International Conference on Space Optics—ICSO 2018

Chania, Greece
9–12 October 2018

Edited by Zoran Sodnik, Nikos Karafolas, and Bruno Cugny

The European CO2 Monitoring Mission: observing

anthropogenic greenhouse gas emissions from space
Bernd Sierk
Jean-Loup Bézy
Armin Löscher
Yasjka Meijer

International Conference on Space Optics — ICSO 2018, edited by Zoran Sodnik,

Nikos Karafolas, Bruno Cugny, Proc. of SPIE Vol. 11180, 111800M · © 2018 ESA
and CNES · CCC code: 0277-786X/18/$18 · doi: 10.1117/12.2535941

Proc. of SPIE Vol. 11180 111800M-1

Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 09 Mar 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 ICSO 2018 Chania, Greece
International Conference on Space Optics 9 - 12 October 2018

The European CO2 Monitoring Mission: Observing anthropogenic greenhouse

gas emissions from space
Bernd Sierka, Jean-Loup Bézya, Armin Löschera, Yasjka Meijerb
a
ESA-ESTEC b; Rhea for ESA-ESTEC

ABSTRACT

Responding to plans of the European Commission for extending the observation capabilities of the Copernicus
programme, the European Space Agency (ESA) has initiated Phase A industrial (technical feasibility) studies for several
new space-borne Earth Observation missions. High priority is given to a constellation of LEO satellites in Sun-
synchronous orbit with the purpose of observing anthropogenic carbon dioxide (CO 2) emissions [European Commission,
2017]. The observing system shall acquire images of CO2 concentration in terms of dry air column-averaged mole
fractions (XCO2), providing complete global land coverage at high spatial resolution (4 km2) within five days. The
demanding requirements call for a payload comprising a combination of multiple instruments, which perform
simultaneous measurements. The XCO2 is inferred from reflectance measurements in the Near-Infrared (NIR) and
Short-Wave Infrared spectral regions (SWIR). This requires at least three spatially co-registered push-broom imaging
spectrometers, measuring spectral radiance and solar irradiance in the NIR (747-773 nm), SWIR-1 (1595-1675 nm) and
SWIR-2 (1990-2095 nm) at moderate spectral resolving power (R~5000-7000). In addition, the observations for CO 2
concentration will be complemented by Differential Optical Absorption Spectroscopy (DOAS) measurements of nitrogen
dioxide (NO2) over the same area. The NO2 measurements in the visible region (400-500 nm) are expected to serve as a
tracer for plumes of high CO 2 concentration resulting from high temperature combustion, which will facilitate plume
identification and mapping. The third component of the payload is a multiple-angle polarimeter (MAP), performing
high-precision measurements of aerosol (and cloud) properties. Its measurements of polarized radiance under various
observation angles are expected to reduce XCO2 bias error and significantly increase the yield of useful retrievals from
the NIR and SWIR spectra. The complex observation architecture, involving multiple instruments and platforms, call for
optimized observational requirements, driven by the primary goal of detecting and quantifying point-sources of
greenhouse gas emissions. In particular, high single-sounding precision is essential for identifying plumes of elevated
CO2 concentration from instantaneous image acquisitions without regional and temporal averaging. This translates into
stringent requirements for Signal-to-noise ratio (SNR), as well as spatial co-registration and spectral stability, which
drive the instrument design. The presentation will introduce the different elements of the candidate Copernicus mission,
in view of the ambitious mission goals. The payload components and observation requirements are addressed with
special emphasis on the derivation of the SNR and spectral resolution requirements, which determine the instrument
sizing.
Keywords: Imaging spectrometer, Copernicus programme, anthropogenic CO2, greenhouse gas observation

1. INTRODUCTION
In 2015 world leaders gathered in Paris to agree on how to combat climate change, as it is now well established that
anthropogenic emissions of greenhouse gases need to be significantly reduced [1]. International treaties and agreements,
aiming at reducing emission from fossil fuel combustion have been singed by many nations. While the targeted emission
reductions are based on voluntary Nationally Determined Contributions (NDCs), there is a future demand of actionable
information on the effectiveness of implemented measures and regulations. In this context, the European Commission
has formulated a system architecture with Monitoring and Verification Support capacity[2]. An earlier report identified
the need of a multi-component observation system for quantification and monitoring of anthropogenic greenhouse gas
emissions requires, involving in-situ and remote sensing observations on various spatial domains [3]. One of the main
pillars, besides on-ground measurements relies on space-borne observations of the concentrations of the most important
anthropogenic greenhouse gases, carbon dioxide (CO2) and methane (CH4). The high-level observation requirements call
for a multi-satellite constellation with imaging capability at high spatial resolution, global coverage and frequent re-visit.

Proc. of SPIE Vol. 11180 111800M-2

Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 09 Mar 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 ICSO 2018 Chania, Greece
International Conference on Space Optics 9 - 12 October 2018

