USE-Lab Seminar -- Building Up: A Satellite-Based Method for Measuring China's Urban Development


Past Event

USE-Lab Seminar -- Building Up: A Satellite-Based Method for Measuring China's Urban Development

May 24, 2023
4:00 PM - 4:45 PM
Event time is displayed in your time zone.
CEEM Conference Room

Speaker: Sebastiano Papini, ETH Zurich

Abstract: One of the biggest challenges in studying the inner urban structure of cities is the lack of high-quality data that includes the vertical dimension required for a systematic analysis of urban change and development. In this study, we present a new method for measuring the floor space (building footprint and height) of 37 major Chinese cities over a period of 5 years using medium-fine-grained satellite imagery. Our multi-task learning approach involves determining the surface area covered by buildings (occupied land) and using a regression model to determine building height from imagery. We used Sentinel-1 and -2 satellite images as our primary data source due to their large cross-sectional and longitudinal scope and unrestricted accessibility. We also analyzed the preprocessing steps of the reference data and their impact on measuring building floor space. In addition, we highlight the superiority of our density measure over the use of night light data often employed in such research setups. Overall, our study represents a first milestone in providing comprehensive and accurate data on building floor space over time for urban studies.

Speaker Bio: Sebastiano Papini is currently pursuing a Doctoral degree at the Department of Management, Technology, and Economics at ETH Zurich. His research focuses on the intricate relationship between urban economic development and transport networks. Specifically, his work delves into the examination of inner city structures during periods of rapid urbanization, as well as the economic ramifications associated with reachability and urban densification. Notably, his research is characterized by the development and utilization of unique datasets driven by data science techniques.