人口空间分布典型数据的时空差异OA

Accurate spatial population distribution data are indispensable for socioeconomic development,urban planning,and public policymaking.With the proliferation of diverse data sources,including statically modeled datasets and dynamic real-time feeds,a systematic and quantitative evaluation of their characteristics,biases,and optimal application scenarios is urgently required to inform reliable,data-driven decisions.This study aimed to address this gap by conducting a comprehensive spatiotemporal analysis of three representative data sources and comparing them with authoritative census data to elucidate their strengths,limitations,and complementary roles.This study employed Quanzhou City,a major coastal urban center in China,as a case study.We constructed two core metrics,Relative Difference and symmetric Mean Absolute Percentage Difference(sMAPD),to quantify the discrepancies.Three data sources were analyzed:static WorldPop data representing modeled long-term distribution;hourly dynamic Baidu Heatmap data reflecting location-based service demand;and mobile phone location data capturing passive device presence.These datasets were subjected to pairwise comparisons across multiple spatial and temporal scales and systematically compared with national population census data.This analysis yielded several key findings.First,regarding scale-dependent performance,all datasets showed high correlation with census data at the township level and above,confirming their utility in capturing broad spatial patterns.At the district scale,static WorldPop data demonstrated superior overall consistency,establishing its strength for macroscale assessments.At the township level,dynamic data showed slightly higher correlation coefficients than those of WorldPop;however,WorldPop and mobile data exhibited lower sMAPD values than those of Baidu data,highlighting a divergence between pattern similarity and numerical accuracy.Second,dynamic data exhibited systematic spatial differences,overestimating populations in central urban areas and underestimating populations in rural zones.sMAPD values were also temporally volatile,with intraday fluctuations reaching significant levels.Between the dynamic sources,Baidu Heatmap data displayed greater temporal volatility and spatial polarization than those of mobile location data.Third,data accuracy was strongly influenced by land use type.Relative Difference values were minimal in residential areas but significantly larger in commercial and public service zones.Fourth,the differences stem from distinct data generation logic:static data reflect institutional population frameworks,mobile data capture the presence of ambient activity,and Baidu data sense explicit demand intensity.This study concludes that no single data source is optimal.Static data are most suitable for macroscale and long-term strategic planning and resource allocation because of their stability.Dynamic data are essential for monitoring short-term population mobility and fine-grained spatial patterns.Specifically,mobile location data are preferable for understanding overall activity patterns and commuting behaviors,whereas Baidu Heatmap data excel in real-time sensing of demand-intensive areas for applications,such as commercial planning or emergency response.We propose a synergistic framework for practical applications using census data as the fundamental benchmark,static data to delineate the macropopulation structure,and dynamic data to monitor and analyze spatiotemporal changes.This study provides a replicable methodological framework for multisource population data evaluation and offers concrete,scenario-specific guidance for data selection,thereby enhancing the scientific basis for urban governance and planning.