Measurements of column averaged dry air mole fractions of carbon dioxide (XCO2) and methane (XCH4) have been
pioneered by the SCIAMACHY instrument on-board ESA’s Envisat satellite[4]. After the launch failure of NASA’s
Orbiting Carbon Observatory (OCO,[5]) in 2009, JAXA’s GOSAT became the first dedicated mission to target CO2 and
CH4 measurements from space[6]. GOSAT performs measurements of top-of-atmosphere (TOA) spectral radiance by
means of a Fourier Transform Infrared Spectrometer (FTIR), with large distances between single soundings (X km). In
August 2014 NASA successfully launched OCO-2 [7] and has since been providing continuous soundings of XCO2 at
high spatial resolution (~ 3 km2). The OCO-2 instrument is a three-band grating spectrometer operated in push-broom
mode. Despite of its narrow swath width (~20 km) and incomplete global coverage, OCO-2 has demonstrated the
capability of imaging point sources, such as power plants [8]. In December 2016, the Chinese TanSat mission was
launched[9], which has similar spatial resolution and coverage, and first results have recently been published[10].
Currently, there are several future missions under preparation, most notably MicroCarb (led by CNES) [[11]] and the
geostationary GeoCarb (University of Oklahoma) [12], which also feature high spatial resolution modes for point-source
observation. Based on the results of previous and on-going missions and instrument studies, there is now considerable
heritage of space-borne greenhouse gas observations to design a system dedicated to anthropogenic CO2 monitoring.
In this context, the European Space Agency (ESA) has initiated technical feasibility (Phase-A) studies with two
industrial consortia led by Airbus Defense and Space and OHB, respectively. The studies are part of a series of Phase-A
studies aiming at extending the capabilities of the Copernicus programme. In this frame, a potential future CO 2
monitoring mission, in the following referred to as CO2M, could be a candidate for a future Sentinel mission. In
preparation and support of the industrial activities, two scientific studies with the objectives of identifying the
observation requirements for a constellation of satellites equipped with a suite of instruments optimized for
anthropogenic CO2 monitoring. The SMARTCARB study performed by a consortium led by Swiss Federal Laboratories
for Materials Science and Technology (EMPA), is addressing potential synergies of measurements for observing and
quantifying CO2 point sources such as large cities and power plants. Also, the impact of different satellite specifications
(e.g. overpass time and spatial coverage) are investigated. The AEROCARB study, performed by a consortium led by the
Space Research Organization of the Netherlands (SRON), investigates the use of auxiliary measurements to better
constrain the atmospheric aerosol and cirrus cloud distribution over the sampled areas, which was identified to be a
significant limitation for accurate XCO2 retrieval.
This paper presents the driving mission and observation requirements identified in these and earlier studies, as well as
the implications for the instrument architecture. It also addresses the translation from Level-2 (concentrating on XCO2),
to calibrated radiance measurements (Level-1b), with special emphasis on the driving signal-to-noise ratio (SNR)
requirement.

2. MISSION OBJECTIVES AND REQUIREMENTS

The principle science goals of past and current space-borne missions for observation of XCO2 and XCH4 are related to
the global carbon cycle and the identification and quantification of natural sources and sinks of greenhouse gases.
Anthropogenic fluxes of CO2 are typically an order of magnitude weaker than biogenic fluxes and therefore more
difficult to quantify and separate from the latter. Emission from fossil fuel combustion can only be detected in the
vicinity of strong point sources. Significantly elevated values for XCO 2 in excess of a few percent of the background
value (~390 ppm) can be detected within down-wind plumes of coal-fired power plants, and potentially large cities.
Since the well-mixed gas CO2 is quickly diluted after emission, the expected signal of enhanced XCO 2 depends on the
size of the observed spatial samples (spatial resolution) and the wind speed.
Space-borne quantification and monitoring of anthropogenic greenhouse gas emission sources has not yet been
established as an operational observing method. However, requirements for a space-borne component of an operational
monitoring system have been identified in an expert study funded by the European Commission [3]. According to the
report, monitoring of anthropogenic carbon emissions require a satellite constellation, observing XCO2 at high precision
(< 0.7ppm) and low systematic bias (< 0.5 ppm) with global coverage every 2-3 days. ESA has initiated scientific
support studies to further consolidate observational requirements with particular emphasis on the precision (repeatability)
and accuracy (systematic bias) of the XCO2 observations. The precision is mainly determined by the signal-to-noise ratio
(SNR) of the measured TOA spectral radiance, and pseudo-noise from spatial co-registration and radiometric scene non-
uniformity. Systematic biases may be introduced by instrumental errors like straylight, polarization sensitivity and
detection non-linearity, as well as by geophysical effects, like the uncertainties in the aerosol distribution above the
scene. The instrumental and geophysical error sources affecting the retrieval of XCO2 from space-borne measurements

Proc. of SPIE Vol. 11180 111800M-3

Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 09 Mar 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 ICSO 2018 Chania, Greece
International Conference on Space Optics 9 - 12 October 2018

of spectral radiance call for an observing system with multiple sets of combined observations. In the following we
summarize the mission requirements for these different sets of observations.

3. PAYLOAD ELEMENTS AND OBSERVATION MODES

3.1 Payload components
Recognizing the challenging task of operational, quantitative monitoring of greenhouse gas sources and sinks and
distinguishing anthropogenic from biogenic emissions, the CO2M mission advisory group recommended a combination
of multiple payloads. The principal instrument will measure Top-Of-Atmosphere (TOA) spectral radiance in three
continuous spectral intervals: The Near-Infrared (NIR) encompassing the oxygen (O2) A-band, and two Short-Wave
Infrared (SWIR) bands covering three CO2 and one CH4 absorption band. The band definition is further detailed in
Section 5.1. The CO2 instrument will be based on a push-broom imaging spectrometer concept for simultaneous and co-
located observation in all spectral bands. The measurement principle is described in [13] and recalled in Figure 1.

4.
Figure 1: Principle of a push-broom imaging spectrometer. The entrance slit of the spectrometer is imaged onto the Earth surface,
defining the swath width in across-track direction. The instrument spectrally disperses the slit image in along-track direction, which is
sampled by the spacecraft motion during integration time. The products for each spatial sample are inferred from radiance spectra
imaged onto the 2D array detector.

