Assessing the impact of a wildfire with satellites

With the recent heat waves and droughts, an unusually high number of virulent wildfires started in various places of the globe, including Central Africa, Western Canada, the South-West of the United States, and Italy. In this article, we will go over an often-used way to assess the damages caused by the recent series of wildfires through the differential in normalized burn-ratio.

Looking at the differential in normalized burn-ratio (dNBR)

The Normalized Burn Ratio (NBR) highlights burned areas and help scientists estimate fire severity. The formula is similar to NDVI, except that it uses near-infrared (NIR) and shortwave-infrared (SWIR) wavelengths.

As with other indexes, NBR uses the different spectral signatures of the features we want highlighted to come up with a simple mathematical formula to apply to the digital values of the pixels to help better visualize said features. In this case, healthy vegetation has a high reflectance in the near-infrared portion of the spectrum, while offering low short-wave infrared reflectance. On the other hand, burned areas have a high shortwave infrared reflectance but low reflectance in the near infrared.


 Image credit: US Forest Service

Image credit: US Forest Service


Using the NBR formula, high NBR values tend to indicate healthy vegetation while recently burned areas would display lower values.

The most commonly used formula for Normalized Burn Ratio is


Where NIR stands for near-infrared values and SWIR for short-wave infrared values.

However, for this specific application, we were able to obtain 2 different SWIR bands and use the following formula instead:


Where SWIR 1 are the digital values of the short-wave infrared 1 band and SWIR 2, the digital values of the short-wave infrared 2 band.


Practical application: Vesuvius

Using the recent Vesuvius wildfire as an example, let’s explore how to use dNBR to better visualize the impact of a wildfire.

Through SkyWatch EarthCache, we extracted two overlapping Landsat-8 products one month prior to the Vesuvius wildfire and one month after the fire. We chose Landsat-8 products because most commercial satellites do not have shortwave infrared (SWIR) bands. These are currently only available on NASA’s Landsat-8, ESA’s Sentinel-2, and Digital Globe’s WorldView-3. When looking at the different options, Landsat-8 returned great images for the area as the extent of the fire was large enough to be highly visible at a 30m resolution, which is why we chose it.


Step 1: Extracting the data using SkyWatch EarthCache API

We needed to find two good images of the area, with low cloud cover, in relatively close time-proximity to the fire, one before, and one after, to get an accurate representation of the changes brought by the fires.

SkyWatch EarthCache pipelines return Earth observation data matching pre-configured parameters that was collected during a specified date range. However, in this case, we were looking for data outside of a date range, the time of the fire. In order to find what we were looking for, we simply set up two pipelines. You can more information on how to set a pipeline for Earth observation data here [link to KB article].

From the pipelines, we retrieved the NIR, SWIR1, and SWIR2 bands.  Once that step was completed, we applied Top of Atmosphere atmospheric correction and coregistered the images into a stack.

Step 2: Top-of-Atmosphere (ToA) atmospheric correction

The radiance measured by satellites can be different from the radiance measured at the surface since light reflected from the ground has to travel through the atmosphere and is partially scattered by atmospheric aerosols before reaching the sensors. A similar scattering effect can be observed when looking at a landscape very far away, on a hazy day.

Top-of-Atmosphere atmospheric correction is a way to transform top-of-atmosphere measurements into a more accurate reflection of surface reflectance. Visually, it translates into clearer looking imagery and studies have shown it improves the accuracy of results when looking at data collected at different times and different places. Most satellites data providers offer easy ways to do TOA atmospheric corrections.

In this specific instance, to convert the digital numbers into reflectance values, we used the following formula:

ToA corrected values = (Band-specific multiplicative rescaling factor * digital numbers) + Band-specific additive rescaling factor

For Landsat-8 data, both the multiplicative rescaling factor and the additive rescaling factor for each band can be found in the metadata file, under REFLECTANCE_MULT_BAND_x + REFLECTANCE _ADD_BAND_x respectively (where x is the band number for that specific band)

Once both images, pre-fire and post-fire were corrected, we coregistered them into a stack.

Step 3: Co-registering the images

Simply overlaying the images based on their geolocation is not always enough to line them up pixel to pixel.  In order to do line both images, we used a cross-correlation based coregistration offered by the Sentinel Application Platform (SNAP).

