In the Harmonization of Multi-Mission High-Resolution Time Series: Application to BELAIR (BELHARMONY) project (see the presentation of Sterckx et al. for the general project overview), the aim is to obtain harmonized time series from Level 1 throughout value-added Level 3 bio-geophysical products for Landsat 8 Operational Land Imager (OLI), Sentinel-2 A/B Multi-Spectral Instrument (MSI), Project for On-Board Autonomy – Vegetation (PROBA-V), and Deimos-1.

The assessment of possible biases in the Level 1 TOA signal provided by the different instruments will be performed over targets with low, medium, and high radiance levels. This distinction is required as, particularly at very low and also extremely high scene brightnesses, the optical systems’ radiometric response might not be linear. Furthermore, deviations from the pre-launch characterized non-linearity curve can happen when in-flight. Finally, it should be noted that the instruments under investigation have been designed for land applications merely or solely and therefore no requirements for the instrument team on the radiometry verification at low radiance levels might exist (which is e.g. the case for PROBA-V). Therefore it is expected that larger biases might occur between the sensors at low radiance levels and that different gains might be applicable throughout the radiance dynamic range.


The work presented focuses on the Level 1 bias assessment between the sensors’ observed TOA reflectances (TOAobs) with simulated reference TOA reflectances (TOAref) over Radiometric Calibration Network (RadCalNet) sites of La Crau (France), Gobabeb (Namibia), and Railroad Valley (USA), as well as over the Pseudo-Invariant Calibration Site (PICS) of Libya-4. At the RadCalNet sites surface and atmospheric data are automatically acquired. The RadCalNet data are distributed through the RadCalNet portal and include among others ‘simulated’ nadir TOA hyperspectral reflectance simulated using the MODTRAN-5 RTM (Berk et al., 2005), which serve as a direct reference to compare with the satellite measured TOA signal (after spectral resampling considering the sensor SRFs).
Over the Libya-4 site, the Optical Sensor Calibration with Simulated Radiances (OSCAR) desert approach (Govaerts et al., 2013) is used. In this approach, simulated TOA Bidirectional Reflectance Factors (BRFs) define an absolute reference against which optical sensors can be cross-calibrated. The simulated TOA BRFs are calculated with the 6SV RTM (Kotchenova et al., 2008), with as input surface BRF values derived from MODIS and MISR observations, meteorological data, and aerosol characterisation data.


For each site, clear-sky scenes were selected based on the respective satellite sensor’s cloud masking and by additional visual inspection of quick look images to avoid adjacency influences of nearby or undetected clouds. For the selected scenes, the ratio TOAref / TOAobs was calculated per spectral band. We will present the assessment results and pay particular focus on the relative differences between the sensors and attributions of these differences and variations to various uncertainty factors, both for the satellite observations and the reference simulations.

Area B - Board 199

Free and open data policy allows access to a big volume of remote sensing data, which together with the advanced cloud computing services, facilitate significantly the analysis of long time series. The consistent study of land surface though implies the correction of the atmospheric impacts when optical observations are involved. However the user communities request nowadays Analysis Ready Data (ARD), meaning already pre-processed and corrected satellite data to a set of requirements for immediate analyses with minimum user effort. Consequently, several entities have already started to generate, or they plan to generate in the short term, ARD at a global scale for Landsat-8 (L-8) and Sentinel-2 (S-2) missions using suitable atmospheric correction (AC) approaches. To that end, European Space Agency (ESA) and National Aeronautics and Space Administration (NASA) initiated the Atmospheric Correction Inter-comparison eXercise (ACIX) in the frame of CEOS WGCV (Committee on Earth Observation Satellites, Working Group on Calibration & Validation).
The first ACIX experiment started in June 2016 with the aim to bring together the developers of state-of-the-art AC processors and to study the variations amongst the different algorithms. The input data were L-8 and S-2A products over various sites of different land cover types around the world, i.e. agricultural, deserts, urban, snow and coastal areas. The description and the conclusions of this first experiment are summarised in Doxani et al. (2018). All the inter-comparison results can be found in the dedicated to ACIX I web page (http://calvalportal.ceos.org/projects/acix). ACIX I was completed in February 2018, but the improved versions of the participating processors and the increasing interest of AC developers to be part of the experiment stimulated the continuation of ACIX and its second implementation (ACIX II).
Following the recommendations of ACIX participants and other Earth Observation data users, an additional inter-comparison of cloud masking assessment was decided to be performed in parallel with ACIX. Cloud masking is a crucial step of the radiometric pre-processing of optical remotely sensed data and an important contributor to the retrieval of accurate SR within an AC process. Therefore, it was considered essential to analyse these two processing chains together. The test sites of the exercises are redefined and more representative cases concerning land surface and atmospheric conditions, e.g. land/water, land cover, aerosols. Particular attention will be given also to aquatic sites, i.e. coastal and inland waters, which will be analysed as a separate sub-category. All the exercises, i.e. CMIX, ACIX-Land, and ACIX-Aqua will run in parallel and follow the same timeline. In this presentation we will focus mainly on ACIX-Land.
In particular, 17 AC developer teams from eight different countries participate in ACIX II-Land, out of 35 teams in total for both ACIX II-CMIX exercises. All the participating processors should run operationally without requiring parametrization when applied on different areas. Therefore, the performance study will be carried out over all the AERONET sites with available measurements in the study time period. Around 150 sites have been identified to fulfil this requirement and the experiment is expected to involve around 7000 L-8 and S-2 (2A and 2B) image scenes. The detailed description of ACIX II-Land implementation will be given in this presentation including the protocol, i.e. test sites, input data, inter-comparison metrics, and some preliminary results. In continuation of the first ACIX experiment’s achievements, the outcomes of the second one are expected to provide an enhanced standardised approach to inter-compare AC processing products, i.e. Aerosol Optical Thickness, Water Vapour and Surface Reflectance, and quantitively assess their quality when in situ measurements are available.

References
Doxani, G.; Vermote, E.; Roger, J.-C.; Gascon, F.; Adriaensen, S.; Frantz, D.; Hagolle, O.; Hollstein, A.; Kirches, G.; Li, F.; Louis, J.; Mangin, A.; Pahlevan, N.; Pflug, B.; Vanhellemont, Q. Atmospheric Correction Inter-Comparison Exercise. Remote Sens. 2018, 10, 352.

Area B - Board 200

Vegetation indices are one of the most widely-used satellite products for studies involving vegetation dynamics and vegetation anomaly monitoring. The Vegetation Index (VI) focus area was established as a new focus area within the Committee on Earth Observation Satellites (CEOS) Working Group on Calibration/Validation (WGCV) Land Product Validation (LPVC) Subgroup in September 2016. The goal of this VI focus area is to establish a set of community-wide, peer-reviewed protocol recommendations for validating and inter-comparing satellite VI products.
The first workshop of the VI focus area was held in conjunction with the CEOS LPV Land Surface Phenology focus area’s workshop in Fort Collins, Colorado, United States on 9-10 November 2016. At the workshop, the following components were identified as those that define VI uncertainty and, thus, that satisfy the user needs: (1) uncertainty of VIs in their units; (2) characterization of VI value changes with respect to changes in actual vegetation condition (biophysical and/or physiological); (3) long-term stability of VI products. Currently, the focus area co-leads are developing a draft framework for product inter-comparison and in situ data-based validation based on past studies and latest developments in this field. To this end, the focus area has planned a second workshop in December 2018 for obtaining inputs and feedbacks on the draft framework from the wide scientific community working with VIs.
This presentation provides the current status of the draft framework concerning the inter-comparison and the evaluation of the long term stability of VI data sets. The scientific questions to be answered in this analysis are: (1) What is the overall similarity?; (2) What is the magnitude of the difference?; (3) What is the spatial pattern of the difference; (4) What is the temporal pattern of the difference; (5) What are the possible causes of the difference, and (6) What is the impact on VI anomalies? The approach will be demonstrated with examples of currently available VI data sets from different sources.
The latest information on the VI focus area is found at the CEOS LPV Vegetation Index Focus Area website (https://lpvs.gsfc.nasa.gov/NDVI/VI_home.html).

Area B - Board 201

In the framework of the Sentinel-2 Radiometric Validation (S2RadVal) ESA project, inter-comparison algorithms were developed to validate Sentinel-2 Level-1C products. One of the algorithms involved in the study used coastal water as a target to validate calibration at low radiance values. The AERONET-Ocean Color (OC) network supplies sea conditions and optical properties in several bands matching some of the Sentinel-2/MSI bands, and its stations’ locations were therefore used as targets. AERONET-OC stations are located in coastal areas, where water surface reflectance is higher to that of open ocean water due to higher suspended particle concentration. The shape, size and concentration of aerosols significantly influence the signal received by the satellite. The work performed in the framework of the S2RadVal project consisted in simulating Sentinel-2/MSI observations acquired over AERONET-OC stations using the 6SV radiative transfer model and comparing the resulting top-of-atmosphere BRF values with Sentinel-2/MSI observations, accounting for their respective uncertainties.

This work extends the results obtained at that time with additional data covering 12 AERONET-OC stations and the time range 2015-2018. A complete modular Python processing chain has been developed for that purpose. Independent databases for AERONET-OC and Sentinel-2 L1C products are linked by a matchup database, which consists in a single object able to fully coordinate processing.

Pre-processing starts with matching AERONET-OC observations and Sentinel-2 L1C products, resulting in the creation of matchups. Sentinel products are downloaded only for valid pairs, which minimises the requested bandwidth, and appropriate tiles are then repacked into netCDF files. Matchups can then be filtered based on criteria depending on the features of the radiative transfer model, e.g. to ignore observations where wind speed is too high to get an accurate estimate of water surface reflectance. For every matchup passing all the tests, the top-of-atmosphere BRF is simulated using a version of the 6SV radiative transfer model modified to leverage AERONET-OC data, in particular spectrally-resolved leaving radiance aerosol optical thickness. The computed TOA BRF is then used to compute an absolute bias estimating the discrepancy between the simulated value and the one obtained by Sentinel-2/MSI.