In support of the XCO2 retrieval, the need of additional measurements of the aerosol and cirrus cloud distribution above
the sampled area was identified. The general observation principle consists in the measurement of polarized radiance
under various viewing directions. This component of the instrument suite is referred to as a Multiple-Angle-Polarimeter
(MAP). The complementary MAP observations can be performed by different types of instruments, which are closer
described in Section 5.6.
In order to improve the capability for emission plume identification and mapping, an additional imaging spectrometer
operating in the visible (VIS) spectral region is required. The objective is to measure column densities of tropospheric
nitrogen dioxide (NO2). Such observations can be used as a tracer for CO2 from anthropogenic high-temperature
combustion processes, as shown in Figure 2 and further explained in Section 5.5. Extensive heritage exists for space-
borne NO2 mapping from previous and on-going atmospheric chemistry missions, like the recently launched Sentinel-5P
[15]. However, the requirements for the CO2M mission call for an unprecedented spatial resolution of such observations
(2 km x 2km).
Finally, the expert team identified the need for cloud-flagging and characterization for all acquired spatial samples.
Strong cloud cover can already be detected from observations of the three previously described payload components.

Proc. of SPIE Vol. 11180 111800M-4

Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 09 Mar 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 ICSO 2018 Chania, Greece
International Conference on Space Optics 9 - 12 October 2018

However, scattering within small, sub-sample sized clouds as well as subvisible cirrus clouds lead to light-path
modulation, which is difficult to distinguish from enhanced CO2 concentration. The strong sensitivity of the XCO2
retrieval to cloud contamination calls for a cloud-imager capable of detecting small tropospheric clouds and cirrus cover
at a spatial resolution exceeding those of the other payload components. The resulting requirements are summarized in
Section.

Figure 2: Example of plume detection for several power plants in East Germany. The simulated images depict XCO2 (upper row) as
well as NO2 columns(lower row) at CO2M spatial resolution (4 km2). The upper left image assumes a random error of of 0.5ppm and
the upper right plot the correposnding image for 1.0 ppm precision. The lower left image shows the strong correlation of the NO2
plumes with those of CO2 emission. The lower right image depicts the same NO2 scene coarser spatial resolution comparable to
Sentinel-5 (7km x 7 km), demonstrating the loss of the capability to distinguish the individual emission sources. Courtesy EMPA.

4.1 Observation modes and coverage

All instruments will perform simultaneous and spatially co-located observations with different spatial co-registration
requirements. Various observation modes under different geometric conditions are foreseen. The most important one is
the nadir mode, in which the field-of-views of all instruments are directed towards the sub-satellite point (apart from the
off-nadir viewing angles of the MAP instrument). The second principal Earth observation mode enables observations in
the Sun-glint geometry. In this mode, the CO2 instrument will be directed pointed the area of specular reflection of
sunlight, which can be achieved either via a scan mirror or by a platform maneuver. Due to the very low reflectance of
water bodies in the SWIR bands, measurements over oceans are only possible in Sun-glint mode. In addition, Sun-glint
observations over land may be useful, especially over lakes and snow-covered or poorly illuminated regions. The two
Earth observation geometries are complemented by calibration modes, which enable in-flight radiometric
characterization using direct observation of the Sun (e.g. via solar diffuser), and geometric calibration using the Moon.

Proc. of SPIE Vol. 11180 111800M-5

Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 09 Mar 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 ICSO 2018 Chania, Greece
International Conference on Space Optics 9 - 12 October 2018

Targeting an operational global observing system, stringent observation requirements for coverage and re-visit have been
established. The expert team of the mission advisory group recommended complete geometrical land coverage within 2-
3 days pole-ward of 40° latitude, and within 5 days at the equator. Since the field-of view, and consequently the swath
width of the above instruments is limited by technical (detector size) constraints as well as retrieval limitations (Solar
Zenith Angle (SZA)), the coverage achievable with a single spacecraft call for a multi-platform constellation.

5. SYSTEM REQUIREMENTS
5.1 Spectral requirements for XCO2 observations
Accurate determination of XCO2 from simultaneous three-band retrieval has been demonstrated with measurements from
OCO-2 and GOSAT [e.g. [16]]. Observations of the oxygen (O2) A-band in the NIR spectral region are required to
determine the effective light propagation in the radiative transfer model used in the retrieval algorithm, making use of the
known concentration of the well-mixed oxygen. In the SWIR spectral region, rotation-vibrational molecular transitions
of the CO2 molecule give rise to absorption lines concentrated in spectral intervals around 1.6 µm, 2.0 µm and 2.04 µm
(see Figure 3). The heritage missions GOSAT and OCO-2 sample these regions in relatively narrow spectral intervals at
a spectral resolution sufficient to resolve individual absorption lines (R>20.000). While high spectral resolution is
advantageous for reducing systematic biases (e.g. from unknown aerosol distribution), it handicaps high signal-to-noise
(SNR) observation because of the lower spectral sampling interval per pixel. For contiguous mapping of CO2
concentration around point sources and their local vicinity, large SNR of the measured radiance, maximizing single-
sounding precision of XCO2, has higher priority than for climate missions primarily targeting the stronger biogenic
fluxes. Low measurement noise is especially important for instantaneous imaging of emission plumes, which depend on
actual wind and aerosol conditions at the time of the over-flight.

Spectral Spectral Resolution Spectral Sampling

Spectral Range [nm]
Band ID Δλ [nm] Ratio
NIR 747 – 773 0.12 3.0
SWIR-1 1590 – 1675 0.30 3.0
SWIR-2 1990-2095 0.35 3.0
Table1: Spectral band definition for the CO2M mission.