The coregistration determines how much the pixels of one image have shifted compared to the other image and warps the shifted image into the coordinate space of the other. Cross-correlation is the mathematical formula used to calculate the exact spatial shift between the two images.

If you are not familiar with SNAP, you can download the most recent version of the toolbox from the ESA website.

Vesuvius, before and after fire.gif

Step 4: Calculating the Normalized Burn Ratio

We then applied the differential normalized burn ratio (dNBR). The dNBR is calculated simply by comparing the NBR pre-fire and the NBR post-fire.

For this specific application, we found formula using both SWIR bands displayed the burn scars more prominently.

dNBR = NBRprefire – NBRpostfire

where NBR is



Step 5: colours!

As a last step, we colour coded the final product to  visually display the burn area and the severity of the damage to vegetation.  In 6 short and easy steps, we were able to obtain a very visual representation of the extent of the burned vegetation on the side of Mt Vesuvius following the wildfires.

 The South side of Vesuvius shows large burn scars from the wildfires

The South side of Vesuvius shows large burn scars from the wildfires

Vesuvius legend.png


The fires were extremely severe on the South side of the volcano, destroying large areas of vegetation and even extending to the residential areas nearby, forcing citizens to flee. Luckily no damage or fatalities were reported. In all, about 18 square kilometers (7 square miles) were affected by the wildfires.

The fires have since been contained but the damage could extend much further. The volcano is composed of alternating layers of hard lava and loose volcanic ash. Many of trees anchoring the top layer of ground burnt, rendering the slopes particularly vulnerable to mudflows during heavy rainfall.

According to David Bressan, “Many streets leading to the volcano were built crossing former erosion channels, characterizing all the slopes of the mountain. If a mudflow forms, it could rush down these channels, posing a risk to the streets. Postfire debris flows are documented from many areas affected by wildfires and periodic storms. Fire not only destroys the vegetation cover but also damages the ground and loosens debris, like large rocks, chunks of soil and dead trees. During heavy rain, this rubble will be washed down the slope and channelized will form a dangerous debris flow. Following the old channels, this material poses a risk also to the nearby residential areas. Depending on the effective damage, to be verified now in the field, the risk of debris flow could last as long as three years until new trees grow enough to stabilize the slopes of Vesuvius.“

This is why it is now critical for the local authorities to actively monitor the regrowth of trees in the area, and, as with dangerous to access areas or large swath of lands, Earth observation technologies can be a cost- and time-effective substitute or supplement to field-based monitoring.


Get more satellite news, straight in your inbox


How government agencies use SkyWatch

The Prairie Pothole Region (PPR) of the United States and Canada is a unique area where shallow depressions, created by the scouring action of Pleistocene glaciation, interact with mid-continental climate variations to create and maintain a variety of wetland classes.

These wetlands possess unique environmental and biotic characteristics that add to the overall regional diversity, including the production of aquatic invertebrates and the vertebrate wildlife that depend upon them as food. Climatic extremes in the PPR have a profound and dynamic influence on wetland hydrology, hydroperiod, chemistry, and ultimately the biota, which is why scientists are so interested in monitoring the area.

 Image of the Prairie pothole region - courtesy of the United States Geological Survey

Image of the Prairie pothole region - courtesy of the United States Geological Survey

For monitoring purposes, government agencies use a combination of satellite and field data to try and get a complete, accurate picture of how the region is evolving over time.

For example, the United States Geological Survey relies heavily on short wave infrared data to create hydrology flow maps. With these data, the USGS creates specific DEM layers that can be combined with statistical data collected from the field to build 3D models and landscape over-time animations.

Prior to SkyWatch, North American scientists monitoring the region would often use the Earth Explorer platform, but they could not always easily find the data required to develop their models, impacting their research on the health of the wetland regions.

One of the teams turned to SkyWatch, intrigued by the ability to access multiple data sets, including atmospheric — which they plan on using for future studies —, as well as the platform's ease of use. Rather than dealing with satellite data vendors on an individual basis, they would now access multiple data sources and set-up data pipelines to receive new Earth observation data by using one single platform. By switching to SkyWatch, this team reduced the time spent looking for Earth observation data by 10X.

By allowing access to multiple data sources through one query location, Skywatch EarthCache has significantly boosted workflow and GIS analysis for scientists. While in the past, they might have had to download extra data or scenes in order to fulfill a request, with the API they can now parse out those unnecessary data points and clip our scenes to their study sites. These capabilities will allow research centers to monitor areas of interest and record changes with ease moving forward.

