The PROBA/CHRIS Mission: ALow-Cost Smallsatfor Hyperspectral, Multi-Angle, Observations of the Earth Surface and Atmosphere

1
The PROBA/CHRIS Mission: A Low-Cost Smallsat for Hyperspectral, Multi-Angle, Observations of the
Earth Surface and Atmosphere
M.J. Barnsley, J.J. Settle, M.A. Cutter, D.R. Lobb, F. Teston
Abstract— The European Space Agency’s Project for On-Board Auton-
Third, by slowly pitching during image acquisition (i.e. through
omy is intended to demonstrate a range of innovations in the design, con-
motion compensation), it is possible to improve the signal-to-
struction and operation of small satellites. It carries a number of scientific
noise performance by increasing the integration times. Finally,
instruments, the most advanced of which is the Compact High-Resolution
Imaging Spectrometer. A typical nadir image is 13km × 13km in size and
it provides a means by which images can be recorded at different
has 18 narrow spectral channels at 17m spatial resolution. When oper-
sensor view angles. The latter is, of course, important for studies
ated at 34m spatial resolution, the instrument can capture data in 62, al-
of the bidirectional reflectance properties of Earth’s surface, and
most contiguous, spectral channels. The platform is highly manoeuvrable:
along-track pointing allows a given site to be imaged five times during a
can also be used to generate digital elevation models through
single overpass, while across-track pointing ensures that the revisit time for
stereo photogrammetric reconstruction. In fact, PROBA-1 can
a site of interest is less than a week. This unique combination of spectral
be pitched sufficiently quickly along-track so that five separate
and angular sampling provides a rich source of data with which to study
CHRIS images can be obtained for a given target area during a
environmental processes in the atmosphere and at Earth’s surface.
single orbital overpass, with each image recorded at a different
I. I
sensor view angle. This multiple-view-angle (MVA) imaging
NTRODUCTION
capability [1], in conjunction with the high spectral and spa-
The Project for On-Board Autonomy (PROBA-1) satellite
tial resolution of CHRIS, provides an enormously rich source of
was launched from Shriharikota, India on 22 October 2001. De-
data for the scientific investigation of Earth’s surface and atmo-
veloped and built by a consortium led by the Belgian company
sphere. Indeed, ESA has recognised the importance of such data
Verhaert, and funded by the European Space Agency (ESA),
to the characterisation of a range of environmental processes op-
PROBA-1 was originally intended to be a short, experimental,
erating at the land surface, and have plans for a similar mission
mission. Its primary objective was to test a number of inno-
(SPECTRA), albeit on a much grander scale and over a longer
vations in platform design, principally relating to attitude con-
period, to be launched later this decade. The lessons learned
trol and recovery from errors, which would enable it to operate
from the PROBA/CHRIS mission will be invaluable in helping
autonomously; that is, with the minimum amount of interven-
to guide the development of this, more ambitious, programme.
tion from the ground. PROBA-1 carries on board a small num-
ber of scientific instruments, intended to demonstrate the use of
II. THE PROBA PLATFORM
the platform for both space and environmental studies. These
A. Overview
include a sensor for detecting space debris, a space radiation-
environment monitor and two digital cameras. The principal
The primary purpose of the PROBA platform is to act as
scientific instrument on board PROBA-1, however, is the Com-
‘proof-of-concept’ for a novel spacecraft technology.
More
pact High Resolution Imaging Spectrometer (CHRIS). This sen-
specifically, it is intended to show the utility of on-board systems
sor acquires high spatial resolution (17–20m or 34–40m) images
that may be used to control the satellite, including its general
of Earth’s surface in up to 62 narrow spectral channels located
operation and the management of resources, spacecraft attitude,
in the visible and near infra-red wavelengths.
data communications and payload operations. The operation of
In addition to their respective scientific and technological ob-
the star trackers, on-board autonomy software and spacecraft
jectives, the combination of CHRIS and PROBA is intended
attitude-adjustment using the four reaction wheels are of par-
to serve as a demonstration of a faster approach to the design,
ticular interest from an engineering perspective, and form the
launch and operation of scientific satellite-sensor missions. In
core of the technology demonstration programme.
this context, ‘faster’ refers to the relatively short time-period
The PROBA project is part of a general move away from the
(two to three years) between the selection of CHRIS as the An-
large, resource-intensive, satellites, which have been character-
nouncement of Opportunity (AO) instrument and the launch of
istic of Earth observation over the past 20 years or so; for these
the PROBA platform. One of most significant benefits of this
missions, the time period between conception and launch is typ-
approach is the potential for greater responsiveness to the cur-
ically much longer than the natural cycle of scientific investi-
rent and future needs of scientific users.
gations. Instead, attention is turning to so-called ‘smallsats’,
A novel feature of the PROBA-1 platform is that it can be
whose design, build and deployment programme is intended to
manoeuvred in orbit using a set of four reaction wheels. These
follow the principles of the ‘faster, better, cheaper’ initiative,
allow the satellite to be pointed off-nadir in both the along-track
which was introduced by NASA to provide more cost-effective
and across-track directions. This agility confers a number of ob-
space missions after the failed Mars Lander attempt in 1993
vious benefits. First, it increases the area of Earth’s surface that
[2]. As a first attempt at implementing this strategy, PROBA
is potentially visible to the on-board imaging instruments dur-
has done reasonably well: the time between the initiation of the
ing a single orbit. Second, it allows the instruments to acquire
project and the launch of PROBA-1 was less than four years, and
images of cloud-free areas of Earth’s surface at the expense of
would have been under three years but for delays to the launch
cloud-covered ones, and to avoid sun glint over water bodies.
caused by factors external to the project.

