Extracting Shoreline Polygons and Polylines from Sentinel-2 Imagery

Introduction

The delineation of shoreline polygons can have a variety of practical applications from shipping and navigation to change detection and mapping coastal erosion. Often shoreline shapefiles are created using a semi-automated method combining automated boundary extraction and manual interpretation. This tutorial will focus on delineating shoreline polygons and polylines from Seninel-2 satellite data using the Naturalized Difference Water Index [NDWI] to define the boundary between water and land. This boundary will then be extracted and manually corrected to produce a final shoreline polygon and polyline file.

The purpose of this tutorial is to show the user how to extract shoreline polygons from satellite data in the Qgis software package. For this tutorial the Copernicus Open Access Hub will be used to access the required satellite imagery. The report will explore how to use this data portal to brows and download the available Sentinel-2 imagery. The satellite imagery will then be imported in to Qgis where polygons and polylines of the shoreline will be created.

Both Qgis and the Copernicus Open Access Hub are free to use, however the Copernicus Open Access Hub does require the user to create an account to download the data. The latest version of Qgis can be downloaded from the following link: (https://www.qgis.org/en/site/ )

Note: QGIS version 3.8.1. was used to create this tutorial. A newer version of the software may be available.

Data

This report will use Sentine-2 satellite data collected by the European Space Agency. The Copernicus Sentinel-2 satellite mission is a constellation of two multispectral polar orbiting satellites ^[1]. The satellites offer data in the form of Level-1C and Level-2A data products. The Level-C1 products are multispectral satellite images that include the top of atmosphere reflectance. This means they do not have any atmospheric correction applied. Level-2A data products are derived from Level-1C data and offer bottom of atmosphere reflectance. In this case atmospheric corrections have been applied to the image to remove some effects of the atmosphere on the image’s reflectance ^[2]. It is worth noting that Level-2A data may not be offered for many regions outside of Europe. For this reason, the study will use Level-1C data.

Both the Level-1C and Level-2A data products have differing spatial resolutions depending on the multispectral bands being viewed. This study will use the Red, Green, Blue and NIR bands. All four of these bands have a spatial resolution of 10m with a swath width of 290m. In addition, when the data is downloaded it is projected in the relevant UTM zone using the WGS 83 datum ^[3].

Tutorial Methods and Instructions

Step 1. Using the Copernicus Open Access Hub

Click on the following link to access the Copernicus Open Access Hub: https://scihub.copernicus.eu/dhus/#/home

Click on the Login icon at the top right of the screen, create an account if you currently do not have an account or login to an existing account (Note: you cannot browse or download the data without creating an account)
After logging in you can now browse all of the data offered on the Copernicus Open Access Hub
To pan the field of view you can click and drag the underlying map you can also zoom in and out using your scroll wheel
You can change the basemap of the data portal by clicking on the Map Layers button on the top right of the screen

You can click and drag the scroll wheel to manually draw a rectangle over your area of interest, when you do this a yellow box should appear over the area you selected (Note: if you click and drag the scroll wheel again you will draw a new area of interest and the previous one will be lost)
Zoom in to the city of Corner Brook, on the west coast of the island of Newfoundland, Canada
Click the scroll wheel and draw an area of interest encasing Cox’s Cove, the bay above Corner Brook as seen in the image below.

Step 2. Browsing the Copernicus Open Access Hub

Click on the List button on the left of the data search bar
You should now see an Advanced Search drop-down menu appear below the search bar as seen below

This menu will allow you to filter all of the data products offered by the data portal and allow you to search for specific criteria and date ranges
Under the Sort By header change the option to Sensing Date, this will show you all specified data products in order of when each image was taken
Under the Sensing Period header select the starting date October 1st 2019 on the left and the end date October 10th 2019 on the right (this will filter the search results to images that were taken on the date range specified)
Note: if you use the Ingestion Period option you will filter images based on the date they were added to the Copernicus Open Access Hub archive not when the images were taken
Check the box beside the Mission: Seninel-2 header to restrict all images to only the Seninel-2 satellites
Under the Product Type header in the Mission: Seninel-2 option select the S2MSI1C option, this will restrict all images to the Level-C1 product type
Note: using the Satellite Platform option will restrict the filtered images to the selected satellite from the Seninel-2 constellation
Note: using the Cloud Cover % option allows the user to specify what proportion of the image will be taken up by could cover, this can be useful to quickly filter out cloudy images from your search
Your search menu should look like the following

After inputting these search options click the Search button

Step 3. Browsing the Copernicus Open Access Hub

You should now see a list of different images that fit your search criteria from above
Notice on the basemap the outline of each image footprint can be seen as a green outline

