There is Never Enough Data — WorldView, Landsat 8 Collection 2 and More
Data collections keep expanding, and you can help as well!
Authors: Monja Šebela, Anja Vrečko, William Ray, Grega Milčinski
With Sentinel Hub gaining popularity, there is an increasing demand to integrate additional data collections in order to support various use-cases. This is not an easy task — we must ensure that data is available in the appropriate format in one of the cloud platforms, where Sentinel Hub is deployed. In some cases we have to agree on the licensing, and almost always there is significant effort required to ensure that data flows through our service without negative impact to quality, so that we can honestly state that one can get original pixel values. Recently, through our partnership with European Space Imaging, we have integrated Maxar’s WorldView data, Landsat 8 collection 2 and several Copernicus data collections. As we believe that many of our users have data which they could contribute to the community, we have designed a process for our users to share their “Bring Your Own Data” collections with others as well.
Maxar WorldView Supplied by European Space Imaging
Sentinel and Landsat are great missions for ongoing monitoring of an area. However, people often want to “dig” deeper into details and so they need very high resolution (VHR) data. In the field of multi-spectral imagery, there are two leading providers of VHR data — Maxar (formerly known as DigitalGlobe) and Airbus. We’ve introduced support for Airbus Pleiades and SPOT a year ago, along with the Planet data. Maxar’s data was missing, though. Due to the sporadic nature of VHR tasking (contrary to systematic observation scenarios of Sentinels, which we have all become accustomed to), it is even more important to have access to the widest possible archives. With extensive support of European Space Imaging, the European partner of Maxar who are able to task the WorldView satellites from their ground station in Germany, we can now finally include VHR data from this constallation into the Sentinel Hub.
WorldView is a constellation of four operational satellites: WorldView-1, Geoeye-1, WorldView-2 and 3. Unfortunately, WorldView-4 satellite is not operational anymore since January 2019, but the data it acquired in its lifetime is still available. Soon, Maxar is launching new satellites, Maxar Legion, which will increase the volume even further.
Five bands are available in Sentinel Hub, a panchromatic band with spatial resolution of 0.5 m and 4 multispectral bands with spatial resolution of 2 m: red, green, blue and near infrared band. It is possible to purchase data for a selected area of interest, which must be at least 5 km². Data is not acquired systematically — instead, acquisitions need to be tasked. However, full archive of data is available.
Purchasing, ordering and accessing data in Sentinel Hub works the same as for other third party data we offer: use our TPDI API examples to search and order data or work with the Requests Builder, if you prefer a graphical user interface. Once the data is imported, you can use all the standard “data access goodies” which are essential ingredients of Sentinel Hub: mosaicking of tiles, access in all the supported coordinate systems, custom processing and visualizing of data using evalscript or batch processing, visualization in EO Browser or simple-to-integrate in your GIS application access using WMS or WMTS services. The data can be used to monitor land cover and vegetation in particular.
Let’s show some data
Enough technical and background information! Let’s check how the data actually looks like.
The eyes of the world were watching the Suez Canal earlier this year. It was obstructed for six long days due to unfortunate grounding of one of the large container ships, Ever Given. Sentinel-2 captured the event on 29 March 2021, just after the ship had been released from the hold of the canal banks.
Two days earlier, on 27 March 2021, WorldView-3 also acquired a cloudless image of the ship.
The difference between spatial resolutions of Sentinel-2 and WorldView images is significant when observing objects of this size. Compare the Sentinel-2 true color images (first row) to WorldView true color images (second row) and WorldView pansharpened images (third row) above. In the last image it is even possible to count how many containers are in a row.
If you want to get familiar with this collection, make sure to participate in the Africa Multi-Resolution Cube Contest, where samples of WorldView data will be available for all participants. We will announce the Contest on our webpage and social media channels in the coming few weeks, so stay tuned!
If interested, contact us or to go to our Dashboard for a purchase.
Landsat 8 Collection 2
Landsat 8 Collection 2 imagery has been added to Sentinel Hub, granting users access to a full global Landsat 8 dataset in level 1 and level 2. Landsat is well known to everyone — the sensor has been collecting data from its OLI (Operational Land Imager) and TIRS (Thermal Infrared Sensor) since February 2013, meaning there is a huge 8 years of archive at 30-meter resolution ready to be accessed.
We have been offering Landsat 8 L1C (Level-1 from Collection 1) data for a while, but this will now be superseded by Collection 2 data and it is important to know that Level-1 data from Collection 1 (currently L8L1C) will soon be deprecated. To adapt your scripts using our Process API you just need to set the “type” field of the “data” object to “LOTL1” for Level-1 (from Collection 2) and “LOTL2” for Level-2 (from Collection 2). Check our forum post to learn how to use it in OGC requests as well.
