Sentinel-2 Semantic Data Cube Austria

Short Description

Starting point / motivation

Based upon the results of two feasibility studies (AutoSentinel2/3 [FFG ASAP] and SemEO [FFG ICT of the Future]), will evaluate and scale automated semantic enrichment of the Copernicus programme’s free and open Sentinel-2 data up to a big image database covering Austria.

Contents and goals

Currently more than 3000 Sentinel-2 scenes cover sections of Austria and the goal is to build an Austrian data & information cube. This project follows a novel and entirely different approach to accessing big Earth observation (EO) image databases, allowing semantic content-based image and information retrieval (SCBIR) through time.

These spatial-temporal query capabilities facilitate searches directly related to scene content or content dynamics, such as changes to any primary land cover category of interest (e.g. water bodies) in a user-specified area-of-interest through time.

Until now, all public EO image databases (e.g. Copernicus Sentinel Data Hub, USGS Landsat) or national initiatives (e.g. Code:DE, PEPS, national ground segments) have only allowed searches based on metadata (e.g. time and place of acquisition), not on image content or data directly, which are stored as flat files requiring download for further analysis.

There are some existing data cube initiatives (e.g. Swiss data cube), which store image data for analysis through time, but without any added information.


In contrast, will enable queries through space and time based on automatically-generated semantic content to demonstrate the first SCBIR system in an operational mode within an Austrian data & information cube. The semantic enrichment used in is based on a physical-model-based, spectral categorisation (i.e. pre-classification) and additionally derived information.

These processes will be fully automated and free of any user parametrisation. The derived information layers will be stored together with the image data in data cubes, implemented in array databases, where the third dimension is acquisition time.

Data cube storage allows indexing through time for fast access and information queries through time, upon which specific services can be developed and applied given semantic enrichment of image data.

Expected results

Demo services, drawn from generic use cases and built upon the data & information cube, will demonstrate the benefits of’s innovation, showing the approach’s broad application range, including:

  • semantically querying images, e.g. identifying cloud-free image data or images with specific events/changes for user defined AOIs instead of simple metadata queries;
  • generating cloud-free mosaics and composites for user defined timeframes;
  • establishing location-based access to data and information through time;
  • enabling object-based investigation, e.g. per-parcel statistics of defined agricultural fields for spectral and semantic profiles through time.

An inference engine for enhanced querying will be programmed as a Web interface in a client-server solution. Collected user requirements will inform this process and allow the implementation of different user interfaces addressing user needs and/or different access levels (e.g. public, user specific views). offers a unique, generic combination of innovations that can serve many different users and services.

Project Partners


University of Salzburg Department of Geoinformatics - Z_GIS

Project partner

  • Agrarmarkt Austria
  • Spatial Services GmbH
  • ZAMG - Zentralanstalt für Meteorologie und Geodynamik

Contact Address

University of Salzburg
Department of Geoinformatics - Z_GIS
Assoc.-Prof. Dirk Tiede PhD
Schillerstr. 30
A-5020 Salzburg