Monitoring Wildfires by Satellite: Hotspots and dNBR

In short
- Thermal infrared bands find hotspots, the places that are giving off heat.
- dNBR turns the degree of burning (burn severity) into a map.
- Early detection and after-the-fact assessment serve different purposes.
At three in the morning, a spark begins to spread somewhere in the mountains. Before smoke is visible to the human eye, a satellite may already be detecting the temperature change at that spot. Conversely, after the fire is out, you need to know "how much burned" to plan recovery, and that requires a different kind of analysis. Even for a single wildfire, what a satellite does is completely different depending on the stage. This post looks separately at the thermal infrared principle for finding hotspots and at dNBR, which puts a number on the degree of damage.

How does a satellite notice that a wildfire has started?
For a person to spot a wildfire from the sky, smoke or light has to be visible. A satellite does not need that. The thermal infrared band (the wavelength range that detects heat given off by objects) measures the temperature of the ground surface itself, regardless of smoke or darkness at night. A burning spot is much hotter than its surroundings, so it appears as a bright point in this band, called a hotspot. Thanks to this, the first sign of ignition can be caught relatively early even at night or under thin cloud, and it is actually used to gain precious time in the early response to a wildfire.
How do you pick out hotspots from the thermal infrared band?
A thermal infrared image gives the brightness temperature of each pixel (radiant energy converted into a temperature value). The usual approach is to mark a pixel as a hotspot candidate when this value is higher than the surrounding background by a certain amount. The problem is that things like hot factory roofs, furnaces and hot desert rock can also be picked up in a similar way. So algorithms look at several bands together or compare a pixel with its neighbors. In real disaster response, satellite hotspot information is not used as is, like a report received by a call center. Local governments and forestry authorities cross-check it with other observation data and use it as reference material to decide whether to dispatch crews.
After the fire has passed, how do you tell how badly it burned?
After a fire is out, you use the Normalized Burn Ratio (NBR) instead of hotspots. Vegetation reflects near-infrared well and absorbs shortwave infrared well, but after burning this relationship is reversed. Using this difference, the pre-fire NBR minus the post-fire NBR is called dNBR (differenced NBR), and a larger value is interpreted as more severe burning. dNBR maps often divide severity into color-coded classes, so you can see at a glance where everything burned and where vegetation survived. This map is used to set recovery priorities and to identify in advance areas where landslide risk has increased.
Why do early detection and after-the-fact assessment play different roles?
For early detection, speed is everything. Even if the resolution is somewhat coarse, the goal is to quickly report "where is it hot right now" using satellites that pass several times a day. For after-the-fact assessment, precision matters more than speed. You take time to calculate dNBR from higher-resolution images, separate the damaged area and severity classes, and put them in a report. It is a problem if the two are confused. Announcing the damaged area from hotspot detection results, or trying to judge "where the fire is now" from a dNBR map, makes the analysis mismatched in timing and purpose.
Can you trust wildfire information seen from satellites as it is?
Satellite data is powerful but not absolute. Thick smoke or cloud can interfere with thermal infrared as well, and dNBR values can also vary with the local vegetation type and differences in acquisition dates. So in practice, before announcing satellite results, you always go through a step of cross-checking them against ground surveys, drone imagery or other satellite data. This is something you only learn by trying: if you divide the same before-and-after images with different thresholds, you can see for yourself that the boundaries of the severity classes change. Releasing a map without such verification can throw off recovery budgets and judgments about risk areas, so care is needed.
Step by step on screen
The figure below is a diagram that summarizes the order of the steps.

- Open the site
Go to a public fire map website such as NASA FIRMS and open the map view - Choose the area
Move the map, or type the place name you want to check in the search box - Turn on the layer
In the layer panel on the right of the screen, turn on a hotspot layer (for example, VIIRS or MODIS) - Set the period
Use the date slider at the bottom of the screen to choose the date range you want to check - Check the hotspots
Click a point shown on the map to see the detection time and brightness temperature information - Compare before and after
In Google Earth Engine, load Sentinel-2 images from before and after the fire and compare their NBR values (menu names can differ by version; if you cannot find them, look at the help or the layer list)
Menu names may differ slightly between program versions.
Prompts you can copy
Draft a wildfire damage report based on dNBR analysis results
You are a staff member in a disaster safety department who summarizes wildfire damage. Write a draft report based on the information below. Wildfire area: [place name] Date and time of ignition and extinguishing: [date and time] dNBR analysis results: severe [area/share], moderate [area/share], low [area/share] In the report, include in this order: a damage summary, an interpretation by class, and the areas that need to be checked first during future recovery. Explain technical terms in one line the first time they appear, and do not guess at unconfirmed numbers or causes; mark them "needs field verification". When you receive the result, always check that the class criteria and the area figures match the actual analysis values.
Replace the bracketed parts for your situation. Check the facts in any answer before you use it.
Try it yourself
- On a public fire map such as NASA FIRMS, look for areas where recent hotspots are shown.
- In Google Earth Engine, calculate and compare NBR values from Sentinel-2 images before and after a fire.
- Query the same fire area with different date ranges and check how the number of hotspots changes.
Common questions
Q. If there is wildfire smoke, can a satellite not find hotspots?
A. Visible light images are blocked by smoke, but the thermal infrared band penetrates smoke to some extent and detects the heat source. However, very thick smoke or cloud can still interfere.
Q. Does a high dNBR value always mean everything burned?
A. dNBR is a relative index, not an absolute criterion, and it is interpreted by dividing it into classes with thresholds for each region. It is safer to check it together with a ground survey.
Q. Does hotspot detection always match an actual fire?
A. No. Heat sources such as hot roofs and factories can also be wrongly marked as hotspots, so additional imagery or a field check is needed.
What comes next
If hotspot detection tells you "where is urgent right now," dNBR is the tool that tells you "how severe it was." Simply using the two for their separate roles changes how reliable a wildfire response report is. A later post in this series looks at how the damage map obtained this way connects to the risk of landslides and soil runoff.
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