When you hover your mouse over the images on the left of the screen the corresponding footprint will darken
Notice how the track of the satellite is broken down into multiple images, very often you will need to download multiple images to get the full coverage needed in a study site
Examine the different images available, you can click on the View Product Details button to get a better look at the image you have selected
When you click the View Product Details button you will be able to see additional information about the image, including the image footprint, time of acquisition and a variety of metadata
Notice each image has an associated name that quickly describes the image and some key pieces of information, here is a breakdown of the most important parts of the naming convention for Sentinel-2 data

For example, you have the image…

S2B_MSIL1C_20191010T145729_N0208_R082_T21UVQ_20191010T164326

S2B: This is the mission ID and satellite that acquired the image (Seninel-2 using the Seninel-2B satellite)

MSIL1C: This is the product level (Level-C1)

20191010T145729: This is the date and start time of acquisition (October 10th, 2019 at 14:57:29pm)

In addition to this each name acts as a unique identifier for the image

Find the two following images in the list of search results and click the Download icon to download a zipfile of both images

S2B_MSIL1C_20191010T145729_N0208_R082_T21UVQ_20191010T164326

S2B_MSIL1C_20191010T145729_N0208_R082_T21UUQ_20191010T164326

Note: if you are having trouble finding these two images, clear your selections in the search window then cut and paste the image name into the search bar to find each of the images one by one

Step 4. Installing Plugins for Qgis

Go to the folder where you downloaded the two images and extract both file folders
Open Qgis
On the top of your screen click on the Plugin tab
A window of all the available plugins for Qgis should now appear
Plugins are different tools created by the Qgis community that add additional functionality to the base Qgis software
Search for and download the Value Tool plugin and the Virtual Raster Builder plugin

After the plugins are installed make sure they are enabled by clicking the checked box next to the tools name

Step 5. Opening Sentinel 2 Data in Qgis

You can now import the Sentinel-2 imagery into Qgis
Click on the Open Data Source Manager button at the top left of your screen
The data source manager window should now pop up, navigate to your folder where the extracted Sentinel-2 images have been stored
In the file folder for the first image navigate to the GRANULE > L1C_T21UUQ_A013550_20191010T145730 > IMG_DATA folder
You should now see a series of .jp2 images that correspond to the bands of the satellite image
In a Sentinel-2 image Band 2 is the Red band, Band 3 is the Green band, Band 4 is the Blue band and Band 8 is the NIR band
Each raster image in this file tells you the band number with the last three characters of the name (Example: B02 is the band number 2 corresponding to the Red band)
Select bands 2,3,4,8 and the TCI image from this folder by holding the CTRL key and clicking each file (Note: select the raster image not the polygon file with the same name, see the image below)

Right click one of the selected bands and select the Add Selected Layers to Project option
The selected bands should now be added to your project under the Layers table of contents
Click through the five bands and look at each (Note: you can change the order of what image is on top by clicking and dragging a raster image under the Layers table of contents)
Bands 2,3,4 and 8 should all be individual greyscale images
The TCI image is the true colour image and has the RGB bands mapped to 2,3,4 respectively this is why it is the only colour image out of the five

Step 6. Combining and Exporting Seninel-2 Data to GeoTIFF Format

The five bands that you imported in step 5 can be opened by Qgis however, most cannot be used with other raster tools in the software
This is why the Create Virtual Raster plugin has been installed, it gives Qgis some extra functionality so that we may create colour composites out of these individual raster images
Frist, right click on the TCI band and select Export > Save As…
Make sure the output format is GoeTIFF and save the file to your working file folder where the two Seninel-2 images are stored, give it a name like "TCI_1.tif"
Keep all the default parameters and click OK
The GeoTIFF version of the TCI image should now be added to your Layers table of contents

Second, we will create a colour composite image from the four individual bands you previously imported
Navigate to the Build Virtual Raster tool by clicking on Raster > Miscellanies > Build Virtual Raster… under the main toolbar
The Build Virtual Raster tool window should now pop up
Under the Input Layers header click the … button and add bands 2, 3, 4, 8 in this order
Keep all default options and click Run

After the tool has completed you should now have a temporary virtual raster added to your Layers table of contents
The image may look slightly orange like in the example below (This is due to the way the image is symbolized, don’t worry we will change this later)
Note that the bands have been renamed in the virtual raster from 2,3,4, 8 to 1,2,3, and 4 respectively (They should still be in the order in which you added them to the Build Virtual Raster tool)
Right click the virtual layer and select Export > Save As…
Make sure the output format is GoeTIFF and save the file to your working folder, give it a name like "ColourComp_1.tif"
Keep all the default parameters and click OK
You should now have a GeoTiff version of the TCI image and a Colour Composite image

