A semi-parametric geographically weighted (S-GWR) approach for modeling spatial distribution of population

Spatial Distribution of Population (SDP) has been recognized as a fundamental indicator of various studies including ecosystem assessment. To estimate SDP with fine resolution at a regional scale, an S-GWR model approach based on a land use map was developed. The model enhances SDP estimation accuracy by considering geo-spatial variation of population density and absolute accuracy in a demographic statistics unit that might introduce significant biases. The model is applied in estimating SDP of Shandong province, China, in 2000 with a resolution of 1 km. It was validated against census data and two common datasets for GPWv3 and CGPD both at the prefecture scale and sub-prefecture scale. The validation revealed that the mean absolute percentage error of SDP based on the S-GWR model (GSDP) is approximately 0 at the prefecture scale, which shows better performance than the other two datasets. The validation at the sub-prefecture scale in Tancheng county shows a mean absolute percentage error of 12.79% for GSDP in 17 townships, which is less than that of CGPD (15.37%) and GPWv3 (18.76%). Furthermore, spatial analysis of the error indicated that the S-GWR model spread the error into the region of Tancheng with the least percentage of towns (35.29%) with a percentage error larger than 15%, where the percentage of CGPD and the percentage of GPWv3 are 47.06% and 58.82%, respectively. The findings from the study demonstrated the great potential and value of the S-GWR model for regional SDP estimation.