Figure 3: Top-of-atmosphere radiance spectra of the three bands used for XCO2 retrieval, both incident monochromatic and at spectral
resolution of the CO2-instrument.
In this context the requirements for spectral bandwidth, resolution and SNR have been derived in order to optimize the
CO2M observations with respect to the primary mission goal of point source monitoring. A large number of retrieval
simulations have been performed at different resolutions, bandwidths and SNR. The results indicate optimum imaging
performance for medium spectral resolution around 0.3 nm (R~5000), while the bandwidth needs to be sufficiently wide
to completely cover the absorption bands and the continuum at their spectral boundaries. The analysis confirmed that the
loss of information by reduced spectral resolving power is compensated by the larger bandwidth (more CO2 lines

Proc. of SPIE Vol. 11180 111800M-6

Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 09 Mar 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 ICSO 2018 Chania, Greece
International Conference on Space Optics 9 - 12 October 2018

measured) and correspondingly higher SNR (see below). The spectral band requirements are summarized in Table 1, and
the spectral radiance at instrument resolution for the geophysical reference scenario is depicted in Fig. 3.
The retrieval simulations performed for optimizing spectral resolution and bandwidth were also used to derive the SNR
requirements, which determine the sizing of the instrument. As mentioned above, the lower spectral resolving power
targeted for CO2M in comparison with heritage missions needs to be compensated by correspondingly higher SNR. The
noise level of the Level-1b radiance measurements directly affects the precision of the Level-2 product (XCO2 in ppm),
which in turn limits the detectability of point sources the accuracy of CO2 flux inversions. Therefore the derivation of the
SNR requirement started from using flux inversion simulations for a range of assumed XCO 2 precisions. Figure 2 depicts
simulated images of XCO2 and NO2 over an area in East Germany, where several coal-fired power plants are located.
The imager in the upper row of the figure represent the same geophysical scenario measured with different precision
(noise) levels in XCO2 at the spatial resolution of the CO2M instrument (4 km2).
The plots give an impression of the spatial extent and variability of typical emission plumes, as well as on the impact of
retrieval noise on the capability to distinguish the plume from the background concentration.

5.2 Signal-to-noise requirements for XCO2 observations

For the purpose of mission requirement derivation, a geophysical scenario reference scenario was defined, corresponding
to an observation geometry with a moderately low solar zenith angle of 50 and relatively low typical for albedo
vegetation. In the following this geophysical reference scenario is referred to as VEG50. With flux inversion simulations
over this scene at the given spatial resolution of 4 km2 it was established, that an XCO2 single-sounding retrieval
precision of 0.7 ppm is required for sufficiently accurate quantification of point source emissions. As this represents the
maximum tolerable random error of the retrieval, a goal precision of 0.5 ppm was established, providing for margin for
other error contributors, like pseudo-noise from spatial co-registration or scene non-uniformity.
These Level-2 precision requirements were linked to the radiometrically calibrated (Level-1b) measurements via a large
number of retrieval simulations with generalized instrument parameters. The basic assumption is that the signal-to-noise
ratio of the instrument can be expressed by a simple combination of two noise components, and for any measured
spectral radiance L is expressed by:
A× L
SNR =
A×L + B , (1)
The parameter A represents the signal-dependent shot-noise, and depends on various instrument parameters:
A = h ×T × Dl ×QE ×tint × Nbin,X , (2)
where η is the etendue of the instrument, T the total transmission; Δλ the spectral sampling interval, QE the quantum
efficiency of the detector. Nbin,X represents the number of detector pixels co-added to form the across-track field of view
of a spatial sampling, and tint the exposure time over which it is acquired (defining the along-track sampling distance).
The second parameter of Eq. (1) encompasses the signal-independent instrumental noise sources, e.g. from kTc noise of
the read-out circuit or digitization noise. In our analysis it is modeled as

B = N bin,X × ( I dark + ITb ) × tint + ( N 2AD + N 2RO + N VC

2
) × Ntemp , (3)
where Idark: is the dark current, ITb: shot noise from background thermal emission, NRO the detector read-out noise, NAD
thedigitization noise, and NVC the video chain noise. In the case of several read-outs per integration time, Ntemp represents
the temporal sampling factor.
While equations (1) – (3) are generally valid for simulating the SNR of any push-broom imaging spectrometer, the
number of parameters indicates a large parameter space, which renders the task of deriving a generalized SNR
requirement a complex task. However, by assuming a realistic range of values for parameters A and B, computed from
Eq. (2) – (3), and using Eq. (1) to simulate the noise of the reference spectra (see Figure 3), we can derive the
dependencies of XCO2 retrieval precision from signal-dependent and –independent noise components. As will be shown

Proc. of SPIE Vol. 11180 111800M-7

Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 09 Mar 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 ICSO 2018 Chania, Greece
International Conference on Space Optics 9 - 12 October 2018

below, from these we can find pivotal tuples of spectral radiances and associated SNR values, which correspond to the
required Level-2 precision.

Figure 4: a) left: Simulated XCO2 precision as a function parameter A, representing the signal-dependent shot noise component, for
various B-values (signal-independent noise component). The blue horizontal lines represent the threshold (0.7ppm) and goal (0.5ppm)
precision requirements. The intersection of the lines with the curves define combinations (A,B), which comply with the Level-2
requirement. Right panel: SNR performance as a function of incident spectral radiance for the identified (A,B) combinations from the
left panel. The curves describe the SNR dependence of compliant instruments with different signal-independent noise (e.g. from
different read-out noise). The curves intersect in one point (indicated by the red lines), which defines the required SNR at a reference
radiance Lref.

