Using Remote Sensing to Compare
Before and After a Wildfire
Embedded Files
Ryan Grammer
GIST 601B
Abstract
Using Landsat 8 imagery, we will compare areas of land before and after a wildfire. We will also use this Landsat imagery to study how the affected areas are bouncing back from the fire. In this project, Landsat imagery was collected of the area before the fire started (October 14th, 2019), after the fire finished burning (October 16th, 2020), and 1 year after the fire was finished burning (October 19th, 2021). Using the same date from each year will minimize our chances that weather and temperature will affect the study. The Landsat images covered a lot of area, so the first step was to clip the images to the extent of the fire. The area affected is mountainous, so to make sure that our study is accurate, topographic correction was used to minimize the variability of observed reflectance. This is done by correcting the effects of the sun's elevation angle and azimuth angle, collected by the Landsat 8 satellite when capturing the imagery. Using the topographically corrected images, the Normalized Burn Ratio (NBR) index was calculated, outlining the areas affected by the fire, and showing the severity of the fire within the burn area. In the affected area, regrowth was very noticeable in some areas, while other areas are struggling to get back to pre-fire levels.
Using Landsat 8 imagery, we will compare areas of land before and after a wildfire. We will also use this Landsat imagery to study how the affected areas are bouncing back from the fire. In this project, Landsat imagery was collected of the area before the fire started (October 14th, 2019), after the fire finished burning (October 16th, 2020), and 1 year after the fire was finished burning (October 19th, 2021). Using the same date from each year will minimize our chances that weather and temperature will affect the study. The Landsat images covered a lot of area, so the first step was to clip the images to the extent of the fire. The area affected is mountainous, so to make sure that our study is accurate, topographic correction was used to minimize the variability of observed reflectance. This is done by correcting the effects of the sun's elevation angle and azimuth angle, collected by the Landsat 8 satellite when capturing the imagery. Using the topographically corrected images, the Normalized Burn Ratio (NBR) index was calculated, outlining the areas affected by the fire, and showing the severity of the fire within the burn area. In the affected area, regrowth was very noticeable in some areas, while other areas are struggling to get back to pre-fire levels.
Introduction
The El Dorado and Apple fires burned during the summer of 2020. Combined, they burned over 56,000 acres in California's San Bernardino County and Riverside County. The Apple Fire started in Cherry Valley, CA on July 31st, 2020. The El Dorado fire, also known as the gender reveal fire, started on September 5th, 2020 near Oak Glen, CA, a neighboring city of Cherry Valley. These two fires ended up joining together shortly after the El Dorado fire started, with the El Dorado fire's eastern front joining with the Apple fire's western front. The image below gives a visual representation of the burned areas and how the two fires are connected. I chose to analyze these fires because they burned very close to my house in Yucaipa. It was surreal seeing the flames just by looking out the windows of my house. In this study, we will examine the burn area and find the areas that were the most severely burned. This will aid in predicting where mudslides/landslides will occur when the storms hit, as well as give the experts areas to focus on when trying to promote regrowth.
What is Normalized Burn Ratio (NBR)?
The NBR is an index used to identify burn areas. It can be calculated using the following formula:
NIR is the Near Infrared band of the Landsat Image (Band 5 in Landsat 8)
SWIR is the Short Wave Infrared band of the Landsat Image (Band 7 in Landsat 8)
NBR is effective because plants reflect strongly in the NIR band, and burned areas reflect strongly in the SWIR band.
NBR will always be between -1 and 1. Values close to -1 represent a burned area or bare ground. Values close to 1 represent healthy vegetation.
Methods
Collect Landsat 8 Imagery:
Pre-Burn Image: Collected October 14th, 2019
Post-Burn Image: Collected October 16th, 2020
1 Year Post-Burn Image: Collected October 19th, 2021
The images above (in order): Pre-Burn, Post-Burn, 1 Year Post-Burn
As you can see, it is hard to distinguish where the burn area is. We will use a different band combination to help differentiate between healthy vegetation and dead vegetation. By setting the RGB band combination to 654 (SWIR 1, NIR, Red), we get an image where we can clearly distinguish between the different land covers.
