Exploring the Correlations Between SNPP-VIIRS Nighttime Light Data and Population From a Multiple Scale Perspective

Abstract

Remote sensing nighttime light (NTL) data are widely used for fine-grained estimation of population due to their effectiveness in monitoring anthropic lights at night. However, there is a lack of research exploring the correlations between population and NTL data at various administrative and grid scales, as well as their related impact factors. Thus, to analyze the spatial effect of NTL on NTL-based population estimation, five models (linear, quadratic, exponential, logarithmic, and power function) were used in this letter to examine the correlations between NTL and population at multiscale by using the Visible Infrared Imaging Radiometer Suite data from the National Polar-orbiting Partnership. Results show that the correlations between NTL and population were significantly related to geographic scales. The power function model had the highest fitting accuracy between NTL and population when the grid scales were less than 5 km. The quadratic model had the highest precision on grid scales larger than 5 km. It showed a sharp increase in R <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> values between 0.5 and 10 km as the scale increase. At different scales, the correlations between NTL and population was negatively impacted by temperature, Normalized Difference Vegetation Index (NDVI), and Road Network Density (RND), but they were positively impacted by the Relief Degree of Land Surface (RDLS). We also found that RDLS had the greatest impact on model accuracy, followed by NDVI and temperature. In addition, RND had a low impact on the accuracy of population estimation on rough grids.