---

2
The first realisation of the PROBA concept, PROBA-1, is a
C. The PROBA Orbit
small platform, weighing approximately 100kg, including pay-
The PROBA-1 satellite was launched from the Indian Space
loads, and measuring approximately 60cm × 60cm × 80cm. The
Research Organisations (ISRO) main launch pad at Shriharikota
spacecraft management (e.g. guidance, navigation and control)
in India, aboard an ISRO Polar Satellite Launch Vehicle (PSLV).
is carried out by a high-performance RISC processor; specif-
The PROBA-1 orbit is sun-synchronous, with an equatorial
ically, a space-qualified SPARC V7 device, which provides
crossing time at launch of 10:30. The orbit is somewhat ellipti-
10MIPS and 2MFLOPS of processing power. The spacecraft
cal, with an altitude varying from about 550km to about 670km.
is designed with a nominal lifetime of greater two years. Ini-
This is useful in the context of the PROBA mission, and espe-
tially, the intention was to operate PROBA-1 in space for one
cially for the space-radiation detection experiment, but is less
year only, but the demand for scientific data from the payload
than optimal for CHRIS activities. In July 2003, the orbital pe-
instruments has been so great that an extra two years of post-
riod was 96.77 minutes, the inclination 97.84◦ and the eccen-
launch support has been funded by ESA.
tricity 0.0084. Atmospheric drag makes it impossible to predict
B. Attitude Control and Pointing
more than about two months in advance whether a given site
will be visible on a particular day, and the relatively low orbital
An important element of the on-board autonomy is the control
altitude restricts the across-track pointing, as noted above.
of the spacecraft’s attitude. The platform uses GPS instruments
(an L1 system with 4 antennae) and star trackers to maintain
D. Power
an accurate record of its position and attitude. The star track-
ers work by recording an image of deep space, comparing the
Power is provided by Gallium Arsenide solar arrays mounted
observed pattern of stars to an internal catalogue of stellar po-
on five of the six faces of the spacecraft. These provide up to
sitions. The image processing load is taken by the star tracker
120W peak output and charge the 196W h lithium-ion on-board
itself. This, combined with efficient matching algorithms, mean
batteries. In terms of the battery operations, there are four typ-
that the star tracker units impose a very small resource load on
ical operational orbits, these being ‘imaging’, ‘transmission’,
the platform itself. The rate of update of the star trackers on-
‘imaging and transmission’ and ‘quiet’. Orbits where imaging
board PROBA-1 is in the range 5–20Hz. At 5Hz, a pointing
takes place discharge the batteries, while the others charge them;
accuracy of a few arc seconds is achievable.
the latter are generally managed autonomously by the on-board
If just one star tracker is used, it will be blinded periodically
mission manager. Any combination of transmission and quiet
by being directed at the moon: the passage of the moon across
orbits is allowed, as this leads to a net charging of the on-board
its field-of-view can last as long as 240 seconds for up to 2 days
battery. Typical image acquisition periods place a higher load
