1. Introduction
Artificial light at night is one of the clearest indicators of human activity available via remote sensing. Images of the Earth at night are therefore an extremely useful tool for research involving human communities and their interaction with the environment. Night light data have been used in the past to study economic variables [
1,
2], socio-economic properties [
3,
4,
5], population [
6,
7] and population density [
8], built area [
9,
10], power consumption [
11], greenhouse gas emissions [
12], gas flaring [
13], atmospheric chemistry [
14], skyglow (light pollution) [
15,
16,
17], the epidemiology of illness related to light exposure [
18], among other analyses [
19].
In 2012, there was a dramatic change in the quality of available nighttime Earth observation data. The new data has greatly improved resolution, accuracy and sensitivity, and the scope of possible applications is therefore likely to increase. The purpose of this paper is to alert researchers to these new data sources and to promote their use. The data sources are highlighted in three case studies that examine patterns of light use and energy consumption. The limitations of the data are discussed within the case studies themselves and summarized in the results and discussion. All of the night light data analyzed in this paper are freely available online.
1.1. Data Sources
Nearly all of the analyses of night light data thus far have been based on measurements from the Defense Meteorological Satellite Program-Operational Linescan System (DMSP). In 2012 a new instrument flown on the Suomi National Polar-orbiting Partnership satellite became operational. The Visible Infrared Imaging Radiometer Suite Day-Night Band (VIIRS DNB) images the entire Earth nightly at a resolution of about 750 meters. The VIIRS DNB has 45–88-times better spatial resolution than DMSP, and 14-bit compared to 6-bit digitization [
20]. Suomi NPP orbits the Earth 14 times per day on a 16-day repeat cycle. The average time at which images are taken is near 01:30 in local solar time, but can vary by over an hour depending on latitude and the particular cycle. Because the overpass time is near midnight, high latitude sites are imaged during night for a greater portion of the year than was possible with DMSP, which had a
∼19:30 overpass time.
Global data products based on VIIRS DNB are produced by the Earth Observation Group of the National Oceanic and Atmospheric Administration. While data from individual passes of the satellite are available, most researchers are likely to prefer using “cloud-free composite” images that are based on many passes of the sensor over the target on cloud-free moonless nights. Instead of having 750-meter resolution, cloud-free composite data are binned into a 15-arcsecond grid spanning from 65 south to 75 north latitude. The pixel size therefore depends on the latitude. The background from atmospheric airglow emission is faint compared to city lights and is not yet subtracted [
21]. Cloud-free composite images are currently available at
http://ngdc.noaa.gov/eog/download.html. As of the moment of writing,DNB “stable light products” with transient sources removed (e.g., aurora, fires, fishing boats) have not yet been published. A detailed description of DNB is presented in [
20].
The second new data source is astronaut photographs of the Earth at night. Astronauts have been taking such images for years, but the quality greatly improved with the installation of the NightPod instrument in 2012. All photographs taken from the International Space Station are available at “The Gateway to Astronaut Photography of Earth”, which is run by NASA:
http://eol.jsc.nasa.gov. Finding a nighttime image of a specific city among the millions of images is, however, a difficult task. The “Atlas of astronaut photos of Earth at night” was developed to solve this problem. The Atlas provides an open directory of geotagged images of cities at night, with links to the original images at the NASA site [
22]. The main page of the Atlas is at
http://www.citiesatnight.org/, and access to the database itself is available at
http://www.nightcitiesiss.org/. The original Atlas was put together by experts, but its future expansion will mainly come through a related citizen science portal for classifying and georeferencing ISS images [
23]. A tutorial on how to use the Atlas database is available at
http://tinyurl.com/qfolkq6.
Figure 1.
Some important sources of light at night. (a) Area lighting with no direct uplight; (b) area lighting including direct uplight; (c) architectural lighting; (d) lit windows; (e) illuminated sign with direct uplight emission; (f) illuminated sign with no direct uplight; (g) auto headlights; (h) searchlight; (i) gas flare. (a–g) by C Kyba; (h) “Tribute in Light” by Mike Hvozda, from Wikimedia Commons; (i) “Hammerfest dusk LNG” by Andreas Rümpel, from Wikimedia Commons.