For their initial study, the team of scientists looked at 180 specific wetland regions to monitor changes over time. Specifically, they requested imagery for each of their 180 areas of interest, every other week if possible, taken between 2013 and 2016. A data collection task take that would have previously taken the research center's technicians, days to complete was done in an hour.


Get more satellite news, straight in your inbox


Locating uranium deposits with satellite data

Current methods for locating new sources of uranium are time-consuming and expensive, often requiring governments and mining companies to fly planes over large swaths of lands in remote areas. With the wide variety of Earth observation data now easily accessible, allowing us to detect more and more from space at a low cost, and the recent increases  in computers’ processing power, which made it possible to sort through the large amounts of data necessary for analysis in a short time, satellites have become an extremely attractive way to help detect features in remote areas.

As a Ph.D. candidate in the UAB School of Engineering, Reda El-Arafy, a geologist by training and now an assistant professor of nuclear geology and remote sensing with Egypt’s Nuclear Materials Authority, worked with mentors Sarah Parcak and Scott Brande to find a more cost-effective way – through sensor systems on Earth-observation satellites – to identify promising targets for uranium exploration in the southwestern Sinai desert.


Establishing ground truth

El-Arafy first selected 30+ areas in the southwestern Sinai in which uranium-rich deposits were suspected. A critical component of the study was to collect geological samples throughout the area for which advanced satellite imagery was available.  Multi-spectral sensors can capture data in several spectral bands, with some sensors recording up to 10 different bands. Hyper-spectral sensors however record data in much narrower spectral bands, allowing us to focus our attention in very small specific parts of the electromagnetic spectrum. This is why El-Arafy chose to study data from the Landsat-8, ASTER, and HYPERION satellites, as minerals of interest in this study are highly recognizable in the short-wave infrared and the thermal-infrared spectral bands detected by multi-spectral and hyper-spectral sensors onboard.

To help with El-Arafy’s research, Egypt’s Nuclear Material Authority also shared data from a gamma-ray spectrometer they flew over the study area, which provided additional information for ground truth interpretation. This new set of data was overlain on satellite images to help identify anomalies or variations in the satellite sensor data and to later optimize algorithms for post processing.

Field investigation (ground truth).JPG


Learning to know your samples

Once the samples and remote-sensor data were collected, El-Arafy divided his time between his two mentors, working alternatively with Parcak – most well-known for her work discovering new archeological dig sites using satellites – learning the intricacies of Earth observation data post-processing, and in the chemistry lab with Brande – a paleontologist and chemistry professor – to help analyze the composition of rock samples that El-Arafy personally collected in the study area.

“Data from this area – the southwestern Sinai – is fairly sparse and these were the first samples collected from some sites. So the team didn’t know exactly what the detailed geochemistry would be.” recalled Brande.


Identifying satellite-visible tell-tale signs for uranium deposits

El-Arafy’s chemical analysis revealed that some secondary uranium mineralization in the southwestern Sinai was specifically associated with zones containing a concentration of iron oxide and clay minerals. While uranium cannot be directly detected from satellite sensors, these associated minerals have a distinct spectral signature that can be identified in satellite data.

The final and most important step of the research was for El-Arafy to develop algorithms tuned to the shortwave and thermal-infra-red signatures of iron oxides and clay minerals as proxy indicators for uranium deposits in the southwestern Sinai. Once these algorithms were in place, the team could analyze satellite imagery in the study area to quickly identify unexplored locations with similar composition, finding new potential targets more quickly and cost-effectively.

Over four years, using the data obtained from the advanced satellite sensors, El-Arafy built a system for identifying new potential targets for uranium exploration in the southwestern Sinai desert with impressive accuracy.

Principal component analysis PCA.jpg


What the future of resource exploration will look like

The success of El-Arafy’s studies parallels similar projects in the field of resource exploration using Earth observation data. Previously, the absence of standardized format and unified points of access for remote sensing data made it very difficult to deliver anything more than stand-alone analysis from remote-sensing imagery. But as algorithms are replacing eyeballs, the amount of information available will continue to grow with the resulting data being used in an increasing number of applications, from resource exploration to smart cities, precision agriculture, or investing.


Get more satellite news, straight in your inbox