Flood Mapping using Sentinel-1 data using SNAP

Objective:
To do flood mapping of a small region of Ganga that occurred during August 2021 in
comparison with the conditions in May 2021 from the Sentinel-1 data using SNAP software.
Highlights:

• Make a spatial subset of the images around the river over the flooded areas
• Apply a “multilooking” in order to reduce the speckle, and this also reduces the
dimension of the images and speed up the processing time
• Apply a calibration which is essential to compare 2 images using a physical quantity
which is in this case is the “sigma0 backscatter”
• Apply a terrain correction to project the images onto a map system and also to correct
for the distortions due to the terrain
• Combine the 2 images in a RGB composite in order to distinguish between flooded
areas and permanent water bodies

STEPS
1. Data Access
1) Download Sentinel-1 data from Copernicus hub (of region somewhere around
Prayagraj)
*Things to keep in mind while downloading data:
o One should be the CRISIS image and other should be an ARCHIVE image
a. CRISIS image – obtained at the point of the event (flood) e.g. here it is
obtained on 09/08/21
b. ARCHIVE image – obtained when the event (flood) hadn’t happened e.g. here
it is obtained on 29/05/21
o Both image should preferably be of the same tile

2) In the “Product Explorer” tab, open the subfolders “Metadata -> Abstracted
Metadata” and double-click on it to open it. The metadata of the images should show
that both the images have following identical features:
a. Product Type: GRD
b. Pass: Ascending
c. Acquisition Mode: IW
2. Opening of Images in SNAP
3) Open both the products from Open Product Icon, viz [1] and [2].
4) Go to Bands -> Amplitude_VV of both the products

1
 5) In order to align both the images side-by-side, select ‘Tile Evenly’ from ‘Window’
pop down menu.

Fig 1: Sentinel-1 imageries downloaded from Copernicus Hub

*There are 2 bands per image, Amplitude and Intensity.

o Amplitude_VV image - A SAR signal contains amplitude and phase information.
Amplitude is the strength of the radar response” (source:
https://sentinel.esa.int/web/sentinel/user-guides/sentinel-1-
sar/productoverview/interferometry).
o Intensity_VV image - Intensity = Amplitude_VV * Amplitude_VV

* Crisis Image Observation:

o Flooded area: appears darker because of a low backscatter due to specular reflection
over the smooth water surfaces: the signal get reflected away from the sensor.
o The surrounding areas: are much rougher and look brighter.

2
Fig 2a: ARCHIVE image

Fig 2b: CRISIS image

3
3. Spatial Subset of Images
Making Subset of the existing image, so that processing becomes faster
6) Identify in the navigation tab an area which is common to both images: zoom and pan
in order to display that common area in the viewers.
7) Click, in the geographical interface, on the image you want to process in order to
8) select/active it, in order to use this image current extent for the spatial subset
9) Raster -> Subset
The subset corresponds to the extent of the viewer. It can visually be checked that in the tool
interface.
10) Keeping the default options as it is, press OK and wait for thumbnail image to load.
11) Repeat the operation for both the images.
The subset of both the images gets loaded in the ‘Product Explorer’ window viz. [3] and [4].
However, they still are just virtual files. They needs to be saved as real product file.

Fig 3: Subset of images

12) Select the subset file [3], and go to File -> Save Product, and save the product to a
desired folder. (or, right click k on the file in the “Product Explorer” tab, then > Save
Product)
13) Say Yes to BEAM-DIMAP format. Click on Save.
14) Repeat the same for file [4].
*Now close all the products from the ‘Product Explorer’ window (File -> Close All Products)
*Now 2 subseted images are present in the working directory.

4
4. Multilooking
15) Open both the subset files from the locations they were saved in.
*Now, in order to reduce the size of the image, we will perform ‘Multilooking’. For High-
resolution mapping of flooded area, Multilooking is not done. Because it reduces the
resolution of the image. However, that would increase the processing time.
16) Go to Radar -> SAR Utilities -> Multilooking
17) Under ‘Processing Parameters’, modify the value of ‘Number of Range looks’ to 3,
implying a multilooking factor of 3x3.
18) Keeping everything else as default, click ‘Run’.
19) Repeat the same for the second subset image.

