A global typology for cross-city learning

The IPCC SR on Cities which will be compiled during the upcoming seventh assessment cycle will focus on climate solutions for different types of cities. However, a thorough and up-todate underpinning of what different categories of cities may learn from each other is missing. Here, I bridge this research gap by developing a novel global typology of cities using Deep Embedded Clustering (DEC), grouping urban areas based on key socioeconomic , infrastructural, and environmental attributes. A major contribution of this typology is the integration of highresolution ODIAC CO2 emissions data and OpenStreetMap (OSM) road network data, enabling a more detailed characterization of urban carbon footprints and infrastructure disparities. Then, by linking the city groups to research topics available from a recently developed database of the city-specific case study literature, I can paint a detailed picture of the cross-city learning potential. My results show that although 3,000 cities with fast population growth have not been covered in city-specific case studies, there are almost 1800 similar cities that have more than one study, particularly related to management practices and emissions, that these cities may learn from. By contrast, an albeit smaller group containing cities with fast-growing emissions are facing knowledge gaps, and consequently a low learning potential regarding research on emissions. More established cities with lower emission growth, where scientific knowledge about emission reductions is arguably less urgently needed, have large learning potentials.