Three (3) different space-based solutions have been attempted to solve the problem of hydrocarbon contamination. The first method involves the use of an oilspill detection tab in SNAP. This method has limitations if the user cannot primarily identify the area where the spill occurred. The second method is the use of a script where all the steps which could be done manually are incorporated within the script. This solution also works well with a limitation of understanding and editing scripts. The last solution involves identifying thresholds of pixel values to detect the spill regions. This solution has produced the most desirable results, however it involves lengthy processes.

Requirements

Data

• Pre-spill and During spill Sentinel 1 SAR data from Alaska Satellite Facility ASF Home | Alaska Satellite Facility .You might need to create an account to enable download
• Click on the “vertex” icon on the left side of the “services” icon. A new page opens
• On the map interface, click drawing tool (with the letter “A” written in it) under the icon “Area of Interest”. Select “Draw a polygon” and draw a polygon around your area of interest.
• Click on “Filters” You can import your shapefile if you wish. Move down and enter your search dates (start and end dates)
• You can add additional filters like the file type. Under the file type, it is advisable to choose the ground products (GRD) which are georeferenced. The Single Look Complex (SLC) products are preferred for Interferometric (InSAR) analyses. You can change the beam mode to IW (Interferometric Wide Swath)
• You can leave the other parameters under “additional filters” as default to avoid limiting your search results.
• Then download product

Software

• The Sentinel Application Platform (SNAP) 11.0 software was used for solutions 1 and 3. SNAP is an open-source software developed by the European Space Agency (ESA) to support the processing and analysing of earth observation data particularly from the Sentinel repository. Link to the Sentinel platform SNAP Download – STEP
• Scroll down and click on either Windows, Mac or Linux depending on your computer’s operating system requirement.
• Google Earth Engine (GEE) was used for solution 2. GEE is an open source, cloud-based platform for processing geospatial data. It works with JavaScript and uses codes to pull out the needed satellite data from different repositories
• Link to GEE platform Google Earth Engine
• Click on “platform” and then “code editor”. This takes you to the code editor interface where you can write and save your scripts or run them.

Physical

• Information from the National Oilspill Detection and Response Agency (NOSDRA)was used to validate the occurrence of the spill

Outline steps for a solution

1 Manually Download Image- The first step is to manually download two (2) Sentinel 1 images from the Alaska Satellite Facility as described above. One image (archived image) before the spill and the other during the spill taking into consideration the temporal resolution (revisit period) of the sensor. The images will be zipped. There is no need to unzip the images

2 Load Image- Load the images onto the SNAP software. Simply drag and drop the zipped images onto the product explorer interface. Click on the icon on the left of the image file and the metadata and band icons will be visible. Click on the “bands” icon to reveal Amplitude VV, Amplitude VH, and Intensities. Load the Amplitude VV bands for both pre and post spill images and compare

3 Subset Image-Create a subset of your Area of Interest by using the subset tool under the icon “Raster”. This helps to reduce processing time as you will be focusing on a smaller area (area of interest),rather than the whole satellite image.

4 Image Pre-processing- Preprocessing the image by Radiometric (speckle filtering and calibrating) and Geometric Corrections in SNAP

4.1 Speckle filtering- This is the removal of background noise, which literally appears as speckles from an image. A speckle filtered image always appears less “grainy” than a non-speckle filtered image. In the Radar toolbar, you will find speckle filtering as the third option, select the single product speckle filter sub-option and use the subset file as input. Under processing parameters, choose amplitude VV, which is the best channel for detecting oil spills. Choose the Lee  filter (3*3) and run. Finally, go to file in the menu bar and select tile horizontally, to have pre-speckle filtered image and post speckle filtered image side by side for comparison. After speckle filtering, the result produced on the product explorer space usually has the extension (.spk)