Fig 4: Multilooking window

* after this operation, the subseted images cannot be synchronized well with the original
images because a size reduction has been applied, reason why all products were closed
previously.
5. Calibration
* Calibration is essential to enable the comparison of the 2 images. It transforms Digital
Numbers (DN) to a physical quantity which is in this case is “Sigma0 backscatter”.
*Backscatter is the normalised measure of the radar return from a distributed target is called
the backscatter coefficient, or sigma nought , and is defined as per unit area on the ground.”
*Sigma Nought (Sigma0) aka Scattering coefficient, is the conventional measure of the
strength of radar signals reflected by a distributed scatterer, usually expressed in dB. It is a
normalised dimensionless number, comparing the strength observed to that expected from an
area of one square meter.”
(Source: https://sentinel.esa.int/web/sentinel/user-guides/sentinel-1- sar/definitions)
20) Go to Radar -> Radiometric -> Calibrate

5
 21) Here leaving the default option of Sigma0 band
22) Click on ‘Run’
23) Repeat for both the images
*From the new calibrated product, Go to Bands and open Sigma0_VV band of both the
images. Open these images. It can be noted that the images are more crispier and sharper
now.

Fig 5: Calibrated images

*It can be seen in the Colour Manipulation Window, that many pixels have very low
backscatter value while there are a few pixels which have high backscatter value.

Colour Manipulation
Window

Fig 6: Colour Manipulation Window with Histogram

6
*Hence, to get a better distribution, pixels are converted from linear scale to non-linear scale
(logarithmic) scale. This is called a decibel band. By doing so, the images have a better
visualisation and histogram (in color manipulation window) can easily be manipulated.
24) Select the first calibrated band name and Go to Raster -> Data Conversion -> Convert
bands to/from dB.
25) Check the output file location and
26) Click ‘Run’.
27) Repeat for other image.

Fig 7: dB conversion of calibrated images

28) From the new converted product, Go to Bands and open Sigma0_VVdB band of both
the images.

Fig 8: Sigma0 dB bands

7
*Notice the change in ‘colour manipulation window’ now.
*If all the products of decibel and non-decibel images are opened simultaneously, the
difference between them becomes eminent.

Fig 9: Simultaneous view of dB and non-dB images

6. Terrain Correction
*Terrain Correction is done to correct the distortions in the images and also make the images
suitable for projection.
29) Go to Radar -> Geometric -> Terrain Correction -> Range-Doppler Terrain
Correction
30) Keeping everything as the default values (Map projection – WGS84),
31) click Run.
32) Repeat the process for the other image.

8
 Fig 10: Range Doppler Terrain Correction window
33) Open the Terrain corrected images.
34) Visualise both the images.
o The image has been projected into a coordinate system: the TC image orientation
has changed!
o In order to be able to compare the non TC an TC images visually, you have to
unsynchronize the views with the dedicated button in the “Navigation” tab (bottom
left panel), and then open each image in a window that you can display side by side.
o You will maybe have to unzoom a lot from one of the two images to adjust the
visualization.
o There is no more distortion in the mountain area
o You could also do a contrast stretch with the “Color Manipulation” tab (bottom
left panel) to highlight only the pixels over the land.

9
 Fig 11: Mulitilooked calibrated terrain corrected images
35) Save these images from File -> Save Product as.
7. Stack creation
*Once the terrain corrected products are saved, we would need to overlay the two images
over each other to see the change in flood area. This is done by creating a stack.
36) Go to Radar -> Coregistration -> Stack Tools -> Create Stack

Fig 12: Steps for image stacking

10
 37) In the ‘Create Stack’ window, click ‘Add Opened’ icon under 1-product set reader.

Fig 13: Create Stack window

38) Now select the images that are NOT the final images(Multilook Calibrated terrain
corrected images) and Remove them.
39) In the 2-create stack tab, select ‘Product Geolocation’ under Initial Offset Method.
40) In the 3-Write tab, remove the part of the name of the file not common to both the
images.
41) Select ‘Run’.
42) Let the Raster data computation complete.
43) Once its done, the final stacked image is loaded in the ‘Product Explorer’ window.
*It can be seen that all the Bands are loaded in the stacked image.