Bias values are then aggregated for different series, filtered by time period and locations. The uncertainty on simulation results is significantly contributed by the aerosol model selection and meteorological conditions. Aggregated values are compared to absolute radiometric requirements, better-suited for low signal intensity, estimated as 5% of the reference level of a typical TOA BRF acquired over vegetated surfaces. This reference level is simulated using 6SV.

In addition to the application of the calibration method, the developed toolchain can be used to perform in-depth analysis and explore potential correlations and dependencies of bias against the aerosol optical thickness or water-leaving reflectance.

Area B - Board 202

With its launch in May 2013 PROBA-V has exceeded without major issues the original operational mission lifetime of 5 years. Due to the lack of on-board calibration devices combined with the complexity of the instrument design (three separate cameras and 18 detector units including 3x3 mechanically butted SWIR detectors) the radiometric calibration of the PROBA-V which relies solely on vicarious methods has been a real challenge.

As vicarious methods have their intrinsic uncertainties, a set of vicarious methods are operationally used in the Image Quality Center (IQC) of the mission to reduce the uncertainty in the in-flight radiometric calibration. The IQC Cal/Val facility contains among others the OSCAR (Optical Sensor CAlibration with simulated Radiance) vicarious calibration methods which exploit the radiance over bright desert surfaces, Deep Convective Clouds (DCC), and Rayleigh scattering. Furthermore, lunar observations are scheduled every month to track the radiometric stability of the center camera. Through observation of oceans during the night the increase in the dark current values over time is monitored. Finally to characterize and to correct for non-uniformities within and between detectors 90° yaw maneuver are performed over desert sites. With this 90° yaw configuration the detector array runs parallel to the direction of the motion and therefore an area on the ground is subsequently viewed by the different pixels of the same spectral band creating a pseudo flat-field signal (Adriaensen et al., LPS2019).

In this presentation we focus on the temporal evolution of the instrument radiometric response over its 6 years in orbit. Results obtained on the basis of the various vicarious methods are inter-compared and discussed. Finally we discuss changes that will be made to the radiometric calibration parameters, such as the application of an improved non-linear degradation model and corrections for residual non-uniformities in the detectors. These changes will be implemented for the reprocessing of the PROBA-V archive at the end of the mission.

Area B - Board 203

Over the last decade there has been a significant increase in the number of satellites launched and countries operating them, including in more recent times commercial providers. This in part reflects the recognition that no-one country of agency has the resources to meet the demand for data. To ensure maximal global societal benefit, agencies need to coordinate activities ideally launching complimentary missions in an attempt to fill observational gaps (temporal, spatial and spectral etc). They also need to ensure that data from different missions and from different agencies are where possible interoperable.

Consistent Cal/Val and assessment/reporting of uncertainties is key to interoperability. To support this CEOS-Working Group on Calibration and Validation (WGCV) and its sub-groups provide a framework for international collaboration- sharing ideas, development of infrastructure and community agreed methods/standards. WGCV is the principle international coordinating body on satellite product Cal/Val (pre-& post-launch) and one of the founding working groups of its mother body, CEOS, the space arm of GEO. It is not only a forum for discussion but is active on establishing international collaborative projects on infrastructure, best practices etc e.g. current initiatives on ARD/data-cubes.

The Infrared Visible and Optical Sensors (IVOS) sub-group for example has organised comparisons of instruments and methods used for post-launch validation of ocean and land surface reflectance, brightness temperature and MTF. It has also established a new open access service, RadCalNet, to provide radiometric calibration coefficients for medium to high resolution sensors typical of those being developed by commercial operators together with a set of ‘reference deserts’ for coarser resolution sensors. The IVOS mandate spans pre-flight Cal to post-launch Cal/Val and although primarily concerned with Level 0 to 1 it also considers various level 2 products such as Sea Surface Temperature, Ocean Colour and Surface Reflectance as well as observational geometric aspects of sensors.

The IVOS team has and encourages active participation from space agencies, academia and industry from five continents with formal plenary meetings taking place around the globe on a rotational and annual basis.

The paper will describe some of the activities and future plans of CEOS-WGCV-IVOS to solicit new members and/or ideas on challenges and activities that would benefit from an internationally coordinated approach. In addition to sharing of ideas and joint research/infrastructure projects the group is able to make consensus recommendations for consideration by senior agency officials.

Area B - Board 204

Agricultural landscapes form the basis for the supply of food, renewable raw materials, and increasingly energy. Against the background of the high demand of a growing world population, it is necessary to make land use decisions to protect the landscape or soil functionality, to optimize the agricultural production and, at the same time, to ensure the preservation or qualitative improvement of ecosystems and their services. Remote sensing data as well as field and laboratory measurements can be used for biophysical monitoring of agricultural landscapes. Essentially therefore is the use of a validation strategy for the quantification of uncertainties.
The Cal/Val DEMMIN (Durable Environmental Multidisciplinary Monitoring Information Network) facility, operated by German Remote Sensing Data Center of the German Aerospace Center (DLR) and the German Research Center for Geosciences (GFZ), collects agrometeorological and soil parameters using an environmental monitoring network of over 40 environmental and 63 soil moisture stations, a crane for forest measurements, a lysimeter hexagon for hydrological observations (SoilCan), and different Eddy-covariance towers for gas measurements. These data are combined with in-situ campaign based field measurements to provide a substantial data base for data and product validation. DEMMIN is also part of the observation network TERENO (TERrestrial ENvironmental Observatories) (www.tereno.net) and since 2017 assigned to the international validation network Joint Experiment of Crop Assessment and Monitoring (JECAM, www.jecam.org).
In 2019 this infrastructure will be complemented by a HYPERNETS instrument tower. The H2020/HYPERNETS project, started in February 2018, is to develop a new lower cost hyperspectral radiometer and an associated pointing system with embedded calibration device for automated measurements of water and land bidirectional reflectance. The instrument will be tested in a prototype network covering a wide range of water and land types under different operating conditions. One of the test sites is the TERENO NE area is DEMMIN, where the installation of a 10m height measuring mast for the HYPERCAM instrument is planned on an agricultural field. In addition, the measuring station will be equipped with sensors for climate and soil variables measurements and an Eddy Flux station.
This work will present the status and future perspectives for the German JECAM site DEMMIN/TERENO-NE with respect to new innovative validation instruments like HYPERNETS for agricultural research.

Area B - Board 205

Remote sensing estimation of biophysical parameters such as FAPAR, Fraction of Absorbed Photosynthetically Active Radiation, and terrestrial chlorophyll content, both related to the state and evolution of vegetation, is important, not only for production estimation or for land use changes assessment, but also and fundamentally, to feed carbon exchange models. These vegetation related indices need to be validated to ensure the correct functioning of the models and improve future predictions related to the carbon cycle.

Direct validation requires the production of a set of accurate independent measurements of the significant quantities at plant level. The Spanish National Space Research Programme Project VARIETAL (Validation of Earth Observation Data and Products at the Valencia Anchor Station) and the Joint ESA-EUMETSAT Sentinel-3 Validation Team (S3VT) project both contribute to the validation of the terrestrial products OGVI (OLCI Global Vegetation Index) and OTCI (OLCI Terrestrial Chlorophyll Index) retrieved from the OLCI (Ocean and Land Colour Instrument) sensor onboard the Copernicus Sentinel-3 mission.

The Valencia Anchor Station has developed a robust infrastructure of FAPAR stations directly measuring the four significant PAR fluxes which, together with an intense transect-based monitoring of chlorophyll content over a 900 m x 900 m area, have allowed to gather a detailed database of these biophysical parameters during the phenological cycle of vines for the last three years. This paper gives an account of the methodology used and the results obtained.

Area B - Board 206

In order to ensure high quality data is provided by the Copernicus Space Component, the Mission Performance Centres are continually making improvements to the algorithms. Generating data products over land in particular, requires special considerations due to terrain and interest in sub-pixel scale features such as fires. Two major improvements in the products from Sentinel-3’s Sea and Land Surface Temperature Radiometer (SLSTR) have been implemented and are discussed.
SLSTR has two dedicated fire channels, in addition to the 3 infrared and 6 visible/short-wave infrared. We present how improvements have been made to the geolocation of the F1 fire channel of SLSTR. Band F1, like band S7, measures radiation at 3.7um but saturates at the higher temperatures required for detecting and making measurement over fire. Comparisons between F1 and S7 resulted in poor colocation of the channels, with offsets of up to 3km, making it difficult to use the channels for the detection of fires, which can often only cover a few pixels in area. By using the true line of sight of the F1 channel, and creating a new specific F1 grid, improvements have been observed.
When geo-locating measurements from a satellite to the ground, the elevation of the surface must be taken into consideration, even as part of the gridding method. For SLSTR, we show how by considering the ortho-geolocation information when re-gridding satellite data onto a uniform grid, the ortho-rectified information must be used. A true-nearest neighbour re-gridding process that uses the ortho-geolocated pixel positions to locate the most appropriate pixel to fill an image grid pixel has been applied. Using this new re-gridding algorithm in SLSTR has shown improvements in positional offsets reduced from up to 10km in high altitude regions, to less than 1km.
These improvements have had a significant impact in the development of the Level 2 Fire Radiative Power (FRP) product and other land users.

