The demand and consequential growth for information and services derived from remote sensing of the Earth by satellite is increasing exponentially. In Europe, this is being fuelled, in part, by the long-term commitment from the Copernicus programme to deliver a comprehensive range of operationally delivered data over at least the next few decades. This is further complimented by the ‘new space’ commercial providers using constellations of micro/nano- satellites. Customers/users in general are interested in services able to deliver ‘on-demand’ in a ‘ready to use’ manner. The DIAS platforms, Analysis Ready Data (ARD) and data cube concepts are vehicles to help provide access to the data and resultant information. However, implicit in this process is the assumption that the input data is ‘fit for purpose’ and can be readily combined from different sources in a seamless and timely manner. To achieve this requires rigorous and consistent quality assurance QA and in particular, a robust understanding/correction of any biases and assessment of uncertainty. This is at present rarely the case! or at least not without significant caveats and/or for a very limited range of applications. That is not to say that individual missions are not providing valuable and useful information but more that the combination of their products to generate temporally seamless (short and long term) data streams is often challenging.

The Quality Assurance Framework for Earth Observation (QA4EO) was developed by CEOS to help address this issue. QA4EO consists of a core principle and a set of guidance documents. The aim being that all data and derived products should have associated with them a fully traceable (to community references ideally SI) quantitative quality indicator i.e. an uncertainty underpinned by documented evidence. Much progress is being made towards this goal, at least for new sensors and ESA is very much seen as an international leader and pioneer, strongly supported by NPL as a representative of the international metrology community.

The harshness of the launch and operational environment in-space can cause significant and unpredictable change in the performance of the instrumentation used for remote-sensing, particularly in the optical domain. Re-assessment of the accuracy of satellite instrumentation post-launch as well as regular validation of its delivered products is considered an obligatory element of most space missions, although not always fully budgeted for. In some cases, some of this can be achieved on-board the satellite but, in many cases, it requires observations of well-calibrated large area natural targets, or at least an independent (in-situ) measure of the same variable/parameter to serve as ‘test-sites’ to calibrate/validate (Cal/Val) against. Of course, the in-situ measurement itself may also be the result of a remote sensing technique. For these to be, reliable, invariant and coherent across the globe requires them also to be traceable to SI units, (instruments and derived measurements). Such Cal/Val sites and networks have and are being established/coordinated by international organisations such as WMO and CEOS but developed and operated with funding for their operation from space agencies such as ESA and similar bodies, such as the European Union and its Copernicus program. However, until recently much of the focus has been on consistency and long-term stability with underpinning metrology and SI traceability a ‘nice to have’ and thus highly limited resource available to achieve it. The exacting needs of climate research have changed that philosophy and the desire is now to strive for robust SI traceability. This has in part led to the creation of the term Fiducial Reference Measurements (FRM), a sub-set of in-situ measurements that are demonstrably documented as traceable to community standards with an evidenced uncertainty budget.

Achieving this overall improvement in traceability and where necessary, reduction in uncertainty on the data products and in the supproting FRMs needed to underpin, places an urgency to address the following challenges:

• Ensure pre-flight calibration (laboratory-based) methods are traceable and flexible to ensure interoperability & harmonisation between different sensors.
• Improve Cal/Val of sensors in the post-launch phase enabling: international interoperabiily, the removal of biases and assessment of uncertainty in long multi-decadal time-series.of observations, and global representativeness of measurements.
• Establish techniques to assess, improve and report on the degree of traceability and associated uncertainties in biophysical parameters based on end-to-end metrological analysis.
• Where necessary develop new instrumentaton/standards to support the above.

Improvement necessitates evolution of laboratory-based metrology transferred to field (and space) situations. Some of the key metrology challenges related to the measurement sensor that have to be addressed are: large scale, operating conditions (e.g. vacuum, external environment), wide dynamic range and inaccessability once deployed. In remotely viewing the Earth such sensors have a wide range of spatial resolutions (1000's to 10s of metres), requiring reference data to validate the measurements on the same scale for any Earth scenes: land, ocean and atmosphere and the means to propagate to the sensor through the atmosphere.

This paper will describe a range of activities and efforts, many carried out as part of international projects, to address some of the metrological challenges. These will span from measurements of vegetation, temperature and colour of the oceans through to the establishment of a ‘metrology laboratory in space’. It will summarise some of the efforts carried out under various ESA funded FRM projects, the IDEAS+ and the EU European Metrology Research Program (EMRP) Metrology for Earth Observation and Climate (MetEOC) series of projects most of which relating to broader international efforts through CEOS WGCV. In conclusion a look to the future and how Cal/Val data from an integrated network of sources might lead to a more comprehensive mechanism to enable interoperability.

