This web application allows you to examine changes in nighttime light emissions (nearly) worldwide, from 1992 up until last month. This text explains where the data comes from, what need to know about the data sources to understand the charts you generate, and who developed the app. If you haven’t already done so, you may wish to use our tutorial to get started looking at the data quickly.

The emissions data come from two satellite sensors. From 1992 to 2013, data comes from the Operational Linescan System of the Defense Meteorological Satellite Program (DMSP) satellites. From 2012 to the present, data comes from the Day/Night Band of the Visible Infrared Imaging Radiometer Suite instrument (VIIRS DNB). The instruments have a number of important differences, and for that reason it is not possible to have a single record running from 1992 to today.

Both of the satellite instruments scanned (nearly) the entire Earth each night, but the DMSP typically passed overhead around 8:30 pm, while VIIRS DNB passes over much later, around 1:30 am. The Earth is brightly illuminated by moonlight when the satellites pass over several days per month, and on any given night, much of the Earth is obscured by cloud. For more than two decades, the team of Chris Elvidge and Kimberly Baugh have produced global annual satellite observed nighttime lights data products. They were based initially at the U.S. National Oceanic and Atmospheric Adminsitration’s National Geophysical Data Center in Boulder, Colorado. In 2019 they relocated the Payne Institute at the Colorado School of Mines, Golden, Colorado. A description of the VIIRS DNB night lights product used in this application is available here. The data used in the app can be accessed from their group’s website.

About the DMSP

Observing night lights was not the main purpose of the DMSP, and the 1992-2013 data record therefore have a number of limitations. The instrument was not calibrated, and therefore the results are displayed in “digital numbers” ranging from 0-63. In city centers, the satellite data was often “saturated” (too bright to measure), and therefore each year will just show a value of 63. The team of Chris Elvidge and Kim Baugh worked to produce approximately radiometrically calibrated data for specific years, and this can be accessed by selecting “Radiance Calibrated DMSP” on the “satellite” dropdown menu.

The spatial resolution of the satellite was several kilometers, and is in fact larger than the pixels displayed by the app. Because the overpass time of the DMSP was so early in the evening, at high latitudes most of the data is acquired during winter. Care should therefore be taken in comparing results from high latitude countries to lower latitude countries.

About VIIRS DNB

The VIIRS DNB was the first satellite instrument intentionally designed to image human lights on the worldwide scale, and is therefore a major improvement over DMSP. The satellite resolution is about 750 meters (or 0.5 square km). This data has been reprojected onto a grid, so in some places (especially high latitudes) the pixels are displayed at finer detail than the satellite can actually see. Because of the higher data quality, it is possible to display data on a monthly timescale.

In 2017, a change in the way the data were calibrated resulted in a change to the baseline "no artificial light" value. Before 2017, this value was generally close to zero, whereas after 2017 it is increased to about 0.2 nW/cm^2sr. We have developed a method that corrects all data to be closer to zero when no artificial light is present. We recommend this option be selected, especially if your analysis region includes large areas without artificial light. To turn this option on, select "VIIRS DNB (zero point correction)" from the "Satellite" drop down menu.

The Earth Observation Group has produced annual composite images based on their monthly stray light corrected composite images. However, the stray light correction is not yet available for 2012 and 2013. For this reason, results at high latitudes (including much of Europe) are systematically different in the first two years, and are therefore not shown in Radiance Light Trends. Furthermore, because the DNB only started regularly reporting data in April of 2012, the annual composites for 2012 will always be systematically different from the other years, because of the three missing winter months.

The Black Marble group at NASA has also developed a product that works a bit differently from the EOG version. Their monthly and annual composites are also included here under the option “VIIRS DNB (Black Marble)”. Whereas the EOG only includes data from moon free nights, the Black Marble uses a moonlight correction model to subtract moonlight out, and therefore is based on more nights per month than the EOG data. A downside is that Black Marble applies a cut at 0.5 nW/cm^2 sr, and values below that are shown as zero. This causes non-linear behavior in dim places (e.g. dimly lit villages or the edge of skyglow around a large city), where values jump between 0 and ~0.5 nW/cm^2sr from month to month, likely causing a slight positive bias. Note also that the atmospheric correction is only for moonlight - there is no atmospheric correction applied to the artificial lights to either the Black Marble or EOG data.