Area B - Board 207

Satellite images are often provided on a regular two dimensional grid which implies the underlying assumption that the original data are sampled at the image grid resolution. In reality, the actual sampling and spatial resolution of the satellite data is dependent on an number of factors including the sensor’s view geometry and the detector response as projected on the ground. This can lead to a mis-interpretation of the image data. The effects become more significant for instruments with large swath widths (several hundred km) where the footprint of the pixel increases with distance from the sub-satellite track. Understanding the extent of how the projected footprint affects the image quality is of great importance to land and coastal users, as there can be significant variability of land features on a sub-pixel scale.
Using field-of-view measurements from the Sentinel-3 SLSTR (Sea and Land Surface Temperature Radiometer) pre-launch calibration and the geometric calibration model, we have derived the instrument footprints and projected these onto the earth’s surface. At the swath centre, the pixel footprint is close to the ideal image grid resolution, while at the edges of the swath the pixel size can be greater than six times the ideal. We then combine these with basic high resolution images to simulate the effects of the variation in the spatial response with view angle on the output level-1 image.
We have analysed the impacts on various features which include fires, clouds and natural land boundaries like coastlines, lakes and rivers. In all cases features appears more distorted further from the sub satellite position, leading to mis-interpretation of the collected data. For example, fires will appear to be less intense leading to an underestimate of fire radiative power if a 1km pixel is assumed. Sharp features such as coastlines will be degraded in quality.
This information is useful for users to make full use of the available information and develop improved data products. Also, the process can be used for the design and development of instruments and their data processors to determine the sensitivity of key instrument parameters on the output data products.

Area B - Board 208

The Ocean and Land Colour Instrument (OLCI) on-board Sentinel-3 routinely generates two L2 land products: the OLCI Global Vegetation Index (OGVI) and the OLCI Terrestrial Chlorophyll Index (OTCI). The OGVI and OTCI are surrogates of the fraction of absorbed photosynthetically active radiation (FAPAR) and canopy chlorophyll content (CCC) respectively. These L2 products are intended to continue the legacy of the Medium Resolution Imaging Spectrometer (MERIS). Validation of land products and the verification of their consistency is essential for end users and scientists to understand the product’s accuracy and uncertainties. This contribution presents results of the direct and indirect validation efforts of OLCI L2 land products.

With respect to direct validation, results are presented from recent field campaigns at the Las Tiesas-Barrax (Spain) and Wytham Woods (UK) sites, which took place within the Fiducial Reference Measurements for Vegetation (FRM4VEG) project. In-situ measurements of FAPAR and CCC were upscaled using Sentinel-2 Multispectral Instrument (MSI) data to produce high spatial resolution (20 m) reference maps, which were then aggregated to the spatial resolution of OLCI (300 m) for comparison. A summary of all direct validation exercises over the last two years and an indication of accuracy of the products is provided.

In addition, indirect validation efforts, including an assessment of the temporal consistency of the OGVI and OTCI with respect to a 10 year MERIS climatology, are presented. To facilitate visualisation, a graphical user interface (GUI) is under development. The aim is to make the web application tool available to the earth observation (EO) community. The main features include near real time interactive verification of consistency across more than 60 sites including the S3MPC core validation sites and CEOS LPV supersites, with accuracy statistics and automatic report generation.

Area B - Board 209

Fiducial Reference Measurements (FRM) for Vegetation (FRM4VEG) is a European Space Agency (ESA) managed project funded under the Copernicus programme. It aims to establish reliable and transparent in-situ measurement and validation standards for Copernicus products derived from Sentinel-2, -3, and PROBA-V. Although the concept of FRM has been evaluated for other satellite-derived variables such as ocean colour and surface temperature, FRM4VEG is for the first time exploring the application of metrological principles to vegetation and land surface reflectance products. The goal is to provide confidence to users in their application of satellite-derived vegetation products by demonstrating and documenting traceable validation methodologies. In particular three bio-geophysical variables are considered: surface reflectance, the fraction of absorbed photosynthetically active radiation (FAPAR) and canopy chlorophyll content (CCC).
The project combined data from laboratory calibration of a variety of commonly used field instruments with two dedicated field campaigns to define uncertainty, error propagation and traceability of in-situ measurements. In-situ measurements were collected from two field sites in the first phase of this project, one of which is a Land Product Validation (LPV) supersite (Wytham Woods, UK, mixed deciduous forest), and another with a legacy of previous ESA calibration/validation activities (Las Tiesas-Barrax, Spain, mixed agricultural). These measurements, and their associated uncertainties, were aggregated at the elementary sampling unit level and upscaled to the Sentinel-3 OLCI and PROBA-V spatial resolutions using contemporaneous Sentinel-2 MSI data. Overall, it is expected that the project will build a consensus for good practices in the measurement of vegetation bio-geophysical variables, resulting in a legacy for the community, and clear CEOS endorsed measurement and validation protocols for future practitioners.

Area B - Board 210

Copernicus/EUMETSAT funded activities aimed to produce a preliminary design of the Copernicus Infrastructure dedicated to Ocean Colour Satellite Vicarious Calibration (OC-SVC) are here considered. A comprehensive vision is firstly presented to establish a reference for the Design phase of the OC-SVC infrastructure. OC-SVC scientific and operational requirements are then evaluated accounting for a wide range of components related to Ocean Colour science and Vicarious Calibration. These include: in situ infrastructures, measurements, and protocols; metrology applied to field radiometry; marine field operations; Ocean Colour operational processing chain; Ocean Colour modelling error sources; radiative transfer modelling and in situ data processing.
The design and the project planning components are undertaken accounting for requirements for the OC-SVC Infrastructure defined in a previous study. A specific emphasis is given to solutions addressed to attain short- and long-term goals of Copernicus OC-SVC infrastructure.
The design proposed for the Copernicus OC-SVC Infrastructure is based on the recently developed MOBY-NET system based on MOBY, this being a successful technology build upon more than 30-year experience capable of offering a technologically proven system within a timeframe in line with Copernicus SVC requirement. The devised Copernicus OC-SVC Infrastructure, called E-MOBY (European Marine Optical BuoY), is conceived to support vicarious calibration for Copernicus current missions (Sentinel-3/-2 A, B,C,D), Third Party ocean colour missions, and Copernicus Ocean Colour data services. The System specification is designed to ensure the value of European contribution to OC-SVC global assets.
Starting from the MOBY-NET concept, the study investigates E-MOBY adaptions to best fit the specificities and the marine optical properties of the European site. To this aim, E-MOBY will be designed to ensure operational and sustainable measurements of water-leaving radiance, the key quantity in the vicarious process, which are SI traceable and meet application requirements. The design approach is committed to the highest radiometry standards and operational readiness.
The Ground Segment will be scoped to acquire and archive E-MOBY measurements, as well as to derive and distribute data products for computing Vicarious Calibration Gains and related uncertainties. A modular structure of the overall system is adopted to be able to update independently individual subsystems (field and ground) without the need to reconsider the whole system architecture.
To meet the objectives and OC-SVC infrastructure requirements, the study includes the following tasks:
1. investigate the optimal location for the E-MOBY deployment by considering a detailed characterization of different candidates.;
2. design the field segment accounting for the surface buoy and optical system, deployment structure as well as associated activities including radiometric characterizations and calibrations;
3. design the ground segment by ensuring reliable solutions for data acquisition, data processing, data products and data services;
4. devise the operations for the field segment (operation procedure for servicing, rotation in the field, refurbishment and replacements) and ground segment (data acquisition scheduling, operational data transmission, data processing, archive and dissemination, maintenance);
5. deliver data products accounting for all the components that contributes to the total uncertainty budget;
6. prepare an Infrastructure Project Plan which will encompass the OC-SVC design with technical specification, development phase, the infrastructure set up and the early operations as well as the cross-institutional cooperation framework;
7. detail the estimated cost of the infrastructure.
The prosed baseline is to deploy the E-MOBY near Lampedusa Island, a small and flat island in the central Mediterranean Sea surrounded by blue waters with very low chlorophyll concertation and no source of sediments. Since 90’s, Lampedusa hosts the Climate Observatory collecting in situ measurements for the detailed characterization of the atmospheric structure and composition, and more recently oceanographic properties. Historical measurements at the site are used, together with ocean color field measurements acquired during oceanographic cruises cruise and satellite data, to perform a detailed site characterization. The study is furthermore aimed to evaluate additional deployment sites, namely: Madeira Island in the Atlantic and Antikithera Island, north of Crete. Past and recent sites characterizations presented in the literature shall be included in the analysis.

Area B - Board 211

Copernicus ocean colour program has now reached its full operational status with redundancy in its Earth Observation capacity on-board Sentinel3 A and B. OLCI (Ocean and Land Colour Instrument) A and B are now providing data with an unprecedented frequency. While redundancy has already been secured in the past for thermal observation with the satellite series like AVHRR, it is the first time the Ocean Colour community can rely on long term data provision from consistent sensors with OLCI A and B. This stream of data should path the way for the development of new and high quality downstream services.
Ocean Colour reaching operational status has brought new challenges. Data quality has to be assessed on a regular basis and most importantly System Vicarious Calibration (SVC) has to be implemented in a very short period of time. Routine product validation can be ensured with Fiducial Reference Measurements infrastructures like BOUSSOLE, MOBY and AERONET-OC. AERONET-OC provides a large amount of data in the coastal region while BOUSSOLE and MOBY provide data in the open ocean. New generation systems like ProVal or HyperNav and networks like HYPERNET-OC should secure both coastal and open ocean routine validation in the future. SVC is a fundamental element of ocean colour data processing and must be implemented prior public product release in order to reach the required level of uncertainty on marine products. Experience from past missions have demonstrated that having two SVC stations like BOUSSOLE and MOBY, several years of data could be needed to derive stable (SVC) gains. This has been challenging in the case of OLCI A and B for which SVC has been implemented with less than two years of data.
This presentation will summarise S3-MPC strategy for routine level-2 product validation as well as the strategy for fast SVC implementation on OLCI A and B products. Latest OLCI-A and first OLCI-B validation status will also be presented.

