Reducing Variability and Removing Natural Light from Nighttime Satellite Imagery: A Case Study Using the VIIRS DNB

Abstract

Temporal variation of natural light sources such as airglow limits the ability of night light sensors to detect changes in small sources of artificial light (such as villages). This study presents a method for correcting for this effect globally, using the satellite radiance detected from regions without artificial light emissions. We developed a routine to define an approximate grid of locations worldwide that do not have regular light emission. We apply this method with a 5 degree equally spaced global grid (total of 2016 individual locations), using data from the Visible Infrared Imaging Radiometer Suite (VIIRS) Day-Night Band (DNB). This code could easily be adapted for other future global sensors. The correction reduces the standard deviation of data in the Earth Observation Group monthly DNB composites by almost a factor of two. The code and datasets presented here are available under an open license by GFZ Data Services, and are implemented in the Radiance Light Trends web application.

Satellite Observations Show Reductions in Light Emissions at International Dark Sky Places During 2012-2020

“…As noted by Hyde et al [2], the data from the different IDSP are not independent observations, due to the nearness of the communities (meaning they are viewed from similar angles and through similar atmospheres). Furthermore, the new airglow correction has a similar value for locations at short distances from each other [4]. This results in correlations in the radiance data [10], but with an unknown degree.…”

“…Nighttime light emissions in the 500-900 nm band were obtained by the Visible Infrared Imaging Radiometer Suite Day-Night Band radiometer [5]. Monthly cloud-and moon-free composites were obtained from the Earth Observation Group [6, https://eogdata.mines.edu/download_dnb_composites.html], and corrected to reduce the influence of atmospheric airglow [4,7]. To reduce variability due to seasonal changes (and especially snow cover), we analyzed only the months of September, October, and November (as in Hyde et al [2]).…”

“…This change in calibration caused the radiance reported for regions of Earth without any installed lighting to increase in 2017 and 2018 compared to the 2012-2016 period. In this article, we re-examine these communities using a newly developed correction that reduces the noise in the satellite data [4], and including additional data from 2019 and 2020.…”

“…Here, we re-examine the first three hypotheses using monthly DNB datasets that have been corrected to center the radiance of unlit regions at zero [4,7]. The correction is derived by observing regions known to have no human settlements, and then interpolating to correct positions between these regions.…”

“…In addition, the correction reduces the noise in the dataset (due to variations in nighttime atmospheric airglow), and the effective sensitivity of the DNB is thereby increased for regions with small amounts of artificial light emission. Readers interested in more details about the correction are referred to Coesfeld et al [4].…”