What Remains Local in Preserved Historic Districts? Multimodal Evidence on Heritage Identity and Multisensory Homogenization in Yunnan

Historic districts are conserved through visible fabric, while locality is also sustained by everyday sounds, language, and social activity. This study examines the spatial patterning of perceived multisensory homogenization across five historic districts in Yunnan, China. We combine street-view imagery for all sites, synchronous field recordings at 340 sampling points on Ximen Street, pairwise human judgments, multimodal model scoring, and explainable machine learning. For the other four districts, auditory variables represent image-derived proxies for potential soundscape conditions, rather than measured acoustics. The results reveal a mismatch between visual preservation and auditory locality: streets that retain historic façades may still be perceived as losing distinctive soundmarks and everyday acoustic identity. In the fitted models, standardized signage, generic decoration, and consumption-oriented frontages are associated with higher visual homogenization predictions, whereas dialect-related cues, resident activity, and diverse everyday sound events are associated with stronger auditory locality. The findings support heritage governance that considers not only material appearance but also the social and acoustic conditions through which place distinctiveness is maintained.