Periodic Reporting for period 1 - ExtremeEarth (From Copernicus Big Data to Extreme Earth Analytics)
Reporting period: 2019-01-01 to 2020-06-30
The technologies to be developed will extend the European Hops data platform of partners Logical Clocks and KTH to offer unprecedented scalability to extreme data volumes and scale-out distributed deep learning for Copernicus data. The extended Hops data platform will run on a DIAS selected after the project starts and will be available as open source to enable its adoption by the strong European Earth Observation downstream services industry. The technologies to be developed will also extend the linked geospatial data systems GeoTriples, JedAI, Strabon and SemaGrow pioneered by project partners UoA and NCSR in the past so that they scale to the extreme volumes of Copernicus data.
The work carried out in ExtremeEarth is of great societal value given the importance of Copernicus data and their applications to Food Security and the Polar Regions.
LogicalClocks and KTH developed important extensions of the Hopsworks data and AI platform to make it the platform of choice for EO applications. The platform currently runs inside the Food Security and Polar Thematic Exploitation Platforms (TEPs) of the European Space Agency and CREODIAS.
LogicalClocks and KTH developed the ExtremeEarth platform infrastructure.
UNITN developed a deep learning architecture for crop type and crop boundary mapping using Long Short Term Memory neural networks. The network will be used in the context of the Food Security use case.
UNITN also developed a large training dataset that have been used for training this deep neural network.
UiT, DLR and KTH developed deep learning architectures based on convolutional neural networks for sea ice classification. These architectures will be used to to provide the accurate sea ice mapping needed in the Polar use case. UiT and DLR also developed large training datasets that have been used for training these neural networks.
Two large-scale training datasets for sea ice classification have also been developed.
The above deep learning architectures for both use cases are currently running on the Hopsworks platform of partner Logical Clocks.
UoA developed two semantic catalogues, one for each of the two use cases of the project using the system Strabo2.
UoA developed the systems GeoTriples-Spark, JedAI-Spatial and Strabo2.
NCSR developed extensions of the system SemaGrow.
NCSR developed a new version of the KOBE open benchmarking environment.
All of the above big linked geospatial data systems run in the Hopsworks platform.
VISTA has collaborated with UNITN in the Food Security use case. The expected result of the use case is irrigation recommendations for farmers.
METNO, PolarView and UKRI-BAS collaborated with UiT, DLR and KTH in the Polar use case. The expected result of the use case is high resolution ice maps for informing maritime users.
Under the lead of coordinating partner UoA, the partners of the project carried out a series of dissemination activities.
The first versions of the innovation management plan and the exploitation plan of the project have been developed under the lead of partner VISTA.
1. The Hopsworks data and AI platform of LogicalClocks has been extended with new functionality that make it the platform of choice for developing big data and deep learning algorithms for Earth Observation.
2. The deep learning algorithms for the Food Security use case and the Polar use case offer precise solutions to the problems of crop type mapping and ice classification.
3. The large-scale training data developed for the deep neural networks for the Food Security use case and Polar use case are the first such publicly available datasets, and it is expected that they will be used by other Remote Sensing researchers.
4. The big linked geospatial data systems GeoTriples-Spark, JedAI-Spatial, Strabo2 and SemaGrow are the most scalable and effective systems currently available with the respective functionalities.
5. The results of the Food Security use case will provide precise irrigation recommendation information for farmers.
6. The results of the Polar use case will allow the semi-automatic creation of ice maps in a fraction of the time that it takes experts to produce them manually today.
The expected results of ExtremeEarth by the end of the project will be the completion of the tasks discussed above under the title ""Worked performed so far"".
The expected impacts of the ExtremeEarth project are the following (listed in the text of the call ICT-12-2018-2020):
1. Increased productivity and quality of system design and software development thanks to better architectures and tools for complex federated/distributed systems handling extremely large volumes and streams of data.
2. Demonstrated, significant increase of speed of data throughput and access, as measured against relevant, industry-validated benchmarks.
3. Demonstrated adoption of results of the extreme-scale analysis and prediction in decision-making (in industry and/or society.
ExtremeEarth has the following additional impacts:
1. Competitive advantage for European industry.
2. Shaping the Integrated Ground Segment of Copernicus and the Sentinel Collaborative Ground Segment.
3. Enable the development of EO services using Copernicus data by European companies that are not consortium members.
4. Bridging the gap between Remote Sensing and Informatics in the academic sector, and the Earth Observation and ICT industry sectors.
5. Enhancing innovation capacity and creating new market opportunities and new jobs in the European EO and ICT sectors.
6. Significant financial impact to farmers due to precise irrigation recommendations given that water savings and optimization of farming measures (e.g. fertilization) are the keys to sustainable practices.
7. Positive impact on maritime navigation and safety due to the provision of accurate and near-real time automated sea ice mapping.
8. Societal and environmental impacts due to the importance of food security and the Polar regions globally.
9. Impact on GEO, GEOSS and EuroGEOSS.
10. Impact on the Big Data Value Public-Private Partnership.
11. Impact on the following research and innovation areas: Remote Sensing and Earth Observation, Big Data and Extreme Earth Analytics, Deep Learning techniques for Remote Sensing, Semantic Web and Linked Data.
12. Impact on OGC and W3C geospatial standards such as GeoSPARQL.
13. Impact on ICT projects INFORE, SmartDataLake and AI4EU and ERC project BigEarth due to our collaboration with them."