Methodologies and Quality 2

An essential part of any climatological study trying to make use of muti-sensor data over long periods, is to ensure that datasets are homogeneous and consistent. In this framework, aA systematic inter comparison exercise is was run over almost the entire envisat ENVISAT mission, from 2004 to 2012, aiming to at understand analysing the evolution of MERIS and SCIAMACHY Level 1 products (i.e. measured radiance) during the mission life-time.

 The two instruments are were simultaneously measuring the solar radiance reflected by the Earth ,under different geometric and spectral conditions with similar geometries and different spatial and spectral resolutions. However, the detailed knowledge of the instrumental spectral functions allows for the direct comparison of radiance.; this paper addresses the methodology used to to inter-compare the measured radiances and presents the results and their evolutions along the mission life-time over systematic collocated acquisitions and over selectable ground test-sites.

The MERIS instrument measureds the solar radiation reflected by the Earth at a ground spatial resolution of 300m (full resolution), in 15 spectral bands, programmable in width and position, in the visible and infrared, with. The a 1150-km swath. is divided into five segments covered by five identical cameras having corresponding files of view with a slight overlap between adjacent cameras.

The SCIAMACHY instrument is was a passive imaging spectrometer, comprising a mirror system, a telescope, a spectrometer and thermal and electronic subsystems, measuring the solar radiation reflected by the Earth from 214 nm up to 2386 nm with different viewing geometry: nadir, limb and occultation mode.  

When in “nadir” mode, the SCIAMACHY swath is was perfectly overlapping  the MERIS’ one and, in principle, measurements could be in principle inter compared, after a delicate careful pre-processing phase necessary to homogenise the geometric and spectral resolution: MERIS resolution shall be degraded to fit into the SCIAMACHY nadir pixel and, vice-versaon the other hand, the SCIAMACHY spectra shall be convoluted with the MERIS filter functions in order to get the correct radiance measurements over the same target area. Moreover, the SCIAMACHY ground resolution is highly dependantdependent on the integration time of each specific spectral clusterrange; this leaded to a complex handling of data in order to inter-compare measurements under the same conditions. 

SCIAMACHY calibrated radiances are obtained by means of SCIAL1C tool, provided by DLR.

The analysed dataset is built considering orbits from both sensors acquired on the first and the fifteenth day of each month, from 2004 to 2012, for a total of about 3000 orbits and, 3 billions millions of collocated measurements. The dataset is then analysed considering measurements taken with a solar zenith angle less than 70 deg and considering target matches with high spatial homogeneity.

low Standard Deviation (STD <10%, Meris radiances contained into a SCIAMACHY observation)

This paper addresses the methodology used to inter-compare the measured radiances and presents the results and their evolutions along the mission life-time on global scale and over and over selected ground test-sites.

Results are critically presented and discussed.  Results over Libya-4 test site are presented and discussed. 

A new permanently instrumented radiometric calibration site for high/medium resolution imaging satellite sensors in the visible/near-IR has been set up in Namibia, near the Gobabeb Research and Training Centre on the edge of the Namib Desert. This site is the European contribution to the Committee on Earth Observation Satellites (CEOS) initiative RadCalNet (Radiometric Calibration Network). The Gobabeb area has been selected based on the analysis of different datasets to estimate surface spatial homogeneity, cloud coverage, temporal variability, atmospheric turbidity and flatness.

A field campaign took place in November 2015 in order to find the precise location of the future permanent instrumentation in the area identified from satellite data (Modis, Landsat8, Sentinel2, Pleiades). This location is the one with the best spatial homogeneity at different scales: instrument field-of-view (tens of centimeters), extent of the instrument-monitored area (tens of meters) and satellite sensor resolution (tens to hundreds of meters). The field campaign also focused on the characterization of the surface reflectance, the surface hemispherical directional reflectance factor and the atmospheric turbidity.

Following this campaign, the permanent instrumentation (CIMEL photometer) has been installed in January-February 2016. The data has been processed by the ROSAS processing software at CNES in order to obtain first results of surface reflectance and to perform the vicarious absolute calibration of optical sensors (Sentinel2, Sentinel3 etc.)

The presentation will focus on the field campaign conclusions and the analysis of the data produced by this new calibration station.

Since its launch in May 2013 the in-orbit radiometric performance of PROBA-V is continuously monitored by the Image Quality Center (IQC). Due to the absence of on-board calibration devices, the in-flight performance monitoring and calibration relies soly on vicarious calibration methods. The design of PROBA-V, with its wide field-of-view, three separate cameras and 18 detector units (including  3x3 mechanically butted SWIR detectors), introduces additional complexity to the in-flight vicarious calibration process.

As vicarious methods methods have their intrinsic uncertainties, a set of vicarious methods are operationally used in the IQC Cal/Val facility. The IQC Cal/Val facility contains among others the OSCAR (Optical Sensor CAlibration with simulated Radiance) vicarious calibration methods which exploit the reflected radiance over bright desert surfaces, Deep Convective Clouds (DCC), and  Rayleigh scattering. Furthermore, monthly lunar observations are scheduled to track the stability of the center camera. By applying different vicarious calibration methods and inter-comparing their respective results, systematic errors are reduced. Due to the relative large random error  associated with a single vicarious calibration result, statistical averaging over  several observations is needed.

