Abstract: Population is a vital indicator of socioeconomic development and urban development planning, especially for developing countries like China. Accurate and high resolution information of population distribution is an important prerequisite to study human-land relationships. However, census data for any given field are inadequately to demonstrate the internal differences of population distribution. In this paper we tried to solve this problem by spatializing the population across Jiangsu Province, which is located in east China. Spatialization of statistical population is one of the vital means to achieve the integration of demographic data and spatial data. Moreover, it tends to reconstruct the spatial features of demographic statistics and supports the sustainable development of the region by providing relative data. In order to link the field between aggregated census data and geo-coded data, various techniques were used to disaggregate the census data. The satellite-measured DMSP/OLS night-time light imagery has been widely used for regional level mapping of socioeconomic activities due to its high temporal resolution, free availability and wide swath. However, because of the coarse resolution and data saturation of DMSP/OLS data, the limitations of applying this data source need to be taken into account. In this paper, population spatial processing is carried out by means of utilizing the theory and technology of RS and GIS. Specifically, data sources include three aspect: DMSP/OLS (night-time satellite imagery of operational line-scan system sensors on board of the defense meteorological satellite program); land-use data (the data was collected from the global nature recourses and it consists of six kinds and 21 subcategories in Jiangsu Province), and the sixth census data. It is proved that all the resources can be used to acquire the achievement, such as population distribution in certain area in China even in the world. Based on the analysis and the methods discussed above, SLM (Spatial Lag Regression Model) was used for population density estimation. We derived a population distribution map at 1 km×1 km grid cells in Jiangsu Province in 2010. In addition, we get the measures of fit (R-squared) of the model 0.93. According to the experiment, the validation of the resulting maps at county-level and town-level showed that average absolute value is comparatively high especially for those with high population density. That is to say, accuracy assessment results show that the DMSP/OLS night-time satellite data and land use data are suitable for restoration the spatial distribution of population and these data could characterize more explicit details. The accuracy of some region that the study observed tends to decline due to the fact that the population density is either too high or extremely low due to spatial heterogeneity. We conclude that a higher accuracy grid would be generated if more ancillary factors associated with population spatialization were incorporated in the future.

Key words: spatialization of statistical population, spatial lag regression model, DMSP/OLS, land use data, Jiangsu Province