Figure 1.
Some important sources of light at night. (a) Area lighting with no direct uplight; (b) area lighting including direct uplight; (c) architectural lighting; (d) lit windows; (e) illuminated sign with direct uplight emission; (f) illuminated sign with no direct uplight; (g) auto headlights; (h) searchlight; (i) gas flare. (a–g) by C Kyba; (h) “Tribute in Light” by Mike Hvozda, from Wikimedia Commons; (i) “Hammerfest dusk LNG” by Andreas Rümpel, from Wikimedia Commons.
1.2. Sources of Light
Many different light are responsible for the light that shines into space. In an aerial survey, Kuechly
et al. performed the first large-scale investigation of the sources of light observed by nadir-viewing satellites [
9]. Their study of Berlin found that the most important areas emitting light toward the zenith were streets, responsible for 32% of the detected light. Other important regions were industrial areas (16%), public service areas (10%), block buildings (8%), city center (6%), airfields (4%) and supply and disposal facilities (4%).
Figure 1 shows examples of some common sources of light at night. Light sources have different angular emission profiles, and the light viewed by a satellite may include light shining directly from a source and light scattered by the ground or buildings. Tall buildings and vegetation can also block a view of the street level. It is important that this be considered when using such data for analyses, like understanding energy consumption or sources of skyglow. In fact, beamed or horizontal sources of light that are readily visible from the ground and contribute to energy use and skyglow may be nearly invisible to a nadir viewing instrument (e.g.,
Figure 1d,f,g,h).
3. Results and Discussion
The three case studies demonstrate that there are large variations in how light is used worldwide and that the data delivered by the VIIRS DNB and astronaut photographs can be used to quantitatively evaluate these differences. The major advantages of the new data over that from older sensors are their much improved resolution, higher sensitivity and the fact that they are calibrated (DNB) [
20] or at least calibratable (photos) [
26]. Night imagery is likely to be extremely useful in fields as disparate as economics, ecology, epidemiology and atmospheric science. The street-level imagery available from the ISS should prove especially useful in epidemiology, and could potentially resolve the question of whether neighborhood light levels are a cause or correlate of diseases like breast cancer [
47]. The case studies have also demonstrated that the data have limitations that will provide some challenges for applying them to some fields (
Table 4). These challenges are related to the great variation in artificial light emission with spectrum, time, emission direction and location. Each of these factors is considered below.
The spectral sensitivity of the DNB is quite different from that of the human visual system, with little sensitivity below 500 nm. This will make comparisons of cities and the study of lighting trends increasingly problematic, because street lighting worldwide is currently being replaced by “white” LEDs, which generally have considerable emission in the range 450 to 480 nm [
29,
48]. Because of this, a transition from low pressure sodium to LED lighting could appear as a large reduction in DNB radiance, even if the luminance at the street level slightly increased. This problem may pose a particular challenge to studies of energy and socio-economic parameters that rely on time series or comparisons between different regions. The problem may be even more critical for epidemiological studies, because the response of the human circadian system peaks in a spectral range that both the DMSP and VIIRS DNB do not image [
49].
The spectral information problem could be minimized to some extent through synergistic application of both datasets. Color information from the ISS photographs could be used to approximately identify the mix of lighting types used in a region corresponding to a single pixel in the DNB stable lights product [
10,
29]. The DNB radiance could then be turned into an estimated spectral radiance for the pixel. Spectral radiances would be of greater use in analyses, such as modeling skyglow [
16,
50], estimating energy consumption [
29] and epidemiological studies [
18].
Table 4.
Opportunities and challenges in the selection of the research fields where the new sources of nighttime light data are likely to be useful.
Table 4.
Opportunities and challenges in the selection of the research fields where the new sources of nighttime light data are likely to be useful.