Area B - Board 212

The Sentinel-3 Mission Performance Centre (S3MPC) is one of the facilities of the Payload Data Ground Segment (PDGS). It is composed of 2 main components: 1) a Coordinating Centre (CC), where the core infrastructure is hosted, and where operators perform the quality control of S3 products and provide support to all experts for more in-depth analysis; 2) a team of highly-qualified Expert Support Laboratories (ESLs) who are involved in calibration and validation activities and who provide specific assessment of the products (e.g., analysis of trends, ad hoc analysis of anomalies, etc.). For these two components, the S3MPC has many interactions with the Land and Marine PDGS teams.

While the scope of S3MPC activities includes altimetry and L2 marine products (in support to EUMETSAT), this abstract focuses on all the operations led by the S3MPC and the calibration and validation of optical land products.

The S3MPC routine operations started with the launch of S3A in February 2016 and now include S3B, launched in April 2018 and which is now in routine phase after the successful IOCR held in mid-October 2018.

Thus, after more than 2 years of operations, we will briefly present all the activities done by the S3MPC which include:
• The QC activities done by the operational component of the S3MPC
• The Monitoring and assessment of instrument performances, and the reporting documents which summarize the analysis done
• The calibration and validation activities done on optical L1 and L2 Land products
• The maintenance and evolutions made on the operational processing chains (IPFs)
• The preparation of new optical products
• The support made to the PDGS coordination desk
• Information (reports, websites,…) prepared by the S3MPC to help users to fully exploit the S3 data.

Lessons learnt during that period, main results and key outcomes will be presented.

Area B - Board 213

The quality of earth observation data, and its usability in further processing, depends critically on the screening of the top of atmosphere signal for conditions suitable for processing. Obstruction flagging is critical, and obstruction can be due to clouds, aerosols, mixtures of these, and the level of obstruction depends on concentration and type of aerosol and clouds. How much these obstructions hinder the downstream processing is often depending on the application. Where for example aerosol optical depth (AOD) retrieval is very sensitive on presence of clouds and thus requires a very strict cloud masking, other applications are capable of handling for example thin cloud layers. On the other hand, the flagging of thick aerosols by a cloud mask is undesirable for AOD retrieval algorithms. Furthermore, the involvement with a great variety of users has shown that the need for unreliable pixel identification is varying depending on the application. This means, water applications for example have other requirements on pixel identification compared to land applications. And the requirements become even more versatile investigation different applications over land and water. The experience has also shown that a greater precision in the identification of easy to confuse features like cloud and snow/ice is needed. Especially for snow and glacier monitoring a precise distinction between snow/ice and clouds is very important.

In the past, most remote sensing products delivered only a binary cloud mask for detection of unfavorable pixels. The above described experience has shown that the requirements for pixel identification have become more and more detailed and are mostly driven by the application and not the data itself. Thus, a simple binary cloud mask is not up to date anymore.

The Sentinel-3 Mission Performance Center (S3-MPC) has put great effort in developing and improving cloud masking algorithms for OLCI, SLSTR and the OLC/SLSTR Synergy (SYN) product and tried to incorporate the need for application driven masking.

OLCI for example provides a three-staged cloud mask with different levels of certainties. Additionally, OLCI downstream processors (Level 2) identify pixels which behave suspiciously during processing and thus give indication for different subsequent applications for being used or disregarded for further processing. Further the S3 Snow project has contributed in developing new methodologies for the identification of clouds over snow/ice.

SLSTR cloud masking has been improved by a Bayesian/Probabilistic cloud masking providing probabilities that can be used by downstream applications depending on their needs besides a classical binary cloud mask.

For the SYN product such an application driven masking was already implemented inside the Level 2 processor, where for AOD retrieval a very strict combination of cloud masks is used but reflectance values are provided including different levels of cloud masking to the end users in the Level 2 product.

These developments have been accompanied by multiple validation activities presented in the frame of Sentinel-3 Validation Team (S3VT) meetings, reflecting the above described outcomes and requirements. The validation activities for OLCI, SLSTR and SYN have been conducted using an expert pixel collection approach covering a great variety of different pixel features from multiple opacities of cloud cover, over a different land cover types with high and low aerosol contaminations and snow coverage. These validation activities will continue with future development of Sentinel-3 pixel identification developments to ensure the highest possible data quality.

These developments are a good step towards an application driven cloud masking. Within the S3-MPC we are developing further this concept, develop and test new way of providing different types of cloud masks at Level 1 (general purpose) and Level 2 (application specific) and using different types of indicators (physically based AOT, COT or probabilistic scale). In this talk we will explain discuss the paradigm shift and report on the recent developments.

The here presented concept of application driven cloud masking is transferable to a variety of Copernicus missions, like Sentinel-2, Proba-V or future FLEX mission.

Area B - Board 214

The CEOS Cloud Masking Inter-comparison eXercise (CMIX) is an international collaborative initiative within the Atmospheric Correction Inter-comparison Exercise (ACIX-II). It is jointly organized by ESA and NASA and aims to inter-compare a set of cloud detection algorithms for space-borne high-spatial resolution (10-30 m) optical sensors. The exercise will focus on Landsat 8 and Sentinel-2 imagery acquired for a range of conditions.

The inter-comparison of cloud masks (CMs) is expected to contribute to the better understanding of strengths, limitations and applicability of different algorithms, and how the derived cloud masks influence the derived Bottom-of-Atmosphere (BOA) products.

The exercise started in Summer 2018 with a call for participation and distribution of a draft protocol for comments. Thirteen algorithms have been registered for participation. In October 2018 a first workshop was conducted in Washington, DC, USA, together with the Atmospheric Correction Intercomparison Experiment, ACIX, where the protocol was discussed in detail.
The CMIX protocol foresees 4 phases:
1) Definition of the inter-comparison protocol (October 2018)
2) Application of the CM processors by the algorithm providers (Deadline: 15 March 2019)
3) Results Analysis Report by the organisers (May 2019)
4) 2nd Workshop of CEOS-WGCV ACIX-CMIX (June 2019 @ ESA)

We will work with two types of datasets: the input data, which will be Sentinel 2 and Landsat 8 L1 data in their original format, and reference data which are classified images or pixels within images. These references will be the Landsat 8 Biome dataset (Foga et al. 2016), the Sentinel 2 Hollstein database (Hollstein et al. 2016), Landsat8 and Sentinel 2 PixBox manually classified pixels (ESA/Brockmann), Goddard L8/S2A/S2B database and the CESBIO classified dataset. Different pixel based and object based statistical metrics will be used to quantify the agreement of the results of the different algorithm with the reference data.

The results will be discussed at the second workshop, and subsequently they will be made publicly available in detail, and a summary will be published in literature.

At the time of the Living Planet Symposium we will be in the phase when the processed data will be sent back from participants to the organisers. We will have some first insight into results, very preliminary though.

Area B - Board 215

The Copernicus programme is a European initiative for the implementation of information services dealing with environment and security, mainly based on observation data received from Earth Observation (EO) satellites. In the frame of this programme, ESA launched the Copernicus Sentinel-2A and Sentinel-2B optical imaging satellites. These satellites deliver a new generation of optical data products designed to directly feed downstream services mainly related to land monitoring, emergency management and security. To ensure the highest quality of service for the Copernicus Sentinel-2 mission, ESA set up the Sentinel-2 Mission Performance Centre (MPC) supported by several Expert Support Laboratories (ESL). This presentation reports results of performance analysis of the Sentinel 2 Level-2A processor Sen2Cor performed by the ESL-L2A.

Sen2Cor Level 2A processing is applied to Top-Of-Atmosphere (TOA) Level 1C ortho-image reflectance products. Sen2Cor L2A outputs are a scene classification image, aerosol and water vapour maps and the ortho-image surface reflection (SR) product. SR is a Bottom-Of-Atmosphere (BOA) reflectance product corrected for the influence of the atmosphere. Scene classification is generated together with Quality Indicators for cloud and snow probabilities.

Many earth remote sensing applications rely on time series of satellite data. However at many locations on the globe it is almost impossible to build time series relying only on cloudless data due the cloudiness of the region. Generally time series get denser if they can include data containing some clouds. However, how accurate and reliable are L2A satellite products of cloudy images? To answer this question, the presentation is focused on providing Sen2Cor performance in dependence of cloud cover of the images.

