From ground to sky: reconciling light pollution measurements from above and below

Image credit: lightpollutionmap.info and Globe At Night
1513 words / 6-minute read

The world at night is changing rapidly. In 2023, scientists estimated that the night sky increased in brightness at a global average rate of 10% per year since 2011. Since night-sky brightness is determined by the amount of light on the ground, these results imply more artificial light at night in the environment. The resulting light pollution is now a worldwide problem with few parts of our planet untouched. And it affects much more than whether we can see the stars at night.

We have written here before about the complexities associated with measuring light pollution. We have global views of 'nighttime lights' from Earth-orbiting satellites. But their data are sometimes at odds with reports from ground-based observers. For example, despite the figure quoted above, a landmark 2017 study based on satellite measurements found a much slower rate of change. Can we reconcile these ways of knowing our planet at night?

Tension between data sources

​Satellite data from 2012-2016 indicated that during that time the radiance, or light intensity, of sources on the ground increased globally at an average rate of about 2% per year. That result relied on measurements made by the U.S. National Oceanic and Atmospheric Administration's Visible Infrared Imaging Radiometer Suite (VIIRS). This workhorse instrument flies aboard several NOAA satellites and sees our entire planet once a day.

The VIIRS "Day-Night Band" (DNB) detects sources of light on the night side of the Earth. But it wasn't designed for that application. As a result, it is insensitive to the blue wavelengths of light emitted in quantity by modern white light-emitting diode (LED) lighting technology. The outcome is that the VIIRS-DNB undercounts the light from newer installations of outdoor lighting compared to that from technologies that preceded LED.

Meanwhile, the Globe At Night program has for almost 20 years collected observations of night-sky brightness from citizen-scientists around the world. It asks participants to estimate the brightness by counting stars in particular constellations. Research has shown that the accuracy of these observations when considered in total is high, and thus they are scientifically valid.

Unlike the VIIRS-DNB, the human eye is sensitive to the blue light emitted by white LED lighting products. Scientists expected that a shift in the colors of light sources would result in a perceived change in the brightness of the night sky. Yet the 10% per year sky-brightness increase still came as a surprise. It suggested that the night sky was getting brighter at a rate few imagined to be so high.

Various sources of artificial light in the nighttime environment and some of the ways that light is sensed and measured. Adapted from Figure 2 in Linares Arroyo et al. (2024).

More than meets the eye

How can this evident tension between satellite radiance and ground-based visual observations resolved? In a recent paper, Spanish professors Salvador Bará (University of Santiago de Compostela) José Castro-Torres (University of Granada) suggested a way forward. Their investigation found that "these different results are not incompatible". They offered an explanation having to do with the colors of the light sources on the ground. That includes light seen directly by satellites and indirectly by human observers as skyglow. And it appears that indeed, ground-based observers are sensing light that the VIIRS-DNB misses.

In particular, the transition from legacy lighting sources to white LED seems to be the cause. But it's not just because VIIRS-DNB misses some of the light that LEDs emit. There is reason to believe that there is more light in the nighttime environment from LEDs than in earlier times. That remains true even after adjusting for population changes.

Chris Kyba (GFZ Helmholtz Centre for Geosciences, Germany) and coworkers found that during the five-year period of their study, published in 2017, the rate at which countries' light emissions increased roughly matched the rate at which their economic output rose. The result implied that countries were emitting more light than before in part because that light became cheaper to consume. That made their observation "inconsistent with the hypothesis of large reductions in global energy consumption for outdoor lighting" because of the introduction of LED.

But Bará and Castro-Torres argue that the change in night-sky brightness could have gone either way. It could "potentially either increase as a result of increased atmospheric scattering of blue light or decrease as a result of improved lighting fixtures that reduce horizontal emission."

These are two competing effects. On one hand, blue light scatters more strongly during its flight through the atmosphere than other colors. That could make the sky apparently brighter at night even if the quantity of light emitted on the ground didn't change.

But we also know that light emitted at shallow upward angles contributes significantly to skyglow. During the transition to LED, many new light fittings were installed that feature fully shielded designs. These decrease light emissions at all upward angles, including those at small angles. But even light directed toward the ground can still end up in the night sky after reflecting from various surfaces. It appears that the net result of all these influences is a real increase in night-sky brightness.

Connecting Earth and sky

One way to know for sure is to get a better handle on the connection between known light emissions on the ground and what the satellites see. We could then learn not only which kinds of sources contribute the most light that ends up in the night sky, but also the times of night when they are doing so. A group of scientists (and citizen-scientists) based in Germany recently did exactly that.

The Nachtlichter ("NightLights") team wrote a mobile device app to guide volunteers through the process of collecting basic information about outdoor lights they saw while walking defined transects through cities. Participants counted almost a quarter-million lights over a combined area of about 22 square kilometers. The recently published results offer great insight into the relationship between the number of lights on the ground and the resulting light emissions measured from space.

Results from the Nachtlichter survey showing the relationship between the counted number of outdoor light fittings per square kilometer (vertical axis) and the brightness of the corresponding location on Earth in VIIRS-DNB measurements (horizontal axis). Three groupings of lighting types are shown. Figure 3 from Nachlichter (2025).

Nachtlichter also provided some support for the idea articulated by Bará and Castro-Torres that some of the light not detected by satellites is due to unfavorable geometry. While it is straightforward to model the effects of street lighting, other kinds of outdoor lighting are not as easy to understand. "Some of these lighting applications, such as decorative and advertising lighting, produce a larger fraction of horizontally propagating light than modern street lighting does," they wrote. These are sources of light emitted at near-horizontal angles that is so important to skyglow formation. Yet satellite detectors can undercount that light. "It is therefore likely that some of the differences between the rates of change for skyglow that we calculate and those estimated from satellite data arise from changes in lighting practices or deployment."

Putting it all together

Bará and Castro-Torres asked whether the ongoing transition to LED, in changing the color of outdoor lighting, might account for the differences between the DNB and Globe At Night results. Their model took into account the physics of light scattering both in the atmosphere as well as inside the eyes of observers on the ground. In a previous paper, they found the latter is an important influence on the perceived brightness of individual light sources. And they speculated that nearby sources of light on the ground might bias the visual observations of night-sky brightness. The result could be inconsistent with satellite measurements.

They found that differences between the Globe At Night reports and DNB data seemed to disappear when observers were dark adapted. That is, their sensitivity to faint light increased by spending time in the dark before making their observations. It implied that there were no nearby, bright sources of outdoor light. To explain the observations "requires the existence of additional light sources that affect the Globe at Night observations but do not show up in the VIIRS-DNB data". This could be sources emitting in directions that the DNB doesn't see. Examples of this include lighted signs and indoor illumination escaping through building windows.

To make that work, the light emissions from sources that not detected by the VIIRS-DNB have to increase at a rate of about 6% per year. This adds to "the estimated 3% per year of the remaining lights deduced from the VIIRS-DNB measurements". It could be that Globe At Night observers were often located near light sources that raised the brightness of the night sky as they perceived it.

The work represents important progress in interpreting both the satellite and ground-based observations. But the case is not quite closed. "Settling this interesting issue requires gathering more detailed data on the lamp substitution processes in different regions of the world," Bará and Castro-Torres wrote. And we also need to know more about light-scattering conditions both in the atmosphere and in the eyes of observers.

With each passing year we learn a little more about how outdoor lighting is affecting nighttime conditions everywhere. Much of what we learn affirms basic principles about how to reduce light pollution. The most effective approaches use it only as needed in the proper places, times, amounts and colors. The recent work by Bará and Castro-Torres, and the Nachlichter team, draws us closer to the fuller picture we seek.