Skip to main content

The Sentinel-1 system consists of two space-based platforms: Sentinel-1A and Sentinel-1B, each carrying C-band synthetic aperture radar (SAR) instruments. Sentinel-1A was launched on April 3, 2014, and Sentinel-1B on April 25, 2016. They orbit 180° apart, together imaging the entire Earth every six days. In addition, this radar imagery is sensitive to standing water, making it an ideal tool for mapping the extent of floodwater covering an area.

Sentinel-1A passed over Houston, Texas, on August 29, 2017, making it possible to create a map that highlights water and flood inundation on the surface caused by Hurricane Harvey. Harvey’s extreme slow movement from August 26 to 30 kept an onslaught of moisture in southeast Texas and Louisiana for days, resulting in catastrophic flooding.

Image of houston texas showing fooded roads with city in background.

Houston flooding as a result of Hurricane Harvey. Credit: CNN.

Prerequisites

System Requirements

  • At least 16 GB memory (RAM)

Many of the steps take a very long time to process. Consider closing other applications and avoid using the computer while products are processing.

Materials List

Generating an Inundation Map

Data Preparation

  1. Download the sample granule from the Prerequisites section or download your own granule using Vertex (Do NOT unzip).
  2. Open the Sentinel-1 Toolbox.
  3. In Sentinel-1 Toolbox, use the Open Product button to add your Sentinel-1 GRD product.
  4. Navigate to the Sentinel-1 GRD product, select the .zip file, and click Open.
  5. In the Product Explorer window on the left, double-click on the Product to expand its information, which includes Metadata, Tie-points grid, and Bands.
  6. View a band by expanding the Bands folder. For each polarization recorded, there are two bands: Amplitude and Intensity. (The Intensity band is a virtual one; it is the square of the amplitude.) Double-click on either the Amplitude or Intensity to view the image.
  7. In the bottom left corner, the World View shows the footprint of the selected scene.
  8. View Product Properties by right-clicking on the Product, which includes information on the mission, acquisition date, pass, etc.
screenshot of houston tx showing amplitude in black and white

Amplitude band of Sentinel-1 GRD Product.

Vertical screenshot showing text at top and map of Texas coast at bottom

Footprint of GRD product in Sentinel-1 Toolbox World View.

Optional: Create a Subset

  1. Open the product phase band and zoom-in on your area of interest using the scroll wheel on your mouse.
  2. Right-click on the image when you are satisfied and select Create Subset from View… in the pop-up menu.
  3. When the Specify Product Subset dialog appears, double-check your Subset scene and click OK to create subset.
  4. The created subset is added as a new product, which is also shown in the World View.
screenshot showing a black and white image of houston

Specify Product Subset dialog.

Map of Texas coast with boxes around Houston area

Subset footprint in Sentinel-1 Toolbox World View.

Pre-Processing

Calibration

  1. In Sentinel-1 Toolbox, select the Sentinel-1 GRD product in the Product Explorer. If you created a subset, select the subset instead.
  2. Navigate to Radar > Radiometric > Calibrate from the Menu panel
  3. A Calibration window will open. In the I/O Parameters tab, specify the Target Product name and output directory. By default, the Target Product name will be appended with _Cal if you do not create a custom filename.
  4. In the Processing Parameters tab, select the VV polarization. This will create a new product with calibrated values of the backscatter coefficient.
  5. Leave all other parameters as default and click Run to calibrate.
Screenshot of calibration dialog box

Calibration dialog with VV selected.

Speckle Filtering

In this step, we will filter out the speckle noise that exists in SAR imagery due to random interference between returns and the scatterers present on a surface. Speckle noise is a critical pre-processing step for detection/classification optimization. In this example, we will be using a standard “Lee” speckle filter; however, you may experiment with the other speckle filters within the Sentinel-1 Toolbox.