per month. To avoid this, and to increase the accuracy gener-
on the battery and may increase the Depth of Discharge (DoD)
ally, PROBA-1 employs two star trackers, pointing in different
beyond acceptable limits (20% is the acceptable limit, 32% be-
directions. The star trackers are situated to avoid blinding by the
ing the maximum recommended). A ‘sun-bathing’ mode is pro-
sun throughout the duration of the mission, whether as a result
vided by PROBA, when the maximum possible area of solar
of seasonal progression or spacecraft manoeuvring. The addi-
panels is exposed toward the Sun during a daytime orbit. The
tional requirement that they are not blinded by Earth-light limits
net charge per orbit when sun-bathing is 0.99Ah. This means
the across-track pointing of the mission slightly, partly due to the
that 20% DoD, corresponding to a net charge per orbit of 1.4Ah,
relatively low orbit into which PROBA-1 was eventually placed:
can be completely recovered in two orbits [4].
it is a minor consideration in a higher orbit.
E. Ground Operations
The spacecraft attitude is controlled by a set of four reaction
wheels, which are integrated to the satellite structure. These al-
ESA’s ground receiving station at Redu has a 2.5m dish,
low the satellite to be manoevered in each of the three planes
which is dedicated to the PROBA-1 mission. PROBA-1 has
(i.e. roll, pitch and yaw). Consequently, the capability to tilt the
1.2Gbit of on-board data storage located within the Mass Mem-
satellite in both the along-track and across-track directions is a
ory Unit (MMU) and its data transfer rates are 4kbit s−1 for up-
significant feature of the PROBA-1 mission. The maximum rate
link and 1Mbit s−1 for download. Thus, about 20 minutes are
at which the spacecraft attitude can be varied is approximately
required to download the contents of the MMU in its entirety,
1◦s−1. This is primarily limited by the ability of the star trackers
equivalent to as many as three overpasses of the Redu ground
to maintain an accurate register of spacecraft position and atti-
receiving station. To achieve the greater throughput required to
tude, rather than the intrinsic capabilities of the reaction wheels.
meet the needs of the scientific mission for CHRIS, additional
As with other missions, such as the HRV and HRVIR sensors
download capacity has been deployed at the Kiruna ground sta-
on board the SPOT series of satellites, across-track pointing in-
tion, while Redu is still used for the data up-link.
creases the frequency with which a given area of ground is vis-
ible to the sensors on-board PROBA-1 [3], and ensures that al-
III. THE CHRIS INSTRUMENT
most all of Earth’s surface is accessible at least once each week.
A. General Description
In its present orbit, certain sites are visible to PROBA-1 on two,
and occasionally three, successive days. Moreover, by tilting
CHRIS is a small hyperspectral imager, designed as a space-
the platform in the along-track direction, its imaging payload is
borne remote sensing instrument. It consists of a telescope and
able to acquire a set of up to five images of the same target area
an imaging spectrometer, attached to a CCD-array detector sys-
during a single overpass.
tem. It weighs approximately 14kg and occupies a volume of

