A Data Science Pipeline for Big Linked Earth Observation Data

Abstract

AbstractThe science of Earth observation uses satellites and other sensors to monitor our planet, e.g., for mitigating the effects of climate change. Earth observation data collected by satellites is a paradigmatic case of big data. Due to programs such as Copernicus in Europe and Landsat in the United States, Earth observation data is open and free today. Users that want to develop an application using this data typically search within the relevant archives, discover the needed data, process it to extract information and knowledge and integrate this information and knowledge into their applications. In this chapter, we argue that if Earth observation data, information and knowledge are published on the Web using the linked data paradigm, then the data discovery, the information and knowledge discovery, the data integration and the development of applications become much easier. To demonstrate this, we present a data science pipeline that starts with data in a satellite archive and ends up with a complete application using this data. We show how to support the various stages of the data science pipeline using software that has been developed in various FP7 and Horizon 2020 projects. As a concrete example, our initial data comes from the Sentinel-2, Sentinel-3 and Sentinel-5P satellite archives, and they are used in developing the Green City use case.