Copernicus: Sentinel-5 (Atmospheric Monitoring Mission) in LEO
Sentinel-5 is an atmospheric monitoring mission within the European Copernicus program, formerly the GMES (Global Monitoring for Environment and Security) program, jointly implemented by ESA and the EC (European Commission). The objective of the mission is the operational monitoring of trace gas concentrations for atmospheric chemistry and climate applications. It will provide accurate measurements of key atmospheric constituents such as ozone, nitrogen dioxide, sulphur dioxide, carbon monoxide, methane, formaldehyde, and aerosol properties.
Table 1: Copernicus is the new name of the former GMES program 1)
The space segment will be implemented as an imaging spectrometer to be flown on EUMETSAT's MetOp -SG (Second Generation) satellites. From a sun-synchronous LEO orbit, Sentinel-5 measurements will complement the Sentinel-4 GEO data over Europe and provide a daily global coverage at an unprecedented spatial resolution of 7 km x 7 km at nadir. 2)
While Sentinel-1, -2 and -3 are completing their development phases and are scheduled to launch until 2014, Sentinels-4/5 are foreseen to provide a synergetic, complementary data set for atmospheric monitoring starting in 2020.
The Sentinel-5 mission is a payload, consisting of a single instrument named UVNS; it will embark on a post-EPS (MetOp) spacecraft, i.e. MetOp-SG, and will be operated by EUMETSAT. 3) In November 2012, the EUMETSAT Council successfully concluded the approval process for the EPS-SG (EUMETSAT Polar System Second Generation) Preparatory Program with all 26 Member States having now firmly committed themselves. 4) 5)
A two-satellite architecture has been selected for MetOp-SG by ESA and EUMETSAT, namely MetOp-SG A and B, flying in the same sun-synchronous orbit. Unlike the current MetOp system of identical satellites operating in a relay, the MetOp-SG system envisages a pair of different satellites, each carrying a different but complementary suite of instruments. This will comprise a mix of instruments offering data continuity with improved performance and new instruments to meet the evolving demands of the meteorological community.
Altogether, the new MetOp-SG system concept features six satellites; the launch of the first one is planned for 2020. The overall system lifetime is 21 years, with each satellite designed to exceed an eight and a half year lifetime.
Table 2: Preliminary concept of the MetOp-SG program with candidate sensor complement 6)
A description of the MetOp-SG spacecraft will be provided when available.
Launch: A launch of Sentinel-5 A is planned for 2020.
Orbit: Sun-synchronous orbit, altitude = 817 km, inclination = 98.5º.
Sensor complement: (UVNS)
UVNS (Ultra-Violet /Visible/Near Infrared/Short Wave Infrared spectrometer):
The instrument is a pushbroom imaging spectrometer, directly imaging the entrance slit onto the Earth's surface. The principle is depicted in Figure 1. The entrance slit is providing the spatial coverage or swath in the ACT (Across-Track) direction, where the projected detector pixels define the spatial samples. The ALT (Along-Track) dimension is acquired by the satellite motion. The light from the slit is spectrally dispersed by a diffraction grating and imaged onto a two-dimensional detector. The spatial sampling along the track is defined by the detector timing. The dwell time, over which typically several exposures are co-added, defines the ALT spatial sampling distance (SSD). The mission will generate Level-1b data consisting of TOA (Top Of Atmosphere) Earth spectral radiance and extra-terrestrial spectral solar irradiance. Division of the Earth radiance spectra by the regularly measured solar irradiance yields the reflectance spectra, which feature the spectral signature of absorbing gases and scattering aerosols. Dedicated retrieval techniques are employed to infer the vertical column amounts of the absorbing and scattering species from these reflectance measurements (Ref. 2).
Figure 1: Principle of the Sentinel-5 pushbroom imaging dispersive spectrometer (image credit: ESA)
The Sentinel-5 mission is targeting a wide range of products for the atmospheric applications. It will provide vertical column amounts of O3 (stratospheric and tropospheric), NO2, SO2, CO, CH4, CH2O, as well as some profile information on O3 and aerosols. Due to the broad spectral range needed to observe all targeted molecular species, the instrument is a complex assembly of various spectrometers dedicated to subbands. In Figure 2, a sketch of an optical concept, developed during the phase 0 study, is depicted, showing a possible implementation with six spectrometers (UV-1, UV-2, VIS, NIR, SWIR-1, SWIR-3) sharing a common telescope.
