Destination–Activity Representation Gaps and Their Built-Environment Signatures: A Multisource Big Data Study of Four Local Urban Systems in Guangzhou

Platform-listed destinations are widely used to represent urban activity opportunities, yet their correspondence with realized activity may vary across local contexts. This study examines destination–activity representation gaps in four bounded local activity systems in Guangzhou, China. Mobile positioning activity, platform-listed destinations, population-positioning surfaces, roads, land cover, and street-view imagery were integrated at a 25 m grid scale. A spatially out-of-fold ridge model estimated activity from destination context and within-area percentiles of observed and destination-predicted activity were used to identify contrasting representation gap types. Destination correspondence was modest overall and spatially uneven across the four cases. Activity-rich/destination-light cells were more prevalent in the two island/peripheral cases, while their intraday profiles resembled those of destination-supported high-activity cells. These cells also showed lower population co-location on average, while contiguous patches revealed substantial local heterogeneity. Roads and land cover provided limited transferable discrimination, while selected raw street-view elements modestly improved annual activity prediction within the central sample. The findings show that platform-listed destinations and realized activity are related but non-equivalent representations and that their divergence varies across local activity systems.