夜间灯光数据在不同尺度对社会经济活动的预测

以广东省为研究范围，采用双对数线性回归模型、香农信息熵模型和差别指数模型，从21个尺度分析NPP-VIIRS夜间灯光数据与经济普查数据中的企业数、企业从业人员数、营利收入以及资产总值的相关性。结果显示：① 夜间灯光数据对经济活动的预测精度随尺度的增大而提升，在预测能力上，城市尺度高于镇街尺度，镇街尺度高于500 m网格尺度。② 在更为精细的19个网格尺度分析中发现，1 km网格尺度出现预测精度的显著提升，在10 km网格尺度预测精度提升相对较高并达到了0.69，在35 km网格尺度预测精度开始趋于稳定，不再随着网格尺度显著增长。③ 在0.5~50 km网格尺度之间，网格尺度越精细，产业类型越呈现单一和不均衡分布的发展特征。随着网格尺度的增加，研究区内的产业类型越趋于多样化和均衡分布的发展特征。

The Ability of Nighttime Imagery in Monitoring Economic Activity in Different Scales

This study uses linear regression model to explore the scale threshold of the prediction of social and economic activities by nighttime data (NPP-VIIRS) from multiple scales such as city, town, and grid in Guangzhou. The third national economic census data is the number of enterprises, employees, total revenues and assets of the enterprise. Furthermore, the research establishes 19 scales, applies Shannon’s information entropy and the index of dissimilarity to analyzes the different precisions of nighttime data in different scales to economic activities from the aspects of diversity, and the balance of industrial structure in the study area. The results show that the R² of nighttime data and total revenues is 0.20 in 0.5 km scale. As the scale increases, the forecasting ability is more grounded. The predicted value in the town street scale is 0.63. It began to stabilize and R² was 0.8 above 35 km scale. In addition, the industrial structure in the study area is characterized by unity and uneven development at a scale of 0.5 km. With the increase of the scale, the industrial structure in the study area has gradually become more diverse and balanced. The outcomes demonstrate that nighttime data can predict economic activities at various scales, and there still exist thresholds for the scale of prediction accuracy. This has guiding significance for understanding the regional industrial structure and optimizing the industrial structure of each region.