China’s urban development is shifting from extensive expansion to intensive improvement, making the identification of inefficient stock land essential for sustainable urban renewal. Yet, existing approaches are often limited by incomplete data sources and low spatial precision. To address these issues, this study proposes a scalable framework that integrates multi-source big data, including land-use surveys, socioeconomic statistics, spatiotemporal trajectories, and ecological metrics. Using Shenzhen as a case study, we developed a multidimensional evaluation system across social, economic, and ecological dimensions, comprising eight specific indicators. Indicator weights were objectively determined using the entropy weight method, and GIS-based spatial analysis (mean-standardization and Moran’s I) was applied to characterize spatial patterns. Results identify 65.37 km² of inefficient land—about 7% of Shenzhen’s construction land in 2019—exhibiting an “edge aggregation and corridor extension” pattern, mainly distributed along urban–rural fringes and administrative boundaries. The spatial configuration is shaped by historical development, ecological constraints, and the city’s spatial structure. This study provides an objective and replicable framework for the precise identification of inefficient land, supporting data-driven urban renewal and spatial governance in high-density cities.

China’s urban development is shifting from extensive expansion to stock-oriented renewal, making the precise identification of inefficient land stock essential for sustainable spatial governance. To overcome the limitations of single-source data and coarse delineation, we propose a multi-source geospatial framework integrating land-use surveys, socioeconomic statistics, spatiotemporal mobility trajectories, and ecological indicators. Using Shenzhen as a case study, we construct an eight-indicator system across social, economic, and ecological dimensions and apply entropy-based objective weighting to support GIS-based weighted overlay evaluation. The results identify 65.37 km² of inefficient land in 2019, accounting for approximately 7% of Shenzhen’s construction land, with a distinctive “edge aggregation and corridor extension” pattern concentrated along urban–rural fringes and administrative boundaries. Inefficient land is highly uneven across districts, with Longgang (21.11 km²) and Bao’an (12.57 km²) contributing 51.5% of the total and exhibiting statistically spatial clustering (p < 0.01). The observed configuration reflects path-dependent historical development and policy–ecology constraints, including the interaction between ecological control boundaries and peripheral expansion. Overall, by integrating multi-source spatiotemporal big data within a multi-dimensional evaluation framework, the framework offers an objective and transferable approach for diagnosing inefficient land stock and informing targeted urban renewal strategies in high-density cities worldwide.

The preservation and spatial reconstruction of historical and cultural heritage are vital for enhancing national cultural soft power and cultural confidence. However, such efforts are often hindered by incomplete or missing historical data, especially when sites no longer exist and can only be inferred from fragmented texts or visual references. To address this challenge, this study proposes a knowledge graph-driven approach to three-dimensional (3D) spatial reconstruction of historical heritage, specifically for scenarios inferred from classical literary works. Our method begins with the semantic extraction of architectural entities and spatial relationships from narratives, emphasizing the "entity-relation" structure. A classical architectural knowledge graph encodes key features such as building types, functions, stylistic elements, and spatial layouts, serving as the basis for generating historically informed models. By integrating textual architectural parameters with contextual environmental features, we construct 3D models that simulate the original spatial form with structural coherence and stylistic authenticity. We further perform spatial clustering based on architectural attributes within the graph, reflecting functional and stylistic groupings that align with historical logic. Moreover, a multi-channel control mechanism guides fine-grained editing and re-rendering of architectural details, informed by spatial structure and aesthetic constraints from historical context. A case study of the Grand View Garden from The Story of the Stone demonstrates our approach: a domain-specific knowledge graph guides 3D reconstruction of both micro-level details and macro-level layouts, effectively restoring the novel's historical ambiance and architectural diversity. This research provides a novel framework for digital reconstruction of cultural heritage under data scarcity and offers a promising direction for integrating literary humanities with computational modeling in cultural preservation and public engagement.

China's urban development is transitioning from an extensive expansion model to an intensive improvement approach. Promoting urban renewal through stock land resources has become a key focus in contemporary urban spatial governance. As cities enter the stock development era, accurately identifying the quantity and spatial distribution of inefficient stock land is crucial for improving land use efficiency. However, existing identification methods for inefficient existing space face challenges including limited data sources, over-reliance on expert experience and insufficient identification accuracy. These factors impede the realization of large-scale, high-precision, and universally applicable identification frameworks. This study proposes an identification method integrating multi-source big data, including land surveys, socioeconomic data, spatiotemporal trajectories, and air quality metrics. Using Shenzhen as a case study, we developed an index system with three dimensions (social, economic, and ecological) and eight indicators. The entropy weight method determined indicator weights, while mean-standardization and Moran’s I index characterized spatial distribution. The study identified a total of 65.37 km² of inefficient existing land in Shenzhen, accounting for approximately 7% of the city's construction land in 2019. Clusters mainly appear in the northwestern part of the city, including administrative boundary zones and urban fringe areas. Longgang District (21.11 km²) and Baoan District (12.57 km²) collectively contribute 51.5% of the inefficient land. Shenzhen's inefficient existing land exhibits an "edge aggregation and corridor extension" spatial pattern, and its formation is subject to the compound influence of historical development patterns, ecological control policies, and urban spatial structure. This study provides a scientific basis for urban renewal and inefficient land redevelopment through the integration of multidimensional big data and spatial statistical analysis. The proposed framework offers replicable technical support for sustainable urban governance and precision regeneration strategies in high-density Chinese cities.