Morphological Persistence of the Principal Street Framework and Changing Commercial Geography in Xi’an’s Ming Walled City, 1611–1995: Implications for Sustainable Conservation and Regeneration

Historic urban conservation requires distinguishing durable spatial structures from changing uses. Focusing on Xi’an’s Ming Walled City, this study treats 1611 as historical context and compares geolocatable commercial records from 1735 (n = 39), 1933 (n = 36), and 1995 (n = 93). Historical GIS, kernel density estimation (KDE), nearest-neighbor ratio (NNR), standard deviational ellipse (SDE), and separate axial-network models are used as complementary descriptive tools. The street networks and commercial locations analyzed here are derived from period-specific plates in the Historical Atlas of Xi’an and were manually digitized in AutoCAD 2020. The study does not claim survey-grade positional accuracy; cross-period comparisons refer to documented geolocatable records and to the descriptive recognizability of principal street alignments. The results show a recognizable principal street framework alongside substantial redistribution of documented commercial activity: southern and southwestern concentrations in 1735, a compact south-central pattern in 1933, and a broader multi-district distribution in 1995. KDE sensitivity checks show that hotspot counts vary with bandwidth, while SDE describes broad differences in mapped spatial extent and mean-center position; these positional differences are not interpreted as calibrated historical displacement. Network statistics are interpreted within individual reconstructions rather than as standardized cross-period measures. On this basis, the study proposes ‘stage-contingent correspondence’ as a descriptive proposition for reading how recognizable long-duration street structure and changing functions coexist within the same historical stage. The main contribution is a boundary-controlled, scale-sensitive workflow that separates structural continuity from period-specific functional intensity and supports cautious conservation diagnosis. Source incompleteness and unresolved 1995 NNR inputs limit stronger inferential claims.