Urban boundary extraction and sprawl analysis using Landsat images: A case study in Wuhan, China

Shougeng Hu, Luyi Tong, Amy E. Frazier, Yansui Liu

Research output: Contribution to journalArticlepeer-review

70 Scopus citations


Periodic monitoring and multi-scale characterization of urban sprawl is essential for improving urban planning and development. However, historical sprawl analysis is not well suited for the neo-urbanization occurring in most cities in China due to the limited data available. This paper proposes a concise and cost-effective method for automating the extraction of urban boundaries (UBs). The method uses integrated land-use information entropy (LUIE) model along with ordinary Kriging based on a gridded land-use map derived from Landsat imagery to extract UBs. Results indicate that overall extraction accuracies greater than 90% were obtained using an 800m-resolution LUIE combined with Kriging. The method was applied to identify UBs in Wuhan, China during 1987-2010, and the UBs were characterized at multiple scales and analyzed using landscape metrics. Results show varied landscape dynamics at local administrative and city scales. The study demonstrates that the method for UB identification and multi-scale analysis has the potential to contribute to sprawl monitoring and measurement at multiple spatial scales. Moreover, the findings from this study can potentially guide policy makers and urban planners tasked with understanding and controlling development occurring under neo-urbanization strategies in China.

Original languageEnglish (US)
Pages (from-to)183-195
Number of pages13
JournalHabitat International
StatePublished - Jun 1 2015
Externally publishedYes


  • China's neo-urbanization
  • Land-use information entropy model
  • Landsat images
  • Landscape metrics
  • Urban boundaries
  • Urban sprawl

ASJC Scopus subject areas

  • Urban Studies


Dive into the research topics of 'Urban boundary extraction and sprawl analysis using Landsat images: A case study in Wuhan, China'. Together they form a unique fingerprint.

Cite this