Abstract: Because of the traditional statistical methods, the time consuming and energy consuming spending on statistical the gross domestic product (GDP) is large. The night lighting remote sensing images can react directly to human social activities and provide a new way of thinking and technical support for the study of social economy. This paper takes Shandong Province as the research area, uses the NPP-VIIRS light image data and the GDP data of 137 counties in Shandong Province to establish many kinds of spatial relation models in order to find the optimal law of the spatial distribution of the two places. The experimental results show that the total GDP and the brightness of the total lighting are all about 0.9 at all levels of the goodness of fit R². The unit area GDP and the average lamplight brightness value are all about 0.85 at all levels of the goodness of fit R².The prediction accuracy of GDP at county level and city level by the optimal model is better than that of the unclassified one.

Key words: NPP-VIIRS, nighttime light image, county GDP, spatial correlation model, GDP prediction analysis