How can cities use machine learning for urban climate change mitigation? A systematic map

In recent years, numerous cities across the world have shown interest in AI approaches to help with achieving carbon neutrality. There is a need for city policy makers and researchers to get overarching guidance or receive an overview of potential application areas and approaches. Currently, however, there is no comprehensive analysis of the existing body of research on the use of machine learning (ML) in urban climate change mitigation.
With this research, we provide an overview of the literature at the intersection between ML research in climate change mitigation and its role in urban governance and policymaking. We conducted this review by applying a systematic map framework that transparently displays mature research areas, as well as more nascent and uncertain ones. Specifically, we created the systematic map through manually coding 3,944 records for relevance, followed by a comprehensive analysis of 1,215 relevant papers published between 1994-2021. In the review, we systematically map the areas of application but also provide an assessment of their climate mitigation potential.
Key findings include that research is taking place across 57 countries and 250 cities. We introduce over 40 impact areas. We find strong alignment between the most research impact areas and IPCC urban mitigation options with high reduction potentials. And we outline three main roles for ML such as using it for optimization towards more efficient processes, demand forecasting as well as simulation of impact of planned measures.
We then interpret findings in relation to governance and climate policy as well as discussing ways through which the potential of ML can be realized at scale and in alignment with high mitigation impact and public sector demand. We conclude with key recommendations for urban policy makers. For example, we call for integration of urban governance considerations into this emerging research nexus.