Area B - Board 216

Since 2015 Sentinel-2 provides a very valuable and huge set of remotely sensed images. However, this satellite imaginary collection needs complementary sources in different applications: combination of Landsat-8 OLI for very frequently observations, completion of the cloud cover gaps, monitoring of the land surface dynamics on long time series using the complete Landsat legacy (MSS, TM, ETM). Consequently, the coherence between different sensor products and additional validations with third source measurements are relevant issues to deeply analyse.
The surface reflectance product provided by ESA, generated by Sen2Cor algorithm [Mueller-Wilm 2017], is widely recognized by research community. Additionally, some other techniques on radiometric correction are accurately designed for specific goals. The radiometric consistency on a combined series of different optical imaging sensor is the main goal of the present methodology. A validation method through in-situ measurements and drone imagery supports its relevance. Two different study regions and the corresponding measurements and imagery processing were conducted.
The location of the first study is the region of the Doñana National Park (DNP), SW Spain, SW Europe, path 202-203, row 034 in Landsat WRS-2; R037-T29SPB and R137-T29SQB Sentinel granules. The field spectroradiometric measurements were obtained by ASD FieldSpec3 that covers the VIS, NIR, and SWIR regions (350 nm–2500 nm). The corresponding sampling areas were designed with a circular shape with a diameter of 20 m; 20 to 40 measurements were made at each sampling plot. The field campaigns were materialized on 4 dates with concurrently at Landsat-8 and Sentinel-2A overpass: May 20th 2016, October 4th 2016 and June 1st 2017, while a field campaign was materialized on April 22nd 2017 with concurrently at Landsat-7 and Sentinel-2A overpass.
The second study region is a 5.26 ha extension centred in 41°56'59.3''N - 1°32'1.4''E, at the region of Catalonia, NE Spain, SW Europe and the acquisition data was April 21st 2018, correspond to path 198 row 031 WRS-2 or R051 T31TCG Sentinel granules. The UAS optical sensor was a multilens MicaSense RedEdge and the UAS platform was a DJI S900 copter. The drone images were radiometrically corrected using field spectroradiometric measurements (OceanOptics USB2000+; 340 nm–1030 nm) and reference targets, empirically fitting the image data to in-situ reflectance data.
The radiometric correction method that is the basis of the current validation work is called PIA-MiraMon [Pons et al. 2014]. This method fits the two unknown parameters (the spectral radiance received by the sensor from an area where there is only atmospheric contribution and the atmospheric optical depth) of a radiometric (atmospheric + topographic) correction model. The solution is obtained using reference and pseudo-invariant values generated through quality filtering and geostatistical analysis process of a large series of Terra-MODIS daily surface reflectance product.
In the first study region, the presented methodology applied to Landsat-7/8 and Sentinel-2 images improves the inter-sensor coherence between them. The coefficient of determination for PIA-MM is 0.964, vs. 0.929 for official agency products (USGS and ESA respectively) considering the four mentioned dates and all optical bands. PIA-MM also maintains a high consistency with field measurements. The mean RMSE for all bands is 3.468 vs. 3.671; we note that the SWIR bands present the highest discrepancies in both methodologies.
In the second study region and using the drone imagery as reference values (previously fitted to field measurements), we achieve a high level of consistency between Sentinel-2A and Landsat-8 images (RMSE blue = 0.852%, RMSE green = 0.940%, RMSE red = 1.147%, RMSE NIR = 1.707%), similar but lower than the official products consistency is (RMSE blue = 0.994%, RMSE green = 1.164%, RMSE red = 1.309%, RMSE NIR = 3.072%). Separately, the validation between drone and Sentinel-2 images corrected by both methodologies is: 1.869 PIA-MM vs. 2.049 official, for mean RMSE at visible bands.
The main conclusion of this work is that under the feasible ranges of uncertainty between different radiometric correction methodologies, a solution that achieves a higher consistency between different sensors is a prioritized option for multisource applications. This consistency is successfully evaluated though different measurements as field spectroradiometry and drone imagery in the present study.

References:
Mueller-Wilm, U. Sen2Cor Configuration and User Manual V2.4; European Space Agency, 2017; pp.1–53. http://step.esa.int/thirdparties/sen2cor/2.4.0/Sen2Cor_240_Documenation_PDF/S2-PDGS-MPC-L2A-SUM-V2.4.0.pdf.
Pons X, Pesquer L, Cristóbal J, González-Guerrero O (2014) Automatic and improved radiometric correction of Landsat imagery using reference values from MODIS surface reflectance images. International Journal of Applied Earth Observation and Geoinformation 33: 243-254.

Area B - Board 217

The geolocation of satellite imagery plays an important role in their applicability, as the horizontal accuracy has a series of implications, eventually affecting the overall reliability of the derived products. Additionally, the coregistration of different satellite image types (e.g. for data fusion), in particular between optical and SAR, is problematic, due the different nature of these products, which render the identification of corresponding pixels very difficult.
A reliable assessment of geolocation accuracy of satellite imagery typically necessitates the existence of a number of Ground Control Points (GCPs). Nevertheless, the recognition of GCPs on satellite imagery with spatial resolution of several meters is not a trivial task. It may be somewhat of an easier procedure on optical imagery than on SAR data, but, in practice, unless very high resolution (sub-meter) data are used, it is virtually impossible to reliably recognize any individual points on either image type.
In this context, the purpose of this study is to validate the geolocation accuracy of Sentinel-2 products, as they have an increased nominal geolocation accuracy, which provides major improvements and overcomes some of the aforementioned issues of previous optical missions (Landsat-5 and Landsat-8), which are also assessed in this paper.
In order to overcome the hurdle of GCPs recognition, we employ an approach that attempts to recognize geometries - rather than specific points – in order to a) validate the geolocation accuracy and b) extract GCPs that would be useful either in georeferencing and/or as an intermediate step for the coregistration of SAR with optical images.
In particular, we use KGNSS (Kinematic Global Navigation Satellite Systems) measurements with GPS and GLONASS and a local adjustment of characteristic geometries (such as crossroads) strategy, in order to identify GCPs in a Geographical Information Systems (GIS) environment. An advantage of this approach is that a large number of GCPs can be extracted, without having to actually identify each and every point on the satellite imagery.
In the results, we primarily discuss the geolocation accuracy of Sentinel-2 when compared against Landsat-5/TM and Landsat-8/OLI data. We also address the overall advantages and limitations of collecting KGNSS data, in order to use them as GCPs, taking into account the possibility of using Galileo GNSS satellites to further increase the accuracy of KGNSS data.
Finally, the prospects of automating the procedure for recognizing geometries and extracting GCPs and the implications on data fusion techniques at data (pixel or even sub-pixel) level are discussed.
This study is part of an experiment on geolocation accuracy and GCP extraction of Copernicus data also performed on SAR images (Sentinel-1, Envisat/ASAR and ERS) over Central Macedonia, Greece (Kalaitzis et al., also presented at this conference).

Area B - Board 218

As the market presence of VHR optical EO data with increasing resolution grows, so does the demand for high resolution, high accuracy orthorectified optical EO data. The pan-European VHR Optical EO data coverage supplied to Copernicus in 2018 is an example of a continental coverage with the most stringent requirements on positional accuracy to date. It has presented a number of challenges and required significant development towards reference data, production methodology and quality assessment and assurance.

As part of our tenure managing the Copernicus data Quality Control (CQC) domain, Airbus UK provided an assessment of the quality of 1% of the data delivered to this coverage dataset.

The assessment provided two main challenges to the team:

1. How to maximise service efficiency in order to provide assessment of the large number of data products required in the given timescale

2. How to provide accurate assessment of geolocation accuracy on orthorectified products with small GSD over such a large AOI

Existing in-house methodologies for measuring positional accuracy were time consuming and relied heavily on off the shelf software products. Additionally, the question remained as to the location of reference data of suitable quality (resolution and orthorectified accuracy) to perform the checks.

In response, we:

1. Developed a semi-automated positional accuracy checking procedure consisting of a new Python-based Positional Accuracy checking Tool (PAT), supplemented by brief operator interpretation via QGIS. The use of this new procedure acted to significantly reduce assessment time per image.

2. Conducted some research into the availability and usefulness of a number of potential sources of reference data including:
a.Two previous Copernicus VHR orthorectified pan-European coverage datasets from 2012 and 2015 – uniformly available across the coverage AOI
b. VHR Aerial data – variable availability on-line by country

We will describe our efforts towards the activities described above and present the related results.

Area B - Board 219

As part of the Copernicus programme of the European Commission (EC), the European Space Agency (ESA) operates the Sentinel-2 mission that acquires high spatial resolution optical imagery. The radiometric consistency of the image time series is ensured by some specific performance requirements such as radiometric stability, routinely monitored by the Sentinel-2 Mission Performance Centre (S2-MPC). This presentation provides a description of the radiometric calibration and validation activities, performed by the S2MPC including several Expert Support Laboratories (ESL). Last measured performances and last analysed results show the good quality of the mission products in terms of radiometry. The current performances for Sentinel-2A and Sentinel-2B are compared to mission product quality requirements.

The Multispectral Instrument (MSI) on board Sentinel-2 platforms A and B measures the Earth's reflected radiance in 13 spectral bands from VNIR to SWIR. It has been built by the requirements for large swath high geometrical and spectral performance of the measurements.
The on-board sun diffuser is a full-field/full-pupil diffuser, called the Calibration and Shutter Mechanism (CSM), which is integrated with MSI for radiometric calibration in order to guarantee high quality radiometric performance. This On-Board Calibration Device (OBCD) collects the sun light after reflection by the diffuser to prevent the instrument from viewing the sun directly in orbit.
Calibration and Validation under S2-MPC responsibility are critical processes in the quality control of generated products, and ensure that the processed data are meaningful to scientific users.

The radiometric calibration is based on the exploitation of the on-board sun diffuser images (for relative gains calibration and absolute radiometric calibration) and images acquired over ocean at night (for dark signal calibration).
The temporal evolution of the equalisation coefficients is monitored with the monthly sun-diffuser acquisitions. After each sun-diffuser acquisition, a new set of relative gains calibration parameters is generated. The time variation of relative gains is weak for VNIR bands: typically maximal variations do not exceed 0.2 to 0.4% between two monthly assessments. Variations are larger for SWIR bands with a change of gain coefficients up to 3% for some pixels and with an inter-pixel variation more pronounced for SWIR bands than for VNIR bands.

For VNIR bands, the absolute calibration coefficients are very stable, while for SWIR bands (especially for Band 10), the trend of decrease is up to −3% in six months. This is properly compensated with monthly update of the coefficients and the sensor sensitivity is completely retrieved with MSI decontamination operation every six months.

Radiometric validation activities assess all radiometric performances related to image quality requirements.
The instrument response non-uniformity is assessed through the Fixed Pattern Noise (FPN) and Maximum Equalisation Noise (MEN) which quantify local non-uniformities in the responses of physical detector pixels across the swath.
For all the VNIR bands, the maximal value of FPN is clearly below the specified limit (0.2% for all bands) typically by an order of magnitude. For SWIR bands, it is not the case: for Bands 10 and 11, the maximal value of FPN can be higher than the specified limit (0.35% for Band 10, 0.2% for Band 11), but the FPN values exceed the limit only for a few pixels as the values of FPN for the 98% quantile are significantly below the specified limit, even for these bands which are more affected by artefacts on the sun-diffuser BRDF characterization than the VNIR bands. The monthly update of equalization coefficients resets the FPN to zero, and ensures that the requirements are met until the next calibration.

