This tutorial describes a method for monitoring wildfires using SCP and the freely available images acquired by Landsat and MODIS.
Monitoring wildfires and assessing burn severity are crucial activities for the protection of forests and more in general for the management of environment and ecosystem services. Remote sensing is very useful for the monitoring of large fires, and several projects provide this information such as the Copernicus Burnt Area at the global level, or the Monitoring Trends in Burn Severity for the USA.
The educational purpose of this tutorial is to monitor burnt area in a semi-automatic way, exploiting the characteristics of satellite images that can be downloaded with SCP. In particular, we are going to calculate a spectral index named Normalized Burn Ratio (NBR) described in this paper A project for monitoring trends in burn severity (Eidenshink, et al., 2007) .
We are going to use MODIS Products that allow for the daily monitoring of fires, at 250m spatial resolution, in addition to Landsat images that have higher spatial resolution.
In this tutorial we are going to study the large fires that occurred in the Madeira Island (Portugal) in August 2016. For more details, please read this informative page by NASA Earth Observatory.
1. NDVI Using MODIS Data
We are going to calculate NDVI (Normalized Difference Vegetation Index) using MODIS Products (250m spatial resolution) in the period before and after the fire. In fact, NDVI values over the burnt area are considerably lower than healthy vegetation (see Spectral Indices).Start a new QGIS project. Before the download of satellite images, it is convenient to define the preprocessing options and calculations that will be applied automatically to downloaded images (read this tutorial From Image Download to NDVI Calculation in One Move ).
Open the tab Band set clicking the button in the SCP menu, or the SCP Tools, or the SCP dock. Check Band calc expression, which is the link to the tool Band calc.
( "#NIR#" - "#RED#" ) / ( "#NIR#" + "#RED#" ) @NDVI_#BANDSET#
NDVI
after @
is the name of the output raster, and #BANDSET#
is a variable to automatically set an output name based on the name of the first band in the Band set.Now we can search the images. Open the tab Download images clicking the button in the SCP menu, or the SCP Tools, or the SCP dock. Select the tab MODIS download and in Search area enter the coordinates of our study area:
- UL X (Lon): -17
- UL Y (Lat): 33
- LR X (Lon): -16
- LR Y (Lat): 32
TIP : In general it is possible to define the area coordinates clicking the button and drawing a rectangle in the map.
- Date from: 2016-08-08
- to: 2016-08-12
Now click the button Find and after a few seconds the images will be listed in the
Image list
.We can download and process all these images. Uncheck the option Only if preview in Layers and click the button to download all the images listed in the
Image list
.After the download and the calculation, the bands and the NDVI for every image are loaded in the map. We can see that NDVI values in the burnt area are decreased after the fire (of course we should exclude cloud cover and water from this calculation).
This is a very rapid way to assess the daily evolution of burnt area.
2. Normalized Burn Ratio Using Landsat Data
Landsat, Sentinel-2, and ASTER images have very high spatial resolutions. However, temporal resolution and cloud cover reduce the availability of images.We are going to download Landsat 8 images and calculate the Normalized Burn Ratio (NBR) which is defined as (Eidenshink, et al., 2007) :
( NIR - SWIR ) / ( NIR + SWIR )Therefore, we can download only these two bands (NIR and SWIR) for the calculation of NBR. In addition, we are going to calculate the NBR difference (dNBR) between two images after and before the fire, which is related to burn severity:
dNBR = NBR_prefire - NBR_postfireWe can perform automatically the NBR calculation after the download. Open the tab Band calc, and enter the following expression to calculate NBR:
( "bandset#b1" - "bandset#b2" ) / ( "bandset#b1" + "bandset#b2" ) @NBR_#BANDSET#
bandset#b2
are the first and the second band in the tab Band set, which are the NIR and the SWIR (we are going to download only these two bands).In Search area enter the coordinates of our study area:
- UL X (Lon): -17
- UL Y (Lat): 33
- LR X (Lon): -16
- LR Y (Lat): 32
- Date from: 2016-07-01
- to: 2016-08-31
TIP : Images selected for this tutorial are LC82080372016224LGN00 and LC82080372016208LGN00 therefore in Filter we could enter:
LC82080372016224LGN00,LC82080372016208LGN00
to filter only these two images.
Image list
. Select the bands 5 and 7 to download (please read this tutorial Data Download for the details about selecting bands for download) then click the button to start the download.The NBR is calculated from the expression in Band calc and the rasters are loaded in the map. Burnt area shows low NBR values, while healthy vegetation shows high NBR values.
Now we can calculate dNBR. Open the tab Band calc, and enter the following expression to calculate NBR:
"NBR__RT_LC82080372016208LGN00_B.tif" - "NBR__RT_LC82080372016224LGN00_B.tif"
For accurate assessment of fire severity we should compare these dNBR values with in-situ measurement. We could monitor vegetation recovery repeating the NBR and dNBR calculation of images acquired several months after the fire.
The same method can be applied to Sentinel-2 images, in which case we would need to download band 8A (or band 8) and band 12 for NBR calculation.
Finally, we could perform land cover classifications (for instance read Tutorial 2) to identify vegetation species and improve the identification of burnt area.
For more tutorials and information about wildfire monitoring, data and tools, you can see this webminar by NASA ARSET .
For any comment or question, join the Facebook group and the Google+ Community about the Semi-Automatic Classification Plugin.