Figure 4a (left plot) shows a plot of simulated XCO2 precision as a function of parameter A for a family of curves
corresponding to different signal-independent noise levels. Each single curve represents an assumed instrument with a
combination of dark- and readout-noise yielding its value of B, and shows the dependence of the resulting XCO2
precision from the signal-dependent parameter A. For a given instrument setup with fixed detection parameters (e.g.
read-out and video-chain noise, etc.), the curves in Fig. 2a identify the required value of A for any desired Level-2
precision. If all other instrument parameters in Eq. (2) are fixed (transmission, spectral resolution, quantum efficiency,
integration time, and over-sampling), the A value can be translated into the pupil size needed to reach the required
measurement precision. The horizontal lines in Figure 4a indicate the targeted threshold and goal precisions (0.7 ppm and
0.5 ppm, respectively). The intersection of these lines with the performance curves yield combinations of A and B, for
which the required precision is reached. Therefore, each pair (A,B), defined by the intersection of the threshold (resp.
goal) line, corresponds to an instrument yielding exactly 0.7 ppm (resp. 0.5 ppm) precision in XCO2.
The compliant instruments identified this way are characterized by their SNR dependence from incident radiance
according to Eq. (1), which is plotted in Figure 4b (right plot). Each such instrument described by a selected (A,B)
combination yields the exact same Level-2 precision for the reference scenario (VEG50). However, those corresponding
to low values of B exhibit slightly higher SNR at lower radiances, but lower SNR at brighter radiances, than the SNR-
curves for large values of B. This is expected, as in the low radiance regime the signal-independent noise component is
more important, while at high radiances the SNR is dominated by shot noise. Instruments with relatively high read-out
noise (large B) require a greater étendue (e.g. large entrance pupil size) to compensate, and therefore perform better at
the bright end of the dynamic range. Because of the described behavior of the SNR-curves in Figure 4b they intersect
with each other at a point, where the signal-independent (read-out noise) dominated regime passes into the shot-noise
domain. At the radiance level Lref defined by this point, all instruments setups selected by the intersection of the
threshold line in Figure 4a measure with the same SNR performance (SNRref). This pivotal point (Lref, SNRref) is used to
define a simple SNR requirement, as any instrument measuring with SNRref at Lref will yield a precision equal to the
required 0.7 ppm (resp. 0.5 ppm).

Proc. of SPIE Vol. 11180 111800M-8

Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 09 Mar 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 ICSO 2018 Chania, Greece
International Conference on Space Optics 9 - 12 October 2018

The derivation of the SNR requirement described above is performed simultaneously for the SWIR-1 (1595nm-1675nm)
and SWIR-2 (1990nm - 2095nm) bands. It turns out that in both SWIR bands the value for SNRref is the same and fixed
by the required XCO2 precision, while the value for Lref depends on the required spectral resolution. The values found
for the CO2M mission are listed in Table X. It is important to note that the derivation of the SNR requirement described
above does not depend on any assumption on the instrument, like a particular detector type (read-out noise), pupil size,
or binning factor. Since the two free parameters A and B encompass all instrumental effects included in the simulation
and Eq. (2) and (3), the requirements in Table 2 ensure the required XCO2 precision for a broad range of possible
instruments with different detector performances and entrance pupil size.

Spectral SNRref (Threshold, SNRref (Goal,

Lref [ph./s/cm2/nm/sr
Band ID 0.7ppm) 0.5ppm)
NIR 4.201012 260 330
SWIR-1 12 345 500
2.1510
SWIR-2 1.851012 345 500
Table 2: SNR requirements for the CO2 instrument of the CO2M mission

5.3 Spatial co-registration for XCO2 observations

Apart from measurement noise, another source of random error, limiting the capability to quantitatively characterize of
green-house gas point sources, is spatial mis-registration between spectral channels. Different parts of the three spectral
bands contribute different information to the retrieval algorithm. The oxygen absorption lines of the O2 A-band measured
in the NIR are used to quantify the ground pressure at the spatial sample, as well as the effective (scattered) photon path
in propagation of the measured sunlight thought the atmosphere. The simultaneously measured absorption bands in the
SWIR-1 and SWIR-2 are used to quantify the amount of CO2 molecules along the photon path, and the combination in
the three-band retrieval yields the concentration in terms of XCO2. Any spatial mismatch between the NIR and SWIR
bands leads to an over- or under-estimation of this concentration, depending on the different columns of air sampled in
the NIR and SWIR channels. This error in XCO2 varies primarily with surface height differences and therefore gives rise
to a pseudo-noise, which is particularly pronounced over hilly or mountainous terrain. The requirement for this error
contribution is therefore driven by surface topography, and translates into the very stringent co-registration requirement
of 5% of the spatial sampling distance (SSD). For a square spatial sample of 2km x 2km this corresponds to a spatial
mismatch of only 100m between any measured spectral channel, observed from an altitude of 800 km. This required
spatial performance is significantly superior to those of comparable space-borne imaging spectrometers (e.g. Sentinel-5
requires 30% of SSD). Flown down to component level, it imposes extremely stringent constraints on image quality,
alignment accuracy and stability, which are difficult to achieve by optical and thermo-mechanical design. It is therefore
likely that alternative approaches for improving spatial co-registration will be deployed in the greenhouse gas instrument
of the CO2M mission.
Possible technologies for improving spatial co-registration involve innovative entrance slit designs based on wave-guides
or optical fibers. For example, the fiber-based entrance slits manufactured in the frame of an ESA-initiated pre-
development activity [14] are likely to significantly reduce co-registration errors, while at the same time allowing for
relaxation of image requirements in terms of keystone and smile distortion. Figure 5 depicts a microscope image of one of
these fiber-slit devices with typical dimensions for space-borne imaging spectrometers. The primary purpose of these
devices is the reduction of pseudo-noise from non-uniform scenes, which is another significant limitation of retrieval
precision. The homogenization functionality and performance has been verified by extensive characterized
measurements using an elaborate a test breadboard [20]. As can be seen evident from Figure 5, the entrance slit formed by
the stacked array of optical fibers exhibits gaps of typically 30 µm between the individual across-track field-of views,
which are due to the cladding around the fiber cores. Since light cannot transmit through the cladding, these gaps will
give rise to non-illuminated stripes on the focal plane. On one hand this constitutes a loss of signal and results in a
reduction of SNR, which needs to be considered and compensated. On the other hand, these “dark stripes” allow for an
unambiguous distinction of the light propagating through each individual fiber, defining the across-track field of view of
one spatial sample. The detected electrons from the illuminated detector pixels are co-added in each spectral channel.
The near-zero signal between the illuminated areas prevents that photons from neighboring spatial samples is co-added

