人口数据空间化研究综述

Research progress in spatialization of population data

人口数据空间化研究旨在发掘和展现人口统计数据中隐含的空间信息,并以地理格网或其他区域划分的形式再现客观世界的人口分布,具有重要的科学意义。人口空间分布数据有助于从不同地理尺度和地理维度对人口统计数据形成有益补充,其应用广泛,相关研究方兴未艾。主要从以下3个方面对人口数据空间化研究进行综述:①主要空间化方法的原理及其适用性;②空间化中用到的建模参考因素,并结合具体应用案例分析其作用机理;③典型人口空间化数据集。在此基础上,分析了现阶段人口数据空间化所运用的输入数据的质量和详细程度、尺度效应及时空分辨率、长时间序列数据集和精度检验等方面存在的问题;并探讨了人口数据空间化未来的研究方向。

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, di- saster response planning, environment impact assessment, and development assistance. Research on spatializa- tion 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 da- ta 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, are- al 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 sens- ing and GIS are most widely used in nowadays, which can be further grouped into three categories for two rea- sons： one is the relationship between population and land use, urban area, traffic network, settlement density, im- age pixel characteristics, or other physical or socioeconomic characteristics, and the other is the calculation strat- egy. 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 fac- tors also were exemplified and analyzed. In addition, we introduced a few widely used spatial population distri- bution datasets or influential population spatialization projects. They consisted of China km grid population data- sets, UNEP/GRID, GPW/GRUMP, LandScan, AfriPop ＆ AsiaPop ＆ AmriPop. The producers, resolution, char- acterization 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 volun- teered 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 enlight- en and promote the future study in this field.