There have been several enhancements made to the newly available Collection 2 data, with several radiometric and geometric improvements using improved processing methods. The advantage of using surface reflectance products means your measurements aren’t interfered with the atmospheric effect of gases and aerosols, producing a better result for your application.
More about the Landsat 8 Collection 2 Level-2 can be found in our Sentinel Hub documentation, and on the U.S. Geological Survey product page — Science Products, and Data.
Note that we are working on adding Landsat 5 next, which will hopefully be available over the next couple of weeks.
Copernicus Digital Elevation Model
Next to the previously supported Mapzen Digital Elevation Model (DEM), Sentinel Hub now also supports global Copernicus DEM in 30- and 90- meter resolution, representing the Earth’s surface, buildings, infrastructure and vegetation. That’s great news, as Copernicus 30 meter DEM has a better resolution in most of the world, where Mapzen only has 90 meters! Copernicus DEM also has generally better quality than Mapzen and is thus more suitable for orthorectification and radiometric terrain correction of Sentinel-1 data. Public COPDEM30 is not 100% global yet — in a few areas where the collection is not available, COPDEM90 is used automatically.
To select Copernicus DEM in your requests, set the data collection to DEM and set the demInstance to COPERNICUS_90 or COPERNICUS_30.
Copernicus DEM collection has one band, called DEM, which holds the values of elevation in meters. Based on this, the users can create visualizations with custom terrain colours and even calculate the contour lines. In EO Browser, there are 3 different pre-prepared visualizations available — grayscale, color and sepia, as displayed on the Copernicus DEM 30 meter images over the Alps below.
As all the Sentinel Hub supported DEM’s are also supported by the processing API, you can even use DEM in data fusion. On the left image below, we can see the contour lines extracted from the Copernicus 30 metre DEM over the Sava river in Slovenia, and on the right, we can see a data fusion visualization of Nyiragongo, where white contour lines over Sentinel-2 imagery display the steep slopes of the volcano.
On the Sentinel-1 image of the Western Australian landscape below, the 30-meter Copernicus DEM was used for orthorectification and radiometric terrain correction. It’s so smooth it almost doesn’t look like Sentinel-1!
Public Collections Catalogue
We have launched a new Public Collections Catalogue, where anyone can contribute their Bring Your Own Data (BYOC) collections! We are in the process of registering all our public collections, and have already published several of them.
Do you remember the Sentinel-2 L2A 120m, which you can use to explore cloudless Sentinel-2 mosaic for the whole world? You can find it in the catalogue, along with other public collections!
For each collection, there’s a dedicated page with useful information, examples and links. To use any collection on the registry, just click on the “details” link and copy the collectionId on the right (don’t forget to check the endpoint as well!), then enter it into your BYOC layer in the Configuration Utility or into your process API request.
You can contribute to the catalogue using git, following the instructions we prepared for you. We invite you to check it out and publish your public collections here, so the catalogue grows and becomes ever more useful to developers, students, researchers and enthusiasts.
Copernicus Services
The Copernicus Programme provides several Copernicus services for a variety of application areas. In partnership with the European Space Agency (part of our Euro Data Cube activity) and CreoDIAS, we have now started to pro-actively on-board these data. For the moment, Sentinel Hub supports 3 different Copernicus services, all freely available as public collections on our Public Collections Catalogue, described above. These are Corine Land Cover (CLC), Global Land Cover (GLC) and Water Bodies, all of them with a 100 meter resolution.
They are integrated into EO Browser, and can be accessed as BYOC layers using a public collection ID, that you can find on our Public Collections Repository. Copernicus services data can then be used to create custom layers in the Configuration Utility and in API requests, so that you can integrate them into your own applications.
For quality and validation information on the CLC see the document on data acquisition methodology and the Validation report part of the CLC website here. For Global Land Cover quality information check this website and for water bodies visit this website.
Corine Land Cover
Corine Land Cover is a vector based data collection with 44 unique classes of land use and land cover, each one with its own unique value. It’s available in most of Europe and updated every 6 years — the first available being 1990 and the most recent one in 2018. The classification is derived from a series of satellite missions. In the majority of European countries, the classification is produced by the interpretation of high-resolution satellite imagery, while in a few countries, semi-automatic solutions, including in-situ data, GIS integration and generalisation are included as well. CLC is commonly used for land use and land cover monitoring, analysis and change prediction for various applications, including agriculture, environment, transport and spatial planning.
The images of Zaragoza (Spain) below display Artificial Surfaces (upper left), Agricultural Areas (upper right), Forest and Seminatural Areas (lower left) and CLC with all the classes combined (lower right).
