What satellites can (and can't) tell us about outdoor lighting changes

Image credit: NASA Earth Observatory/Goddard Space Flight Center/J. Stevens/M. Román​
1820 words / 7-minute read

The launch of the first artificial satellite into space in 1957 was a great leap for human technology – at least as important as the “giant leap” to the Moon that followed a decade later. But perhaps the greatest discovery of the Space Age was not about the cosmos. Rather, in a sense it was about humanity discovering itself and its place in space. In particular, the first view of our home planet as a contained system inspired some humility at a time when Cold War tensions ran high. Reflecting on the famous Apollo 8 “Earthrise” photograph 50 years later, astronaut William Anders wrote, “We set out to explore the moon and instead discovered the Earth.”

Only by flying instruments high above the Earth could we capture information about the Earth on regional to truly global scales. Aerial photography could only reach so far. Remote sensing of the Earth from orbit was required to truly view the system in a stark and obvious form. They are indispensable to our modern understanding of light pollution, detecting the light of our cities escaping the atmosphere. Satellite observations of “night lights” have been applied to diverse scientific questions, illustrating everything from land use patterns to energy consumption to disaster recovery to levels of urbanization.

There is only one workhorse satellite remote sensing platform for nighttime lights that offers anything like an ideal combination of characteristics. These include global coverage, nightly observations, and useful ground resolution. The instrument is called the Visible Infrared Imaging Radiometer Suite Day/Night Band (VIIRS-DNB), and it flies aboard three U.S. National Oceanic and Atmospheric Administration (NOAA) polar-orbiting satellites: the Suomi National Polar-orbiting Partnership (NPP) satellite, and the NOAA-20 and NOAA-21 weather satellites. It's the basis for a lot of published scientific papers — 667 in 2024 alone, according to Google Scholar.

Every scientific instrument has its shortcomings, and the VIIRS-DNB is no different. Given what we know about it, do we really understand what it’s telling us? More importantly, do we know what it can possibly tell us? For that, here we will consider a recent example.

Case study: 'cool pavements' and skyglow

A useful application of satellite remote sensing of nighttime lights is to test the effects of making changes to the built environment that might influence light pollution. If we want to know whether, for example, changes to outdoor lighting reduce light emissions directed into the night sky, looking down from overhead is a good option. Some of the light emitted by lighting installations reflects off the ground before it travels upward into the night sky and may be detected by satellites. What happens if the optical properties of the ground itself change?
​
We set out to answer that question in a particular context. Cities around the world are increasingly looking for ways to mitigate the effects of climate change on their residents. In hot environments like deserts, concerns include extreme summertime heat that can be deadly. Often full of materials like concrete, cities tend to absorb sunlight during the daytime and re-radiate that energy in the form of heat during the overnight hours. This urban “heat island” effect can keep air temperatures high overnight, taxing air conditioning systems.

​One idea to offset this effect is to make surfaces in the built environment more reflective to sunlight. Materials that absorb less sunlight during the day don’t heat up as much, and at least in principle they help keep overnight temperatures lower. Although the intent is to reduce the heat island effect by making surfaces more reflective to heat, they also reflect some visible-wavelength light. We wondered whether that might result in brighter night skies over cities. The logic is simple: brighter surfaces reflect more light,  so more street lighting goes up into the night sky, causing skyglow to increase. But is this what really happens in practice?

Overhead view of cool pavement overcoatings being applied to a residential neighborhood street in Phoenix, Arizona. Image courtesy of the City of Phoenix.

VIIRS-DNB data might be able to tell us. We modeled the effect of adding reflective materials to roadway surfaces, commonly referred to as “cool pavements”. These materials have a paint-like consistency and are applied directly atop existing street pavements. There is some evidence that they work, lowering air temperatures in treated areas by up to a few degrees Celsius. Based on field measurements of the cool pavement materials applied to some streets in Phoenix, Arizona, U.S., we calculated the expected change in nighttime brightness of the roadway surfaces at between about 2-6%. Although the materials appear bright to the eye, much of the radiation they reflect is in the infrared part of the spectrum, some of which we perceive as heat.

To test this model, we looked at VIIRS-DNB measurements of the nighttime brightness of Phoenix made over more than a decade, including the time when the City of Phoenix applied cool pavement materials in certain neighborhoods. We compared the measurements to those of nearby neighborhoods whose streets did not receive the pavement treatments. The results were recently published in the Journal of Quantitative Spectroscopy and Radiative Transfer. 

What we found was surprising and a little disappointing: we couldn’t tell whether application of the cool pavement treatments made any difference in the satellite data. We saw unexpectedly high variations in the brightnesses of all of the neighborhoods from one month to the next. Those variations were so large that the brightness changes due to the cool pavement treatments our model predicted were lost in the noise. In the end, we couldn’t rule either in or our real changes that might affect the brightness of the night sky over Phoenix.

Satellite-detected nighttime brightness (“radiance”) of a Phoenix neighborhood receiving cool pavement treatment in October 2021. Filled circles represent monthly average radiance values from 2012 to 2024. The solid line tracks the slowly varying, long-term trend. Figure 3, Barentine (2025).

