Ocean pollution due to oil spills remains a major environmental hazard. Although oil tanker accidents are well known, they are not the main cause for this type of event. Illegal discharges from ships or offshore platforms, drilling rigs, pipeline accidents or natural leaks amongst others bring together most of the sources for oil pollution in the ocean.
On August 10th, 2017, an oil spill was reported in the south of Kuwait, near the Al Khiran area where the Al Khafji offshore oil field is located. While the cause of the incident was not clear (tanker offshore, pipeline damage), almost 132500 liters have been leaked based on conservative estimations made by SkyTruth, a non-profit organization based on the United States.
Note: This exercise was developed by the Serco EO Training Team within the framework of the RUS contract. The final outcome of this tutorial has not been validated and the information contained in this tutorial does not purport to constitute professional advice.
1 Sentinel-1A IW GRD image [acquired on 10/08/2017]:
Reference folder structure
Folder structure referenced in the exercise is as follows:
- /Original – should contain your downloaded input products
- /Processing – Contains all intermediate and final results of the processing
Data download – ESA SciHUB
In this step, we will download the Sentinel-1 scene we will use for the exercise, from the Copernicus Open Access Hub using the online interface. Launch your Internet Browser and go to https://scihub.copernicus.eu
Go to “Open HUB”, if you do not have an account please register by going to “Sign-up” in the LOGIN menu in the upper right corner.
If you had to fill-in the registration form, you will receive an activation link by e-mail. Once your account is activated or if you already have an account, “LOGIN”.
Switch the rectangle drawing mode to pan mode by clicking on the icon in the lower left corner of the map (Green arrow) and navigate to Kuwait (approximate area – blue rectangle).
NOTE 1: that data older than 1 year may be placed to the long term archive (LTA) and need to be requested before they can be downloaded. To see how to request the data see: https://scihub.copernicus.eu/userguide/LongTermArchive.
We need to download 1 Sentinel-1 image acquired during the spill. Switch to drawing mode and draw a search rectangle approximately as indicated below. Open the search menu by clicking to the left part of the search bar and specify the parameters below. Then click on the “Search” icon.
Sensing period: From 2017/08/10 to 2017/08/10
Check Mission: Sentinel-1
Product type: GRD
Sensor Mode: IW
The search returns one result. Download the following scene by clicking on the download icon:
Image ID: S1A_IW_GRDH_1SDV_20170810T024714_20170810T024738_017855_01DEF7_F48C
The product will be downloaded in the Downloads folder of your browser. Move the downloaded data to …/OCEA03_OilSpill_Kuwait/Original (or your desired location).
SNAP – open and explore data
Open SNAP (Applications → Processing). To import the Sentinel-1 image, click File → Open product, navigate to the downloaded product and open the product by double clicking on it.
The opened product will appear in Product Explorer. Click + to expand the contents of the file, then expand the Bands folder and double click on the Amplitude_VV band to visualize it.
To reduce the processing time of the algorithm, we subset the image to our area of interest. Click on Raster → Subset. In the Spatial Subset tab, set the following parameters in the Pixel Coordinates tab and click OK.
Scene start X: 3065 Scene start Y: 3659 Scene end X: 8861 Scene end Y: 7208
The subseted product will be created immediately but it is not saved on your hard disk. Right click on the subset product index  and select Save Product. Set the Output folder to: …/OCEA03_OilSpill_Kuwait/Processing
Click Save. If a window pops-up, click Yes. Then, click + to expand the contents of the file, expand the Bands folder and double click on the Amplitude_VV band to visualize it
To reduce the usual salt and pepper like texturing of SAR images (See NOTE 1), a speckle filter is needed. Click on Radar → Speckle Filtering → Single Product Speckle Filter.
NOTE 2: Speckle noise-like feature is a common phenomenon in SAR systems. It confers to SAR images a granular aspect and random spatial variation. The source of this noise is attributed to random interference between the coherent returns. The principle of speckle filtering is to reduce the variance of the complex speckled scattering and improve the estimate of the de-speckled scattering coefficient.
Then in the I/O Parameters tab, select as input the subset product created previously (index ) and set the output folder to …/OCEA03_OilSpill_Kuwait/Processing/ In the Processing Parameters tab, all the settings remain as default. Click Run and display the result afterwards.
Click + to expand the contents of the file (index ), then expand the Bands folder and double click on the Amplitude_VV band to visualize it. The blackish area corresponds to the oil spill (see NOTE 3).
NOTE 3: The all-weather and day-and-night sensing capabilities, spatial coverage, revisit time, and scattering of the SAR signal are some of the features that allow the use of Sentinel-1 as source of information for an oil spill surveillance program. The backscatter of the SAR signal over the ocean is mainly a result of sea roughness (i.e. short gravity-capillarity waves). Oil films decrease the sea surface roughness and hence the backscatter. This cause spills to appear darker in SAR images than spill-free areas. However, the contrast between polluted and non-polluted areas depends on different parameters such as wave height, wind speed, type of oil and sensor characteristics (wavelength, polarization, incident angle).
Oil spill profile plot
To visualize how an oil spill affects the reflectance of the SAR signal we can display a profile of the sigma naught (σ0) value in the VV polarization mode. Click the line drawing tool icon (see screenshot below) and draw a line through the oil spill that starts and ends in a non-oil spill area.
Click on Analysis → Profile Plot. Change the Box size parameter to adjust the graph and analyze it.
Oil Spill Mapping
To identify oil spills in the ocean, we will use Sentinel-1 data and the dedicated tool that SNAP offers for this purpose. However, it has to be highlighted that only ‘possible oil spills’ are detected since some specific oceanic conditions can generate similar visual patterns to the ones of an oil spill (See NOTE 4).