In the evalscript, it’s easy to classify the classes as needed. On the example of Berlin below, we grouped all the Artificial Surface classes (values 1–11) and colored them red, and all the Forest and Seminatural Areas (values 23–29) and colored them green. We returned other pixels white. This visualization gives us an understanding of the distribution of built-up areas in comparison to naturally vegetated ones.
CLC can also be used in comparison with other collections, such as Sentinel-2. On the example below, we compare the Ulyssys Water Quality visualization (which displays chlorophyll and sedimentation in water bodies) with the CLC water bodies layer to examine the health of Veneto lagoons (in green, value 42).
See more information and the information on all the classes here. The classification and visualization of the CLC layer in EO Browser is made using this nomenclature document. The classification script can be found on our Custom Script Repository. The other 5 layers in EO Browser were obtained by integrating classes into a higher level classification based on the classification here.
Global Land Cover
Global Land Cover is a classified product with land use and land cover classification developed based on the UN-FAO Land Cover Classification system LCCS. There are 23 land cover classes available for the Discrete Classification Map band. There are several other bands available, including the continuous fractional layers. The available bands, values, classes and colors used can be checked here. The collection is global, making it possible to aid policy decisions on various issues, including agriculture and food security, biodiversity, climate change, forest and water resources, land degradation & desertification and rural development on a global level. Data is updated each year, available since 2015, with the most recent year 2019.
In EO Browser, 2 classification layers are pre-prepared — the Discrete Classification Map layer, displaying all 23 land cover classes and the Forest Type layer with 6 classes, displaying different forest cover types. Both visualizations follow color coding specifications described in the product manual. The Discrete Classification Map script can be accessed on our Custom Script Repository.
On the image below, you can see the Discrete Classification Map of Tanzania. Forest cover can be observed in green colors, with different forest types clearly visible. Agricultural areas can be observed in yellows and oranges, water areas in blues and built up areas in reds and pinks. Such classification is highly valuable for any country, as it takes huge effort to classify land cover, especially in underdeveloped areas of the world.
Fractional layers are available for several class types and are very interesting, as they provide the continuous probability that a certain area has a certain land cover. They give us the entire range of values from 0 to 100, and make it possible to visualize classes in RGB composites.
We have achieved the following visualization of urban areas in the US by simply dragging the Change_confidence_layer (the layer indicating the changes from the previous year) into the red channel and the fractional built-up layer (BuiltUp_CoverFraction_layer) into both green and blue channels, to visualize urban areas in cyan. The background is black, because there was no change detected since the previous year in this area.
In the example below, we examined the changes in forest cover between 2015 and 2019. To do so, we returned the tree cover fractional layer (Tree_CoverFraction_layer) in green and blue channels, coloring forests cyan. In the red channel, we returned the fractional layers of grasslands (Grass_CoverFraction_layer), agricultural areas (Crops_CoverFraction_layer) and built-up areas (BuiltUp_CoverFraction_layer). We combined them into a single category by addition.
As we can see, the forest cover probability decreased in some areas, but also increased significantly in other areas since 2015. The increase is likely caused by the Chilean extensive reforestation project, launched following devastating forest losses due to wildfires. Such information can give us invaluable insight into what is happening with our forests around the globe. This is just one example — imagine all the potential use-cases with this amazing collection!
Water Bodies
The Water Bodies product contains 2 bands: the main Water Body Detection layer, which visualizes water bodies using a Modified Normalized Difference Index (MNDWI) from Sentinel-2 L1C, and the Quality layer, which provides information on the seasonal dynamics of detected water bodies. Quality layer is generated from the statistical computations from previous monthly products. The Quality layer’s water body occurrence ranks from very low occurrence to permanent. The collection is global and available since October 2020 with monthly updates. The product is used for monitoring changes in water bodies due to drought, floods, excess irrigation or climate change.
Check the values of each band here and the visualizations for the 2 layers in EO Browser here. The visualizations used in EO Browser can be found on our Custom Script Repository.
Using the Water Bodies layer in EO Browser, where detected water bodies are displayed in blue, we can observe density and distribution of water bodies on a global level. On the example below, we can observe water body distribution in Australia.
Using the Water Bodies Occurrence layer, we can get an idea of where the water is always present and where it only appears occasionally. On the example below, we can see the occurrence differences of the Irravaddy river in Myanmar. The meandering river is prone to flooding and seasonal variations. The dark blue color indicates permanent occurrence, while violet indicates rare occurrence.
It’s wonderful that so many remote sensing collections are free to use for everyone! And with Sentinel Hub, integrating EO data into applications is as easy as ever. We’re hoping many real-world solutions will arise from these collections, helping us mitigate real global issues.
We’re looking forward to seeing your awesome contributions!
