Population spatialization at building scale based on residential population index—A case study of Qingdao city

Abstract

The study of population spatialization has provided important basic data for urban planning, development, environment and other issues. With the development of urbanization, urban residential buildings are getting higher and higher, and the difference between urban and rural population density is getting larger and larger. At present, most population spatial studies adopt the grid scale, and the population in buildings is evenly divided into various grids, which will lead to the neglect of the population distribution in vertical space, and the authenticity is not strong. In order to improve the accuracy of the population distribution, this paper studied the spatial distribution of population at the building scale, combined the digital surface model (DSM) and the digital elevation model (DEM) to calculate the floor of buildings, and proposed a new index based on the total floor area of residential buildings, called residential population index (RPI). RPI is directly related to the number of people a building can accommodate, so it can effectively estimate the population of both urban and rural areas even if the structure of urban and rural buildings is very different. In addition, this paper combined remote sensing monitoring data with geographic big data and adopted principal component regression (PCR) method to construct RPI prediction model to obtain building-scale population distribution data of Qingdao in 2018, providing ideas for population spatialization research. Through field sampling survey and overall assessment, the results were basically consistent with the actual residential situation. The average error with field survey samples is 14.5%. The R2 is 0.643 and the urbanization rate is 69.7%, which are all higher than WorldPop data set. Therefore, this method can reflect the specific distribution of urban resident population, enhance the heterogeneity and complexity of population distribution, and the estimated results have important reference significance for urban management, urban resource allocation, environmental protection and other fields.

Modeling the Spatial Distribution of Population Based on Random Forest and Parameter Optimization Methods: A Case Study of Sichuan, China

“…The goal of feature selection is to choose the most relevant or important features from the original feature set to enhance model performance, reduce dimensionality, and mitigate the risk of overfitting. Mao et al [52] employed the Pearson correlation coefficient (PCC) to select Points of Interest (POI) information, considering values with an absolute correlation coefficient greater than 0.5 as indicators of social factors influencing Residential Population (RPI). Lu et al [53] utilized the Pearson correlation coefficient (PCC) to assess the correlation between Numerical Weather Prediction (NWP) variables and wind power.…”

Partition refinement of WorldPop population spatial distribution data method: A case study of Zhuhai, China

“…The selected accuracy evaluation indices include the Standard Deviation (SD), Normalized Root Mean Square Error (%RMSE), and Coefficient of Determination (R 2 ). The calculation formulas are presented in Eqs ( 8)- (10). The SD evaluates the dispersion of the data, %RMSE reflects the overall accuracy level of the model error, and R 2 measures the goodness of fit of the estimated results.…”

“…The spatial distribution factor was established by fusing multi-source data, which reflects the spatial distribution of the population more accurately than single data. Dond et al [8], Zhang et al [9], and Mao et al [10] investigated the spatial population distribution and geographical elements, obtained the weight coefficient of the grid population distribution, and distributed the population of the grid. Yong et al [11] and Peng et al [12] used geographic and multi-source information fusion data to achieve refined population allocation and simulation.…”

“…Previous studies found that WorldPop has the highest overall accuracy and spatial resolution in the public population spatial distribution dataset [10,22]. However, the land cover data reflect the natural attributes of land rather than the social and economic attributes of land generated by human activities combined with land.…”