Under the Layers table of contents select the five bands previously imported and the temporary virtual raster by holding the CTRL key and clicking each image
Right click one of the selected rasters and select Remove Layer to remove them from your project

Step 7. Merging Raster Images

Repeat steps 5 to 6 using the second Seninel-2 image giving the output GoeTIFFs different names like "TCI_2" and "ColourComp_2"
You should now have two colour composite images and two true colour images, one for each half of the full Senienl-2 Scene

We can now merge the two halves of both the true colour and colour composite scenes using the Merge raster tool
Navigate to Raster > Miscellanies > Merge the merge raster tool window should now appear
Select the two TCI rasters then click run

This will produce another temporary raster so you will need to export and save the merged file to a new GeoTIFF file using Export > Save As…
Repeat the merge process with the two halves of the colour composite image
You should now have two complete rasters, one a true colour image and one a colour composite image
Right click the unmerged rasters and remove them from the project using the Remove Layer option

Step 8. Symbolizing Raster Layers in Qgis

In Qgis you are able to change the way raster layers are symbolized
To change a raster’s symbolization right click the layer and select Properties
Under the various headers there are options for changing the image’s saturation, contrast, brightness, minimum value, maximum value and band mapping
We will now create a false colour image from the colour composite merge layer
Right click the colour composite merge raster and select Properties
Under the band rendering window, you should be able to see what bands have been assigned to the Red, Green and Blue bands (Note: the band numbers have changed in this image after creating the first colour composite image back in step 6 Band 1 in this image is the Red band from the Sentinel 2 data, Band 2 is the Green band and Band 3 is the Blue band, Band 4 is NIR)
A common false colour image combination used to view vegetation and water is NIR mapped as (red) , Green mapped as (blue) and Red mapped as (green)
This combination highlights vegetation in red because vegetation reflects large amount of NIRlight
Change the assigned bands under the Band Rendering header so that Red is assigned to band 4, Green is assigned to band 3 and Red is assigned to band 2, click OK
The image should now appear as a false colour image seen below

You can adjust the brightness on both the false colour and true colour image to your liking using the brightness slider in the Layer Properties window

Step 9. Creating a NDWI Image

The Naturalized Difference Water Index is spectral ratio that is used to accentuate the presence of water in a satellite image, it is calculated by the following formula

(Green – NIR) / (Green + NIR)

This formula is applied to each pixel of an image and uses the reflective properties of water in the NIR and Green bands derive a value from -1 to 1 that represents the water content of that pixel
NDWI is very useful for determining the boundary between land and water in an image
To create a raster of the NDWI navigate to the Raster > Raster Calculator tool
The raster calculator window should now appear, under the Bands header you should see all the available bands of each raster image loaded into your project (Note: the number after the @ symbol corresponds to the band number)
In the text box under the Raster Calculator Expression header type the following expression:

("ColourComp_M@2"-"ColourComp_M@4") / ("ColourComp_M@2"+"ColourComp_M@4")

Give the output layer a save location and name it "NDWI.tif" then press OK
You should now have a greyscale raster of the NDWI as seen below

Step 10. Finding the Water Land Boundary

Now that we have an image of the NDWI it is possible to find the boundary between water pixels and land pixels in the image
First click on the Identify Features button then click on the Value Tool button to activate the tool (Note: to turn off the tool click the identify feature button again)
Zoom in to the image and find an area the water meets the land, click around the image to see the corresponding pixel value at each location in the popup window on the right of your screen (Note: light pixels are water, dark pixels are land)
Look at the different values between the water pixels and land pixels to see if you can find a general value that separates the two (Note: this value will change between different satellite images and even within the same satellite image)

In the example above the value separating land and water appears to be around -0.169 so this will be the value where we draw the line between water and land
To generate a raster of the land and water boundary navigate to the raster calculator tool under Raster > Raster Calculator
Enter the following expression for the tool and specify the save location (Note: NDWI@1 is the name of the NDWI raster that was created, this name may change depending on what you named it)

(("NDWI@1">= -0.16)=1) AND (("NDWI@1"<=-0.16)=0)

The resulting raster should be a black and white raster showing black as land and white as water

Step 11. Creating an AOI

The next few steps would be very labour intensive and take a long time to run on the full Sentinel-2 image so we will create a smaller subset of our image to work with going forward
To extract an area of interest or AOI from the scene we first need to create a polygon around the area we are interested in
To create an empty shapefile layer, navigate to Layers > Create Layer > New Shapefile Layer
Give the new shapefile a name like AOI and specify a save location then change the geometry type to Polygon and press OK
You should now see a polygon layer under the table of contents called AOI
To draw your AOI right click the new polygon layer and select Toggle Editing
Click on the Add Polygon Feature button
You can now click and draw the four corners of your AOI rectangle, when you have finished the fourth corner right click press OK (Note: if you click cancel you can delete the currently drawn polygon and start again)
You should now have a rectangle over the are you drew the polygon, you may choose a different location then the tutorial if you wish