Figure 2: Sketch of a proposed phase 0 concept of the Sentinel-5 UVNS instrument, showing various individual spectrometers sharing a common telescope (image credit: ESA)
1) Geometric requirements:
One of the most outstanding characteristics of the Sentinel-5 mission are manifested by an unprecedented spatial coverage and sampling. The scientific requirements demand full coverage of the entire globe with a daily revisit frequency. Figure 3 depicts the observation geometry and coverage. For a LEO mission at MetOp SG's envisaged orbit altitude of 817 km, daily coverage requires an average swath width of 2715 km (slightly varying due to the oblateness of Earth and orbit eccentricity). As can be seen from upper right panel, small gaps occur at the equator resulting in a revisit period of about two days, whereas daily revisit is reached beyond 14° latitude (see lower panel). The enormous swath width translates at instrument level into a large cross-track FOV of 108.4° (see upper left panel).
Figure 3: Sentinel-5 coverage requirements (image credit: ESA)
Legend to Figure 3: Upper left: Daily global coverage requires a wide swath (2715 km) and across-track field of view of 108.4°. Upper right: Swaths of two consecutive orbits, indicating small gaps at the equator. Below: Resulting revisit frequency map, showing the daily coverage within less than one day above 14° latitude.
Another characteristic of Sentinel-5 is the high spatial resolution of the trace gas measurements. At nadir, the area over which all targeted species are determined, will have a maximum size of 7 km x 7 km. Figure 4 shows a comparison between the ground pixel size of Sentinel-5 and its predecessor missions GOME, SCIAMACHY, and OMI. The high spatial resolution will enable more accurate detection of emission sources and provide an increased number of cloud-free ground pixels. The ground samples will be observed under equal sampling angles, which means that the across-track SSD (Spatial Sampling Distance) will vary over the swath and reach 35 km at its edges. The along-track (ALT) component of the SSD is determined by the dwell time and remains constant at 7 km. The MetOP-SG platform will perform continuous yaw steering maneuvers over the orbit to compensate for image distortion and misregistrations induced by Earth rotation.
Figure 4: Comparison of spatial resolution (ground pixel size) of Sentinel-5 (S5) with heritage missions (image credit: ESA)
Some atmospheric products are inferred from simultaneous measurements in different spectral bands, which have to be acquired over identical spatial samples (or ground pixels). This results in stringent co-registration requirements, allowing for inter-band deviation of only 10% of the ground pixel size (< 700 m at nadir) and driving optical as well as mechanical designs to minimize such errors. Apart from optimizing image quality (minimize keystone errors) this involves a careful design of the polarization scrambler, which can induce coregistration errors depending on the polarization state of the measured radiance.
2) Spectral requirements:
The large number of targeted atmospheric constituents can only be measured over an extremely broad spectral bandwidth, spanning from the UV (starting at 270 nm) to the SWIR spectral regions (up to 2385 nm). The simultaneously measured spectral bands are depicted in Figure 5. The spectral resolution varies from 1 nm in the UV1 (270-300 nm), used for retrieval of stratospheric O3 profiles, over 0.5-0.4 nm for the visible and NIR range, respectively, to 0.25 nm in the two SWIR bands. A spectral resolution element is sampled by 2.5-3.0 detector pixels in order to avoid spectral aliasing.
The broad and partially discontinuous spectral range dictates, that the light collected by the instrument's telescope must be spectrally split and distributed to a number of individual spectrometers. For each of them, a spectral knowledge of 1.5 pm (picometer) is demanded; the stability requirements (15 pm) call for a very robust design and an accurate on-ground calibration. The shape of the ISRF (Instrument Spectral Response Function) has to be known within 1% throughout the mission lifetime (7.5 years), which calls for an in-flight monitoring utilizing a spectral stimulus device. Technology solutions for meeting these demanding specifications are currently under investigation.
