Introduction to the openEO Platform
openEO Platform — a federated platform for unified access and analytics of Earth Observation (EO) data from different back ends. Sounds too good to be true? Learn more about it in this blog post.
Welcome to the first blog post in our blog post series about openEO Platform. In this series we share with you the latest development updates, interesting new approaches and experiences on the way from the openEO API specification to an operational and open EO platform.
In this first blog post, we introduce you to the predecessor project openEO — the foundation on which openEO Platform is being built on, why we need a federated platform, where we currently stand, and what our goals and visions are for the coming months.
How it all began: openEO Horizon 2020
As dozens of terabytes of new satellite data are acquired each day, it has become less and less feasible for users to download and process the data locally, especially over larger areas or longer time periods. To ease this burden on the user, companies have developed big EO cloud back ends — a back end in this case provides all (EO) data access and processing capabilities — that take care of all the data storage and provide processing capabilities so that the user only needs to pass their code/requests to the back ends and download intermediate or final results for viewing or post-processing. Shifting this work to the back end has the major advantage of requiring less data to be transferred, much less processing resources to be consumed on the user side, and avoiding idle resources, all resulting in lower costs for the user.
Since different companies develop and operate their own back ends, processes and application programming interfaces (APIs) for accessing data based on different business models, it has been very time-consuming to compare back ends in terms of functionality, cost and results. openEO set out to solve this problem.
With the successful development of a functional openEO API, the original Horizon 2020 project ended in 2020. However, openEO did not cease to exist with this and lives on in the form of openeo.org. Since openEO is open source, anyone can use it for free and contribute to its further development. It is used and further developed by an active community and a variety of projects.
Since openEO is free and open source, it is possible to develop services based on the API. The most advanced and ambitious development in this regard is certainly openEO Platform which is advancing the API into an operational, open and easy-to-use platform.
openEO Platform
Defining a common API within openEO was only the first part of the story. openEO Platform has now set out to add the next big chapter by developing a production-ready platform. openEO Platform is a joint effort of a consortium led by EODC in collaboration with Vito, Sinergise, the University of Münster (WWU), EURAC and EGI with in-kind support by GEO.
Where openEO offered the ability to connect to multiple EO cloud back ends (one at a time with separate accounts needed per back end), openEO Platform will transform this into an easy-to-use platform offering a unified access supported by the capabilities of all participating back ends (currently these are TerraScope (Vito), EODC, Sentinel Hub (Sinergise) and CreoDias — experimental). As such, the user will
- only need a single account to access all participating EO back ends,
- gain direct access to a growing number of EO data collections — currently already more than 50! (different back ends offer different processing levels and extents, as well as different set of collections) and
- get the union of processing capabilities and resources.
This gives the user a transparent (open source code, reproducible results that support scientists’ best practices, clear and prior cost estimation, confidentiality of intellectual property rights) and simple (dynamic resource allocation, federated cloud environment, pythonic programming syntax) solution that enables processing of data from pixel to continental scale.
The development of the platform started in late 2020 and since then we have already made great progress. In early 2021, we demonstrated on-demand processing and use of CARD4L-compliant SAR data and ARD optical data within the platform (use case 1). This was followed 6 months later by two additional use cases (use case 3 and use case 6). In the first case, a crop type classification was performed using features calculated within the platform based on Sentinel-1 and Sentinel-2 data. The classification was performed using two different approaches, a simple rule-based approach and a more sophisticated one based on a random forest model. The result was a large-scale (5 countries) crop type stratification map. The ability to create a crop type map over no less than 20 MGRS tiles is a significant advance in large-scale processing and an important step for many use cases focused on the continental scale. In use case 6, the platform was extended to include time- series fitting capabilities, which were demonstrated showing near real-time forest change detection. We are preparing a blog post specifically addressing these use cases, so stay tuned.
In November 2021, one month after completing use cases 3 and 6, we celebrated the first major public milestone by opening the platform to early adopters during ESA Φ-week 2021. Since then, the platform has been open for interested members of the EO community to explore, test, and provide feedback. It is still possible to sign up for an early adopter package (90 days) or get a free trial account (30 days) — more information here.
All information about the platform and its capabilities can be found on our homepage openeo.cloud and for those testing it, the full technical documentation at docs.openeo.cloud.
Outlook
We are currently preparing the platform to support two more use cases. One will provide the ability to create a (European) dynamic land cover classification (use case 9). The added workflow will support custom control of input parameters such as user-defined class legends, area of interest, and time interval, giving the user a wide range of choices for creating their final result. The other use case addresses the generation of fractional canopy cover maps based on a regression model trained on calibration data derived from commercial very high-resolution (VHR) data and open Sentinel-1 and Sentinel-2 data (use case 8).
In parallel, we are preparing for the Living Planet Symposium in Bonn (May 23–27, 2022), where we will be present with openEO Platform.
You are welcome to meet us there and learn more about openEO Platform
- in the demo area of the ESA exhibition booth,
- attend one of our presentations during the science sessions,
- participate in our user consultation event,
- or get some hands on experience in using the platform during one of our classroom trainings.
To keep up to date with the latest developments follow us on Twitter, LinkedIn and check out the news section on our homepage.
2500 € research grant opportunity!
Researchers who want to test openEO Platform for their studies can apply for a research grant of 2500 € until 15.03.2022 . All you need to do is submit an application outlining your motivation, goal and the scope of the research you would like to conduct using the platform. You can find more information here.
openEO Platform is funded by the European Space Agency (ESA).