NOTE 4: SAR imagery over oceans usually contains oceanic and atmospheric phenomena referred to as look-alikes that can cause false alarm detections. They dampen the short waves and create dark patches on the surface, originating problems to distinguish them from oil spills. Look-alikes include natural films/slicks, grease ice, areas with specific wind speed, rain cells, internal waves, etc.
Click on Radar → SAR Applications → Ocean Applications → Oil Spill Detection. The tool includes some preprocessing steps such as land masking and calibration and the required algorithm to identify possible oil spills (See NOTE 5).
NOTE 5: The oil spill detection tool includes two preprocessing steps: mask out the inland areas and radiometric calibration so that pixel values truly represent the radar backscatter of the reflecting surface. After those preprocessing steps, dark spots are detected using an adaptive threshold algorithm where the local mean backscatter level is estimated using pixels in a large window. Then, a threshold is set to ‘k’ decibel below the local mean calculated before. Pixels within the window with values lower than the threshold are detected as dark spot. Finally, the detected pixels are clustered into a single cluster and those with sizes smaller than a predefined size selected by the user are eliminated.
Follow the instructions to complete the parameters of each tab.
- Read → select the speckle filtered product (index ).
- Land-Sea-Mask and Calibration → all the parameters remain as default.
- Calibration → Select only the Sigma0_VV as source band. Set the Background Window Size to 1400 and the Threshold Shift (dB) to 3.5
- Oil-spill-clustering → the parameter remains as default
- Write → set the output name to ‘Oil_Spill_detection_spk_1400_3_5’ and set output Directory: …/OCEA03_OilSpill_Kuwait/Processing
The output product is created as a binary mask that can be found on the ’Bands’ or ‘Masks’ folders of the product . Expand the bands folder and open the Sigma0_VV band. To improve the visualization and contrast, we can transform the pixel values using the decibels scale. For that, right click on the band Sigma0_VV and select Linear to/from dB. In the pop-up window, click Yes. The image will be created and store as a virtual band. To save it, right click on the band Sigma0_VV_db and select Convert band. Then, double click on it to visualize it.
To have a better visualization of the oil spill mask, display it on top of the SAR image. For that, open the Layer Manager (Layer → Manager), expand the ‘Masks’ folder and select the Sigma0_VV_oil_spill_bit_msk_detection band.
After the oil spill detection is completed, we can reproject our data into a specific coordinate reference system. To perform this step, we will use the Ellipsoid Correction (See NOTE 6).
NOTE 6: Amongst the different options to perform geometric corrections in SNAP we use in this case the Ellipsoid Correction and not the Range Doppler Terrain Correction. Since our study area is over the ocean, there are not topographic variations that can lead to geometric distortions of the SAR backscatter.
Click on Radar → Geometric → Ellipsoid Correction → Geolocation Grid
On the I/O Parameter tab, select the oil spill detection product (index ) and make sure to select the desired output path: …/OCEA03_OilSpill_Kuwait/Processing. Then, click on the Processing Parameters tab, select the UTM / WGS 84 (Automatic) projection and click Run.
Follow the same procedure as before to produce the final visualization.
Visualization in QGIS
To export the result to QGIS, we will first change the color of the oil spill mask. Expand the reprojected product (index ), open the ‘Bands’ folder and open the Sigma0_VV_oil_spill_bit_msk. Click on the Colour Manipulation tab (bottom left corner) (or View → Tool Windows → Colour Manipulation), select the Table editor and set all the colours to red.
Once the colour has changed, right click over the image. Select Export View as Image. Select the …/Processing folder to save the product, choose the option Full Scene in the image region section and write Oil_Spill_Detection as name. Then, click Save.
Minimize SNAP and open QGIS (Applications → Processing → QGIS Desktop). Press the Add Raster Layer button. Navigate to the …/Processing folder and select the Oil_Spill_Detection GeoTIFF file. Click Open.
For further analysis, you can add auxiliary data such as ocean, land and highways layers. Press the Add Vector Layer button and add the additional shapefiles (highways, administrative boundaries and ocean: you may request them to the Serco EO Training Team).
You can also use the ‘OpenLayers plugin’ (See NOTE 7) to display the result using OpenStreetMap as background map. Click on Web → OpenLayers plugin → Google Maps → Google Streets. In case Google Satellite is not available, use a different layer, e.g. Bing → Bing Aerial.
NOTE 7: If you do not have the ‘OpenLayers’ plugin installed, click on the menu Plugins → Manage and install plugins. Select the ‘All’ section on the left side panel, write ‘openlayers’ on the search box, select the ‘OpenLayers Plugin’ and click install.
Further reading and resources
- Sentinel-1 and oil spill detection – ESA
- European Maritime Safety Agency (EMSA)
- Brekke, C., & Solberg, A. H. S. (2005). Oil spill detection by satellite remote sensing. Remote Sensing of Environment, 95(1), 1–13
- Schistad Solberg, A. H., Storvik, G., Solberg, R., & Volden, E. (1999). Automatic detection of oil spills in ERS SAR images. IEEE Transactions on Geoscience and Remote Sensing, 37(4), 1916–1924
- Stathakis, D., Topouzelis, K., & Karathanassi, V. (2006). Large-scale feature selection using evolved neural networks (Vol. 6365, pp. 636513–636519)
- Topouzelis, K., & Singha, S. (2016). Oil Spill Detection: Past and Future Trends. Living Planet Symposium, SP-740 (July)