AI Strategy for Earth System Data

The KI:STE developments will facilitate the use of ML and AI methods for spatial data analysis applications.

Artificial intelligence (AI) methods currently experience rapid development and increasing use in the context of environmental data. However, this often happens in isolated solutions. Environmental and earth system sciences have yet to establish the systemic use of modern AI methods. In particular, discrepancy exists between the requirements of a solid and technically sound environmental data analysis and the applicability of modern AI methods such as Deep Learning for researchers.

The KI:STE project strives to facilitate and evaluate the use of AI for remote sensing Earth Observation data for a range of applications. The fields studied in the project range from air quality to clouds and radiation, to snow and ice propagating, as well as water that drives vegetation, then closing the loop with air quality again. A core focus is not only to adopt and apply AI concepts to these areas, but also to train several PhD students and build an e-learning platform. This will ease and facilitate access to the algorithms and tools developed for a wider audience – from scientists to practitioners.

52°North develops the Spatial Research Data Infrastructure (SRDI) that will supply the AI processing platform with data. A requirement analysis provides the basis for defining and developing interfaces for data acquisition and provision. The platform must react flexibly to the requirements of the AI algorithm requesting data in order to be able to provide them in a format optimized for the required processing. We work on the SRDI in close collaboration with the Ambrosys GmbH.

During 2022, our team drafted the SRDI architecture based on the results of the requirement analysis, which we carried out previously. We have also set up the first components stemming from the architecture. Our solution builds on the open source project GeoNode. GeoNode provides tools to collect and harmonize different data sets from different services. A searchable catalog presents all integrated data that can be assessed through machine readable APIs. Different data adapters and visualizations still need to be developed and tailored to the KI:STE project’s data needs. In close collaboration with our KI:STE-partner University Bonn, we re-built their wilderness use-case as an example for a reproducible workflow. The Open Geospatial Consortium (OGC) Testbed initiative has also co-funded this effort in 2022. We have documented our findings of the workflow automation and how to make it easily reproducible. These findings will feed into further activities at OGC. Furthermore, 52°North largely contributed to the joint OpenGeoHub and KI:STE summer school held from August 29th to September 2nd, 2022. Our team contributed lectures, a reprocucibility hackathon and joined the discussions throughout the week.

Forschungszentrum Jülich GmbH, Germany

Jülich Supercomputing Centre (JSC) und Institut für Bio- und Geowissenschaften – Agrosphäre (IBG-3), Germany

Universität zu Köln, Institut für Geophysik und Meteorologie, Germany

Universität Bonn, Institut für Geodäsie und Geoinformatik, Germany

RWTH Aachen, Aachen Institute for Advanced Study in Computational Engineering Science, Germany

Ambrosys GmbH Gesellschaft für Management komplexer Systeme, Potsdam, Germany