Fast Earth Observation Data Exploration Platform to Optimize Value-adding and Climate-positive Applications

Paper number

GLOC-2023,T,IP,x75375

Author

Mr. Hannes Baeuerle, Fraunhofer - Institut für Kurzzeitdynamik, Ernst-Mach-Institut (EMI), Germany

Year

2023

Abstract

Today's active Earth observation (EO) satellites collect an increasing number of terabytes every day. However, the potential of this data to support measurable, reportable, and verifiable climate-positive actions has hardly been exhausted on a global scale. The growing amount of non-intercompatible EO data and a rising number of data providers, leaves possible users to an elaborate data procurement and processing. Furthermore, even the identification of interesting EO data is still non-transparent. Consequently, most value-adding products use only a fraction of the EO data available and rely on few well-known, public data sources (e.g., Sentinel or Landsat missions). However, especially more data from various sensors, reduces effective revisiting times and translates into actionable insights of increased value.
Therefore, the Fast and INtuitive Data Retrieval (FINDR) platform is developed, to minimize the complexity of identifying suitable EO data of interest, close gaps in previously missing data, increase market transparency and thus, considerably reduce entry barriers for new applications or help to create improved research results. In General, FINDR includes a comprehensive and transparent overview of available EO data from major free and commercial providers or sub-providers. On top, an implemented fast sensor coverage analysis methodology, developed at the Ernst-Mach-Institute (EMI), enables the user to quickly assess the data availability for a given set of coordinates and requirements in the past as well as in the future. Additionally, FINDR offers a homogenization approach of all input imagery based on research of the German Research Centre for Geosciences (GFZ), to allow for obstacle-free integration of EO data from different providers. During the homogenization, the data is radiometrically, spatially, and spectrally adapted to a common specification depending on the requirements of subsequent analysis steps. Radiometric homogenization compensates for varying image acquisition and illumination geometries, as well as for atmospheric effects with bottom-of-atmosphere-reflectance as the output radiometric unit. For spatial homogenization, the input data is georeferenced and spatially re-sampled to a common coordinate grid and geographic projection. In the spectral domain, machine learning techniques allow for adjusting the spectral band positions and widths as well as predicting unilaterally missing spectral information.
Hence, FINDR offers to rapidly identify suitable EO data in addition to generate output images with a unified tiling scheme and data format, independent of the source.

Abstract document

GLOC-2023,T,IP,x75375.brief.pdf

Manuscript document

GLOC-2023,T,IP,x75375.pdf (🔒 authorized access only).

To get the manuscript, please contact IAF Secretariat.