A gentrificação é caracterizada pela substituição da classe social predominante em uma área por outra de maior renda, em função de melhorias que acarretam a valorização da área, tais como o aumento da acessibilidade. Tal fenômeno pode ser preocupante, principalmente nas grandes cidades de países em desenvolvimento, nas quais não se anseia que a oferta de infraestrutura fortaleça padrões de segregação socioespacial. A principal motivação deste trabalho consistiu em verificar se as transições de uso do solo poderiam apontar a possível ocorrência de processos associados à gentrificação, sob influência de variáveis relacionadas à acessibilidade tais como a malha de transporte público, equipamentos de educação e saúde e a disponibilidade de emprego. Como estudo de caso, elegem-se distritos no sudoeste da cidade de São Paulo, área de ocupação heterogênea, com predomínio do uso residencial de baixa renda. Para representar um fenômeno que varia no tempo e no espaço, o estudo foi conduzido através da modelagem dinâmica via autômatos celulares, validando simulações de períodos passados para avaliar a influência das variáveis explicativas sobre as transições observadas. O trabalho foi beneficiado pela disponibilidade de dados espaciais para o horizonte de estudo (2000-2016), que, além de contemplar informações sobre o uso do solo, também categorizam as construções residenciais pelo seu padrão e porte. A probabilidade de cada transição celular foi calculada a partir do método de pesos de evidência, baseado no Teorema de Bayes. O processo de modelagem demonstrou que determinadas faixas de valores das variáveis apresentaram associação com o aumento da presença de estabelecimentos de comércio e serviços, verticalização, ocupação de terrenos vagos e redução do uso industrial. Verifica-se, de um lado, a consolidação da ocupação informal de baixa renda através do aumento dos padrões de tais construções, de outro, a expansão do desenvolvimento imobiliário voltado às rendas baixa e média. Os resultados satisfatórios na validação das simulações corroboram o sucesso do desempenho das variáveis eleitas como explicativas das transformações da forma urbana na área de estudo. Verifica-se grande potencial para aplicação da modelagem como ferramenta de planejamento, subsidiando análises fundamentadas para as políticas de uso do solo urbano.

The gentrification phenomenon is characterized by the replacement of the prevailing social class living in a residential area by another one with a higher income due to improvements in technical and social infrastructures, such as upgrades in the accessibility conditions. Gentrification is a matter of great concern, especially in big cities of developing countries, where infrastructure provision should not reinforce patterns of spatial segregation. The main motivation of this work is to verify if land use transitions could lead to possible gentrification processes, under the influence of accessibility-related variables such as the public transportation network, education and health equipments and the availability of employment. The case study comprises districts in the southwestern sector of São Paulo city, Brazil, an area marked by the heterogeneity of its land use, with a considerable predominance of low-income dwellings. In order to represent a phenomenon that varies over time and space, the study was executed by means of dynamic modeling via cellular automata, using the validation of past simulations to measure the suitability of each accessibility variable to explain the observed land use transitions. The work benefited from the availability of spatial data for the study horizon (2000-2016), which besides providing information on land use, also enable the categorization of residential and retail buildings according to their standard and size (one-storey x multi-storey). The probability of each cell´s transition was calculated using the weights of evidence method considering the explanatory variables, based on the Bayes´ Theorem. The modeling process showed that for different ranges of each variable, transition trends usually associated with gentrification took place, such as the increasing presence of retail and services, the construction of new buildings in previously non-residential areas, the occupation of vacant land and the reduction of industrial use. There is a consolidation of informal low-income occupation by the raising standards of those constructions, at the same time in which is possible to see the expansion of low and middle-income real estate development. The satisfactory results in the model validation confirmed the success of the explaining variables selected to model the urban form transformations within the study area, revealing a great potential of cellular automata models as a planning tool to support urban land-use policies. The gentrification phenomenon is characterized by the replacement of the prevailing social class living in a residential area by another one with a higher income due to improvements in technical and social infrastructures, such as upgrades in the accessibility conditions. Gentrification is a matter of great concern, especially in big cities of developing countries, where infrastructure provision should not reinforce patterns of spatial segregation. The main motivation of this work is to verify if land use transitions could lead to possible gentrification processes, under the influence of accessibility-related variables such as the public transportation network, education and health equipments and the availability of employment. The case study comprises districts in the southwestern sector of São Paulo city, Brazil, an area marked by the heterogeneity of its land use, with a considerable predominance of low-income dwellings. In order to represent a phenomenon that varies over time and space, the study was executed by means of dynamic modeling via cellular automata, using the validation of past simulations to measure the suitability of each accessibility variable to explain the observed land use transitions. The work benefited from the availability of spatial data for the study horizon (2000-2016), which besides providing information on land use, also enable the categorization of residential and retail buildings according to their standard and size (one-storey x multi-storey). The probability of each cell´s transition was calculated using the weights of evidence method considering the explanatory variables, based on the Bayes´ Theorem. The modeling process showed that for different ranges of each variable, transition trends usually associated with gentrification took place, such as the increasing presence of retail and services, the construction of new buildings in previously non-residential areas, the occupation of vacant land and the reduction of industrial use. There is a consolidation of informal low-income occupation by the raising standards of those constructions, at the same time in which is possible to see the expansion of low and middle-income real estate development. The satisfactory results in the model validation confirmed the success of the explaining variables selected to model the urban form transformations within the study area, revealing a great potential of cellular automata models as a planning tool to support urban land-use policies.