This paper discusses the accuracy of the operational Copernicus Sentinel-3 Ocean Land Colour Imager (OLCI) Level 2 land products corresponding to the Fraction of Absorbed Photosynthetically Active Radiation (FAPAR) and rectified channels. These products are estimated using OLCI spectral measurements acquired at the top-of-atmosphere, using a physically based approach. The daily products are operationally available at the full spatial resolution, i.e. at 300 m, and at the reduce resolution, i.e. 1.2 km. The quality assessment of the OLCI FAPAR products capitalises on the availability of two years of data acquired by Sentinel-3A and 6months of Sentinel-3B.
The actual evaluation exercise is performed in two steps: first, a comparison using MERIS climatology provides information on expected seasonality over various land cover types, second, with a direct comparison of the FAPAR time series against ground-based estimations as well as similar FAPAR products derived from other optical sensor data, such as MODIS and Sentinel-2A and Sentinel-2B MSI. An analysis of the accuracy of the FAPAR algorithm itself with respect to 3D-RT modelling data is also performed.

Albedo, defined as the ratio of the hemispheric reflected solar radiation flux to the incident flux upon the surface, is a key climate forcing variable that governs the absorption of incoming solar radiation. Albedo contributes significant uncertainties in the simulation of climate changes; and as such, it is defined by the Global Climate Observing System (GCOS) as a terrestrial essential climate variable (ECV) required by global and regional climate and biogeochemical models.

Satellite-derived albedo products from the Moderate Resolution Imaging Spectroradiometer (MODIS), the Visible Infrared Imaging Radiometer Suite (VIIRS), MEdium Resolution Imaging Spectrometer (MERIS), Spinning Enhanced Visible and Infrared Imager (SEVIRI), Landsat 8 Operational Land Imager (OLI), and Sentinel-2A/B MultiSpectral Instrument (MSI) have been routinely evaluated using in-situ and tower measurements. The satellite products agree well with the field data acquired from spatially representative field stations. The daily retrievals from MODIS and VIIRS capture the seasonal variation, ephemeral snow, and snow melt effects on surface albedo, as measured by tower-based albedometers.

The Land Product Validation (LPV) sub-group of the Committee on Earth Observation Satellites (CEOS) Working Group on Calibration and Validation (WGCV) aims to address the challenges associated with the validation of global land products. The validation of global surface radiation/albedo products is one of the LPV subgroup activities. The albedo validation protocol aims to develop the validation strategy using a reference dataset which covers various land types to assess the accuracy of satellite albedo products. The reference dataset includes in situ data (Baseline Surface Radiation Network (BSRN), FLUXNET, and National Ecological Observatory Network (NEON) etc.) and airborne measurements.

To develop enhanced multi-angular remote sensing techniques for the retrieval of accurate surface albedo estimates, the CEOS-LPV is conducting a series of pathfinder missions known as the Multi AngLe Imaging Bidirectional Reflectance Distribution Function small-UAS (MALIBU). The MALIBU instrument package includes two multispectral imagers oriented at two different viewing geometries (i.e., port and starboard sides) to capture the intrinsic surface reflectance anisotropy and albedo. MALIBU provides a new resource of reference data, in particular over heterogeneous surfaces, to improve global LPV efforts.

PICSCAR: The CEOS Initiative to better characterize the PICS

The PICSCAR initiative was born in November 2015 at CEOS/IVOS Meeting 27 with the objectives to coordinate the communities work on Pseudo Invariant Sites (PICS) and improve the characterisation of these PICS. Following this meeting, a roadmap has been defined addressing the following priorities for the PICS characterization: BRDF of the surface, spectral characterization of surface reflectance, the atmosphere properties, the temporal stability, methods to combine multiple site calibration results…. First, efforts have been made to fully understand the PICS usage through a questionnaire sent to spatial agencies and private operators. Request for information deals first for the PICS names and exact locations used by the different teams for the post-launch calibration of space-based optical imaging sensors, and second for the identification of the PICS properties on which focus needs to be put on to improve the PICS knowledge. As a result, among the 30 sites that are used by the different teams, the top six rely on the 6 PICS endorsed by CEOS at IVOS 19 in 2008. The results will be shown.
Together with the questionnaires, in the first phase of this initiative, we have collected more than 50 datasets of TOA reflectances acquired by different sensors over the Libya 4 PICS. Based on these datasets, the Libya4 site stability has been analysed thanks to the CNES PARASOL BRDF model that has been used to normalize the data to a common geometry. Temporal stability of the site will be demonstrated.
In order to compare the methodologies used to monitor the intercalibration of sensors, the PICSCAR team has set up three exercises, aiming at comparing the results of L8/OLI and S2A/MSI, intercalibration, MODIS and MERIS intercalibration, and last the assessment of the BRDF correction on MERIS times series. The results of these exercises will be shown, knowing that they have been discussed within the working group which met regularly every six weeks.
In order to capitalize the results obtained through this initiative, a PICSCAR internet site has been developed and provides the information collected on PICS, the results of the PICSCAR exercises, and tools that could be used for the data normalization.