The Global Climate Observing System (GCOS) has specified the need to systematically produce and validate Essential Climate Variables (ECVs). The Committee on Earth Observation Satellites (CEOS) Working Group on Calibration and Validation (WGCV) and in particular its subgroup on Land Product Validation (LPV) is playing a key coordination role leveraging the international expertise required to address actions related to the validation of global terrestrial ECVs. The primary objective of LPV is to set best practices for validation methods and reporting in order to provide traceability and reliable uncertainty estimates to science and stakeholders. LPV covers 11 focus areas of land surface variables, each coordinated by two international co-leads including leaf area index (LAI), fraction of absorbed photosynthetically active radiation (FAPAR), vegetation phenology and vegetation indices, surface albedo, fire disturbance, snow cover, land surface type, soil moisture, land surface temperature (LST) and above ground biomass. LPV has identified four hierarchical validation levels corresponding to increasing spatial and temporal representativeness of reference samples used to perform validation. Best practice validation protocols (1) provide the definition of variables, ancillary information and uncertainty metrics, (2) describe available data sources and methods to establish reference validation datasets with SI traceability, (3) describe evaluation methods and reporting that account for product representativeness. This talk will present a global land product validation framework, and summarizes best practices for those variables with established protocols, whereas for other variables the current status and prospects will be described. The CEOS LPV validation level for each variable will be presented, identifying bottlenecks and opportunities. Finally, CEOS LPV validation supersites will be presented, identifying gaps and opportunities to expand and consolidate this network over Europe.

The Fiducial Reference Measurements (FRM; from the Latin verb ‘fidere’ – to trust) programme implemented through ESA is designed to develop reliable and transparent measurement and validation standards and processes for satellite data products. This includes tracing information back to its source and accounting for uncertainties in the data as well as conducting independent validation assessments of land surface products following best practice guidelines. The FRM4Veg project seeks to apply the FRM concept to the validation of surface reflectance and other key vegetation parameters such as fAPAR (fraction of absorbed photosynthetically active radiation) and canopy chlorophyll content (CCC) produced by ESA satellites over vegetated field sites. The aim is to create and test a validation protocol which facilitates the production of FRM ready validation data that is suitable for direct comparison with the satellite products of interest - namely those produced by PROBA-V, Sentinel-2 and Sentinel-3 OLCI. This includes the collection and processing of in-situ data as well as the treatment of those satellite products.

Surface reflectance is the base product from which several other Level 2 products such as albedo, LAI, and fAPAR are derived. This means it is vital that surface reflectance quality is demonstrated over a range of land covers. Validation over vegetated areas is particularly challenging because of its unique spectral and structural characteristics. However, confidence in the surface reflectance information for vegetated land covers is crucial due to the requirement for analysis ready data in forestry and agricultural applications.

The FRM4VEG project focused on two field sites: an agricultural field site (Barrax, Spain) and deciduous forest (Wytham Woods, UK). The measurement campaign and ancillary data collection considers subtle differences between the measurement configuration employed by the satellite and that attainable in the field. For example, we correct for differences between the stated product definitions (often the Bidirectional Reflectance Factor) and field conditions as well as for the differences in viewing geometry. We also discuss the sampling trade-offs that were made in order to conduct the measurements within an acceptable time period around the satellite overpass. This trade-off is mitigated by potential solutions using unmanned aerial vehicle (UAV) technology, which we evaluate at a separate field site.

The validation procedure considers all the main uncertainty sources such as sampling design, instrument calibration, product definitions, instrument configuration and scaling. The uncertainty derived during the instrument and panel calibrations are propagated through the processing chain and combined with ancillary data streams along the way. This leads to measured values which correspond directly with the satellite product, and are SI traceable.

The Fraction of Absorbed Photosynthetically Active Radiation (fAPAR) quantifies the fraction of the solar radiation absorbed by live leaves or any photosynthetic organism, for the photosynthesis activity. This biophysical land parameter is directly related to the primary productivity of photosynthesis, and it is used to estimate the carbon uptake by vegetation. Therefore, fAPAR is recognized as a terrestrial Essential Climate Variable by the Global Climate Observing System.
fAPAR is operationally provided by several Earth Observation (EO) satellites as a global product at different spatial and temporal resolutions. In particular, Sentinel-3/OLCI satellite constellation and the contributing Proba-V mission are providing frequent global coverage of fAPAR at a spatial resolution of 300 m. In this context, it has become crucial to assess the quality of these fAPAR satellite products through validation activities which are mandatory to satisfy potential user demands. Fiducial Reference Measurements (FRM) are defined as a suite of independent ground measurements that delivers to users the required confidence in data products, in the form of independent validation and satellite measurement uncertainty estimation.
This paper presents the results of the validation of Sentinel-3/ OGVI (fAPAR) and Copernicus Global Land Service PROBA-V fAPAR products performed in the context of the ESA Fiducial Reference Measuremenst for Vegetation (FRM4VEG) project. In 2018, two sites were selected to collect FRM of fAPAR: one cropland site in Barrax (Spain) sampled in June and one deciduous forest in Wytham (UK) sampled in July.
A protocol was defined to sample the spatial heterogeneity of the site, and to sample the Elementary Sampling Unit (ESUs), using a calibrated Digital Hemispherical Camera. The Wytham Woods PAR network was also used to provide additional fAPAR data. The raw DHP data was processed using CAN-EYE to retrieve green fIPAR, a very good approximation to green fAPAR, while the PAR network provides four-flux, total fAPAR. The raw data was processed using CAN-EYE to retrieve green fIPAR, a very good approximation to green fAPAR. The quality of field data was checked, certifying that it is suitable for satellite validation. The second step was to upscale the FRM collected over ESUs to the site extent using Sentinel-2 imagery and robust regression and machine learning techniques. The up-scaled map is derived with a quality flag and uncertainty estimates. Finally, the validation is performed considering the equivalent point spread function of the satellite product to match-up the high-spatial resolution FRM with the satellite product. Validation statistics (RMSD, Bias, r2, slope, offset) are provided and discussed.