Figure 5: Sentinel-5 spectral band definition (spectral radiance of SWIR bands not in scale), image credit: ESA
3) Radiometric requirements:
The signal-to-noise (SNR) requirements are driving the instrument pupil size, and have been derived from sensitivity analyses for the retrieval of each individual molecular species. Which spectral ranges are driving the aperture size of the instrument depends on the chosen concept. A notoriously challenging region, especially in terms of straylight, is the UV2 region due to a variation of radiance levels over 3 orders of magnitude within a short spectral range (indicated in Figure 5). Special techniques, like graded filters may be necessary to cope with the enormous dynamic range. The high accuracy requirement for retrieval of NO2 columns is driving the SNR in the VIS region to 1500. In the SWIR regions, where CO, CO2, CH4 and H2O provide strong absorption signals, a known limiting factor is the readout noise of the MCT (Mercury Cadmium Telluride) CMOS detectors.
While SNR limits are in principle always achievable by increasing the pupil size (and with it size and mass) of the instrument, the demanding requirements for relative and absolute radiometric accuracy are pushing technology to the limits of feasibility. The relative spectral radiometric accuracy (RγRA) describes spurious features in the measured spectra, which propagate into the retrieval error. There are many contributors to this error, including speckles from the sun calibration diffuser, spectral straylight and polarization scrambler effects. One major contributor is also the instrument's sensitivity to the strongly varying polarization of the signal.
In the NIR region, the degree of polarization of the TOA Earth radiance can vary rapidly with wavelength from nearly 0 (in the continuum) to almost 100% (in the center of the O2 A-band). The Sentinel-5 requirements limit the polarization sensitivity of the instrument to below 0.5% in the UV, Vis and NIR, which is only achievable by utilizing a spatial pseudo-depolarizer as an optical component, which de-polarizes the light collected by the telescope.
Another contributor to spectral errors analyzed within the S5 feasibility studies are radiometric artifacts arising from the spatially heterogeneous nature of the radiance emanating from the Earth surface (due to irregular cloud cover or albedo variations within a spatial sample). The varying radiance levels across the slit (in ALT direction) result in an inhomogeneous illumination of the entrance slit, and consequently in a distortion of the ISRF (Instrument Spectral Response Function).
The upper right panel of Figure 6 depicts a simulated ISRF, which is deformed by realistic scene heterogeneity as inferred from MODIS imager data. The radiometric error resulting from the absence of knowledge in the ISRF distortions is referred to as PN (Pseudo Noise), because it behaves like random noise in the spatial direction whereas it correlates strongly with the measured signal along the spectral direction. The simulated PN in the NIR-2 for the analyzed scene is also plotted in Figure 6. Error levels of several percent are incompatible with the requirement on RγRA, which allow maximum errors of only 0.05% in the VIS and NIR bands.
Figure 6: Pseudo noise and distorted ISRF due to across-slit scene heterogeneity (image credit: ESA)
Therefore, dedicated correction techniques have to be developed. A significant part of the PN error results from a spectral shift of the ISRF barycenter, seen in Figure 6, and can largely be corrected by suitable spectral calibration techniques. However, the distortion of the ISRF shape still exceeds the 1% required for Sentinel-5 and additional measures may be necessary for PN mitigation. A prediction of the distorted ISRF is possible by means of temporally highly sampled radiance measurements, which provide sub-SSD spatial resolution in the ALT direction. The resulting possible software correction is supported by dedicated requirements for Sentinel-5, requesting so-called "small-pixel" data. Alternatively, it is possible to correct the non-uniform slit illumination by means of a slit homogenizer.
Special hardware developments:
SH (Slit Homogenizer):
Special components, referred to as slit homogenizers, are being developed in order to mitigate radiometric errors arising from naturally occurring (across-slit) scene heterogeneity. One possible realization of a slit homogenizer, shown in Figure 7, is a three-dimensional slit, consisting of two parallel, highly reflective mirrors. The incident beam at the SH entrance (coming from the telescope) is reflected several times in ALT (across slit) direction between the mirror surfaces, which "scrambles" the spatial information in this direction. In order to maintain the ACT spatial resolution, an anamorphosis has to be introduced, focusing the ALT component of the beam from the telescope on the entrance of the SH and the perpendicular ACT component on its exit.
