Exploiting Big Earth Data from Space – First Experiences with the TimeScan Processing Chain

The European Sentinel missions and the latest generation of the United States Landsat satellites provide new opportunities for global environmental monitoring. They acquire imagery at spatial resolutions between 10 and 60 m in a temporal and spatial coverage that could before only be realized on the basis of lower resolution Earth observation data (>250 m). However, images gathered by these modern missions rapidly add up to data volume that can no longer be handled with standard work stations and software solutions. Hence, this contribution introduces the TimeScan concept which combines pre-existing tools to an exemplary modular pipeline for the flexible and scalable processing of massive image data collections on a variety of (private or public) computing clusters. The TimeScan framework covers solutions for data access to arbitrary mission archives (with different data provisioning policies) and data ingestion into a processing environment EO2Data module), mission specific pre-processing of multi-temporal data collections (Data2TimeS module), and the generation of a final TimeScan baseline product (TimeS2Stats module) providing a spectrally and temporally harmonized representation of the observed surfaces. Technically, a TimeScan layer aggregates the information content of hundreds or thousands of single images available for the area and time period of interest (i.e. up to hundreds of TBs or even PBs of data) into a higher level product with significantly reduced volume. In first test, the TimeScan pipeline has been used to process a global coverage of 452,799 multispectral Landsat–8 scenes acquired from 2013 to 2015, a global data-set of 25,550 Envisat ASAR radar images collected 2010–2012, and regional Sentinel–1 and Sentinel–2 collections of ∼1500 images acquired from 2014 to 2016. The resulting TimeScan products have already been successfully used in various studies related to the large-scale monitoring of environmental processes and their temporal dynamics.