An Enhanced Hierarchical Framework for Mapping Urban Building Age Using Long-Term Landsat Time-Series Imagery

Accurate building age mapping is crucial for urban renewal, energy consumption estimation, and risk assessment. Recent studies used medium-resolution satellite image time series to estimate building ages by identifying breakpoints in spectral/index time series, making great advancement beyond the traditional single-year building façade analysis. However, their accuracy remains limited, which is primarily caused by the temporal mismatch between spectral/index time series breakpoints and actual building completion years, and the conventional binary scheme (renewal vs. non-renewal) which cannot separate unchanged old buildings from sprawl areas. To address these challenges, we propose a three-layer hierarchical temporal recognition framework. First, a new set of long-term spectral/index change features, combined with a supervised classifier, is used to acquire more detailed building change categories: unchanged, sprawl, and renewal. Second, high-confidence construction time windows for sprawl and renewal buildings are identified by integrating empirical index-based breakpoint detection with a multi-year tolerance strategy. Third, building completion years within these windows are further refined using short-term spectral/index variation features and a data-driven modeling approach. Experimental results in Guangzhou, a city characterized by complex building age structures, demonstrate state-of-the-art performance: the new method outperforms the traditional breakpoint-based method by 16% in single-year accuracy and attains 87.16% accuracy under a 2-year tolerance, matching the latter's performance only at a 3-year tolerance. The proposed framework produces a highly accurate and spatially detailed building age map, offering both a technical advancement in remote sensing urban analysis and a reliable data foundation for urban renewal planning and sustainable management.