The key factors of the level of urban resilience: an analysis of 13,000 cities worldwide

Climate-related events have been increasing in frequency and uncertainty. The majority of the world’s population is urban and, since future growth will occur in cities, people and resources will be even more concentrated, enhancing their vulnerability to shocks and stresses. In order to address the impacts of climate change and promote a sustainable development, cities must build resilience. Therefore, the main objectives of this research were to understand the factors (independent variables) that influence the level (dependent variables) of urban resilience, to what extent they can be measured by existing indicators and databases available, what is the level of urban resilience as calculated by a composite index, and what are the key factors influencing it. Despite this being a concept without a common definition and hard to be operationalized, a thorough literature review resulted in the definition of the variables and indicators for an integrated framework to measure urban resilience, which is divided into five dimensions ‘ economic, environmental, institutional, physical and social. When comparing this theoretical framework with existing databases, it was possible to notice that none of them would allow a complete assessment, evidencing the importance of better data availability. However, since the GHS Urban Centre Database 2015 had indicators available for all dimensions, besides a sample of 13,000 cities worldwide, it was chosen for the purposes of this research. To build a composite index from the dependent variables, a technical procedure of supervised learning techniques ‘ such as multivariate analysis, normalization, weighting and aggregation, and uncertainty and sensitivity analysis ‘ was performed. The result was a robust score that is able to rank cities’ performance. Its interpretation lead to the understanding that indeed a comprehensive assessment can provide a different perspective on what it means to be resilient. Parameters like emissions of CO2 and PM2.5 emissions, green area per capita, built area and flood exposure carried a significant weight on the level of urban resilience. Some unsupervised learning techniques, such as dendrograms and cluster analysis, provided allowed the identification of some groups of cities with similar scores. Additionally, evaluating the relation of the independent variables with the composite index through regression analysis, it was evidenced that temperature, precipitation, land use efficiency and development level are the main factors influencing it. As a result, it was clear that the integrated framework proposed can provide urban planners and local governments an important decision-making tool. Keywords: Urban resilience; framework; indicator; supervised techniques; unsupervised learning