Inferring the heterogeneous effect of urban land use on building height with causal machine learning

Abstract

ABSTRACTMachine learning has become an important approach for land use change modeling. However, conventional machine learning algorithms are limited in their ability to capture causal relationships in land use change, which are important knowledge for planners and decision makers. In this study, we showcase the usefulness of causal machine learning to understand the heterogeneous causal effect of changing land use on building height through a case study in Shenzhen, China. Also, by leveraging the power of causal machine learning, we identify the key conditions under which greater building height change would occur after land use interventions. The results suggest that building height would increase by 3.68 floors and 1.61 floors on average if industrial land is converted to residential and commercial, respectively, and 2.35 floors if commercial land is changed to residential land. The heterogeneity of causal effect is also captured for different land use change scenarios. The factor analysis based on the decision tree algorithm reveals the key conditions on which greater building height increase would occur by changing land use. Overall, this study can contribute to literature by providing an effective approach to counterfactual land use change modeling with enhanced explainability.

Keywords

• Causal inference
• machine learning
• urban land use
• building height