The datasets contain low resolution image data [1] that can be used to approximately remove natural light from the monthly composite nighttime images produced by the Earth Observation Group (EOG) using the Visible Infrared Imaging Radiometer Suite (VIIRS) Day-Night Band (DNB) [2]. Natural light includes airglow and reflected light from stars, as well as polar light at high latitudes. The low-resolution images can be expanded and then subtracted from the EOG monthly composites, using python scripts included in this data publication [1]. More details are provided in the data description and in Coesfeld et al. (2020). The background correction is based on the VIIRS-DNB monthly composites data from April 2012 to December 2019 without stray light correction produced by the EOG, and the associated file containing the number of cloud-free observations (Earth Observation Group, 2015). Locations away from human settlements and artificial lights were selected using the Global Human Settlement Layer (GHSL) population density map from 2015 [3] and the VIIRS-DNB 2015 vcm-orm annual composite by the EOG (Earth Observation Group, 2015). The python script we used in Coesfeld et al. (2020) to select locations on an equally spaced grid over the global DNB extent (from 75N to 65S) is included. The six images of the EOG composite are processed separately. This code identifies locations on a 72x28 grid (2016 locations total). In order to optimize the location of the specified grid points, the Global Human Settlement Layer (GHSL) population density and VIIRS-DNB annual composite of 2015 are consulted.