Mapping urban construction sites in China through geospatial data fusion: Methods and applications

Abstract

The rapid increase in Urban Construction Sites (UCSs) due to urbanization has become a global trend. UCSs are crucial for timely tracking of urban expansion and renewal progress, understanding settlement environments and human activities, and achieving Sustainable Development Goals (SDGs) 3 and 11. However, distinguishing UCSs from other land covers remains challenging, whether using spatial texture and spectral features or time-series characteristics. There is an urgent need for a universally applicable UCS mapping method at the national scale, a gap that current research has yet to fill. In this study, we proposed a method combining geospatial data with remote sensing data for national UCS mapping under medium spatial resolution. Additionally, we combine the UCS mapping results with SDGSAT-1 GLI data to evaluate the utilization status of new construction areas, thereby supporting SDG 11.3. The results showed that, for six representative cities, the F1-Score and Matthews Correlation Coefficients (MCC) for exposed UCS mapping results ranged from 98.83 % to 99.49 % and from 0.64 to 0.77, respectively. Variable importance detected in the Random Forest (RF) model highlighted that the key to identifying UCSs lay in geospatial information describing UCS spatial distribution, including distance to roads, city boundaries, and dust-proof nets. The assessment of the utilization status for new construction areas highlights the differences in the utilization status with which cities at various stages of development utilize these new areas. We then compared the ability of UCS distribution with existing impervious surface products in reflecting the dynamics of urban construction. The results showed that UCS spatial distribution could reflect urban construction patterns more timely and accurately, providing key insights for urban planners. Overall, this study provides a universal methodology that can be referenced for mapping land covers that have low separability in spectral and textural features in complex urban environments. The proposed method offers a cost-effective and reliable way to map nationwide UCS distribution, providing clear and timely spatial information for urban planning and achieving SDGs.