4.2 Calibration- This is done to improve radiance and reflectance to ensure that digital numbers (DN) accurately represent the reflectance of the physical characteristics of the SAR image. Speckle filtering and calibration are types of radiometric corrections. The calibration tool can be found under the “radar” icon in the SNAP software. Convert the result (sigma vv) from linear to db (decibels). This results in a virtual band. Right click (sigma-vv-db) and convert to band to have a real band. After speckle filtering and calibrating, the result produced on the product explorer interface usually has the extension (.spk cal)

4.3 Terrain Correction- If the Area of Interest was only the ocean, we might not need to do a terrain correction. However, it will be needed in this case as we are more concerned with the terrestrial environment. Go to “Radar” then “Geometric”, then “Terrain correction”. Choose the Range doppler terrain correction option. Set the options for your Digital Elevation Model (DEM)and run

After terrain correction, the result on the product explorer interface will have the extension (Spk.Cal.TC-1)

The three pre-processing steps documented above are required for solutions 1 and 3 

Next step for Solution 1

Oilspill detection Tool: Within the Radar toolbar in SNAP, there is a “SAR Applications” window which further opens into “Ocean Applications” and then the oil spill detection tool can be found. After entering the parameters and running the tool, it creates a mask of the spill area and the geometry to aid spill area calculation.

Solution 2

GEE Platform

The steps listed below are the steps coded in the script to delineate areas with oil spills

1 Define Area of interest (AOI): The region of interest must be defined either by using a shapefile or geotag.

2 Define Temporal Scale: The revisit time (temporal resolution) of the Sentinel 1 SAR sensor is twelve (12) days. Dates must be entered to accommodate this temporal resolution

3 Load Sentinel-1 Ground Data:The Sentinel-1 data with the ‘vertical-vertical’ polarization (VV) which is best for oil spill detection is loaded,

4 Speckle filtering to reduce noise: Most Sentinel images come with speckles which may obscure the viewing of important information. Speckle filtering helps to provide a clearer image and better feature extraction.

5 Oil spill Change Mask and Thresholding is applied: This is done using the differences in the band sensitivity between the oil and water areas to delineate the oil spill areas.

6 Validation using the Nigerian Oilspill Detection and Response Agency (NOSDRA) database:

Code will be submitted in a separate document.

Solution 3 (Thresholding)

After the preprocessing steps explained above, the next steps are as follows

1 Layer Stacking-This is done to perform advanced analysis on the bands. Go to Radar, under “Radar”, you find coregistration then “stack tools” then “create stack”. Choose the bands you want to layerstack, this should be the (sigma0-vv-db bands) of the pre spill image and the post spill image.

2 Open RGB image- Now that the bands are stacked, you can do a band combination to have an idea of the areas where the spill occurred. Right click on the stack and “open RGB image”, put the post spill data in red and green and leave the pre spill data in blue, then run. This returns an image with the areas where the spill likely occurred in the image.ee RGB image window below.

The green areas are the likely spill areas. The shape file covers the region of interest but we are interested in the surrounding areas as well.

Histogram Generation to obtain threshold values- Thehistogram shows the region of the oilspill.Load the histogram of the post spill image by highlighting the image in the product explorer window (sigma00-vv-db), click “Analysis”, scroll down and click histogram, then generate histogram. This gives the range of pixel values where the spill can be detected. See histogram below

Here, we see that the histogram range is from 0 to -25. However, from -15 to -25 looks like a false peak. So the threshold value chosen is -15 to -17

Also, from the image, if you hover around the green areas and check the pixel info on the “pixel info” tab next to “product explorer”, it ranges from -15 to -25, it is explained above why -15 was chosen.

The last step is to use this chosen value or range of values as the case may be to create an expression using the band maths tool to obtain the final spill area.

Band Maths Expression -Access the band maths tool which is  the first option under  “raster”. Create an expression with your chosen threshold values as seen in the figure below.

In the band maths expression above, the pre-spill data (24th May 2023) is subtracted from the post-spill data (17th June 2023) to obtain the spill area.

Results

There is an image with likely spill areas

Describe the impact this solution has on the ground