Abstract: Readily available and accurate data on spatial population distribution is essential for understanding, and responding to, many social, political, economical and environmental issues, such as humanitarian relief, disaster response planning, environment impact assessment, and development assistance. Research on spatialization of demographic data plays an important role in grid transformation of social-economical data. Furthermore, as gridded population data can be effectively interoperate with geospatial data and remote sensing images, it is a useful supplement to census data. This paper reviewed spatialization methodologies, predictive modeling factors and typical datasets in the literature of population data spatialization research. Shortcomings of demographic data and advantages of spatial population distribution data are compared and summarized firstly. The spatialization methodologies are grouped into three categories, i.e., population distribution models from urban geography, areal interpolation methods and spatialization methods based on remote sensing and GIS. Population models from urban geography include the Clark's model and allometric growth model. The areal interpolation methods had been distinguished by point based method and area based method. Spatialization methods based on remote sensing and GIS are most widely used in nowadays, which can be further grouped into three categories for two reasons: one is the relationship between population and land use, urban area, traffic network, settlement density, image pixel characteristics, or other physical or socioeconomic characteristics, and the other is the calculation strategy. Various methods mentioned above have their own application environment and limitations. We reviewed the principles and applicability of every method in detail. After that, we generalized the frequently used factors in the spatialization process, involving land use/land cover, traffic network, topography, settlements density, night light, texture variable, and spectral reflectance. In the meantime, some typical research cases about the factors also were exemplified and analyzed. In addition, we introduced a few widely used spatial population distribution datasets or influential population spatialization projects. They consisted of China km grid population datasets, UNEP/GRID, GPW/GRUMP, LandScan, AfriPop & AsiaPop & AmriPop. The producers, resolution, characterization year and generation method of each one were presented exhaustively. Based on the above review, we discussed the current research problems and outlined research priorities in the future. The problems include the temporal inconsistency of input data, coarse resolution of demographic data, lack of in-depth study on scale effect, the scarcity of time series products and few validation works. To deal with these issues, more studies should be conducted to the following aspects: comprehension of population distribution mechanism, calculation of consistency and validation of existing datasets, application of multi-sources remote sensing data and volunteered geographic information, continuous space-time simulation of population distribution in the typical areas, sub-block-level population estimation, self-adaptive spatialization method which integrates multiple elements and multiple models. In summary, the research on spatialization of demographic data has made breakthroughs in the past two decades. Meanwhile, there are a few problems that need to be solved immediately. Since these two aspects had been reviewed as comprehensively as possible, we hope issues discussed in this paper could enlighten and promote the future study in this field.

Key words: datasets, modeling factors, population data, spatialization methodology