On-demand, semantic EO data cubes – knowledge-based, semantic querying of multimodal data for mesoscale analyses anywhere on Earth

With the daily increasing amount of available Earth Observation (EO) data, the importance of processing frameworks that allow users to focus on the actual analysis of the data instead of the technical and conceptual complexity of data access and integration is growing. In this context, we present a Python-based implementation of ad-hoc data cubes to perform big EO data analysis in a few lines of code. In contrast to existing data cube frameworks, our semantic, knowledge-based approach enables data to be processed beyond its simple numerical representation, with structured integration and communication of expert knowledge from the relevant domains. The technical foundations for this are threefold: Firstly, on-demand fetching of data in cloud-optimized formats via SpatioTemporal Asset Catalog (STAC) standardized metadata to regularized three-dimensional data cubes. Secondly, provision of a semantic language along with an analysis structure that enables to address data and create knowledge-based models. And thirdly, chunking and parallelization mechanisms to execute the created models in a scalable and efficient manner. From the user’s point of view, big EO data archives can be analyzed both on local, commercially available devices and on cloud-based processing infrastructures without being tied to a specific platform. Visualization options for models enable effective exchange with end users and domain experts regarding the design of analyses. The concrete benefits of the presented framework are demonstrated using two application examples relevant for environmental monitoring: querying cloud-free data and analyzing the extent of forest disturbance areas.