Proc. of SPIE Vol. 11180 111800M-9

Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 09 Mar 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 ICSO 2018 Chania, Greece
International Conference on Space Optics 9 - 12 October 2018

to the signal, even in presence of large image distortion or detector alignment errors (e.g. rotation). In this way, nearly
perfect co-registration can be achieved across all spectral bands, as long as they share the same entrance slit. As an
additional advantage, the illumination gaps on the focal plane may be used for in-flight monitoring straylight between the
spatial samples across the entire swath width and spectral range.

Figure 5: Microscope image of spectrometer entrance slit based stacked optical fibers. The rectangular fiber cores define the slit width
in along-track direction (~100 µm), and the spatial sampling distance in across-track direction. The scene heterogeneity is scrambled
during propagation through the fiber (~20 mm). The spatial gaps between the fibers is due to the cladding around the cores, and serves
to improve spatial co-registration. Courtesy CeramOptec.

5.4 Accuracy (bias) requirements for XCO2 observations

The above sections offered a detailed discussion of requirements, which are deemed to be driving the primary mission
goal of CO2M to quantify anthropogenic greenhouse gas emissions from spatially resolved images of CO4 and CH4
concentration. Since anthropogenic carbon fluxes are underlying much larger biogenic ones, the latter also have to
accurately measured over larger distances of country- and continental scales. As for the heritage missions like GOSAT
and OCO-2, this imposes stringent requirements for low bias observations (< 0.5 ppm). These in turn translate into
challenging instrument specifications for straylight performance, polarization sensitivity and detection non-linearity. The
approach to constraining these effects is building on heritage from the CarbonSat studies. For the Earth Explorer
candidate mission, bias accuracy was specified in terms of a figure-of merit called the Effective Spectral Radiometric
Accuracy (ESRA), which is defined as the scalar product of error spectra with so-called gain vectors. Spurious spectral
features are simulated for the optical design, e.g. from straylight models, and the deviation from a perfect measurement
defines the error spectrum. The gain vectors are computed from a radiative transfer model and the Jacobian matrix
composed of the partial derivatives of the TOA radiance w.r.t. the estimated parameters (like CO2 column density). The
gain vector approach was explained in detail in a previous conference paper [14]. The gain vectors presented therein
have been updated for CO2M instrument, and are plotted in Figure 6. The gains can be interpreted as representing the
sensitivity of the retrieved XCO2 to radiometric error. The spectral regions with large gain amplitude contribute
proportionally stronger to the total resulting bias in XCO 2. Consequently, systematic radiometric biases spectrally
correlating with the gains are particularly harmful in the retrieval.

Proc. of SPIE Vol. 11180 111800M-10

Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 09 Mar 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 ICSO 2018 Chania, Greece
International Conference on Space Optics 9 - 12 October 2018

Figure 6: Gain spectra for the retrieval of XCO2 from spectral radiance measurements at CO2M spectral resolution. These gain spectra
or vectors are used to define the requirement constraining spectral features from various instrumental effects, including straylight and
polarization sensitivity.

5.5 Requirements for NO2 observations

In addition, the support studies for the CO2M mission identified the necessity of auxiliary measurements, which
significantly enhance the ability to separate anthropogenic CO2 emissions from natural ones. The wind direction and
speed, which are required for flux inversion, can be constrained by co-located observation of nitrogen dioxide (NO2).
Therefore, a spectral interval covering the NO2 absorption features in the visible (VIS) has been added to the band
definition of the CO2M instrument. The measurements in the VIS band are not used in the three-band retrieval for
XCO2, but rather in the subsequent flux-inversion by means of the NO2 total column product. Therefore there is no
stringent co-registration required between the VIS on the one hand, and the NIR and SWIR bands on the other. This
allows for implementation of the VIS spectrometer as a separate instrument on the CO2M platform. However, in order to
be able to re-sample the NO2 images onto the spatial grid of the CO2 observations, very high spatial sampling of the
observations is specified at 1 km2. The spatial resolution, defined as the area over which the SNR and integrated energy
requirements are fulfilled, is the same as CO2M at 4 km2 (2 km x 2km for square samples). This is roughly 6 times
higher spatial resolution than the currently highest resolved NO2 product from the Tropomi instrument of the Sentinel-5P
mission [15]. Also the SNR and the spectral resolution requirements are comparable to the heritage missions for
atmospheric chemistry. The most important requirements for the VIS band are summarized in Table 3.