---

3
C. Spectrometer
Light from the slit is dispersed and re-imaged by the spec-
trometer optics onto the detector, which is an area-array CCD.
The spectrometer has a magnification close to 1 for each
wavelength in both the along-track and across-track directions
(i.e. orthogonal to and parallel to the slit direction, respectively),
so that the overall focal length of the combined instrument is the
same as that of the telescope (746mm). The spectrometer in-
Fig. 1. Optical plan of the CHRIS instrument.
cludes two dispersing prisms, so that each point in the entrance
slit is re-imaged as a spectrum line, each line being orthogonal
to the slit image. Thus, a monochrome image of the entrance slit
0.79m × 0.26m × 0.20m. Between the telescope and the spec-
is formed on each row of detector elements, while the spectrum
trometer lies the spectrometer entrance slit, which is treated as
of each point in the slit (and hence on the ground) is imaged
part of the telescope assembly in the context of this paper. The
along a detector column.
telescope forms an image of the distant Earth scene onto the en-
The re-imaging function of the spectrometer optics is per-
trance slit, so that the telescope and slit define a slit-shaped in-
formed by an arrangement of three spherical mirrors. The pri-
stantaneous field-of-view (IFOV). The spectrometer re-images
mary and tertiary mirrors are concave; the secondary is convex.
the collected light, disperses it in the direction perpendicular to
The mirror configuration is very similar to that of a classical
the slit image, and focuses it on to the area-array detector. The
Offner relay, with all centres of curvature close together and the
CHRIS deployed on PROBA-1 is limited to channels in the vis-
radii of curvature of the concave mirrors approximately double
ible and near infra-red wavelengths, but the design allows for
that of the secondary. The prisms are placed in the path between
an extended range of wavelengths through the use of an ad-
the entrance slit and the first mirror, and between the third mirror
ditional detector system and minor adjustments to the general
and the detector. The prism surfaces are spherical to allow cor-
instrument layout. The result is a low cost, light weight, in-
rection of their optical aberrations in beams that are diverging
strument with no moving parts, capable of delivering high res-
and converging.
olution spectra throughout the visible and near-infra-red wave-
The prisms are made of fused silica. The mirrors are made of
lengths (400–1050nm).
Schott BK7 glass. Again, this is to provide a good CTE match to
the titanium structure. With different coatings and minor design
adjustments, the system can provide a well-corrected spectrum
B. Telescope
image for the complete spectral range from UV to 2500nm.
The special merits of the spectrometer design are its simplic-
The telescope has a pupil diameter of 120mm, operating at
ity, ease of manufacture and alignment, and small dimensions.
f /6, and has a conventional two mirror configuration. After
The simplicity arises from the use of just three mirrors and two
reflection from the primary and secondary mirrors, the beam
refracting elements, which also limits costs and permits good
passes through a central hole in the primary mirror to reach a
control of out-of-band stray light. The ease of manufacture and
focus at the entrance slit (Fig. 1). Only spherical surfaces are
alignment are achieved by avoiding aspheric surfaces, difficult
used. A large meniscus lens at the entrance pupil of the instru-
materials and optically contacted surfaces. The design is also
ment is employed to correct for spherical aberration. This also
very compact for the achieved spatial and spectral resolution and
provides a convenient method for mounting the secondary mir-
relative aperture.
ror, which is cemented to the inner face of the meniscus. The
telescope optics are completed by two small lenses in the con-
D. CCD
verging beam in the entrance slit assembly, which correct for
some minor telescope aberrations and allow the telescope to be
The re-imaged light emerges from the spectrometer with the
approximately telecentric.
slit dispersed spectrally, at right angles to the slit direction. The
CCD array in the spectrometer focal plane is aligned so that it
The meniscus lens, the secondary to which it is cemented,
acquires spatial data in one dimension (columns) and spectral
and the two small lenses are made of fused silica. The primary
data in the other (rows). The detector used is an e2v CCD-
mirror is made of Schott BK7 glass, to provide a good CTE
type (CCD25-20), a frame transfer device with 770 columns
match to the titanium structure. Lens elements are multi-layer
and 576 rows. It is thinned, back-illuminated and has a single-
anti-reflection coated to reduce stray reflections below 1.5% per
layer anti-reflection coating, which is uniform over the image
surface at all wavelengths in the nominal instrument range. The
area, and quarter-wave effective thickness at 1000nm.
This
telescope mirrors are multi-layer dielectric coated to give 98%
gives high quantum efficiency, including good performance in
reflection per surface over the wavelength range.
the deep blue spectral region (20% at 400nm) and better than
The slit is 16.7mm long by 0.0255nm wide and the focal
7% at 1000nm. The element size is 0.0225mm × 0.0225mm.
length of the telescope is 746mm, so that at apogee (670km) the
The individual detectors have a dump gate on the output reg-
ground area instantaneously imaged by the slit and telescope is
ister, which enables rapid dumping of signals from unwanted
15km across-track by 20m along-track; these figures reduce to
image areas. The row length of 770 devices provides a useful
about 13km by 17m at perigee.
compromise between the need for a fast line-transfer time and a

