How to Monitor the Fires in the Amazon Rainforest on a Daily Basis

The news of the fires in the Amazon rainforest is alarming.
There are several monitoring services such as FIRMS by NASA that allows for accessing the archive of global active fires (from MODIS and VIIRS) in near real-time.

This tutorial aims to describe how to easily monitor the evolution of fires in the Amazon forest, using satellite images acquired daily.
In particular, we are going to use MODIS and Sentinel-2 images to view changes and calculate the NDVI index, using the Semi-Automatic Classification Plugin (SCP).
MODIS images are acquired daily, but have 250m resolution, while Sentinel-2 images are acquired every 5 days but have 10m resolution, therefore the latter are very good for photo-interpretation and analysis of fire severity.
Of course we could also use Landsat images that have 30m spatial resolution or Sentinel-3 images.

NDVI calculated from a Sentinel-2 image (Copernicus data)

ATTENTION: for MODIS images EOSDIS Earthdata credentials (https://urs.earthdata.nasa.gov) are required for download. Login using your EOSDIS Earthdata credentials or register for free at https://urs.earthdata.nasa.gov/users/new . Before downloading MODIS images, you must approve LP DAAC Data Pool clicking the following link,
For Sentinel images a free registration is required at https://scihub.copernicus.eu/userguide/1SelfRegistration  

First we need to search the images. Open the tab Download products clicking the button in the SCP menu, or the SCP Tools, or the SCP dock (you can access the whole SCP user manual here). 
Enter your credentials in the tab Login data.


Now in the tab Search we can use the button  to define the search area interactively in the map.
In this tutorial enter the following coordinates in Search area:
  • UL X (Lon): -55
  • UL Y (Lat): -8.9
  • LR X (Lon): -54.9
  • LR Y (Lat): -9

Select the Products MOD09GQ that provides daily reflectance at 250m spatial resolution from Terra MODIS. 
Now we set the search dates, in general we look for one image before the fire and one after:
  • Date from: 2019-08-01
  • to: 2019-08-25


Now click the button Find .
After a while, search results will be displayed in the table.
In this tutorial, select the following images (that have low cloud cover), and click the button to add the preview to the map:
  • MOD09GQ.A2019231.h12v09.006.2019233095508 acquired 2019-08-19
  • MOD09GQ.A2019225.h12v09.006.2019227032552 acquired 2019-08-13
  • MOD09GQ.A2019218.h12v09.006.2019220032409 acquired 2019-08-06

The search and the previews of MODIS images

Now select the Products Sentinel-2, with the previous other settings, and click the button Find .
This way, also Sentinel-2 images will be added to the table.
In this tutorial. select the following images and click the button  to add the preview to the map:
  • L2A_T21LYL_A012677_20190810T140059 acquired 2019-08-10
  • L2A_T21LYL_A012820_20190820T140333 acquired 2019-08-20

The preview of Sentinel-2 images

We can download and process all the images in one step. Also, the NDVI can be calculated automatically (i.e. a useful index because low values mean the absence or low health of vegetation).

First, we need to set a few parameters.

1. Open the tab Preprocessing and select the tab Sentinel-2.
Check the option  Add bands in a new Band set. 
This way every downloaded image will be added to a new Band set, allowing for viewing a color composite.


Sentinel-2 preprocessing

2. Open the tab Preprocessing and select the tab MODIS.
Check the option  Add bands in a new Band set as we did for Sentinel-2.


MODIS preprocessing

3. Then open the tab Band calc and click the menu Index calculation and select NDVI to enter the NDVI expression automatically.
This calculation will be automatically performed after image download.


The NDVI expression

4. For the NDVI calculation, open the tab Band set and check the option  Band calc expression .
This way the NDVI calculation will be performed for every created Band set that is every downloaded image.

The parameter for calculating the NDVI for every image

5. Now, open the tab Download products and select the tab Download option.
To reduce the download time, select only the Sentinel-2 bands 2, 3, and 8, that are sufficient for calculating NDVI and displaying a color composite.
Click RUN to start the download and process all the images having the preview loaded in QGIS.


After the download and the processing we can see the images loaded in the map.

MODIS and Sentinel-2 images

To display a color composite open the tab Band set; as you can see there are several Band sets created.
Click any of the new Band set such as Band set 2. This Band set corresponds to the image acquired on 2019-08-19 (the Band set order may vary).



In the Working toolbar enter the composite RGB = 1-2-1 (this image has only 2 bands).
The map should now display the MODIS image (NASA data).
If we zoom over the center of the image we can see vegetation in green, and dark areas that are probably caused by fires.

MODIS images acquired on 2019-08-19
If we repeat the same steps for the other Band sets, we can see the evolution of the area.
In the image acquired on 2019-08-13 (before the image we have already seen) we can see the violet dark areas, about the same size of the image acquired on 2019-08-19.
Of course, white pixels are clouds.

MODIS image acquired on 2019-08-13

In the image acquired on 2019-08-06 there are fewer dark areas than the other images (excluding the clouds and their shadows), meaning that those fires started after this date.

MODIS image acquired on 2019-08-06

We can display the NDVI rasters assigning a symbology where low values are red.
Comparing the NDVI of the 2019-08-06 image and the NDVI of the 2019-08-19 image we can see the increase of the red areas related to fires.
Please note that also clouds and water bodies appear red because have low NDVI values.

NDVI of the MODIS image acquired on 2019-08-06

NDVI of the MODIS image acquired on 2019-08-19

Now, if we zoom to the area covered by the Sentinel-2 image (Copernicus data) we can appreciate the higher resolution.
To display a color composite open the tab Band set; and open the Band set 5. This Band set corresponds to the image acquired on 2019-08-20.

The Sentinel-2 band set

In the Working toolbar enter the composite RGB = 3-2-1 (we have downloaded only 3 bands).
With this color composite vegetation appears red, and burned areas appear very dark or black.
Also we can see white clouds and the black shadows thereof.

Sentinel-2 image acquired on 2019-08-20

If we display the Band set 6 acquired on 2019-08-10 we can see that the large dark patch on the left side is not present, meaning that the fire started between the 10 and 20 of August. We can also see that some of the fires cover areas that seem already harvested or without dense forest cover.

Sentinel-2 image acquired on 2019-08-10

We can also display the NDVI of these images (red areas represent low NDVI values), noting the large burned areas on the left that increased between 10 and 20 August.

NDVI of the Sentinel-2 image acquired on 2019-08-10 

NDVI of the Sentinel-2 image acquired on 2019-08-20

This tutorial aimed to illustrate an easy and rapid method to use satellite images for monitoring fires.
We have monitored only a small part of the Amazon Rainforest, therefore we should repeat all the above steps to cover also other parts.
Of course this method have several limits. There is uncertainty on data because of cloud cover and cloud shadows, therefore one should collect more than one image to confirm the fire. Spatial resolution of MODIS images makes the photo-interpretation quite difficult.
Also, we don't know the cause of the fires, or the land use of burned area. We would need additional information to understand this, such as field data or forest cover data.

Other indices can be calculated for assessing fire severity. Please read this previous tutorial  https://fromgistors.blogspot.com/2017/01/wildfire-monitoring.html.
Following the video of this previous tutorial.



I really hope that the fires will be soon extinguished, preserving this fundamental environment that is the Amazon Rainforest.

For any comment or question, join the Facebook group about the Semi-Automatic Classification Plugin.

Newer posts Older posts

Follow

RG