The stability of the dark current values is monitored through nighttime observations over dark oceans in a prolonged image capture mode. The performance of the equalization coefficients to correct for inter-pixel variations over the Field Of View (FOV) is regularly analyzed through acquisitions over large homogeneous snow and desert sites.

At the conference we will give an overview of the calibration results obtained with the different vicarious radiometric calibration methods with a focus on the instrument response temporal evolution over its 3 years in orbit. Besides an overview of the IQC calibration acitivities, the results of an independent validation activity will be presented, in which PROBA-V observations are cross-compared against historical MERIS observations. Finally we will give an overview of  the changes made for the first reprocessing of the PROBA-V archive.

Quantifying and understanding long term environmental variability, including climatic trends, requires the comparison of observations made decades apart by different sensors. The reliability of such comparisons depends on ensuring the long-term stability of the data record, and a detailed understanding of the sources of uncertainty in historic, current and future sensors. Approaches from the discipline of metrology can assist Earth observation practitioners in developing defensible quantitative characterisation of errors and long-term stability.

Here we describe work within the H2020 project FIDUCEO (Fidelity and uncertainty in climate data records from Earth observations) project (www.fiduceo.eu) to generate both metrologically rigorous uncertainty budgets for the measurements made with historical sensors to determine Fundamental Climate Data Records (FCDRs): long term records of sensor level 1 products (radiance, reflectance or brightness temperature) and Climate Data Records: long term records of sensor level 2 products (geophysical parameters).

In particular we show, by looking at AVHRR, how level 1 product uncertainties can be understood using a bottom-up approach. We then consider, from this perspective, the correlations between the different sources of uncertainty and how this feeds into different correlation structures for the sea surface temperature climate data record (SST-CDR). We then compare this to the correlation-based uncertainty classification of the ESA-CCI SST project. This includes fully random effects (uncorrelated from pixel to pixel), fully systematic effects (biases that affect all measurements, though may degrade slowly over long timescales) and locally systematic effects that have strong correlation for neighbouring pixels, but less correlation as the distance between pixels increases. 

This complete understanding of uncertainty and correlations enables an accurate uncertainty analysis for the comparison of measurements made at different times, or in different locations, and for averaging over areas or times. Because systematic effects average differently from random effects, and can be (partially or fully) cancelled out in comparisons, it is essential to understand this correlation structure. 

In preparation of Sentinel-3 launch, historical tools developed for ocean colour validation have been upgraded to support OLCI calibration and validation activities. The synergy of MERMAID (http://mermaid.acri.fr) and ODESA (http://earth.eo.esa.int/odesa) have proven to be powerful tools to support ocean colour Cal/Val by providing easy access to in situ to satellite matchups, level 2 processing source codes and analysis tools to generate statistics and validation plots. MERMAID and ODESA processing and analysis packages have therefore been updated to support both land and water products Cal/Val.

Similar to water product validation, a series of permanent land monitoring station (for example, FluxNet and NEON, about 40 and 20 stations respectively) will be operated in MERMAID to allow land products validation. Throughout OLCI lifetime, in situ data coming from temporary field measurements will also be added to the database and therefore enrich the diversity of land cover representation in the validation framework. Core land validation sites as defined by Sentinel-3 MPC ESLs (New forest, Barrax, Valencia Alicante and Nebraska) will be continuously monitored. As previously implemented for water product validation the usual fiducial in situ stations (BOUSSOLE, MOBY, AERONET-OC, EGSL network) will be operated in MERMAID in addition to which in situ data coming from oceanographic cruises will be progressively added to enrich the diversity of in situ products. Apart from BOUSSOLE which also provide HPLC analysis of chlorophyll and IOP measurements, in situ data coming for fiducial acquisition station indeed mainly measure radiometry. Oceanographic cruises are therefore of prime importance if we are to validate on a larger scale products like IOPs, chlorophyll and suspended sediment load which are nominal OLCI products. For both land and water processing, ODESA software will provide users with the capacity to implement their own products, some of whom might be implemented in collaborative ground segments. It is therefore of prime importance to open MERMAID to a large variety of in situ measurement if we expect to have alternative products properly validated. Among the new born in the series of automated in situ measuring devices are the bio-Argo floats which measure chlorophyll fluorescence on drifting trajectory. This data are already operationally used for Level-3 product validation (http://seasiderendezvous.eu). They will also be operated in MERMAID to support Level-2 CalVal. In a closed future, similar devices will deploy radiometric sensors, opening the path to brand new datasets other areas of the ocean only scarcely ever instrumented. As performed for Land product validation, a selection of case 1 waters going from oligotrophic (South Pacific Gyre, South Indian Ocean … ) to eutrophic environment (Benguela and Peru upwellings …) will be constantly monitored.

A significant drawback of MERMAID system in the past was its time response to refresh the database. The system has now gained in reactivity with the automated download of in situ data. In situ data from permanent land and water stations are, when feasible, automatically downloaded and formatted to be ingested in the extraction and matchup generation routines. CalVal experts will therefore be provided with MERMAID extraction other land and water within a very short delay.