The vicarious validation of the radiometric measurements is performed using DIMITRI toolbox and the Radiometric Calibration Test Site (RadCaTS) dataset. Fourteen desert, ocean and snow CEOS PICS are used to achieve this activity. The validation results are consistent over all test sites and over the different methods.
The multi-temporal relative radiometric uncertainty validation results show a good stability of MSI, and the measurements are compliant with the mission requirements of less than 5% error, except for Band 05 due to DIMITRI limitation in Top-Of-Atmosphere (TOA) simulation of spectral bands with significant absorption from water vapour and dioxygen.

The ground reflectance-based approach is used over the RadCaTS to perform the absolute radiometric vicarious calibration over the whole spectral range of MSI. The results indicate a very good agreement with DIMITRI toolbox results for the VNIR bands. The RadCaTS data are provided by the NASA Landsat Cal/Val Team as part of the ESA expert users effort.

The radiometric validation suggests that MSI-A measured reflectances are slightly brighter that MSI-B ones by about 1% (except for Band 12). This small discrepancy could be due to uncertainties in the ground characterization of the sun-diffusers albedo. However, this is not a concern as this difference is within the method’s uncertainty magnitude.

The Signal-to-Noise Ratio (SNR) is well inside the requirement (with more than 20% margin) for all bands and appears to be very stable.

Area B - Board 220

High resolution satellite images such as Sentinel-2 include a material containing rational polynomial coefficients (RPC) for three-dimensional geopositioning. However, three-dimensional ground positions calculated by using only the RPC from stereo pair have inevitable systematic errors. Thus, it is necessary to correct the systematic errors of the RPC using several ground control points (GCPs). However, the use of GCPs results in the overall application costly and time consuming. Since it is not appropriate to be applied in inaccessible areas, various ways to alleviate the use of GCPs have been researched. However, the existing proposed methods are too complicated and low accuracy.

To overcome this limitation, we propose an efficient method to automatically register high resolution satellite image by using tie points obtained from stereo images and a digital elevation model (DEM) without any additional GCPs.

This method includes four steps: 1) generation of a large set of tie points, 2) determination of the ground coordinates of the tie points, 3) refinement of the ground coordinates using a DEM, and 4) RPC adjustment model parameter estimation. We will analyze the accuracy of the results from our algorithm using Sentinel-2 stereo pair and GCP.

Area B - Board 221

SENTINEL-2 observations are particularly well suited for crop management due to the short revisiting time and decametric spatial resolution. In this context, Green Area Index (GAI) and Canopy Chlorophyll Content (CCC) are variables of particular interest since they both relate to (1) the canopy photosynthetic potential and as such can be assimilated into crop functioning models and (2) to the amount of electromagnetic radiation reflected into space which make them accessible from S2.
The SENTINEL2 SNAP toolbox proposes a generic algorithm to derive LAI and Chlorophyll maps from SENTINEL-2 acquisitions based on 1D radiative transfer model inversion. This algorithm is also currently implemented at the THEIA land data service. Although it was already implemented for other sensors, it was mainly evaluated for GAI estimations and against a limited number of situations, including SENTINEL2 pre-launch in-situ campaigns. The objectives of the P2S2 project are threefold: (1) to provide an extensive calibration and validation dataset over a large range of crops observed under contrasted growth conditions at several phenological stages representing the whole vegetation cycle, (2) to develop generic and crop specific algorithms based on machine learning techniques trained with in-situ data, (3) to compare the performances of the in-situ empirical approach with 3D radiative transfer model inversion.
We present here the validation database that was gathered in 2018 during the first year of the P2S2 (Produits Pour Sentinel2) project and compile existing datasets with ground measurement campaigns that are conducted during the project (9 crop types, at 4 or 5 phenological stages and 4 locations, leading to around 800 SENTINEL2 samples). The sampling protocol of the measurements follows the recommendations of the CEOS LPV (Land Product Validation) group. Chlorophyll was measured with the SPAD instrument. An automatic processing of the hemispherical images used to estimate GAI was specifically developed based on deep learning techniques. The whole data base collected will be made freely available to the scientific community. Finally, performances of the several algorithms to estimate GAI and CCC are compared.

Area B - Board 222

In recent years, several moderate resolution optical Earth Observation (EO) sensors were developed to operationally acquire data for a variety of land monitoring applications. The growing use of EO data by the scientific community has reinforced the importance of obtaining consistent long-term datasets for multi-temporal applications such as vegetation monitoring and land-cover land-use classification and change.
The joint USGS and NASA Landsat program is the longest running program for provision of global EO optical data freely available to the public and considerable effort has been undertaken for data preservation thanks to the Landsat Data Continuity Mission (LDCM) and the Landsat Global Archive Consolidation (LGAC).
In the last five years, the ESA Data Service Initiative (DSI) has continuously performed collection/consolidation of Level 0 (L0) products and data reprocessing up to Level 1 (L1) products in view of data preservation. In this context, the Instrument Data Evaluation and Quality Analysis Service (IDEAS+) has been responsible for product / algorithm specification and design, as well as data quality control and validation. Within this service, the ESA L1 products from the Landsat 1 to Landsat 5 Multi-Spectral Scanner (MSS) missions acquired in the period 1972 - 2001 have been reprocessed. The Landsat MSS products represent the oldest archive at ESA, with a geographical coverage from Greenland to Continental Europe and North Africa.
Due to the age and performance decay of the MSS sensors, these activities have been particularly challenging and the Instrument Processing Facility (IPF) software and internal Quality Control (QC) tool have been significantly improved by the IDEAS+ service. Notably, improvements were achieved for radiometric / geometric accuracy, image data recovery and additional quality assurance information.
Over 600,000 L1 MSS products have been generated and are expected to be reprocessed with the latest updates, complementing the existing 1 million Thematic Mapper (TM) and Enhanced Thematic Mapper Plus (ETM+) products.
This paper presents the format of the reprocessed MSS products and provides detailed information on the updated and additional metadata parameters and the generation of a Quality Assurance Band (BQA) designed to simplify pixel selection and discard low-quality pixels. The geometric modelling and radiometric processing algorithm, applied to improve the overall data quality and usability, are also discussed in the paper.

Area B - Board 223

Most satellite-based snow cover products provide only snow extent - a “binary” map, whereby each pixel is classified as either “snow” or “not snow”. Fractional snow cover (FSC) - the viewable fraction of the surface covered with snow in a remotely sensed image pixel - is necessary for atmospheric correction and to retrieve various satellite products characterizing the land surface, including albedo, temperature, soil moisture, heat fluxes, and vegetation parameters on a global scale.
Snow extent and fraction are planned to be provided by Sentinel-2 and Sentinel-3 satellite constellations of the Copernicus Space Component for a large number of applications and user needs. The current status and potential future modifications to improve validation of the both snow products are considered in the presentation.
An increasing role of various inter-comparison projects (modeling, algorithms, forecasts) characterizes a current stage of developing theoretical and applied studies. For example, quantitative estimate of snow products provided on the basis of different approaches is an important component of the CCI+ SNOW project. However at the present time there is no a common validation protocol agreed by the snow community.
The emphasis in this presentation is put on the description of recommended robust methodologies to estimate the quality of retrieved snow cover products and the comparison of the approaches developed by different authors.
Optical remote sensing provides the most reliable information on snow extent including fraction and physical snow properties. The validation procedures are designed to quantitatively estimate the errors in moderate resolution snow products on the basis of a comparison with high-resolution (Sentinel-2 and Landsat) remote sensing data providing reference information that is used as an effective source to establish reliable ground truth.
The difference between the aggregated fine resolution data and the moderate resolution products is considered as an indicator of an algorithm uncertainty. However such a typical approach to estimate the errors in snow products is not without significant shortcomings. In many cases the comparison of the results provided by the same algorithm to the same image at different resolutions is used to estimate the uncertainty in snow retrieval. The validity of such an approach is very doubtful, because performance of practically any algorithm will be excellent if estimated by this “methodology.” It is not even a cross-comparison between algorithms, but could be classified as auto-cross-comparison providing unrealistically low magnitudes of algorithm errors.
The following approach to analyze and use high-resolution satellite observations provided by Sentinel-2 and Landsat is considered the most general and proposed for possible detailed exploration by the Sentinel Validation Teams and the Copernicus Control activities.
(A)First of all, it is necessary to determine two types of high-resolution pixels: characterized as completely covered by snow and snow-free. It is known that these pixels are easily identified in false cover images presenting a combination of different bands increasing the spectral separation and interpretability of the observations.
The location of the snow and non-snow pixels in n-dimensional space with n axes corresponding to reflectances and brightness temperatures measured from a satellite might be certainly located in the multi-dimensional sample space without doubt if a novel coloring scheme is used that replaces colors representing a relative likelihood of probability density by colors corresponding to the false color image.
(B)This step provides comprehensive information not only on the most probable magnitudes of measurements for each bands, but also about related variability (including standard variation) characterizing snow and snow-free high-resolution pixels. The task of snow cover classification is to distinguish those two types of pixels - for binary classification and to estimate snow fraction. The task of estimating subpixel fraction as well as binary classification essentially requires the determination of snow and non-snow endmembers.
(C)There are two different methods to define snow and non-snow endmembers: using reflectances or their combinations. In the latter case the two major combinations are (1) a ratio of reflectances and (2) normalized indexes, such as the Normalized Difference Snow Index (NDSI). Both of these combinations include two reflectances. The most probable magnitudes and standard variations for the combinations are calculated on the basis of information on average reflectances and standard variations predetermined in step (B) for individual bands.
(D)The most probable magnitudes of snow and non-snow endmembers and their standard deviations could be used to estimate the signal-to-noise ratio measuring the ratio of signal level (the difference between magnitudes of snow and non-snow endmembers) to the level of noise in endmembers characterized by their variabilities. The choice of the channels or their combinations with the largest signal-to-noise ratio will lead to the higher quality of snow products.
The enhanced NDSI-based algorithm can explicitly accounts for changes in the endmembers from one image to another using the equation
FSC = (NDSI - NDSInon-snow) / (NDSIsnow - NDSInon-snow),
where NDSIsnow and NDSInon-snow are scene-specific snow and non-snow endmembers calculated in the algorithm on the fly.
The comparison of snow products derived from remote sensing information with the reference data serves as a very valuable source of understanding shortcomings in a snow retrieval algorithm and choosing ways to improve its performance. The careful detailed validation for a variety of conditions provides important information on the reasons of retrieval errors and helps identify the directions of studies to improve the snow products.
Further enhancements of snow algorithms and their validation take into account the following features
(E)Though reflectances in different bands are linearly related they are not necessarily proportional to each other;
(F)Estimate of snow products quality depends on a spatial scale reflecting regional surface property and analytical algorithm characteristics;
(G)Reliable validation of snow products might be done on the basis of considering stratified performance depending on numerous factors and calculated for varying snow fractions within the entire range of reference fraction changes;
(H)The statistics for very variable in space and time reflective snow properties need to be determined for each individual scene separately.
The last feature could be considered as one of necessary requirements to any snow algorithms, which allows us to emphasize that the optimal way to derive moderate resolution optical remote sensing information on snow cover extent and fraction should take into account the variability of snow and non-snow properties within a scene-specific algorithm, which is implemented in a modifies NDSI-based approach.
All the above stated features were included into consideration to compare the quality of different potential snow product algorithms tested for applicability to Sentinel-3 demonstrating advantages of modified NDSI-based approach.
One of original mode of validation was applied to compare results of the NDSI-based algorithm with the widely used tie-point algorithm - a simple version of a linear spectral mixture technique to a visible band. The comparison implemented the best achievable realizations of NDSI-based and tie-point algorithms characterized by the smallest possible root mean square errors for each individual image. It demonstrates advantages of the NDSI-based approach characterizing by twice smaller dispersion errors.
A large scale international snow products inter-comparison and validation project (htttp://snowpex.enveo.at) initiated by the European Space Agency (ESA) and completed at the Environmental Earth Observation company in Innsbruck, Austria, confirmed the advantages of the NDSI-based approach finding that the snow fraction algorithm of Salomonson and Appel has advantages in comparison with other tested methods, both for open land and in forested areas for two reference data sets created using different techniques.
The inter-comparison of results provided by different techniques is a necessary stage in algorithm assessment and thorough validation is required at the earliest possible stages of algorithm development. Collaboration between researchers with experiences supplementing each other can be a very promising way to tackle the problem of optimizing quality of snow algorithms. Such collaboration should be considered as a key recommendation for developing snow algorithms and other related activities.

