Enlightening the long-term future of cities through a prospective approach : what our models say about the Bordeaux metropolitan area ; Éclairer le long terme des villes par une approche prospective : ce que disent nos modèles de la métropole de Bordeaux

Cities are at the heart of the organization of the human habitat and concentrate a growing share of the world's population. Logically, they must face environmental and social challenges that humanity is experiencing. They are thus victims of the environmental crisis, suffering the consequences of air pollution, heat waves, floods, hurricanes, and other intense events. But they are also actors of these problems as they often bring together economic, political and social powers. In addition, the construction of urban systems is a long-term process, with buildings remaining in place for several decades or even centuries, and urban shapes drawing a grid not easily reversible. In this context, it is essential that parts of cities that we build today integrate both the short-term aspects (financial, aesthetic, adapting to the needs of owners or clients) and their long-term impacts. Several visions of future city, at least how it should tend to be, have emerged over the last few decades, from a technological city (or "smart city") to a sustainable city through a collective city. Within this framework, this work proposes to apply the principles of prospective modeling to urban planning issues in order to contribute to the consideration of environmental issues. To feed our models, we describe a replicable methodology for creating quantitative archetypes of urban shapes and buildings from French public data. This approach, based on classical techniques (principal component analysis and k-means), confirms some urban characteristics such as the difficult definition of the peri-urban area. Our prospective model then consists of a bottom-up optimization model based on the TIMES framework whose originality lies in the representation of the soil and its uses, a second spatial projection model using constraint maps and finally a convolutional neural network used to estimate the relative impacts on temperature of the calculated urban modifications. This modeling chain is then applied to a few prospective scenarios for the ...