The first PFS (Surface Reflectance, Surface Temperature and Normalised Radar Backscatter) were authored by experts from different agencies.
LSI-VC is actively trying to involve the commercial sector in the development of new CEOS ARD PFS.
If you would like to contribute either to the annual review cycle of the endorsed PFS or to the development of future PFS, please contact the LSI-VC Secretariat.
The CARD4L PFS do not prescribe how a CARD4L product should be generated – that is for the data providers to decide, given their existing processing workflows and preferred processing software to use.
The PFS documents are nonetheless useful as guidance when defining the CARD4L processing workflows because they describe in detail what metadata information, ancillary per-pixel metadata products, accuracy requirements, etc. need to be taken into account.
If the data cannot be processed it is not clear how it can become CARD4L compliant. However for key datasets, in particular historical datasets, work may continue to develop the necessary methods. For example USGS are continuing work to process Landsat MSS data with an ultimate goal that it can be CARD4L compliant.
The CEOS Interoperability Terminology is a CEOS list of definitions seeking to achieve a level of clarity across CEOS initiatives. It is being developed as an internal document, although once completed the Terminology will be made publicly available via the CEOS website. If you are interested in the Terminology please contact the Working Group on Information Systems and Services (WGISS): http://ceos.org/ourwork/workinggroups/wgiss/contacts/
Self-assessment summary tables will be available on the CEOS ARD website.
Data providers can download a copy of the relevant PFS on the CEOS ARD website here. Each row of the PFS table should be reviewed and responses added to the appropriate columns. Once complete, the document should be submitted to the LSI-VC Secretariat to initiate the peer review process.
We are agnostic to the processing method per se., and therefore would not peer-review software. The CARD4L framework is based on assessing the products, knowing that the algorithms will continuously evolve and improve. This allows the accuracy of the products to continuously improve, rather than be merely consistent.
The latest versions of the PFS are those linked on the CEOS ARD website here. These are considered ‘living documents’ and will be updated as needed, but no more than is necessary and at most annually. LSI-VC has put in place an “annual review cycle” to take into consideration feedback from the broader community as well as feedback generated through the CARD4L product assessments.
The CEOS ARD Product Alignment Assessment has two key steps:
1. Self-assessment: data providers self-assess their products using the CARD4L Product Family Specifications (PFS*) to understand how well their products meet the CEOS Analysis Ready Data specifications.
2. Peer Review: an independent assessment to confirm the CEOS ARD product level (Threshold or Target).
* The PFS are technical documents and provide details on what data providers need to do to deliver CEOS Analysis Ready Data.
A CARD4L PFS for interferometric radar is currently under development. We expect it to be ready in Q1 2021.
Peer review is itself the quality assurance step for compliance with the specifications. User feedback will play a key role in driving continuous improvement in the quality of ARD datasets.
The CEOS ARD approach is intended to support direct analysis using multi-sensor data. E.g., for surface reflectance products, sensor calibration is not required at the Threshold level, but it is required at the Target level. For this specific example, both Sentinel-2 and Landsat 8 apply calibration.
The driver is to provide non-radar-specialist users with simple radar data products that do not require expert knowledge or dedicated SAR processing software to ingest and analyse.
Please see the results of a 2018 survey here. Additional feedback is gathered through our regular participation in international meetings and fora related to CEOS, GEO, Open Data Cube, Digital Earth Africa, etc.
CEOS ARD will simplify the process. If data providers produce and distribute CEOS ARD it will be directly ingestible into a Data Cube. Data Cubes are a key driver for ARD due to the need to have data that is comparable through both space and time.
We have several “pilot” type activities in work that are aimed at doing this. For example we expect that CARD4L will facilitate the future harmonisation of Landsat and Sentinel-2.
As this is a fairly new process, at this time we only have two datasets that conform to the Threshold level of CARD4L compliance (USGS Landsat Collection 2 Surface Reflectance and Surface Temperature products). ESA’s Sentinel-2 Level-2A Surface Reflectance product is also currently in the peer review process for Threshold level CARD4L compliance and several other agencies are currently preparing their self-assessment packages. Both the USGS Landsat Collection 2 Surface Reflectance and Surface Temperature products have also been assessed at being partially compliant at the Target level, and additional information will soon be available on which requirements still need to be met by future USGS Landsat product releases.
USGS Landsat Collection 2 sample products are available here
The CARD4L PFSs are a result of community consultation including experts and data providers. There are a range of views on what corrections are necessary, and practical, to produce a useful measure of land surface reflectance.
The Target level of CARD4L indicates corrections that are considered to be desirable but difficult to achieve. Demand from users for higher levels of correction will help to drive improvements. If you are aware of requirements that are not met by current specifications, please let us know.
CARD4L requirements look into radiometry and geometry in an attempt to provide analysis ready data, and asks for pixel flags regarding data quality, such as cloud-affected pixels. Some of the [to be] providers might use Scene Classification Maps or have them as an additional input/output – as can be seen in the Sentinel-2 production chain and products.
After extensive consultation with a broad user community, it has been widely adopted and accepted that Level-2 processing (biophysical measurement) is the minimum that is considered CARD4L. While higher levels of processing may also be deemed “ARD”, at this point and with the available resources, Level-2 processing is the current focus.
Essential Variables (EVs) are further ‘down-stream’. The CARD4L Framework aims to produce the most basic physical measurements of the land surface, such as surface reflectance and temperature.
EVs are more derived and are targeted at particular needs. A ready supply of CARD4L data should allow EVs to be derived more quickly and reliably.