Area B - Board 224

The Radiometric Calibration Network (RadCalNet) portal (hosted at https://www.radcalnet.org) is the infrastructure created by the RadCalNet Working Group under ESA contract to distribute the data acquired by a network of instruments for the absolute radiometric calibration, intercalibration, and validation of Earth-observing sensors.
At this time, there are four calibration test sites equipped located in Gobabeb (GONA, Namibia), La Crau (LCFR, France), Railroad Valley Playa (RVUS, USA), and Baotou (BTCN, China) which collect BOA reflectances and atmospheric parameters. Therefore, a processing of the data is performed by NASA to estimate TOA reflectances in 30 min intervals for a nadir view between 9:00 and 15:00 local time.
A portal dedicated to the data distribution has been developed and is opened to the public users (after registration) since July 2018. At that time (Nov. 2018), there are 222 registered users located in 28 countries, and the number is growing each month, including representatives of national space agencies, private operators, and universities.
Dataset spatial coverage is large, starting from March, 2013 for RVUS, (634 days of data available), December 2014 for LCFR (536 days of data available), January 2017 for GONA (281 days of data available) and March 2016 for BTCN (172 days of data available)
The users can download directly the totality of the dataset for all sites and all days, but can also select from a calendar the days of interests for calibration. For all days for which TOA reflectances are available, the atmosphere parameters are also provided (water vapor content, ozone content, pressure and aerosol optical thickness and angstrom parameter).
Graphic displays of all data (reflectances and atmospheric parameters) are available for each date, helping the user to select or not the data.
Besides the data, the portal provides also the documentation relative to the sites, the processing, data format, uncertainties and calibration inter-comparison reports.
A forum of discussion is also available (https://forum.radcalnet.org/). It provides announcements and news, and all topics related to RadCalNet data, sites, and documentation.

Area B - Board 225

DIMITRI is an ESA tool developed to monitor the radiometry of sensors using sites located in the ocean and deserts in the first version, and now over snow and clouds. Historical long time series of satellite data (MERIS, MODIS/Aqua, AATSR, ATSR-2, VEGETATION, and PARASOL) are available in the database. The application has been updated to read S2A/MSI, S3A/OLCI and S3A/SLSTR, Landsat 8/OLI.

The software contains functionalities to ingest data and perform the vicarious validation of the Level-1C products over ocean dark sites located in South Indian Ocean, and South Pacific Gyre, over ocean bright targets when sunglint is available in the acquisitions, and over bright desert sites. The long list of sensors and sites is available on DIMITRI website.

In DIMITRI V3, two types of methodologies were implemented. The first one, called “Sensor to Sensor comparison”, allows to compare TOA reflectances acquired in the quasi identical geometrical conditions for two sensors. The second one, called ‘Sensor to Simulation comparison”, allow to compare radiative transfer simulations of TOA reflectances to TOA measurements. These comparisons are performed over ocean sites for the Rayleigh method and sunglint over ocean, and desert sites.

In DIMITRI V4, two new methodologies have been added to the Sensor to Simulation module: the monitoring of radiometry over Deep Convective Clouds and over bright snow sites located at Dome C, Antarctica. The results will be presented for Sentinel 3A data.

DIMITRI software is developed in IDL language. It is well structured to add easily new methods, new data and new sites.

Area B - Board 226

Global, operational water quality remote sensing now extends from oceans and seas to estuaries, lakes and rivers. The cycle of instrument calibration, algorithm validation and subsequent product validation still takes a number of years which hampers development of novel algorithms and solutions. Sustained in situ observations, in particular of water-leaving reflectance are therefore required at strategically located sites representing a range of water types under varying atmospheric conditions. This requirement is particularly urgent for optically complex waters such as lakes/reservoirs, transitional systems (rivers, estuaries, lagoons) and the near-coastal environment.

The H2020-funded MONOCLE is developing a range of solutions for in situ (spectro)radiometry to obtain atmospheric thickness and water-leaving reflectance from ships, remotely piloted aircraft systems, buoys and handheld devices, including solutions suitable for operation by citizen scientists and remote locations. Progress towards development of these solutions will be presented. The presentation will also cover the data services in development for sensor-to-service communication - designed to guarantee near real-time data availability to operational satellite data processing services, continuous quality control and remote sensor calibration.

Developments covered by MONOCLE include: extending radiometric measurements for satellite validation in optically complex waters; improved interoperability and synchronization of new and existing water quality sensors; connecting automated, piloted and citizen operated sensors with interoperable data interfaces; DIY kits and open-source blueprints to accommodate community development.

For more information on MONOCLE, the project consortium, and regular updates please visit monocle-h2020.eu or follow the project on twitter @monocle_h2020

Area B - Board 227

A network of hyperspectral radiometers is being developed for radiometric validation of satellite missions. This network follows closely the AERONET-OC federation concept [Zibordi et al, 2009] but uses the TRIOS/RAMSES hyperspectral radiometer and a more extensive multi-look pointing scenario. The instrument system consists of one radiance and one irradiance sensor on a pointing robot, controlled by a microprocessor and supplemented with GPS, inclinometer and video camera data feeds. The measurement protocol is based on the abovewater method of [Mobley, 1999 and 2015], but includes additional scenarios for different viewing zenith and azimuth configurations. The system is being deployed initially in Belgian coastal and inland waters, then at sites in Argentina and France before full international expansion. The network will provide water reflectance data for the radiometric validation of all visible and near infrared bands of all optical missions, including Sentinel-2A&B, Sentinel-3A&B, PROBA-V, MODIS-AQUA&TERRA, VIIRS, Landsat-8, Pléiades, CHRIS-PROBA, MSG-SEVIRI, PlanetDove … ENMAP, PACE, MTG and ... any future optical missions, including nanosatellites. This presentation will provide results from prototype testing on Aqua Alta Oceanographic Tower in July 2018 and a first operational deployment in 2019 in Belgian North Sea waters, including a comparison with the corresponding AERONET-OC data from these sites. Research in progress on the abovewater measurement protocol and the operational plan for characterisation of all radiometers will also be outlined.

Area B - Board 228

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 hyperspectral data from each site which can be used for many optical missions (S2, S3, PROBA-V, MODIS, VIIRS, L8, Pléiades, ENMAP, PRISMA, SABIA/MAR,etc.) gives a huge economy of scale. The existing networks (e.g. AERONET-OC, RADCALNET, ...) are based on multispectral instruments requiring modelling associated uncertainties to cover all spectral bands of all sensors. Therefore, the general objective of the HYPERNETS project is to ensure that high quality in situ measurements are available at all spectral bands and for a wide range of water and land types for the validation of the surface reflectance data issued from all earth observation missions. With its global network of land and water sites providing hyperspectral data, HYPERNETS is expected to fill the very important gap in the current Sentinel-3&2 Validation plans and the in situ component of the Copernicus program. HYPERNETS will become the main source of surface reflectance validation data for all spectral bands of all optical missions for at least the next 10 years.



