Plot-scale population estimation modeling based on residential plot form clustering and locational attractiveness analysis

In many regions, urbanization has advanced to a stage that requires urban renewal, making precise population data essential for effective regional renewal and sustainable development. Therefore, this paper aims to disaggregate Jiedao-level (an administrative unit under the district) census population data to the Plot level. From an urban morphology perspective, the Gaussian Mixture Model (GMM) clustering algorithm was applied to classify the form of residential plots, assigning a type parameter for each type: the per capita housing area, to describe population density differences among the types. We then used Pearson correlation analysis to assess the relationship between POI density and population density at various bandwidths, identifying the optimal bandwidth for different POI types and calculating the overall POI density for each plot to evaluate its locational attractiveness. A regression model was established using per capita housing area, POI density, and total building area to derive population weight layers for estimating population at the plot level. The results of accuracy assessment show that using the morphological type parameter can effectively improve the estimation accuracy at plot scale, especially in areas with diverse land-use patterns and lower population density. However, our optimized locational attractiveness calculation method shows only a slight improvement to the method using a fixed bandwidth. This study develops a more accurate population estimation method of plot-level based on morphological classification， and highlights the population distribution characteristics of different types of residential plots, aiding urban decision-makers in developing targeted strategies for housing optimization and community resource allocation.