Fine-scale population spatialization data of China in 2018 based on real location-based big data

Abstract

Accurate location-based big data has a high resolution and a direct interaction with human activities, allowing for fine-scale population spatial data to be realized. We take the average of Tencent user location big data as a measure of ambient population. The county-level statistical population data in 2018 was used as the assigned input data. The log linear spatially weighted regression model was used to establish the relationship between location data and statistical data to allocate the latter to a 0.01° grid, and the ambient population data of mainland China was obtained. Extracting street-level (lower than county-level) statistics for accuracy testing, we found that POP2018 has the best fit with the actual permanent population (R² = 0.91), and the error is the smallest (MSE_POP2018 = 22.48 <MSE_WorldPop = 37.24 <MSE_LandScan = 100.91). This research supplemented in the refined spatial distribution data of people between census years, as well as presenting the application technique of big data in ambient population estimation and zoning mapping.