The images above (in order): Pre-Burn, Post-Burn, 1 Year Post-Burn
If we were to run our analysis on these images, we would get good results, but we can use topographic correction to increase the quality of our results. To be able to compute topographic correction, we need to obtain the slope and aspect of each 30m square in our raster dataset. We can download SRTM (Shuttle Radar Topography Mission) data that gives us a Digital Elevation Model. Through processes, we can obtain a plot of the slope and aspect in our study area.
The images above (in order): Shaded Relief, Slope, Aspect
Now that we have our slope and aspect, we can start the topographic correction process. Using the sun elevation and sun azimuth collected by the satellite when the image was captured, we can run a code that will correct the shadows in our images.
The topographically corrected images above (in order): Pre-Burn, Post-Burn, 1 Year Post-Burn
Results
Here are the results that we get when calculating the NBR of the 3 images.
We can see that the dark purple areas in the mountains represent healthy vegetation, while the urban areas are white, representing the middle of the spectrum. Ignore the orange area in the middle of the map, there was a previous fire there.
Post-fire, we can see many dark orange areas where the fire severity was high. We should watch these areas when looking at our NBR 1 year post-fire to check how the regrowth is going.
1 year post-fire, we can see many dark orange areas where the fire severity was high in the post-fire NBR maps are now light orange, while most of the light orange areas from the previous map have turned white, showing some regrowth.
Discussion
We can see that the more severe fire occurred in the Northwest section and the central section of the burned area. The most severe areas do not have any roots left in the soil, which makes them more susceptible to landslides/mudslides. Working with local government, we can figure out areas that would have the highest chance of landslides/mudslides, and coordinate with emergency services to send out evacuation orders when we know a storm is coming. There is a mountainous highway running through the burn area that would be very susceptible to a mudslides in the event of a storm because of the steep terrain that borders it. These maps can provide information beforehand to make an informed decision whether to close the highway for the duration of the storm.
A home in Forest Falls that was caught in a mudslide that stemmed from the El Dorado fire - picture taken 2 years after the fire on September 15, 2022
(The Mercury News)
Conclusion
We saw that 1 year after the fire, the land has made a significant comeback. There is not a lot of burned area left, mostly barren ground now. By now, if the trend has continued, the less severely burned areas should be back to almost pre-fire vegetation levels, and the more severely burned areas will not be far behind. Our results can help local government and emergency services plan for natural disasters, such as mudslides, as well as help find areas that need help in the regrowth process.
References
Butler, Kevin. “Band Combinations for Landsat 8.” ArcGIS Blog, 25 Nov. 2019, www.esri.com/arcgis-blog/products/product/imagery/band-combinations -for-landsat-8/.
Landsat Missions. “Landsat 8.” Landsat 8 | U.S. Geological Survey, www.usgs.gov/landsat-missions/landsat-8. Accessed 6 Dec. 2023.
Marcus, Matthew Dr. "Lab 2: Satellite Data Exploration and Visualization in R." GIST 601B Lesson 2. University of Arizona.
Marcus, Matthew Dr. "Lab 4: Remote Sensing-Radiometric Normalization." GIST 601B Lesson 4. University of Arizona.
Marcus, Matthew Dr. "Lab 5: Topographic correction in R." GIST 601B Lesson 5. University of Arizona.
Percy, Nathaniel, and Quinn Wilson. “Person Still Missing, 30 Homes Destroyed or Damaged by Mudslides in Southern California.” The Mercury News, The Mercury News, 15 Sept. 2022, www.mercurynews.com/2022/09/15/1-still-missing-30-homes-destroyed-or-damaged-by-flooding-in-forest-falls-oak-glen/.
Wasser, Leah, and Megan Cattau. “Work with the Difference Normalized Burn Index - Using Spectral Remote Sensing to Understand the Impacts of Fire on the Landscape.” Earth Data Science - Earth Lab, 1 Mar. 2017, www.earthdatascience.org/courses/earth-analytics/multispectral-remote-sensing -modis/normalized-burn-index-dNBR/.
“Normalized Burn Ratio (NBR).” UP42 Documentation, docs.up42.com/processing-platform/spectral-indexes/nbr#:~:text=NBR%20ranges% 20between%20%2D1%20and,have%20values%20close%20to%20zero. Accessed 6 Dec. 2023.
Page updated
Report abuse