Surface modeling of human population on subdistrict scale using SPOT5 image and census data: A case study of Xiamen, P. R. China

Abstract

From local scale to globe one, the spatial distribution of human population is essential to the research in the fields crossing Demography, Humanities and Social Sciences, Environmental Science, etc. The census data of human population traditionally used inevitably have the disadvantage of low spatial and temporal resolution, and it's difficult to integrate the census data with multi-source ones as well due to the different survey units in space dimension. While the spatial distribution of statistical demographic data based on RS can solve these problems, therefore nowadays it has received much attention in the area of spatial distribution of human population. However, previous research work mainly focused on the regional scale, and little was on urban scale. In our work, firstly gathering census data on the subdistrict scale of 2007 year in the study area of Xiamen City, Fujian Province, P. R. China, meanwhile acquiring the information of residential settlements and that of road transportation by interpreting the fused SPOT5 image (with spatial resolution of 2.5m) of the year, we identified the gravity center of human population of subarea in Xiamen, and then calculated the population density of every buffer zone of each district. Using the SPSS software, we built a population density model and simulated the spatial distribution of total population of every district in Xiamen of 2007 year. Additionally, we estimated the precision by the calculated relative error with census data on the subdistrict scale. Finally, we acquired the distribution map of human population density (with grid size of 100 m). Compared with previous research, in the work, we used preferable spatial distribution indicators of human population derived from high resolution remotely sensed image, including buildings and road transportation information. Based on which and according to urban spatial structure, we make the human population simulation much more appropriate for the real world. The integration of RS and GIS can provide useful reference for the research on human population spatial distribution with high precision on the urban subdistrict scale. And the research result can also be referenced by the studies in the areas, such as crime management, economic decision-making, disaster warning, infrastructure construction and so forth.