<div class="se-component se-image-container __se__float-center"><figure style="width: 1000px;"><img src="https://ep-insight.s3.ap-northeast-2.amazonaws.com/public/41/en-50-Years-of-Wildfire-Outbreak.png" alt="" style="width: 1000px; height: auto;" origin-size="2000,1001"></figure></div> According to the Korea Forest Service, there are currently about 567 wildfires occurring in South Korea on average over a 10-year period (2014 to 2023). As the years go by, the frequency of large wildfires is increasing, leading to a significant rise in wildfire damages. To prevent and respond to wildfires effectively, South Korea is expanding the scope of utilizing remote sensing technologies such as drones and satellites every year. <blockquote class="__se__custom__vertical-blockquote"> Monitoring Wildfires using Satellite Imagery🔎</blockquote> With satellite images, we can identify active wildfires due to smoke or heat, assess the extent of damage post-fire, and get precise help for the affected areas. So, would you like to explore the process through which satellite imagery is utilized? <blockquote class="__se__custom__normal-blockquote">Detecting the Area of wildfire Damage With Satellite Photos </blockquote> There are many ways to use satellite images to find forest fire damage. Lately, thanks to better artificial intelligence, there's also a lot of work on creating models that can detect these disaster areas. <div class="se-component se-image-container __se__float-none"><figure><img src="https://ep-insight.s3.ap-northeast-2.amazonaws.com/public/41/en-190405-Forest-Fire-RGB.png" alt="" style="" origin-size="2001,1001"></figure></div> The wildfire that started in Gangneung-si at 11 PM on April 4, 2019, was put out a day later. The RGB satellite images we often use can really help figure out when and where a forest fire happened. <div class="se-component se-image-container __se__float-none"><figure><img src="https://ep-insight.s3.ap-northeast-2.amazonaws.com/public/41/en-190405-Forest-Fire-RGB-FireLayer.png" alt="" origin-size="2001,1001" style=""></figure></div> Nara Space used the Sentinel-2 satellite's short-wave infrared (SWIR)<span style="font-size:10pt;font-family:'Nanum Gothic',sans-serif;color:#000000;background-color:transparent;font-weight:400;font-style:normal;font-variant:normal;text-decoration:none;vertical-align:baseline;white-space:pre;white-space:pre-wrap;">* band, along with the vegetation index (NDVI)<span style="font-size:10pt;font-family:'Nanum Gothic',sans-serif;color:#000000;background-color:transparent;font-weight:400;font-style:normal;font-variant:normal;text-decoration:none;vertical-align:baseline;white-space:pre;white-space:pre-wrap;">* and normal moisture index (NDWI)<span style="font-size:10pt;font-family:'Nanum Gothic',sans-serif;color:#000000;background-color:transparent;font-weight:400;font-style:normal;font-variant:normal;text-decoration:none;vertical-align:baseline;white-space:pre;white-space:pre-wrap;">*, to zoom in on wildfires. This way, they can identify the areas that heat up during a blaze. <ul><li style="font-size: 13px;">SWIR (Short wave infrared) : This is a part of the infrared, between 3.5 to 4.0 micrometers.</li><li style="font-size: 13px;">NDVI (Normalized Difference Vegetation Index) : This is used to measure how plants are healthy.</li><li style="font-size: 13px;">NDWI (Normalized Difference Water Index) : This shows how much water is in plants or on the ground.</li></ul> <div class="img-compare"><img-comparison-slider class="slider-with-animated-handle"><img src="https://ep-insight.s3.ap-northeast-2.amazonaws.com/public/41/en-190326-Forest-Fire-RGB-Before.png" alt="" style=""><img src="https://ep-insight.s3.ap-northeast-2.amazonaws.com/public/41/en-190420-Forest-Fire-RGB-After.png" alt="" style=""></img-comparison-slider></div> By compositing and enhancing the bands of satellite imagery, similar to the example mentioned above, it becomes possible to identify areas of uncontrolled wildfires after their occurrence. In disasters such as large-scale wildefire, detecting the area damaged by disaster is considered one of the essential processes for establishing future damage recovery plans.In addition, it can be used to calculate the area of disaster-damaged areas. Although RGB (True Color) satellite images can show the disaster area with a purple color, it only shows simple visual information and it is not easy to determine the exact damaged area. Therefore, as mentioned earlier, accurate detection of the extent is important in order to calculate the damaged area. <blockquote class="__se__custom__normal-blockquote">What is the Normalized Burn Ratio (NBR)? </blockquote> <div class="se-component se-image-container __se__float-none"><figure><img src="https://ep-insight.s3.ap-northeast-2.amazonaws.com/public/41/en-Burn-Severity-vs-Fire-Intensity.png" alt="" style=""></figure></div> In order to not only detect wildfire damage areas but also classify the degree of damage, the Normalized Burn Ratio(NBR), which was index designed by identifying the reflective spectrum of forests blackened by fire and healthy vegetation, is used. NBR is an analysis method that uses the ratio of reflectance of near infrared (NIR) and short wave infrared (SWIR) to identify damaged areas and damage intensity. It is widely used as it can be obtained in a relatively simple way using images taken after suppressing wildfire. <div class="se-component se-image-container __se__float-none"><figure><img src="https://ep-insight.s3.ap-northeast-2.amazonaws.com/public/41/en-exploiting-spectral-response-curves.png" alt="" style=""></figure></div> Wildfire ash changes the ground’s reflectance in the near infrared (NIR) and short wave