Right click the AOI polygon layer under the table of contents and select Save Layer Edits then right click the layer again and select Toggle Editing to stop editing
Navigate to Raster > Extract > Clip Raster by Mask Layer…
The Clip Raster by Mask Layer tool window should now be open
Under the Input Layer specify your black and white land water boundary raster, under the Mask Layer specify your AOI polygon
Specify a save location for the Clipped (Mask) and give it a save location then press Run
You should now have a black and white land water boundary raster clipped to your AOI extent

Step 12. Creating Land/Water Polygons

To create a polygon layer from the clipped black and white raster navigate to the Raster -> Conversion > Polygonize (Raster to Vector)… tool
Make sure the input is your black and white raster (Note: the output will be a temporary file so you do not need to specify a save location just yet)
The output will be a polygon version of the raster clipped black and white raster (Note: each polygon still has the associated pixel value of the pixel that was used to create it)

To select the polygons that only correspond to land first click the "Vectorized" polygon layer to highlight it in blue, then click on the Select Feature by Value button
A window should now pop up, under the DN heading type 0 then press the select features button
The land should now be highlighted in yellow, you can now click Close to close the window

You can now export the land to a new permanent shapefile by right clicking the "Vectorized" layer under the table of contents and selecting Export > Save Selected Features As…
Make sure the output format is ESRI Shapefile then specify a save location and name for the output file
You should now have land only polygon layer

Step 13. Correcting the Land Polygon

Turn off all rasters except for the false colour image and bring the land polygon to the top of the table of contents
Zoom in and pan around the image, you will notice many areas were the shoreline is incorrectly drawn
This is the result of many different errors like shadow, cloud, shallow water and even selecting the wrong value for the NDWI boundary in step 10
This is why manual correction is always needed before finalizing shorelines and it highlights one of the many downfalls to extracting shoreline from satellite imagery
You can manually correct the shoreline by right clicking the polygon layer in the table of contents and selecting Toggle Editing
In the image below we can see that the shoreline has incorrectly isolated some small pixels classifying them as land, in addition some areas have incorrectly been identified as water
You can use the false colour and true colour image to compare what should be shoreline and what should be water by turning the layers on and off in the table of contents

The easiest way to correct this section of the image is to delete all of the isolated pixels then go back and redraw the shoreline manually
To delete polygons from the layer use the Select Features by Area or Single Click button then click on the polygon you want to delete (Note: you can click and drag to draw a rectangle over an area to select multiple polygons at once)
The selected polygon should now be highlighted in yellow and you can press the delete key to delete it
To undo any previous actions, use the hot key CTRL + Y
To redraw the shoreline, navigate to the Edit > Reshape Features tool
Your cursor should now be a crosshair icon
To redraw the polygon first click inside the polygon then trace the outline of the shore, clicking along the way to add nodes (you should see a red line showing you where you have drawn)
Once you have redrawn the shoreline click inside the polygon once more to add a node inside the polygon’s boundary, then right click to finish redrawing the polygon
To add a new polygon over an island or area that has been missed click the Add Polygon Feature button and draw a new polygon, right click and press OK to finish drawing the new polygon
To delete incorrect holes in the shoreline polygon, navigate to the Edit > Delete Ring tool
Then click the hole to remove it from the polygon
You can then pan around the image and correct different areas where the shoreline has been incorrectly drawn
Remember to save your edits and toggle off Toggle Editing before moving on
Overall manually correcting an AOI can take hours to days, even weeks depending on the size of the area being mapped

References

^[4] ^[5]

^[6]

↑ [1],Retrieved: 2019-11-05
↑ [2],Retrieved: 2019-11-05
↑ [3],Retrieved: 2019-11-05
↑ [4],Sentinel-2, Retrieved: 2019-11-05
↑ [5],Data Products, Retrieved: 2019-11-05
↑ [6],Resolution and Swath, Retrieved: 2019-11-05

[1] [1],Retrieved: 2019-11-05

[2] [2],Retrieved: 2019-11-05

[3] [3],Retrieved: 2019-11-05

[4] [4],Sentinel-2, Retrieved: 2019-11-05

[5] [5],Data Products, Retrieved: 2019-11-05

[6] [6],Resolution and Swath, Retrieved: 2019-11-05

[1]

[2]

[3]

[4]

[5]

[6]