Fig 14: Product explorer window showing all the created bands in the stack
44) Load the images Sigma0_VV of two varying dates (Crisis and Archive).

11
 Fig 15: Stacked images
By sliding the sliders in the colour manipulation window, the contrast of the images can be
changed. The contrast of image changes as the slider is moved.
Now, In order to overlay the two images in the same viewer and to create a RGB composite
of the two images,
8. RGB Creation
45) Go to Layer Manager -> Click on the + icon -> Select ‘Image of Band/Tie Band Grid’
-> Next

Fig 16: Add layer screen during RGB creation

12
 46) Now select the other image (not currently opened in the viewer), in this case it is 29
May image.and select Finish.
47) As soon as this is done, both the images are loaded in the layer manager window. And
these can simultaneously be viewed by selecting or deselecting the boxes next to
them.
48) Alternately, both layers can also be viewed by selecting the layer and changing the
transparency slider below.

Fig 17: Adjusted contrast (notice the range in the histogram)

*Now, in order to differentiate between the permanent water bodies in the area and ‘flooded’
regions, a RGB composite is created.
49) To do that, we first select the stacked product image from the ‘Product Explorer’
window.
50) Now, Go to Window -> Open RGB Image Window
51) In the window that pops up, select the ARCHIVE (29May) image as the Red band and
the CRISIS (09Aug) image as the Green and Blue Band.
*The reason we do that because in the ARCHIVE image, land having low backscatter
tendency will have high Radar response in the Red Channel. However, the flooded area will
have low backscatter in Green and Blue bands. All the surrounding areas will appear grey and
water bodies will have uniformly dark return in both crisis and archive images as they too
will have low backscatter in red, blue and green bands. The flooded areas will appear in red
as they will have high response in Red band and low response in Green and Blue bands.

13
 Fig 18: Selecting RGB image channels
52) Click OK.
53) The RGB image gets loaded in the viewer.

Fig 19: Final RGB flood mapped image

54) This is the final Flood Map with the flooded area marked in Red.
*Interpretation
o Flooded areas: Appears in red because, given the selected RGB composite above,
where floods occur, the Archive image (in the red channel) has higher backscatter (no
flood) than the Crisis image (in the green and blue channel) (low backscatter for
flooded areas). So for flooded areas, there is a high value in the red channel and low
value in the green and blue channels.

14
 o Non-flooded land: Appears in tones of gray as there is approximately the same
backscatter for the 2 images (no difference), and that that “same” information is
associated to all 3 channels.
o Permanent water bodies: Appears as uniform dark areas as there is a low backscatter
values for both Archive and Crisis images, both associated to the 3 RGB channels
o Some part of the land are cyan: This translates a higher response in the green and blue
channels corresponding to the crisis image, than in the Archive image. This may be
due to particular ground cover which is not related to flood.
9. Export
55) This image can be exported in another format e.g. geotiff or kmz file.
56) To convert to a geotiff file, go to File -> Export -> GeoTIFF
57) To view it as a Google earth image, right click on the image and select Export view as
Google Earth KMZ

Fig 20: Exporting to kmz file

58) Save it to the desired location.
59) Once saving is finished, browse to the folder in which it was saved and open the .kmz
file in Google Earth.
60) If you have Google Earth Pro installed on your computer, you can simply double-
click on this file and it will automatically open in Google Earth at the right position.
61) If you do not have Google Earth Pro installed on your computer, you can get it from
the URL: https://www.google.com/intl/fr_ALL/earth/versions/
62) You can then compare the flood map with Google Earth Pro imagery and also check
the registration (georeferencing) of the flood map.
63) The flood map obtained is found to be a perfect match with the Google Earth optical
imagery.

15
 Fig 21: Visualising the created RGB layer in Google Earth.

References:
o Video Tutorial: https://www.youtube.com/watch?v=derOXkPCH80
o Transcript to the video: https://orbi.uliege.be/handle/2268/240620
***

16