infrared (SWIR) bands compared to when there is vegetation.Ground with healthy vegetation has high NIR reflectance but low SWIR reflectance. In contrast, areas damaged by wildfire exhibit low NIR reflectance and high SWIR reflectance. <div class="img-compare"><img-comparison-slider class="slider-with-animated-handle"><img src="https://ep-insight.s3.ap-northeast-2.amazonaws.com/public/41/en-Satellite-Imagery-of-forest-fire-damage-severity-Before.png" alt="" style=""><img src="https://ep-insight.s3.ap-northeast-2.amazonaws.com/public/41/en-Satellite-Imagery-of-forest-fire-damage-severity-After.png" alt="" style=""></img-comparison-slider></div> By looking at the difference in the Normalized Burn Ratio (NBR) before and after a wildfire, you can recognize which areas were damaged by wildfire. This difference is known as dNBR (Difference Normalized Burn Ratio). <div class="img-compare"><img-comparison-slider class="slider-with-animated-handle"><img src="https://ep-insight.s3.ap-northeast-2.amazonaws.com/public/41/en-dNBR-Satellite-Images.png" alt="" style=""><img src="https://ep-insight.s3.ap-northeast-2.amazonaws.com/public/41/en-Extracting-forest-areas-from-a-Land-Cover-Map.png" alt="" style=""></img-comparison-slider></div> By overlaying the land cover map on the dNBR satellite image, which highlights changes in the NBR, you can identify forested areas within the fire-damaged zones. Applying the land cover map removes non-forest areas such as lakes and seas, which may appear as fire-damaged zones in the dNBR image, leaving only true forest areas. However, using dNBR requires satellite images taken before and after a wildfire. This technique will not work if you do not have images from required dates or if the satellite lacks a SWIR band. <blockquote class="__se__custom__normal-blockquote">Identifying Wildfire Damage With Deep Learning </blockquote> Lately, there's been a surge in using deep learning for detecting disaster-affected areas. Nara Space is on it too, working on a deep learning model to detect wildfire damaged areas. <div class="img-compare"><img-comparison-slider class="slider-with-animated-handle"><img src="https://ep-insight.s3.ap-northeast-2.amazonaws.com/public/41/en-Composing-Red-Green-and-NIR-Satellite-Band.png" alt="" style=""><img src="https://ep-insight.s3.ap-northeast-2.amazonaws.com/public/41/en-Extracting-Wildfire-Damaged-area-Using-Deep-Learning.png" alt="" style=""></img-comparison-slider></div> Nara Space is developing a deep learning model to identify wildfire-damaged areas, as part of a recent trend of using deep learning to detect disaster-affected areas. This approach identifies wildfire-affected zones using four common satellite bands: Red, Green, Blue, and NIR (near-infrared). By composing the Red, Green, and NIR bands, you can make images that highlight the area damaged by a wildfire. With those images, develop a technical model and then use deep learning to identify the fire-damaged areas. <div class="img-compare"><img-comparison-slider class="slider-with-animated-handle"><img src="https://ep-insight.s3.ap-northeast-2.amazonaws.com/public/41/en-Damaged-areas-detection-using-Deep-learning.png" alt="" style=""><img src="https://ep-insight.s3.ap-northeast-2.amazonaws.com/public/41/en-Damaged-areas-detection-using-dNBR.png" alt="" style=""></img-comparison-slider></div> The results of detecting wildfire damage areas using dNBR and deep learning differ slightly but are generally similar. Unlike dNBR, which needs satellite images from before and after a wildfire, deep learning can effectively identify damaged areas with just one image. This makes it adaptable for detecting forest fire damage under various conditions. <blockquote class="__se__custom__normal-blockquote">Classifying Suspected Damage and Recovery Areas After the WildFire </blockquote> Since wildfire damaged areas detected via satellite images are based on the burned area, it is difficult to detect damage caused by the burning of plant roots or pillars over time using only images taken immediately after the forest fire. Therefore, it is necessary to monitor the progress after the wildfire. What changes have occurred over the last two years since the Gangneung forest fire on April 4, 2019? <div class="se-component se-image-container __se__float-none"><figure><img src="https://ep-insight.s3.ap-northeast-2.amazonaws.com/public/41/en-Extracting-forest-areas-from-a-Land-Cover-Map-Process.png" alt="" style=""></figure></div> Two years after the wildfire, on April 14, 2021, we extracted the damaged area using a satellite image of the same location as the Gangneung, using the wildefire damage detection method described earlier. <div class="se-component se-image-container __se__float-none"><figure><img src="https://ep-insight.s3.ap-northeast-2.amazonaws.com/public/41/en-Two-Years-after-the-wildfire.png" alt="" style=""></figure></div> Red indicates areas still seen as damaged, and blue represents areas that were not detected in image immediately after the wildfire but is suspected of additional damage over time . Green areas are expected to be recovering, with vegetation growth and restoration underway. The estimated sizes are: 16.816 km² for fire-damaged areas, 2.786 km² for areas with suspected additional damage, and 1.493 km² for areas showing signs of vegetation recovery. With climate change impacting not just Korea but the entire world, we can not afford to ignore disasters anymore. It is crucial to focus more on finding solutions. While tackling these varied disasters will not be simple, satellite analysis could pave the way for quicker, more accurate responses.

Tackle Forest Fire Damage in One Go with Satellite Imagery!

You May Like