---

4
matic images of the slit fall on straight detector rows, and line
12
spectra of resolved ground areas fall on detector columns. De-
11
partures from these two conditions are known, respectively, as
10
9
‘smile’ and ‘frown’. In design, the worst case smile, in terms of
8
wavelength errors, is 0.00035mm at 1050nm, corresponding to
7
a wavelength error of 0.2nm. The worst-case frown is 1% of the
6
pixel width.
5
Bandwidth, nm
4
F. Stray Light Control
3
The low-cost telescope design presents some interesting chal-
2
lenges in terms of stray-light control. The main source of stray-
1
light error is believed to be produced by low-angle scatter at op-
400 500 600 700 800 900 1000 1100
tical surfaces, arising from imperfections in the polish and coat-
Central wavelength, nm
ings. In particular, some light from the scene can reach the en-
trance slit by transmission through the three lens elements, with-
Fig. 2. CHRIS spectral resolution against central wavelength.
out reflection at either mirror. An over-sized secondary mirror,
and a sequence of baffles are deployed to mitigate these possibil-
wide image swath.
ities. An analysis of stray-light for the instrument confirms the
The CCD offers the ability to make radiometric measure-
expectations from the optical design, namely that radiance errors
ments in the spectral range 400–1050nm with a spectral reso-
equivalent to 0.5% of the average scene radiance are achievable.
lution that varies from 1.25nm to 11.25nm across the spectrum
IV. CHRIS ON PROBA: MISSION CHARACTERISTICS
(Fig. 2). As has been noted previously, the design of the spec-
trometer is such that this could be expanded out into the short-
The combination of a programmable imaging spectrometer
wave infrared (up to about 2500nm) by the addition of a sec-
(CHRIS) and an agile, pointable, satellite platform (PROBA-1)
ond detector sensitive to this part of the spectrum. A particular
offers considerable flexibility in terms of image data acquisi-
strength of the CCD package is its flexibility in terms of line
tion. As a result, it is possible to tailor the data acquired by
transfer and read-out. This allows a wide variety of different
PROBA/CHRIS to address the particular needs of selected sci-
imaging configurations. These are considered later in this pa-
ence problems. More specifically, the CHRIS instrument on
per. The spectral and spatial definition of any given mode of the
board PROBA-1 offers some control over the spatial resolution
CHRIS instrument may, therefore, be controlled by combining
of the image data (by a factor 2), the number and spectral band-
or neglecting CCD elements, and by changing the integration
widths of the wavebands in which these data are recorded, and
time.
the view zenith and azimuth angles at which they are acquired.
In practice, given that this is primarily an experimental mission,
E. Theoretical Performance and Tolerances
a single mode of angular sampling is employed, and the spectral
and spatial sampling options (i.e. those realised by particular on-
The telescope gives design resolution better than 0.01mm in
chip averaging configurations) are limited to a few operational
the focal plane, over the whole spectral range and field of view.
modes for almost all acquisitions.
Alignment tolerances are generally coarse, so that the spatial
resolution is, in practice, limited mainly by the finite sizes of
A. Sampling Characteristics
the entrance slit (along track) and the detector pixels (across
track).
The design performance of the spectrometer optics,
The spatial sampling characteristics (17m or 34m nominal
in combination with the telescope, is better than 0.012mm in
sampling interval and 13km nominal swath) have already been
the across-track direction, so that spatial resolution is, in the-
outlined. The main purpose of having the coarser spatial resolu-
ory, limited almost entirely by the finite detector element size
tion is to increase the number of bands that can be read out from
(0.0225mm). The design resolution of the spectrometer is better
the CCD array. Thus, there is a trade-off between spatial and
than 0.011mm in the spectral-dispersion direction for most of
spectral resolution. A half-swath option at full spatial resolution
the wavelength range, so that, in practice, the spectral resolution
is another of the operational modes; this again trades some spa-
is similarly limited mainly by the dimensions of the slit and de-
tial information for enhanced spectral information. In theory, it
tectors. The half-width of the instrument line-spread functions
is also possible to accept the coarser spatial resolution over the
in the column direction is approximately 0.0255mm (i.e. one
half-swath, giving just 190 or so pixels across-track, to increase
detector width), although charge dispersion in the detector de-
the number of spectral bands still further (over 100, in fact). The
grades this slightly at near infrared wavelengths. Spectral reso-
spatial sampling interval varies by ±8% around the orbit, and in
lution varies with wavelength, following the typical dispersion
time at any one location, because of the varying altitude of the
provided by refraction through fused silica prisms. The theoret-
satellite.
ical half-width of the line-spread functions correspond to spec-
tral resolutions of 1.25nm at 415nm, increasing to 11.25nm at
A.1 Spectral Sampling
1050nm (Fig. 2).
In practice, the operational modes of CHRIS are characterised
The complete optical design is optimised so that monochro-
in terms of different combinations of spatial resolution, swath