CEOS PFS are certainly an approach to standardise, and we think it is essential to have clear specifications around analysis-ready data or it will become ‘all things to all people’, and the value proposition will be lost.
However, we are approaching standardisation through building community understanding and consensus rather than by going directly to formal Standards.
The PFS were developed on the base of the supporting Agencies. Each CEOS Agency has a chance to delegate members who are familiar with their users’ base needs. As of now, CEOS LSI-VC is actively working on outreach and has adopted a “best practices” approach, including working with organisations dealing with geospatial data (e.g., OGC Testbed - Community Document).
Driven by the challenge of Big Data, the industry is converging toward interoperable and machine-readable formats, such as Cloud Optimized GeoTIFF (COG) and Spatio-Temporal Asset Catalogues (STAC). We are now looking at including advice on these formats in the PFS.
Sample CARD4L datasets will be available for the endorsed products. We aim to have these finalised by October. Once CARD4L data becomes widely available, for example Landsat Collection-2, sample data may no longer be required. Download links will be provided on ceos.org/ard
This is an interesting concept, and we would like to learn more to fully understand the definition and what is meant by “Interpretation Ready Data”. Please contact us through the LSI-VC Secretariat to discuss this further.
CEOS ARD for Land (CARD4L) is not a data format. CARD4L Specifications detail the corrections and metadata needed to produce geophysical measurements of the land surface. This has been intentionally kept separate from file formats.
We are currently exploring publishing a best practices (or similar) document through OGC.
The CEOS ARD Framework does not mandate particular formats or technical solutions like STAC. However, the industry is converging toward interoperable and machine-readable formats, such as Cloud Optimized GeoTIFF (COG) and Spatio-Temporal Asset Catalogues (STAC).
We are now looking at including advice on these formats in the PFS so that data providers are aware of these possibilities which may make their data more readily consumable by users.
The Product Family Specifications can be applied to mosaics including temporal composites. The provenance of each pixel in the end-mosaic would need to be retained.
We have connected with EARSC (European Association of Remote Sensing Companies) in the EU for example, which includes many of the EU cloud platform providers. Moreover, as we gear up our efforts on community outreach we expect more interest, involvement and connection from such providers.
To become a distributor: the simplest way is to harvest freely available CARD4L data from the various data producers (USGS, ESA, JAXA, etc.) and re-share it. Data distributors can also generate their own CARD4L from underlying data, after completing an accompanying CARD4L assessment. The major cloud providers are actively interested in housing analysis ready data and CARD4L.
We have had discussions with other (industry) data providers/distributors where they are considering implementing a CARD4L workflow to provide their own CARD4L data.
These developments will be driven by user demand, industry (data hosting and cloud providers), and global coordination initiatives.
As this is a fairly new process, at this time we only have two datasets that conform to the Threshold level of CARD4L compliance (USGS Landsat Collection 2 Surface Reflectance and Surface Temperature products). ESA’s Sentinel-2 Level-2A Surface Reflectance product is also currently in the peer review process for Threshold level CARD4L compliance and several other agencies are currently preparing their self-assessment packages. Both the USGS Landsat Collection 2 Surface Reflectance and Surface Temperature products have also been assessed at being partially compliant at the Target level, and additional information will soon be available on which requirements still need to be met by future USGS Landsat product releases. USGS Landsat Collection 2 sample products are available here
Information regarding which data providers or distributors are implementing the CARD4L Framework is planned for future enhancements to the CEOS ARD website; however, CEOS LSI-VC is also sensitive to data providers and distributors desires to be included on this website or not.
While it isn’t yet decided, the USGS is evaluating each specific requirement that is not meeting Target level of compliance as a candidate to be included with USGS Landsat Collection 3 for its Surface Reflectance and Surface Temperature products.
JAXA is planning free public release ALOS PALSAR (all observation modes) and ALOS-2 PALSAR-2 (ScanSAR mode) at a future date. Public release of PALSAR-2 Fine Beam data is still TBD.
These developments will be driven by user demand, industry (data hosting and cloud providers), and global coordination initiatives.
It is also an active line of discussion in GEO and CEOS.
Furthermore, programs like Digital Earth Africa are demonstrating operational CEOS ARD ‘pipelines’ to allow universal access to ARD for many countries.
The USGS is currently planning for a September/October 2020 release of its USGS Landsat Collection 2 Surface Reflectance and Surface Temperature products, which are 100% Threshold level CARD4L-compliant and partially Target level CARD4L-compliant.
Yes. In addition to the global USGS Landsat Collection 2 Surface Reflectance and Surface Temperature UTM scene-based products, the USGS will be generating Landsat Collection 2 Albers Equal Area projected, tile-based U.S. ARD products following generation of the scene-based products.
Yes. With Collection 2, Surface Temperature products will systematically and globally be generated for nighttime scenes as well, given the same constraints as with daytime scenes (i.e., sufficient emissivity auxiliary data and solar angle constraints).
The simplest and most sustainable model is where the data producer also takes on the role of CARD4L supplier. However, this is a decision for each data provider and is not ‘forced’ by CEOS. Various other models (combining data suppliers, industry, and data users) are also possible. For example, Australia produces its own surface reflectance data from Landsat and Sentinel-2.
Several Product Family Specifications are expected to be finalised in 2020/2021:
Q4 2020: Aquatic Reflectance
Q1/Q2 2021: Interferometric SAR
Q1/Q2 2021: Geocoded SLC SAR