Can Night Satellite Images Find Power Outage Areas?

A night satellite image records the strength of light leaking up from the ground at night. When a power outage occurs, the brightness at that spot drops sharply, so comparing images from before and after a disaster gives clues about outage and damage areas.
In short
- Night satellite images show the brightness distribution of city lights
- Comparing brightness before and after a disaster helps estimate outage and damage areas
- Without correction for moonlight, cloud and season, it is easy to misread brightness changes
On the morning after a typhoon passes, what is the fastest way to find out which areas lost power? Even before reports reach the power company or field reports come in, one satellite image taken overnight can first tell you, "this neighborhood is unusually dark."
After major typhoons and earthquakes, cases of estimating the extent of outages from night-light images are reported steadily. But it is worth looking separately at what exactly these images capture and how far you can trust them. Even the same electricity use looks different in an image on a bright moonlit night and on a cloudy one, and in a snowy winter and in summer.

What exactly does a night satellite image capture?
A night-light image is an image in which a sensor records the light leaking upward from the ground at night. It mainly captures artificial light such as street lamps, building lights, road lighting and factory lights, and the brightness value is stored for each pixel (the smallest square unit that makes up an image). A sensor called VIIRS is known as a representative instrument that steadily takes such night images. The important point is that this image does not directly measure "electricity use" but measures "the amount of light leaking into the night sky." When power is cut, lights go out and the light decreases accordingly, so an outage can be inferred indirectly. For a disaster manager, the right use of this image is not as evidence that confirms an outage on the spot, but as reference material for deciding where to check first.
Can you see outage areas right away by comparing brightness before and after a disaster?
The basic principle is simple. Overlay the usual nighttime brightness image from before the disaster and the nighttime brightness image from right after it, for the same area, and calculate the difference. Places where pixels with sharply reduced brightness values are clustered are suspected to be possible outage areas. What matters here is comparing with the "usual pattern," not with "the previous day's image." If you look at only one or two days of brightness, it is hard to tell a rural area that is always dark from a city center that lost power. So a commonly used method is to take the average brightness over several days before the disaster as a baseline and look at the drop from that baseline. The "brightness decrease area" map extracted this way is used as reference for deciding which areas to survey first for damage, before outage reports start to pile up.
Could moonlight or cloud lead to a wrong judgment?
Yes. Besides outages, there are several other causes that lower brightness in a night image. On a full-moon night, the moonlight itself lights the ground and raises the overall brightness values, and on a cloudy night, lights on the ground are hidden by cloud and appear darker than usual. Right after a disaster it is often cloudy or raining, so pixels that look dark because of cloud, even though power is actually on, get mixed in. So before analysis, it is common to pick only clear, cloud-free images or to go through a correction step that filters out cloud. To reduce the effect of moonlight, comparing days with a similar moon phase (how full or thin the moon is) is also used. If you announce outage areas from a brightness drop alone, without such correction, you risk mistaking areas that are dark only because of the weather for damaged areas.
Doesn't a difference in season make the brightness comparison meaningless?
Seasonal differences are also a variable you must take into account. In winter the sun sets early and snow cover can raise brightness values above normal because of light reflected off the snow, and in summer thick foliage can sometimes block light. In addition, periods such as holidays and festivals when more lights are on than usual, and areas with policies to reduce late-night lighting, have a "usual brightness" that itself goes up and down by season and day of the week. So when setting the pre-disaster baseline, it is recommended to gather images from the same season and, as far as possible, the same days of the week and average them. If you do not control for the seasonal variable in this way, when analyzing outage areas from a winter heavy snowfall you cannot separate snow-reflected light from an actual drop in lighting, and you may draw the extent of damage wrongly.
How far can you trust it, given the resolution limits?
In a night-light image, one pixel often represents a few hundred meters on the ground, so it is hard to pinpoint whether a single building or a single block lost power. It is useful for seeing a trend such as "this area has become unusually dark" over a wide area such as a large village or an administrative district, but you cannot confirm from this image alone whether an individual apartment complex or a particular hospital lost power. The imaging cycle is also often about once a day, so an outage restored during the day is not reflected in the next day's image and can be missed. So in practice it is safer to set a wide priority survey area with night images and then cross-check with other data such as the power company's intake records, field patrols and signals from mobile network base stations. If you announce outage areas from a single satellite image without this cross-checking, the resolution limits can produce a map that differs from the actual damage.
Step by step on screen
The image is a diagram of the order of steps. The tool used here is NASA Worldview or a similar public web map, and its menu names may differ from those described.

- Get the images
On NASA Worldview or a similar public map site, turn on the night image layer while changing the date to before and after the disaster. - Choose the layer
In the layer list, find the item named for night lights (Night Lights, VIIRS and so on) and tick it. - Compare dates
In the date selector at the top, enter the day before the disaster and the day after it, and switch between the two scenes. - Check cloud
Turn on an ordinary satellite image layer (the true-color view) for the same date and check whether cloud covers the area. - Mark the brightness change
Zoom in on your area of interest and mark the zones that became darker than usual with a screen capture or a note. - Cross-check with other data
Compare with outage reports or news coverage for the same area to see whether the zones of reduced brightness overlap with actual damage.
Menu names may differ slightly between versions. The figure above is a diagram that summarizes the order of steps.
Prompts you can copy
When you want to organize night-light image analysis results into a report
You are a geospatial information analysis assistant in a disaster safety department. Based on the content below, write a draft report on suspected outage areas caused by [target disaster]. Input: disaster date [date], night image dates used in the analysis [two comparison dates], list of areas where brightness dropped sharply [list the area names] The report should include: 1) a one-paragraph summary of the analysis method 2) a table of the degree of brightness decrease by area 3) stated possible error factors such as cloud and moonlight 4) a proposal of priority areas that need field checks Do not write as if these were confirmed facts; use expressions such as "estimate" and "reference material" Keep the length to about one A4 page
Replace the bracketed parts for your situation. Do not use an answer as it is; check the facts first.
Try it yourself
- On a public map site such as NASA Worldview, turn on night images for dates before and after a past typhoon or earthquake and compare the brightness difference by eye
- Open the same area on a full-moon day and on a day close to a new moon, and check whether the brightness values differ
- Switch between a cloudy-day image and a clear-day image and compare how much cloud hides the lights
Common questions
Can outage areas be officially announced from night satellite images alone?
It is not recommended. Resolution limits and the effects of cloud and moonlight can introduce errors, so it is safer to cross-check with field reports or power company data and use the image as reference material.
Can an outage be seen in satellite images taken in the daytime?
Lights do not show up in daytime images, so they do not directly show whether there was an outage. However, traces of accompanying damage such as fire or flooding can be checked in daytime images, so the two kinds of images are often viewed together.
Is outage analysis impossible in heavily clouded areas?
It is hard to judge from the image of that moment alone. You can check an additional image from the day after the cloud clears, or use other data that passes through cloud, such as radar-based images.
What comes next
A night satellite image is not a tool that measures electricity directly but indirect data that lets you infer an outage from "light leaking out at night." It becomes useful only when you keep four variables in mind, moonlight, cloud, season and resolution, and view it together with other field data. A later post in this series may cover how to use such night images together with communications data.
This is an English adaptation of a post in the Korean GeoAI basics series. The original is here, and more learning materials are on GeoAI for Disaster Safety.
This post was drafted with the help of AI tools and reviewed by the author. Tool features and data portals change over time, so check the official documentation when you practice.
Comments
Post a Comment