Limitations of satellite data

Night lights data can tell many interesting stories, but it is important to keep in mind exactly what it represents. For example, the data presented here should not be understood as (directly) representing energy consumption, community wealth, or light pollution. While each of those are often related to light emissions, the satellite does not measure them directly.

The satellites measure light in the spectral range of about 500-900 nm. This is not the same as human vision, which runs from 400-700 nm. The satellite data is therefore less sensitive to white light than the human eye, and has sensitivity at infrared wavelengths that the human eye cannot see. For this reason, when communities switch from orange high pressure sodium lamps to white LED, the satellites often report a drop in brightness even if a human would say that the white LEDs look brighter. Since cities worldwide are changing to white LEDs, it is important to keep this in mind.

Even far away from artificial lights, the world is not entirely dark. It is lit by starlight, aurora light, and “airglow”, and the satellite has some sensitivity to this light, especially around the polar regions. In areas that are mainly unlit, changes in airglow or the satellite calibration may affect all of the pixels similarly. Fires can also produce very bright temporary light emissions, and false signals caused by solar radiation appear over a certain region near the southern Atlantic. It is therefore not advisable to include large areas that have little or no installed lighting in your analyses. To some extent, this problem can be reduced by using a zero point correction or applying a mask. For example, if you select “2016 vcm-orm-ntl”, this will remove areas that were not lit consistently during 2016. The zero point correction we are using is described in this paper.

The emissions observed from any given area change from month to month for a number of reasons. To some extent, and over the long term, this is due to actual changes in the installed lighting. But on shorter time scales, variation is due to a number of other reasons, such as the imaging angles that made up the monthly composite. Many areas, especially those at high latitudes, also display seasonal cycles. In areas that receive snow, it is advisable to restrict analyses to months that are unlikely to have snow on the ground. In addition, at high latitudes there is no data during the summer due to stray light shining on the satellite sensor.

Some of the light emitted upwards can scatter off of molecules and particles in the atmosphere, producing “skyglow”, a type of light pollution. Because of this, the satellites often report light emissions from unlit areas near bright light sources. This can be very easily seen in cities that are on coasts. Because this light is regularly present, it will not entirely be removed by the “vcm-orm-ntl” mask. You may also be interested in viewing a map of predicted clear night sky brightness (see “ATLAS 2015” on that page).

Additional information

Analyzing large areas can take several seconds, and during this time it will not be possible for other webapp users to make an analysis request. For this reason, we currently limit analyses to 10,000 square kilometers.

The summed radiance reported when a polygon is selected is the sum of the individual pixels. It is not weighted by area, and that means that the sums will be larger near the poles than at the equator for an equivalently lit area. If you would like to compare two locations with an approximate correction for area, you can do so by multiplying by the reported sum by the cosine of the latitude of the polygon center.

Using images or screenshots from this page: The charts generated by the webapp have no restrictions on sharing, and may be used without contacting us. With regard to the image layers, images with the "Lights layer opacity" set to 100% have no restrictions on sharing, but if used should include the text "Image and Data processing by NOAA's National Geophysical Data Center." If images containing DMSP data are used, you should also provide the credit: "DMSP data collected by the US Air Force Weather Agency." If any of the basemaps are included in your image, then your reuse is subject to Microsoft® Bing™ Maps Platform APIs' Terms Of Use.

Funding and development

This web application was partially funded through the European Union’s Horizon 2020 research and innovation programme ERA-PLANET, grant agreement no. 689443, via the GEOEssential project, with additional funding and project direction from the GFZ German Research Centre for Geosciences. The code underlying the web application is available under the EUPL license.

If you have further questions about the data or application, please direct them to Christopher Kyba

The application was programmed by:
Deneb, Geoinformation solutions, Jurij Stare s.p.
Adamičeva ulica 4
1000 Ljubljana
Slovenia