CEOS Open Data Cube Lessons Learned and Way

CEOS Open Data Cube Lessons Learned and Way Forward GEO Data Technology Workshop Brian Killough / 23 Apr 2019 / Vienna, Austria @Open. Data. Cube / Brian. D. Killough@nasa. gov

What is the Open Data Cube? Satellite Analysis-Ready Data Open Data Cube Infrastructure Data Cube Core Code, API and Database Applications and Tools Application Library User Interface

What are the benefits of the Open Data Cube? • Focused on Analysis Ready Data (ARD) to reduce data preparation • • • time and complexity. . . easy for users Based on free and open source software Enables data interoperability and efficient time series analyses Easily deployed in a cloud or a local computer Existing customizable web-based user interface and a library of application algorithms Proven concept. . . Working in Australia, Colombia, Switzerland, Vietnam and 5 countries in Africa! 14 more countries under development and 33 more countries expressing interest.

How do we feed the Open Data Cube with satellite data? • We promote the use of pre-processed Analysis Ready Data (ARD) from Landsat, • • Sentinel-1 (S 1), Sentinel-2 (S 2) and ALOS. Landsat data must be ordered and downloaded from USGS (e. g. surface reflectance) and then uploaded and indexed in the cloud. By the end of 2019, USGS will have their global ARD on AWS. Sentinel-1 radar. . . there is no global cloud source of ARD (backscatter intensity) but we are working on solutions. CREODIAS, Alaska Science Facility, and AWS hold partial or global level-1 archives. Sentinel-2 optical. . . global cloud-based ARD (surface reflectance) exists on CREODIAS and AWS since Dec 2018. Archive data back to 2016 is needed, but no easy solution exists. We are working on a number of options. ALOS PALSAR radar. . . annual mosaics (ARD format) are available globally but they require some effort and knowledge to move into cloud formats for data cubes.

How can we use the Open Data Cube? • Cloud-free Mosaics: Recent Pixel, Median, • • Geomedian, Max-NDVI Spectral Indices: NDVI, NDBI, NDSI, NDWI, SAVI, EVI, Fractional Cover Land Classification: K-Means, Random Forest, FAO 8 -class decision tree Water: Landsat WOFS (Australia), Sentinel-1 WASARD (NASA), Landsat Water Quality- TSM (Australia) Land Change: Spectral Threshold Anomaly, Coastal Change, Py. CCD (USGS)

Lessons Learned • Analysis Ready Data (ARD) is no longer a desire of global users, but • • is now becoming an expectation. Production and access to global ARD is not easy. Avoiding egress (download and uploads) is critical as file sizes are becoming increasingly larger for satellite data. The internet is not always ”good” so developing countries still want local solutions. Clouds in optical data (e. g Landsat or Sentinel-2) are an issue for many global users, so there is a growing interest in radar data. Few users understand its benefits or know how to use this data

Way Forward • Grow the Open Data Cube (ODC) community. . . more developers and users • Enhance capacity development to bring along more users and show impact • Communicate what makes the ODC unique, as there are many other options • • for using satellite data Be agile and aware of new technologies and move toward better solutions Grow and test our library of application algorithms. . . focus on UN-SDGs and use ground and in-situ data for validation Refine the CEOS Analysis Ready Data for Land (CARD 4 L) definitions and develop efficient data flows to support data cubes Demonstrate interoperability between multiple datasets. . . optical and radar

Contact Brian Killough NASA, CEOS Systems Engineering Office Brian. D. Killough@nasa. gov www. opendatacube. org @Open. Data. Cube