The optical measurements of the Copernicus Sentinel-2 and Sentinel-3 missions over ocean and inland waters constitute a crucial source of Earth-Observation (EO) data for a wide range of marine applications, ranging from very local and short-term aquatic monitoring to the production of long-term and global data records for climate studies. In order to be fit for purpose in different usages, these measurements shall be delivered by space agencies with associated quality levels. System Vicarious Calibration (SVC) based on high-quality and stable Fiducial Reference Measurements (FRM) is a first mandatory mean to ensure SI-traceability of the measurement and correct for systematic error at sea level, at least in ideal conditions (open waters under clear atmosphere). Validation against other independent FRMs then allows assessing the overall mission performance on broader conditions and check whether the mission requirements are met. Quality assessment at pixel level is a further step for an optimal use of the EO data, e.g. for optimal spatial and temporal averaging at Level-3, multi-mission data merging, proper detection of anomalies in time-series, effective data assimilation in numerical models. The per-pixel Quality and Science Flags (QSF) are generally useful for a conservative data screening, but their binary definition does not provide a quantitative level of confidence on the measurement itself (e.g. what is the quality of the marine reflectance whether the pixel is flagged or not for sun glint or atmospheric correction failure).

From a metrological point of view, any data at pixel level shall be associated with an uncertainty, which depicts quantitively the confidence we can have in the data in its specific measurement conditions (solar illumination, observation geometry and camera’s detector, weather conditions, atmospheric content, surface state, etc.). The present work, conducted in the frame of ESA Sensor Performance, Products and Algorithms (SPPA, IDEAS+ consortium), revisits the complete Ocean Colour processing in view of:
• Identifying all sources of uncertainties: radiometric (Level-1b), model, ancillary data, sub-pixel
• Providing output per-pixel uncertainties to each Level-2 product
• Designing new inverse methods more robust to the input uncertainties
We apply the rigorous uncertainty methodology of the Quality Assurance Framework for Earth Observation (QA4EO) and the Guide to the expression of Uncertainty in Measurement (GUM). The first virtue of this approach is to clarify the metrics sometimes misunderstood in EO data quality assessment (uncertainty, error, random and systematic errors, etc.). The traceability chain and measurement equations of the Level-2 processor are then described. For optical sensor, the structure of the uncertainty is crucial: spectral correlation exists at various stages and must be handled in the uncertainty propagation (variance-covariance matrix). Examples on Sentinel-3/OLCI data are presented for three main steps: SVC, atmospheric correction (AC) and chlorophyll-a estimate. We show how this uncertainty framework is effective not only to derive output quality levels, but to design algorithms making the best use of the optical information. For instance, we show that the standard OLCI AC relying on only two bands in the near-infrared (NIR), historically developed on theoretical simulations, cannot achieve in practice its expected performance due to the various sources of EO uncertainties; more robust inverse algorithms need a larger spectral baseline, implying a trade-off between the number of spectral information and uncertainty of the model. For chlorophyll estimate, we show similarly how spectral matching algorithm can outperform the classical blue/green band ratio algorithm, as well as presenting a sound mathematical formalism for uncertainty estimates. Importantly, uncertainties of the optical measurements must be validated in their turn. This is achieved in a statistical way, by corroborating the uncertainty estimates with the distribution of the observed errors against FRMs.

We finally propose a roadmap and discuss open issues for operationalising such metrological method in Sentinels optical data processors and mission performance tools.