  1. In Sentinel-1 Toolbox, select the calibrated product in the Product Explorer.
  2. Navigate to Radar > Speckle Filtering > Single Product Speckle Filter from the Menu panel.
  3. The Single Product Speckle Filter dialog will appear. Specify Target Product name and output directory. By default, the Target Product name will be appended with _Spk if you do not create a custom filename.
  4. In the Processing Parameters tab, choose “Lee” for the filter.
  5. Change the Filter Size X and Filter Size Y to “7”.
  6. Click Run to begin image filtration.
  7. A new product will be created in the Product Explorer. Open the band in the newly created product to compare the non-filtered and filtered versions.
Screenshot of speckle filter dialog box

Single Product Speckle Filter dialog with 7x7 filter.

screenshot showing comparison of speckle and non-speckle images

Juxtaposition of speckle and non-speckle filtered images.

Determine Threshold

To separate water from non-water, a threshold can be applied. For this, we will analyze the histogram of the filtered backscatter coefficient.

  1. Select the speckle-filtered product you created in the Preprocessing — Speckle Filtering step (above).
  2. Open the Sigma0_VV band in the Product Explorer.
  3. On the left-side panel, select the Color Manipulation tab.
  4. The histogram of the image will show up. Select the logarithmic display button (Log10).
  5. The histogram will show one or more peaks of different magnitude depending on the data. Low values of the backscatter will correspond to water, and high values will correspond to non-water classes.
  6. To separate water from the rest of the image, pick a threshold at or near the minimum between the two modes of the distribution, assuming you have a bimodal data distribution (e.g., at 4.2E-2). Drag the sliders within the histogram for reference. Keep a note of this threshold, as you will use it in later steps.

Note: This step assumes that your images have a bimodal distribution of pixel values in the image. If you are using different data, the class values will vary and it may take several iterations to determine the appropriate thresholds. For the process of determining the appropriate thresholds, it is good practice to consider one image representing the maximum flooding and another image representing minimum flooding. Knowledge of the geographical coverage of seasonal flooding helps confirm whether the results are correct.

Screenshot with data on left and black and white image on right

Histogram of GRD product with threshold selected.

Geometric Correction

Before creating the water mask, we will terrain-correct the image, which will, in turn, reproject the image into latitude and longitude as well as mask out areas with no elevation.

  1. Select the speckle-filtered product you created in the Preprocessing — Speckle Filtering step (above).
  2. Navigate to Radar > Geometric > Terrain Correction > Range-Doppler Terrain Correction from the Menu bar.
  3. In the I/O Parameters tab, specify the output name and directory. By default, the name will be appended with _TC if you do not create a custom filename.
  4. In the Processing Parameters tab, leave the default settings and click Run to begin correction.
  5. A new product will be created and will appear in the Product Explorer.
screenshot of text selection window

Range-Doppler Terrain Correction dialog.

Binarization

  1. Select the terrain-corrected image created in Geometric Correction step and open the band.
  2. Using the determined binary threshold in Determine Threshold step, right-click on the band and select Band Maths… from the pop-up menu.
  3. The Band Maths window will open. Type a name for the new band, for example, “Water Mask.”
  4. Uncheck the Virtual Band option.
  5. In the Band Maths Expression box, enter the following formula: if Sigma0_VV<4.22E-2 then 1 else NaN
  6. Click OK to create your water mask.
screenshot of band math text entry

Band Maths dialog with water mask expression.

Visualization in Google Earth

  1. Open the band you created in the Binarization step above.
  2. On the left side panel, select the Color Manipulation tab.
  3. Within the Color Manipulation tab, select Table.
  4. Click on the third Color box and change to a color of your choice (e.g., blue).
  5. In the Product View, right-click on the image and select Export View as Google Earth KMZ.
  6. In the Export dialog, specify a name and directory to output your KMZ file.
  7. Once saved, open your new image in Google Earth for visualization.
screenshot with text on left and water mask image on right
Water mask with Color Manipulation: Table adjusted to blue.

Sample Image

Water mask image of Houston during flooding

Water mask of the Houston area overlain in Google Earth; Contains modified Copernicus Sentinel data 2017; processed by ESA (European Space Agency).

This data recipe was adapted from the United Nations Platform for Disaster Management and Emergency Response — UN-SPIDER program.

Details

Last Updated

Published on

Data Centers

Alaska Satellite Facility DAAC (ASF DAAC)