, respectively, the results indicate that the pace and magnitude of slum expansion in different regions of the country present distinct temporalities and spatiality opposing already consolidated areas to new fronts of urban poverty expansion. Furthermore, considering the transitions between classes, we estimated the urban sprawl of SA over native vegetation areas over time and compared it with the same type of transition in areas of formal urbanization. As a result, we found that SA corresponds to the lowest total and proportional occupations of native vegetation.