GPSS Doctoral Student Publishes Study on Environmental Drivers of Urban Shrinkage in the Tokyo Metropolitan Area

GPSS doctoral student Zheng Hao, together with his supervisor Professor Zhang Runsen and their colleagues, has published a new study in Habitat International titled “Environmental drivers of urban shrinkage in the Tokyo metropolitan area: A machine learning analysis of nonlinearity and spatial heterogeneity.”

The study investigates how environmental conditions shape urban shrinkage in the Tokyo Metropolitan Area. By integrating nighttime light data with an interpretable machine learning framework, the research offers a new perspective on how shrinkage and growth coexist within Japan’s largest metropolitan region. 

Fig. Analytical workflow of this study

Highlights of Methodology and Findings

Using VIIRS nighttime light data from 2012 to 2022 together with a systematic set of built and natural environmental factors, the study combines Theil–Sen trend analysis, Random Forest, Geographically Weighted Random Forest, and SHAP analysis to identify shrinkage patterns, examine nonlinear relationships, and reveal spatial heterogeneity.

The study found that:

(1) Urban shrinkage and growth coexist in the Tokyo Metropolitan Area, forming a three-tier ring-like pattern with a highly active core and more vulnerable outer areas.

(2) Among the 24 environmental factors examined, 10 factors, comprising 4 natural factors and 6 built-environment factors, were found to significantly and independently influence urban growth and shrinkage.

(3) Several key environmental drivers show clear nonlinear effects, indicating that their influence on urban vitality changes across thresholds rather than following simple linear relationships.

(4) The effects and importance of environmental factors vary substantially across different locations, highlighting strong spatial heterogeneity in the mechanisms of urban shrinkage within the metropolitan area.

By revealing the environmental drivers behind metropolitan shrinkage, this study provides a data-driven framework for urban planners and policymakers. The findings highlight the need for spatially differentiated planning strategies to respond to demographic decline and changing urban conditions in large metropolitan regions.

The full study can be accessed via DOI:

https://doi.org/10.1016/j.habitatint.2026.103795.

Readers are invited to explore this research and engage in discussions on urban shrinkage, environmental conditions and machine learning.

Fig. Spatial distribution of the impacts of different environmental factors on urban shrinkage: (a) FAR (floor area ratio); (b) LSG (land & sea gradient); (c) RI (residential intensity); (d) TMS (train & metro station); (e) FLD (flood risk); (f) II (industrial intensity); (g) EQ (earthquake risk); (h) CF (commercial facilities); (i) BS (bus stop); and (j) SLP (slope)