Field | Opportunities | Challenges |
---|
atmospheric science | visible band remote sensing e.g., aerosol properties | understanding sources variable air mass |
ecology | higher sensitivity in “dark” areas | relationship between upward emitted light and environmental exposure |
energy | no saturation calibrated data variable overpass times | changes in lamp properties (spectra & angle) albedo/tree cover |
epidemiology | street-level information | spectral information angular emission direction radiance calibration overpass time |
light pollution | highly reliable data | all of the above |
socio-economic parameters | higher spatial resolution higher sensitivity no saturation | changes in lamp properties angular distribution |
Case Study 1 showed that city light is not constant, but changes dynamically over the course of the evening. Skyglow studies have observed such changes, with associated changes in spectra [
51]. Researchers using future DNB “stable lights” products should keep in mind that the satellite overpass time is variable and occurs after local midnight. Stadiums and other sources that operate lights on a curfew may potentially not appear lit in stable light products or may have radiances that represent an average over several days and overpass times. Data from ISS photos may provide additional information by determining trends in lighting shut-offs. Some areas reduce light emission as the night progresses, and these effects are likely to become stronger as more areas introduce legislation of light at night [
44]. Finally, seasonal changes can cause great variation in the fraction of light that can escape to space (e.g., snow and foliage cover). This is of particular importance at high latitudes, where “yearly” DNB composites will have a winter bias, due to summertime overpasses during twilight not contributing to the average.
In general, neither VIIRS nor the ISS photos are taken directly at nadir, but rather at some angle
θ. As
θ increases from zero, the light propagates through increasing air mass. This increases the optical depth and red-shifts the spectrum of the direct beam. In addition, the light emitted upwards by cities is not Lambertian, but, rather, dependent on
θ [
52,
53]. This dependence will change as older lamps are replaced, because newer lamps tend to emit a smaller fraction of light directly upwards. At angles much larger than zenith, light is strongly shaded by buildings. Radiance measurements from DNB and the ISS should therefore be expected to have strong dependencies on both the nadir and azimuth direction. Astronauts should place a priority on imaging cities that pass near the station’s nadir.
Case Study 2 demonstrated that there are considerable differences in light emission, both between and within countries. Some analyses, for example, the retrieval of aerosol properties, are sure to require a careful characterization of the light sources (spectra, intensity and emission direction) specific to the location. Lack of saturation in DNB will improve understanding of urban lighting and will expand the range of possible socioeconomic studies. The higher sensitivity of DNB will expand the regions of Earth over which faint levels of artificial light are visible. This may be particularly important for understanding the ecological impacts of artificial light [
54,
55].
Lighting accounts for about 19% of total electrical energy consumption [
56], and despite improvements in luminous efficiency, per capita consumption of light has remained near 0.7% of GDP [
57]. The combination of increasing GDP and improving luminous efficiency has led to increases in the amount of light shone into space [
13,
58,
59,
60,
61,
62]. The major differences in light emitted upward by Germany and the USA suggest that such a trend need not necessarily continue, as the difference in light use between the countries is considerably larger than the difference in per capita GDP. Future studies should examine the roles that factors, such as population density, automobile use, street width, vegetation and building height, play in determining how bright cities appear from space.
By examining and contrasting cities that have anomalously large and small amounts of uplight, more effective strategies for minimizing waste light could be developed. The DNB data could also be used to examine the effectiveness of light pollution laws [
44] or city planning strategies, like the Berlin
Lichtkonzept [
32], at reducing waste light.
4. Conclusions
Artificial light at night is often far brighter than natural and reflected celestial light, so nighttime visible band images of Earth highlight human activity in a way that daylit scenes do not, especially in urban areas. Newly available imagery from the International Space Station and VIIRS Day-Night Band has the potential for application across many scientific disciplines. The intensity and spectra of artificial light at night has rapidly changed over the last century, and the new imagery will allow global tracking of these changes for the first time.
Lighting management appears to impart a cultural footprint upon nighttime views of cities and nations. The case studies presented here show that there are large differences in current patterns of light use, even between countries with highly-developed economies. American cities emit far more light than German cities of equivalent size, and the brightest areas of Central European capitals are fainter than those in the West. Even within a single country, historical forces influence lighting: cities in the former East Germany are more brightly lit than those in the former West. Further study of the light sources, for example characterization of the mix of lighting technologies in use, will allow more in-depth understanding of these differences and will make night light data even more reliable and useful for researchers.