Figure 7: Sketch of a slit homogenizer for mitigation of across-slit (ALT) scene heterogeneity (image credit: ESA)
A perfect SH would equally distribute the radiation incident at the entrance across the exit of the device, resulting in a uniform illumination pattern at the exit. The homogenizing effect can be quantified by transfer functions, which represent the across-slit intensity distribution resulting from a Dirac-type stimulus at the entrance of the SH. For an ideal device, the transfer functions are boxcar shaped for any across-slit position of the incident beam. In reality, they deviate from this behavior due to interference between mirror reflections. The left panel of Figure 8 shows a 2D color representation of the across-slit energy distribution (vertical) as a function of the position of the stimulus at the SH entrance (horizontal). A cross section of this plot depicted in the right panel of Figure 8 illustrates the non-ideal transfer functions. Nevertheless, first results indicate that SH devices reduce PN levels by roughly an order of magnitude.
Figure 8: Transfer function of a slit homogenizer (image credit: ESA)
Legend to Figure 8: The 2D color plot on the left panel shows the slit exit illumination (vertical axis) for a Dirac input illumination as a function of its position across the input of the homogenizer (horizontal axis). The right panel shows a few the slit exit energy distributions, which are cross sections of the left plot. To obtain the ISRF from these illumination patterns they have to be convolved with the spectrometer and detector point spread functions.
The stringent requirements on polarization sensitivity and RγRA necessitate the deployment of a pseudo-depolarizer or scrambler. The most suitable type is the DBCP (Dual Babinet Compensator Pseudo-depolarizer), consisting of four wedges of birefringent material arranged in pairs (Figure 9). The wedges of a pair are bonded with crossed crystal axes, and the first pair is rotated by 45° w.r.t. the second one. Light beams passing through this device experience polarization dependent phase delays, which vary over the pupil and the resulting polarization states at the exit largely average out.
The ideal position for the scrambler would be in front of the telescope so that all optical components are illuminated with de-polarized light. However, this is impossible for the Sentinel-5 instrument, because of the large ACT FOV and the corresponding large incidence angles (up to 54°) resulting in non-acceptable spectral oscillations in the residual polarization sensitivity. The telescope mirrors placed before the scrambler will therefore contribute to the instrument's polarization sensitivity.
A known feature of DBCP scramblers is that polarized light passing through the four wedges is split into four beams due to birefringence. This effect is a direct consequence of the depolarization and cannot be avoided. Due to the geometry of the device, the four beams are arranged in a parallelogram, resulting in a "diamond spot pattern" at the detector. The stronger the de-polarization power of the DBCP, the greater is the separation of the four spots. As a consequence of the beam separation, a detector element traced through the instrument is imaged four-fold onto the Earth surface, which effectively reduces the spatial resolution. The latter is described in terms of the integrated (or ensquared) energy, representing the fraction of photons originating from a given ground pixel. The requirements for polarization sensitivity and integrated energy are in conflict, one calling for high and the other for low depolarization power, and a balance between them has to be found by careful design.
Figure 9: Sketch of a Dual Babinet Compensator Pseudo-depolarizer (polarization scrambler), image credit: ESA
The energy distribution between the four spots of a classical DBCP scrambler's diamond pattern depends on the polarization state of the incoming light, which varies strongly not only with the viewing geometry, but also with wavelength. This gives rise to a spectrally dependent effective pointing, inducing a co-registration error between spatial samples probed at different wavelength. In order to comply with the stringent co-registration specifications for single- (10% of SSD, 700 m at nadir) and multi-band (20% of SSD) retrievals, modifications of the DBCP scrambler design are being developed by the Sentinel-5 industrial primes to reduce polarization independent pointing.
Diffraction gratings are the preferred dispersing components for wide-swath push-broom spectrometers and for atmospheric applications requiring a spectral resolution around 0.5 nm. However, they are also a significant source of straylight and polarization sensitivity. Therefore special care has to be taken regarding the choice of grating technology for each individual spectral band. In the UV, Vis and NIR, a number of solutions are available, including convex blazed holographic, binary transmission, Echelle and prism gratings. For the two SWIR regions, the resolution of 0.25 nm in combination with the stringent limitations on volume and mass make the use of immersed gratings mandatory. These devices are reflective gratings formed on a prism, where the incident light is dispersed inside the substrate material.