All we could say with statistical certainty was that those brightness changes could not have exceeded +14 percent. That meant that in addition to being unable to test our model predictions, we also couldn’t rule out the possibility that there was no change at all due to the cool pavements application.

Satellite data interpretation: a tricky business

What might be going on here? Other researchers have also noted big variations in VIIRS-DNB measurements in cities (e.g., here and here). In writing this post, we reached out to Professor Chris Elvidge, Director of the Earth Observation Group at the Colorado School of Mines. Elvidge is among the forefathers of satellite night-lights observations, researching and writing about the subject for over 25 years. There are a lot of reasons why brightnesses seen by satellites may vary: changing view angles, clouds, differing amounts of dust in the atmosphere, snow cover, and illumination of the ground by moonlight.

But Elvidge suspects a particular effect that explains the source of the urban variations:  electrical “ficker”. Electric lighting is susceptible to output changes from the alternating current used to power it. In 2022 his group published a paper in which they pointed out the possibility that flicker could explain the variations seen in VIIRS-DNB data. Those changes can happen with a frequency that is different from the frequency of VIIRS-DNB measurements. While we can account for many other influences on the brightness measurements, he wrote, “none of these adjustments will reduce the radiance instability introduced by flicker.”

Does lighting flicker explain what we saw in the cool pavements study? It could well affect measurements made from one night to the next. But we averaged together many observations over the course of each month and considered changes only to the monthly averages. We also found that while variations in brightness were matched in nearby neighborhoods, they were located far enough from each other that coordinated flickering of light sources isn’t very likely. So it leaves a bit of a mystery as to why low-density residential neighborhoods lit mostly by streetlighting during the overnight hours seem to vary so much in brightness from one month to the next.

The need for a dedicated night lights satellite mission

All remote sensing measurements are obtained in imperfect conditions. We can control these circumstances to some extent; for example, the effect of moonlight on nighttime lights observations can be limited by acquiring images only when the Moon is below the local horizon. Some influences, like weather, can’t be completely avoided. Long-term observations can help even out the extremes in order to get at the true underlying trends.

But some of the difficulties we experience are due to satellite designs that are less than ideal for this kind of application. Facilities like the VIIRS-DNB are almost afterthoughts, added to satellites for reasons other than studying night lights. While they provide a trove of information about the distribution of artificial light at night on our planet, they’re not really built specifically for that reason. 
Consequently, some of their characteristics fall short of what we might want. A prime example of this has to do with the “blindness” of the VIIRS-DNB to blue light. The instrument is only sensitive to light with wavelengths of between about 500 and 900 nanometers (one-millionth of a meter). But a lot of modern LED lighting emits considerable light just below the 500-nanometer cutoff. 
​
This presented a problem for light pollution researchers from the very beginning when the first copy of the VIIRS instrument was launched in 2012. It was around that time that white LED light was beginning to replace earlier lighting technologies. As lighting modernization projects proceeded, VIIRS-DNB observations saw changes that could easily be misinterpreted. Cities seemed to be getting darker in the satellite images, whereas measurements on the ground clearly indicated that they were no darker than before the lighting retrofits; see the example below. We can account for effects like this to some extent, yet we are left to guess the exact amount of light that the VIIRS-DNB simply doesn’t see.

Three views of the city of Milan, Italy, from Earth orbit. The panels labeled “A” and “B” are astronaut photos taken aboard the International Space Station in 2012 and 2015, respectively, courtesy of the Earth Science and Remote Sensing Unit, NASA Johnson Space Center, with identification and georeferencing by the European Space Agency, the International Astronomical Union, and Cities at Night. They show the city center before and after the conversion of street lighting from high-pressure sodium to white LED sources. Panel “C” shows the change in radiance detected by the VIIRS-DNB from 2012 to 2016. Cooler colors indicate an apparent radiance decrease. Figure 5 in Kyba et al. (2017).

While a few satellites have been launched whose designs are better suited to observing (all the light from) lighting installations on the ground, none is ideal. Some have high resolution and good sensitivity to light, but they only observe limited parts of the Earth. Others don’t regularly pass over any given location, leaving us unable to tell how things are changing. All of these missions leave us inadequately informed about how light pollution is changing around the world.

Some research groups have argued for new satellite missions dedicated to measuring nighttime light. We published one such argument in 2021, beginning with a strong scientific case for the kinds of measurements we need to address research questions. From this we specified the kinds of characteristics future satellite missions would need to answer those questions. So far, none of the spacefaring countries have taken up this proposal. The VIIRS-DNB remains our best data source, even with its shortcomings.

For now, the question in our paper about cool pavements and skyglow remains imperfectly answered. It may be that there’s nothing to worry about, but for now we can’t say that with any certainty. We developed the method for being able to tell from the data whether changes to surfaces like the application of cool pavements make a difference in their nighttime brightness. But for now we lack the complementary machinery of remote sensing measurements that can enable us to know for sure.