Over the last decades, operational ground-based monitoring sites have been established as part of wider thematic observation networks, providing a wealth of highly relevant information for validation and quality assurance purposes. However, the uptake of these valuable in situ observations is hampered by a number of issues such as access restrictions, insufficient standardization, inconsistent quality, or lack of spatial or temporal coverage.
GBOV (Ground-Based Observations for Validation, http://gbov.copernicus.acri.fr), as part of the Copernicus Global Land Service (CGLS), aims at facilitating the use of observations from operational ground-based monitoring networks and their comparison to Earth Observation products through the following objectives:
• Collection of multi-year ground-based observations of high relevance for the understanding of land surface processes from existing global networks. These ground-based observations will be collected over a series of selected sites operated by international research networks (such as AERONET, ARM, BSRN, FluxNet, NEON, OZFlux, SurfRad, TERN, USRCN, …).
• Upgrade of existing sites with new instrumentation or establishing entirely new monitoring sites to close thematic or geographic gaps.
• Implementation and maintenance of a database for the distribution of reference measurements (RMs) and the corresponding Land Products (LPs). The focus of GBOV is on providing the information required for in depth quality assessment of the products offered through the global component of the Copernicus Land Monitoring Service (CLMS), i.e. the Copernicus Global Land Service (CGLS).
The GBOV service is primarily directed at the quality control of selected CGLS products (top-of-canopy reflectance, surface albedo, leaf areas index, fraction of absorbed PAR, fraction of vegetation cover, land surface temperature and soil moisture), providing collections of multiple-year ground-based RMs as well as derived LPs. Obviously, the information offered through GBOV may also be used for a wide range of applications beyond the CGLS. Sentinel-2 and Sentinel-3 land products validation is indeed within the scope of GBOV. Such external use will be supported by the GBOV data policy granting free and unrestricted access after online registration.
This talk will present the GBOV infrastructure and provide an overview of the services and products offered.

In support of a range of operational applications, the Copernicus Global Land Service (CGLS) systematically produces global bio-geophysical products from moderate to low spatial resolution earth observation (EO) data. Recognising the need for quality assurance of these products, the European Commission’s Joint Research Centre (JRC) recently initiated the ‘Ground Based Observations for Validation (GBOV)’ service, which aims to develop and distribute robust in-situ datasets for the purposes of direct validation. In the first component of the project, raw observations from a range of existing networks have been collected and processed to provide datasets suitable for validating global albedo, land surface temperature, soil moisture, surface reflectance, and vegetation products. In this contribution, we focus on the considered vegetation variables, namely leaf area index (LAI), the fraction of absorbed photosynthetically active radiation (FAPAR), and the fraction of vegetation cover (FCOVER).

In the case of all three vegetation variables, in-situ reference measurements (RMs) are derived from raw digital hemispherical photography (DHP), and are provided with associated quality indicators and estimates of uncertainty. Using high spatial resolution imagery from both the Landsat 8 Operational Land Imager (OLI) and Sentinel-2 Multispectral Instrument (MSI), quality controlled RMs are then upscaled to provide land products (LPs) suitable for validating moderate spatial resolution EO data. Upscaling is acheived following the guidelines of the Land Product Validation (LPV) sub-group of the Committee on Earth Observation Satellites (CEOS) Working Group on Calibration and Validation (WGCV). LPs take the form of 3 km x 3 km reference maps and include quality indicators and estimates of uncertainty, both of which are provided on a per-pixel basis.

The 744 LPs produced for each variable during the first year of the project cover 20 sites between 2013 and 2017, representing a wide range of vegetation types (including cropland, deciduous forest, evergreen forest, grassland, and shrubland). Using these LPs, we evaluate the performance of several operational vegetation products, including those produced by the CGLS from PROBA-V data, in addition to the Moderate Resolution Imaging Spectroradiometer (MODIS) and Visible Infrared Imaging Radiometer Suite (VIIRS) LAI and FAPAR products. Whilst similar performance is achieved by the MODIS and VIIRS products, which share the same retrieval scheme, the CGLS products demonstrate the best agreement with the GBOV dataset, exhibiting reduced bias, increased precision, and increased overall accuracy. Finally, potential improvements to the GBOV dataset are discussed.