Fiducial Reference Measurements (FRM) are a suite of independent, fully characterized, and traceable ground measurements that follow the guidelines outlined by the GEO/CEOS Quality Assurance framework for Earth Observation (QA4EO). These FRM provide the maximum Return On Investment (ROI) for a satellite mission by delivering, to users, the required confidence in data products, in the form of independent validation results and satellite measurement uncertainty estimation, over the entire end-to-end duration of a satellite mission (https://earth.esa.int/web/sppa/activities).
The FRM measurements must a) have documented SI traceability (e.g. via round-robin intercalibration of instruments) using metrology standards; b) be independent from the satellite geophysical retrieval process (noting the exception of L2 product vicarious adjustment that fundamentally depends on FRM ground based measurements); c) have an uncertainty budget for all FRM instruments and derived measurements available and maintained; d) have defined and openly published measurement protocols and community-wide management practices (measurement, processing, archive, documents etc.); d) be openly and freely available for independent scrutiny.
Following these principles, the FRM4SOC project, funded by the European Space Agency (ESA), has been supporting of the evaluation and improvement of the state of the art in ocean colour validation through a series of comparisons under the auspices of the Committee on Earth Observation Satellites (CEOS) Working Group on Calibration & Validation (WGCV) and in support of the CEOS ocean colour virtual constellation. FRM4SOC activities have also contributed to fulfilment of the International Ocean Colour Coordinating Group (IOCCG) in situ ocean colour radiometry white paper objectives and the relevant IOCCG working groups and task forces (e.g. the working group on uncertainties in ocean colour remote sensing and the ocean colour satellite sensor calibration task force). The project achievements have made a fundamental contribution to the European system for monitoring the Earth (Copernicus) through its core role of working to ensure that ground-based measurements of ocean colour parameters are traceable to SI standards. This helps to ensure high-quality and accurate Copernicus satellite mission data, in particular Sentinel-2 MSI and Sentinel-3 OLCI ocean colour products. The FRM4SOC project has also contributed to the work of ESA and EUMETSAT to ensure that these instruments are validated in orbit.
The structure, objectives, principles of methodology, major events, comparison exercises, and results from the FRM4SOC project are introduced.

Networking of automated instruments on unmanned platforms has proved to be the most effective way to provide validation data for earth observation missions. The re-use of data from each site for many optical missions (S2, S3, PROBA-V, MODIS, VIIRS, L8, Pléiades, ENMAP, PRISMA, SABIAMAR,etc.) gives a huge economy of scale. The existing water and land networks (e.g. AERONET-OC, RADCALNET, ...) are based on multispectral instruments requiring modelling associated uncertainties to cover all spectral bands of all sensors. To fill this gap in hyperspectral validation data, the HYPERNETS project is working on the development of a new lower cost hyperspectral radiometer with an associated pointing system and embedded calibration device. The final objective of the project is to provide high quality in situ measurements for the validation of the surface reflectance data issued from all earth observation missions. Meanwhile the development of a state-of-the-art hyperspectral instrument, the HYPERNETS project also aims to improve the measurement protocols by taking advantage of the automated pointing system.

The present study focuses on providing guidelines for above-water measurements from fixed platforms for the validation of ocean color data. When deriving the water-leaving reflectance with above-water sensors, the contribution of sun glint and sky light that is reflected at the sea-surface into the direction of the sensor needs to be estimated and removed from the total upwelling surface radiance measured by the sensor. When the sea-surface is relatively flat, the sky-light may be accurately estimated by simultaneously measuring the total upwelling radiance and the sky radiance in the direction of the region of the sky that reflects into the sea-viewing sensor. However, even small surface waves will render the estimation of the fraction of the sky-light that is actually reflected by the water surface to the sensor very challenging. This is the most critical aspect when deriving the water leaving reflectance from above-water measurement. In addition to sky and sun glint, nearby structures (e.g., masts and other colored poles) or the platform itself may also affect the measured total upwelling radiance by shading the water surface or by inducing adjacency effect. Using simulated data, guidelines are suggested in terms of viewing geometry to improve the estimation of the sky and sun glint. Errors related to the platforms and nearby structures are also investigated and illustrated with data from two validation platforms in the Belgian waters.