Estimating Chinese residential populations from analysis of impervious surfaces derived from satellite images

Abstract

Gridded population datasets are essential for displaying spatial distributions of residential populations. They are widely used in urban planning, decision-making, disaster assessment, and public health. However, the grid resolution may affect the accuracy of population distributions, and this issue should be further explored to obtain a clearer understanding. Therefore, it is crucial to determine appropriate grid sizes for ascertaining the spatial characteristics of population distributions on a large scale. The choice of the grid resolution for a population dataset generally depends on the source datasets and the requirements of a specific project. While previous studies on grid resolutions were conducted predominantly in small study areas, this study focused primarily on the population distribution of the whole of China at 14 different scales, from 100 m to 1 km (with a 100-m interval), and from 1 km to 5 km (with a 1-km interval). Population spatialization was conducted using census data from 351 cities in China at the city level and impervious surface data derived from satellite images. Dasymetric mapping method was employed to estimate the population distribution, and the scale effects of the population estimates were examined at different scales of impervious surface data. The results of an accuracy assessment of the population estimates using county-level census data demonstrated that the impervious surface data were useful and effective when estimating residential populations with dasymetric mapping. The scale effects had varying degrees of accuracy of the estimated populations derived at different scales of impervious surface data, and a scale of 2–4 km was deemed optimal for estimating the residential population distribution based on impervious surfaces while using the dasymetric mapping method.