Requirement Value

Spectral band 405nm-490nm

Spectral
0.60 nm
resolution
1.351013
Lref
photons/s/cm2/nm/sr
SNR@Lref 500(T)/1000(G)
Table 3.: Main requirements for radiance measurement of the VIS band

5.6 Requirements for Multiple-Angle Polarimeter (MAP) observations

Another set of required auxiliary observations are measurements of spectral radiance and degree of polarization over a
broad spectral range and various viewing angles. From measurements of a Multiple-Angle-Polarimeter (MAP) the
aerosol distribution above the area under investigation can be inferred, which in turn improves the ability to constrain the

Proc. of SPIE Vol. 11180 111800M-11

Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 09 Mar 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 ICSO 2018 Chania, Greece
International Conference on Space Optics 9 - 12 October 2018

effective photon path in the XCO2 retrieval. Such measurements have been acquired in the context of instruments
targeting aerosol studies, such as POLDER [17] and are proposed for the future 3MI mission [18]. These instruments
provide observations of radiance and the Degree of linear polarization (DoLP) for a set of spectral bands and a few tens
of viewing angles. This observation strategy with bandwidths of typically > 10 nm in the NIR, VIS and SWIR regions is
referred to as MAP-band method.
More recently, a new approach to aerosol observation has been proposed for the PACE mission. The technique, based on
the SPEX instrument concept [19], employs a radiometric modulation of an acquired, continuous radiance spectrum,
whereby the modulation depth depends on the degree of linear polarization (DOLP). In comparison with the MAP-band
method, the modulation technique (MAP-mod) required less viewing angles (less than ten), but higher spectral resolution
on the order of 2-4 nm. The required bandwidth of the continuously sampled spectra encompasses the VIS and NIR
spectral regions.
Since both techniques, MAP-band and MAP-mod, are regarded as equally suitable approaches to constraining aerosol
distribution for enhanced XCO2 retrieval, two different sets of requirements have been established for CO2M
constellation, which are summarized in Table 4 and 5. Due to the distinctiveness of the two measurement techniques, the
spectral and radiometric, as well as the geometric requirements (in terms of required viewing angle) are quite different.
The two major common features are the spatial sampling distance of the re-sampled Level-1b data (4 km2) and the
polarimetric accuracy and precision (both 0.0025 of DoLP).

Channel Central Spectral DoLP

Wavelengt Resolution
h
(nm) (nm)
MAP-1 410 20 Y
MAP-2 443 20 Y
MAP-3 490 20 Y
MAP-4 555 20 Y
MAP-5 670 20 Y
MAP-6 (G) 763 10 N
MAP-7 (G) 765 40 N
MAP-8 865 40 Y
Table 4: Spectral requirements of radiance and polarimetric measurements for the MAP based on the bandpass concept (MAP-band)

Observable Spectral Spectral

Range Resolution
(nm) (nm)

1.1 Radiance
measurements 385-770 5

1.2 DoLP 15 @ 385 nm

measurements 385-770 40 @ 755 nm

Table 5: Spectral requirements of radiance and polarimetric measurements for the MAP based on the modulation concept (MAP-mod)

Proc. of SPIE Vol. 11180 111800M-12

Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 09 Mar 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 ICSO 2018 Chania, Greece
International Conference on Space Optics 9 - 12 October 2018

6. SUMMARY AND CONCLUSIONS

We have presented the payload components and most important observational requirements of the CO2M mission. The
payload is comprised of a suite of instruments, performing simultaneous, co-located observations. The primary CO2
instrument measuring in the three spectral bands is complemented auxiliary payloads consisting in a visible spectrometer
for NO2 observation, a multiple-angle polarimeter and a cloud imager. The objective of detecting and quantifying
anthropogenic emissions drives the instrument to high single-sounding precision (low random error) of the Level-2
images of XCO2, Dedicated analysis emphasizes the importance of minimizing random retrieval noise to ensure
detectability of emission plumes with faint differential absorption signals. We have presented a method, which translates
the required Level-2 retrieval precision into a simple requirement for signal-to-noise ratio of the CO2 instrument. The
requirement derivation establishes a general link between Level-2 and Level-1b observations without assuming
particular instrument setups. The need for low bias observations of XCO2 are comparable to previous and planned
missions, imposing low systematic errors of the measurements.

