Kolloquium Geoinformatik Big Earth Observation Databases: infrastructure and spatiotemporal analysis

Current scientific methods for extracting information for Earth observation data lag far behind our capacity to build sophisticated satellites. These satellites produce massive amounts of data, but only a fraction of that data is effectively used for scientific research and operational applications. This presentation will address a key scientific problem: How can substantially improve the extraction of information from big Earth Observation data sets in an open and reproducible way? The presentation will focus on the innovative techniques of large-scale databases, that can hold thousands of images joined in space and time. Users are then able to develop algorithms that can seamlessly span partitions in space, time, and spectral dimensions. Array databases are specially well suited to be combined with spatiotemporal data analysis, which enable producing new types of land change information. In this work, we introduce the Fields data type as a possible way to represent large data sets. We propose a generic data type for fields that can represent different types of spatiotemporal data, such as trajectories, time series, remote sensing and, climate data. We show how to represent existing algebras for spatial data with the Fields data type. We also argue that array databases are the best support for processing big spatial data and show how to use the Fields data type with array databases.