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

Area of Science:

• Geographic Information Systems (GIS)
• Spatio-temporal data analysis
• Population geography

Background:

• Accurate population distribution data is crucial for urban planning and resource allocation.
• Existing census data often lacks the fine-scale resolution required for dynamic human activity analysis.
• Location-based big data offers unprecedented opportunities for detailed population mapping.

Purpose of the Study:

• To develop a high-resolution ambient population dataset for mainland China using big data.
• To establish a robust methodology for allocating traditional statistical population data to a fine spatial grid.
• To validate the accuracy of the generated population data against independent street-level statistics.

Main Methods:

• Utilized Tencent user location big data as a proxy for ambient population.
• Employed a log-linear spatially weighted regression model to link big data with county-level statistical population data.
• Allocated statistical data to a 0.01° grid to create the POP2018 dataset.
• Validated POP2018 against street-level population statistics and compared it with WorldPop and LandScan datasets.

Main Results:

• The POP2018 dataset demonstrated a strong fit with actual permanent population data (R² = 0.91).
• POP2018 exhibited the lowest Mean Squared Error (MSE) compared to WorldPop and LandScan datasets.
• The methodology successfully generated fine-scale ambient population data for mainland China.

Conclusions:

• Location-based big data, specifically Tencent location data, is a valuable tool for ambient population estimation.
• The developed method provides a significant improvement in the spatial resolution and accuracy of population distribution data.
• This research contributes refined population data between census years and demonstrates a practical application of big data in population geography.