Evaluating low-carbon urban planning strategies across urban typologies using causal machine learning

With climate change and rapid urban migration, cities face the question of where to locate new residents to minimize travel and related CO2 emissions. While the IPCC suggests general planning concepts such as compact or transit-oriented development, a place-specific understanding of how this translates to the local context is lacking.
Here, we introduce a novel approach to assess the induced transport CO2 emissions for different urban planning strategies and determine the city-specific, optimal locations for low-carbon residential development. We use double machine learning to estimate the non-linear effect of the built environment on travel-related CO2 emissions for each neighborhood while controlling for residential self-selection. Using a selection of global cities with diverse urban typologies, we demonstrate how optional low-carbon urban planning differs across city typologies. Based on population projections for 2050, we assess the global mitigation potential compared to a business-as-usual scenario characterized by increasing urban sprawl. Our results underscore the importance of evidence-based, spatially differentiated compact development to decarbonize the transport sector.