REFERENCES

[1] "Paris Agreement”, United Nations Treaty Collection, C.N.92.2016.TREATIES-XXVII.7.d, July 2016
[2] Pinty B., G. Janssens-Maenhout, M. Dowell, H. Zunker, T. Brunhes, P. Ciais, D. Dee, H. Denier van der Gon,
H. Dolman, M. Drinkwater, R. Engelen, M. Heiman, K. Holmlund, R. Husband, A. Kentarchos, Y. Meijer, P.
Palmer and M. Scholze: An Operational Anthropogenic CO2 Emissions Monitoring & Verification Support
capacity - Baseline Requirements, Model Components and Functional Architecture, doi:0.2760/08644,
European Commission Joint Research Centre, EUR 28736 EN, 2017
[3] Ciais, P., D. Crisp, H. Denier Van Der Gon, R. Engelen, M. Heimann, G. Janssens-Maenhout, P. Rayner, and
M. Scholze, Towards a European Operational Observing System to Monitor Fossil CO2 emissions, Final
Report, European Commission, Oct. 2015
[4] Buchwitz, M., R. de Beek, J. P. Burrows, H. Bovensmann, T. Warneke, J. Notholt, J. F. Meirink, A. P. H.
Goede, P. Bergamaschi, S. Körner, M. Heimann, and A. Schulz, Atmospheric methane and carbon dioxide from
SCIAMACHY satellite data: Initial comparison with chemistry and transport models, Atmos. Chem. Phys., 5,
941-962, 2005
[5] Crisp, D., et al., The Orbiting Carbon Observatory (OCO) mission, Adv. Space Res., 34, 700-709, 2004.
[6] Kuze, A., Suto, H., Nakajima, M., and Hamazaki, T.: Thermal and near infrared sensor for car- bon
observation Fourier-transform spectrometer on the Greenhouse Gases Observing Satellite for greenhouse
gases monitoring, Applied Optics, 48, 6716–6733, doi:10.1364/AO.48.006716
[7] Crisp, D., Miller, C. E., and DeCola, P. L.: NASA Orbiting Carbon Observatory: measuring the column
averaged carbon dioxide mole fraction from space, Journal Of Applied Remote Sensing, 2, 23 508,
doi:10.1117/1.2898457, 2008.
[8] Nassar, R., Hill, T. G., McLinden, C. A., Wunch, D., Jones, D. B. A., & Crisp, D. (2017). Quantifying CO2
emissions from individual power plants from space. Geophysical Research Letters, 44, 10,045–10,053.
https://doi.org/10.1002/ 2017GL074702
[9] Liu Yi and the TanSat Science Team, "The Pre-launch Status of TanSat Mission," 12th International Workshop
on Greenhouse Gas Measurements from Space (IWGGMS-12)," Kyoto University, Kyoto, Japan, June7- 9,
2016, URL: https://www.omc.co.jp/iwggms12/pdf/Session_9/42_Yi_Liu.pdf
[10] Dongxu Yang, Yi Liu, Zhaonan Cai, Xi Chen, Lu Yao, Daren Lu, "First Global Carbon Dioxide Maps Produced
from TanSat Measurements," Advances in Atmospheric Sciences, Vol. 35(6), 2018, pp: 621–623,
https://doi.org/10.1007/s00376-018-7312-6, URL: https://link.springer.com/content/pdf/10.1007%2Fs00376-
018-7312-6.pdf
[11] Pasternak, F., Bernard, P., Georges, L., Pascal, V.,: The MicroCarb instrument, Proceedings Volume 10562,
International Conference on Space Optics — ICSO 2016; 105621P (2017) https://doi.org/10.1117/12.2296225
[12] O'Brien, D. M., Polonsky, I. N., Utembe, S. R., and Rayner, P. J.: Potential of a geostationary geoCARB
mission to estimate surface emissions of CO2, CH4 and CO in a polluted urban environment: case study
Shanghai, Atmos. Meas. Tech., 9, 4633-4654, https://doi.org/10.5194/amt-9-4633-2016, 2016

Proc. of SPIE Vol. 11180 111800M-13

Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 09 Mar 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 ICSO 2018 Chania, Greece
International Conference on Space Optics 9 - 12 October 2018

[13] Sierk, B., Caron, J., Bézy, J.-L., Löscher, A., ; Meijer, Y., Jurado, P., The CarbonSat candidate mission for
imaging greenhouse gases from space: concepts and system requirements, Proceedings Volume 10563,
International Conference on Space Optics — ICSO 2014; 105633I (2017) https://doi.org/10.1117/12.2304154
[14] Sierk, B., Löscher, A., Caron, J., Bézy, J.-L., Meijer, Y., CarbonSat instrument pre- developments: Towards
monitoring carbon dioxide and methane concentrations from space, Proceedings Volume 10562, International
Conference on Space Optics — ICSO 2016; 1056260 (2018) https://doi.org/10.1117/12.2322406
[15] Veefkind, P., et al.: Overview of Early Results from TROPOMI on the Copernicus Sentinel 5 Precursor, 20th
EGU General Assembly, EGU2018, Proceedings from the conference held 4-13 April, 2018 in Vienna, Austria,
p.12216
[16] Wu, L., Hasekamp, O., Hu, H., Landgraf, J., Butz, A., de Brugh, J. a., et al. (2018). Carbon dioxide retrieval
from OCO-2 satellite observations using the RemoTeC algorithm and validation with TCCON measurements.
Atmospheric Measurement Techniques, 11(5), 3111-3130. doi:10.5194/amt-11-3111-2018.
[17] P.-Y. Deschamps, F.-M. Bréon, M. Leroy, A. Podaire, A, Bricaud, J.-C. Buriez, and G. Sèze, “The POLDER
mission: Instrument Characteristics and Scientific Objectives,” IEEE Transactions On Geoscience and Remote
Sensing 32(3), pp. 598-615, May 1994
[18] I. Manolis, S. Grabarnik, J. Caron; J-L. Bézy; M. Loiselet; M. Betto; H. Barré; G. Mason; R. Meynart, “The
MetOp second generation 3MI instrument,” in Proc. SPIE 8889, Sensors, Systems, and Next- Generation
Satellites XVII, October 2013
[19] van Amerongen, A., Rietjens, J., Smit, M., van Loon, D., van Brug, H., van der Meulen, W., Esposito, M.,
Hasekamp, O.: SPEX: the Dutch roadmap towards aerosol measurement from space, International Conference
on Space Optics (ICSO) 2016
[20] Amann, S. Duong-Ederer, Q., Haist, T., Sierk, B., Guldimann. B., Osten, W.: Characterization of fiber-based
slit homogenizer devices in the NIR and SWIR, International Conference on Space Optics (ICSO) 2018

Proc. of SPIE Vol. 11180 111800M-14

Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 09 Mar 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use