Refraction at the exit surface further increases the angular dispersion, which allows reaching high spectral resolution with a much smaller grating than in a conventional setup. As a rule of thumb, the reduction in volume resulting from using an immersed grating w.r.t. a conventional one, is on the order of n3, with n being the refractive index of the substrate material. This advantage is essential for Sentinel-5 and its stringent envelope limits (1.2 m x 1.1 m x 1.0 m). Among possible substrate materials with high refractive index in the SWIR, silicon (n=3.4) offers the most mature technology and is the baseline for the Sentinel-5 Precursor mission.
The final key component encountered by the light is the focal plane assembly housing the detector. The broad targeted spectral range requires individual technology solutions for each spectrometer. In the UV, VIS and NIR regions, frame transfer CCDs (Charge-Coupled Devices) based on silicon are usually the technology of choice. Scientific CCDs feature high quantum efficiency and full well capacity, and good linearity over a wide dynamic range. The technology is mature and available in a variety of formats, which can be shaped according to the needs. From the principle of the pushbroom spectrometer concept, it is clear that detectors with fast readout capability are required, to maximize the detection time and avoiding gaps in spatial coverage. In case of CCD detectors, this is achieved by a frame transfer, in which the image acquired over the dwell time is transferred into a storage region.
From the storage region the image is successively read out while the next image is already acquired in the exposure region. The readout can be performed both in spectral (across-slit) and spatial (along-slit) direction, which in turn has implications on the video chain layout. Since the detector pixels are continuously illuminated during the charge transfer, the frame transfer has the disadvantage to create a smear effect on the signal, which needs to be corrected.
Since silicon is not responsive for wavelengths beyond 1000 nm, alternative detection media have to be considered in the SWIR spectral regions, like InGaAs or HgCdTe (MCT). MCT CMOS devices are a viable solution due to their high industrial maturity. While U.S. companies have accumulated a considerable advance in this technology, European detectors are only available in relatively small formats (256 x 1024 pixels) and with large pixel pitch (30 µm), which imposes constraints on the optical design. ESA has initiated pre-development activities with European companies on new generation SWIR detectors offering formats of 1024 x 1024 pixels with pixel sizes < 20 µm. Although these pre-developments are not yet completed, they are currently baselined as the detector solution for the Sentinel-5 mission.
1) “Copernicus: new name for European Earth Observation Programme,” European Commission Press Release, Dec. 12, 2012, URL: http://europa.eu/rapid/press-release_IP-12-1345_en.htm
2) Bernd Sierk, Jean-Loup Bézy, Jérôme Caron, Roland Meynard, Ben Veihelmann, Paul Ingmann, “The GMES Sentinel-5 mission for operational atmospheric monitoring: Status and Developments,” Proceedings of the ICSO (International Conference on Space Optics), Ajaccio, Corse, France, Oct. 9-12, 2012, paper: ICSO-069
3) Eleni Paliouras, “The GMES Space Component and Service Support Activities,” Padua, Italy, Sept. 16, 2010, URL: http://www.fp7-space.eu/infoday2010/16/05_Paliouras-ESA_GMES.pdf
4) “EUMETSAT Council approves EPS-SG Preparatory Programme and extends Indian Ocean service,” EUMETSAT Press Release, Nov. 16, 2012, URL: http://www.eumetsat.int/Home/Main/News/Press_Releases/825429?l=en
5) Paul Ingmann, Yasjka Meijer, Ben Veihelmann, “Status Overview of the GMES Missions Sentinel-4, Sentinel-5 and Sentinel-5p,” Proceedings of ATMOS 2012 - Advances in Atmospheric Science and Applications, Bruges, Belgium, June 18-22, 2012, ESA SP-708, Nov. 2012, URL of presentation: http://congrexprojects.com/docs/12m06_docs2/4_the-atmosphere-related-sentinels_ingmann_asc2012.pdf
6) “MetOp Second Generation instruments,” ESA, March 9, 2011, http://www.esa.int/esaLP/SEM95PXTVKG_LPmetop_0.html
The information compiled and edited in this article was provided by Herbert J. Kramer from his documentation of: ”Observation of the Earth and Its Environment: Survey of Missions and Sensors” (Springer Verlag) as well as many other sources after the publication of the 4th edition in 2002. - Comments and corrections to this article are always welcome for further updates.
The Sentinel series: