基于卫星遥感和POI数据的人口空间化研究——以广州市为例

以广州市为例,基于NPP/VIIRS夜间灯光、土地利用、POI（Points of Interest）等自然地理和社会经济因素,构建了人口空间化指标体系,采用主成分赋权法确定人口分布权重,利用GIS技术对人口统计数据进行了空间化处理。结果显示：综合考虑了自然地理和社会经济因素的人口空间化结果与真实的人口空间格局相吻合,空间分辨率为30 m,且相对误差绝对值＜25%的乡镇有62个,所占比例为36.47%;而不加POI数据得到的人口空间化结果,相对绝对值＜25%的乡镇有33个,所占比例约为20%,精度明显降低。结果表明：1）综合考虑NPP/VIIRS夜间灯光、土地利用、POI等自然地理和社会经济因素,有助于实现精度较高的人口空间化结果;2）将能够反映微观细节信息的POI数据引入人口空间化研究,扩展了人口空间化的数据源,并且可以提高人口空间化结果的精度。

Population Spatialization Based on Satellite Remote Sensing and POI Data:Guangzhou as an Example

The fine-scale spatial distribution of a population can provide assistance for decision-making regarding research, such as urban planning, economic decision-making, disaster prevention, and crime management, which is of considerable importance to the study of human activities. Considering Guangzhou as an example, this study constructed a population spatialization index system based on NPP/VIIRS night light intensity, land use, points of interest(POI), and other natural geographical and socio–economic factors. Further, the principal component weighting method was used to calculate the weight of the population distribution. According to which, geographic information system(GIS) technology was used to assign census data to each pixel. Finally, an accuracy test was performed by comparing the result of the population spatialization with the real population of townships of Guangzhou. The result of the population spatialization, considering natural geography and socio-economic factors,is consistent with the real population spatial pattern, and the spatial resolution is 30 m. The accuracy test result showed that the absolute relative error of 62 townships was less than 25%, which accounted for 36.47% of all townships in Guangzhou, indicating that the result of the population spatialization is highly accurate. The population spatialization result without POI data showed that the absolute relative error of 33 townships was less than 25%, which accounted for 20% of all townships in Guangzhou, showing that the accuracy of population of the spatialization result was significantly reduced. Compared with other studies on population spatialization, this study included the following characteristics: 1) a comprehensive consideration of NPP/VIIRS night light intensity,land use, POI, and other natural geography and socio-economic factors, which are helpful for achieving high-precision results of population spatialization and 2) POI data that can reflect surface features, such as urban land use types and population activity preferences. Additionally, POI data are easy to acquire, rich in data volume,have a high positioning accuracy, and have a better resolution regarding micro details. Introducing POI data into population spatialization research can expand the data source of population spatialization and improve the accuracy of population spatialization results.