The HYPERNETS instrument design of the new lower cost hyperspectral radiometer for automated measurement of water and land bidirectional reflectance is presented as well as the associated pointing system and embedded calibration device. Sensor and system designs of the instrument complies with requirements gathered from the water and land community. In order to test and demonstrate the performance of the developed system and instrument within HYPERNETS project, 24 sites have been selected covering a wide range of water and land types and a variety of environmental (cold/hot, wet/dry, calm/windy, biofouling, etc.) and logistical conditions (accessible/remote, standalone/heavily instrumented site, well protected/vulnerable to humans and/or animals). Figure 1 shows the location of 22 sites selected for the HYPERNETS prototype network. The prototype network design is presented. Site characteristics, including spatial and temporal specificities, are described and challenges with the system set-up are discussed.

Area B - Board 229

The Environmental Mapping and Analysis Programme Hyperspectral Imager (EnMAP HSI) will allow to acquire images of the Earth surface in a push-broom configuration. Spectral coverage includes 230 wavelength bands between 420 nm and 2450 nm which are simultaneously recorded with a ground resolution of 30 m x 30 m. This contribution reports development, test and implementation of a setup for fine alignment and characterization of the EnMAP HSI.

The instrument optics unit (IOU) of the HSI comprises a telescope assembly, an instrument spectrometer unit (ISU) and cameras. The ISU has two independent spectral channels: visible and near-infrared (VNIR) and short-wave infrared (SWIR). The telescope assembly, each spectral channel and cameras are aligned separately. After final integration, an overall characterization is performed with a required spectral calibration accuracy of 0.5 nm for VNIR and 1 nm for SWIR channel and lateral calibration of the full field-of-view with sub-arcsecond accuracy.

Alignment and characterization of the state-of-the-art hyperspectral instrument require to establish measurement setups that outperform the test object in all relevant performance aspects to achieve the required measurement accuracies. At the same time, technical as well as economical considerations yield to develop a measurement equipment that can support multiple use-cases throughout the process of the Alignment, Integration and Test (AIT) of the Instrument.

The setup consists of several modules. The core component is a highly stable diffraction-limited 200 mm collimator including a movement system and a scene generator. The equipment supports the following measurement cases:
• Double-pass wave front measurement for alignment of the HSI-Telescope to the ISU;
• Line-of-sight characterization of the HSI;
• Scanning knife-edge modulation transfer function (MTF) measurement of the HSI in the entire field-of-view;
• Spectral response characterization of the HSI in the entire spectral range.

The overall system has been set-up and successfully implemented for several use-cases by AIT team at OHB System in Wessling (Munich). Validation of the design requirements as well as setup performance are reported.

ACKNOWLEDGEMENTS
The work presented herein was performed on behalf of the German Space Agency DLR with funds of the German Federal Ministry of Economic Affairs and Technology under the grant No. 50 EP 0801.

Area B - Board 230

The main objective of the 4th (A)ATSR Reprocessing is to generate (A)ATSR (ATSR-1, ATSR-2 and AATSR from ERS-1, ERS-2 and Envisat) L1B products in a new format that closely mimics the format adopted by SLSTR (Sentinel-3) products. This new format will allow for users to continue their analysis of high quality measurement data (SST and LST) from the (A)ATSR missions and into the SLSTR missions, in a more harmonised way.

The new format for the 4th (A)ATSR Reprocessing will allow for the inclusion of the following new L1B product features: radiometric calibration uncertainty estimates, Bayesian and probabilistic cloud masks, and meteorological information (ECMWF Re-analysis). The 4th (A)ATSR Reprocessing will also implement the following product data (and dataset) improvements: extended dataset coverage (due to data recovery), the use of orthogeolocation, and improved surface classification masks (using the Sentinel-3 surface classification masks).

It is expected that the official generation of (A)ATSR L1B products is to start in the first half of 2019. All (A)ATSR L1B products generated will be subjected to QC by the AATSR IDEAS+ QC team. Note that the 4th (A)ATSR Reprocessing will be generating (A)ATSR L1B products only; the ESA CCI SST and LST projects will be responsible for generating L2 products, making use of up-to date expert knowledge.

The paper will demonstrate the implementation some of the aforementioned new product features and data (and dataset) improvements, using the latest sample of (A)ATSR L1B products generated by the officially verified L1B processor.

Area B - Board 231

The harsh environment of space and the rigours of launch cause the degradation of earth observation satellite sensors and mean that post-launch calibration is essential to ensure satellite accuracy, stability and interoperability. In-orbit recalibration uses both onboard references and vicarious processes. Vicarious methods include the use of instrumented field sites (e.g. the automated sites of RadCalNet and one-off measurements of ground field campaigns), the use of Pseudo-Invariant Calibration Sites (PICS) and the use of natural phenomena (e.g. Rayleigh scattering, deep convective clouds and sun glint). The Moon is a special kind of PICS that has many advantages. It is a similar brightness to Earth scenes, while being extremely stable long-term. However, to use the Moon as a vicarious calibration reference requires a model to account for the familiar monthly phase cycles, as well as the 18 year Saros libration cycle.
Previous lunar irradiance models exist, the most notable being the ROLO (Robotic Lunar Observatory) model. This model is based on measurements obtained over a period of 8 years of images using two telescopes at the ROLO observatory in Flagstaff Arizona, from March 1996 to September 2003 (Kieffer H. S., 2005) .The observations covered a wide range of observable libration angles and phase angles and the resulting model is now the most extensive characterisation of the Moon’s absolute irradiance to date. ROLO was brought into the “GSICS Implementation of ROLO” (GIRO), which produced a community-accessible tool for implementing the ROLO model in a consistent manner. Although the ROLO/GIRO model has been found to provide good results, as a model describing relative radiometric changes, ROLO still has a nominal absolute uncertainty of 5 % - 10 %. Some comparison exercises have suggested there could also be systematic errors inherent in the relative model.
Here we present preliminary work from an ESA-funded project to make new lunar measurements and to extend the ROLO model and reduce its associated uncertainty. This project uses a CIMEL 318-TP9 lunar and solar photometer located at Teide Peak in Tenerife (3555 m altitude) during the summer and in Izana atmospheric observatory (2400 m altitude) during the winter. Teide Peak in Tenerife benefits from subsiding air from the descending branch of the Hadley cell and therefore has extremely low aerosol levels. The Moon is observed for ~2 hours each night. The logarithm of the observed measurements (at different air masses) are fit to a straight line to extrapolate to a top-of-atmosphere signal using the Langley plot method.
Before installation the instrument was extensively characterised to confirm its linearity and to determine the spectral irradiance calibration coefficients. These tests were carried out at NPL and are thus fully traceable to relevant primary standards, ensuring the lowest possible uncertainties for the calibration. An assessment of the thermal sensitivity of the instrument was also carried out at UVa.
Here we present the measurement protocol, the instrument calibration and initial model results. The final step in this project is to use the determined TOA lunar irradiance on each night obtained over an 18 month period to refine the parameters used in the current ROLO model. This presentation will include an explanation of this process and how the associated uncertainties will be determined.

Area B - Board 232

New Zealand represents an attractive target for satellite calibration/validation studies for many disparate reasons. These include the following:
(a) It has a small population (only 4.9 million) in an area much the same as Italy, Poland, or the United Kingdom, so that it is relatively unpolluted whilst having many research institutions already operating environmental, ecological and agricultural data collection projects rendering ground-truth records stretching back some years.
(b) In terms of climate, it possesses regions ranging from typical for northern Europe through to sub-tropical.
(c) Being in the southern hemisphere, the seasons in NZ are half-a-year out of phase with the north, making it possible to obtain two seasonal cycles of cal/val data collection per annum.
(d) In terms of terrain, NZ is highly differentiated with the North Island being formed by past (and ongoing) volcanic eruptions whilst the South Island contains alpine forms along with coastal plains resulting from the collision between the Pacific and the Australian tectonic plates.
(e) NZ has highly-active geology, with frequent earthquakes causing major land shifts, active volcanoes, land slips and major erosion events often occurring, along with floods, storms and other geomorphological changes of interest to scientists; in 1991 the top 30 metres fell off NZ's tallest mountain.
(f) NZ has many other geographical and geological peculiarities; for example it has more than 3,000 lakes over a hectare in size, including the two most-transparent natural bodies of water on Earth.
(g) Human occupation began in NZ only about 800 years ago, and European colonisation less than two centuries ago, so that land use changes are relatively recent and the era of farming has been well-recorded.
(h) Agricultural use of land has resulted in many different types of ground coverage ranging from grass and grapes through to fruit orchards and pine plantations.
(i) There are parts of NZ that are still wilderness, essentially untouched by human activity, including virtually impenetrable bush and fjords in the far southwest.
(j) Apart from the main populated islands often thought of as comprising ‘New Zealand’, the ‘Realm of New Zealand’ is far more extensive, stretching from the South Pole almost to the equator; NZ has one of the largest Exclusive Economic Zones globally due to its many scattered islands; NZ’s search and rescue responsibility covers the seas from the Antarctic to the equator, and half way across the Pacific Ocean; this speaks to calibration and validation opportunities far more extensive than those represented by the main islands alone.
(k) NZ has few governmental layers, making it relatively straightforward to get things done.
(l) Obtaining clearance to fly airborne sensors and UAVs in NZ is comparatively easy compared to nations with more congested airspaces.
(m) NZ is well-connected in terms of air services and internet availability.
(n) There is a burgeoning space industry in New Zealand ready and able to collaborate on Earth observation programmes.

Area B - Board 233