Containing urban sprawl in China: A cross-city evaluation of urban development boundaries using U-Net deep learning

Abstract

Urban Growth Boundaries (UGBs) have been used in developed countries to manage urban sprawl, but China's introduction of Urban Development Boundaries (UDBs) is a more recent development. However, the effectiveness of UDBs in China, with its vast territory and spatial disparities, remains unclear. This study evaluated UDBs' effectiveness in containing urban growth by combining historical data from Landsat images with predictive simulations from the U-Net Deep Learning model, focusing on eleven cities that piloted UDBs in 2014. The results show that UDBs are expected to promote more compact urban growth in the long term (2022–2035) compared to the pilot phase (2014–2022). Starting from the year of introducing UDBs, many cities continued to experience leapfrogged and scattered development patterns. But U-Net simulations predict that UDBs will encourage infill development and curb piecemeal and leapfrog sprawl by 2035. While UDBs improve urban aggregation and regularity, irregular and fragmented boundary delineation under stringent land quotas may limit their ability to curtail sprawl. Cities like Beijing and Suzhou, with limited developable land within UDBs, may experience spillover growth. These findings highlight the need for continuous assessment and context-specific policies to maximize UDB effectiveness. The study demonstrates that the U-Net model offers a scalable and adaptable approach for simulating urban growth, providing valuable insights for managing urban containment.