---

5
1
0.8
0.6
0.4
Atmospheric
0.2
transmittance
0
400
500
600
700
800
900
1000
0.7
0.6
0.5
0.4
0.3
Spectral
reflectance
0.2
0.1
0
400
500
600
700
800
900
1000
MERIS
Bands
400
500
600
700
800
900
1000
Wavelength (nm)
Fig. 3. CHRIS Mode 1 Band Set (62 spectral channels) compared with the nominal atmospheric transmittance (top), the spectral reflectance of vegetation (middle;
solid line) and MERIS spectral bands (bottom).
1
1
0.9
0.9
0.8
0.8
0.7
0.7
0.6
0.6
0.5
0.5
0.4
0.4
0.3
Reflectance/Transmittance
0.3
Reflectance/Transmittance
0.2
0.2
0.1
0.1
0
0
400
500
600
700
800
900
1000
400
500
600
700
800
900
1000
Wavelength (nm)
Wavelength (nm)
(a) Mode 2 Band Set (Water)
(b) Mode 3 Band Set (Land/Aerosols)
Fig. 4.
CHRIS Mode 2 (Water) and 3 (Land/Aerosols) band sets, with typical atmospheric transmittance (continuous line) and vegetation spectral reflectance
(dashed line) curves superimposed.

---

6
width and spectral band sets, as outlined below. Other modes
TABLE I
have been used for specific tasks such as wavelength calibration.
PROBA/CHRIS IMAGING SEQUENCE.
A.1.a Mode 1: Full Swath, Reduced Spatial Resolution. In this
Tag Number
FZA
Scan Direction
mode, CHRIS acquires image data across a full swath (nominal
3
+55◦
N–S
13km) at reduced spatial resolution (nominal 34m), in 62 spec-
1
+36◦
S–N
tral channels (Fig. 3). Most channels are 5–10nm wide, and the
0
0◦
N–S
set is largely contiguous.
2
−36◦
S–N
A.1.b Modes 2–4: Full Swath, Full Spatial Resolution. This is
4
−55◦
N–S
the most common mode of operation, in which CHRIS is used
to acquire image data at the highest spatial resolution (nominal
17m) over the full swath and in 18 spectral channels. The lat-
and AATSR) have dual-view capability, with data recorded at
ter are obtained by binning the signals from sets of CCD rows,
nadir (actually, between 0◦ and 20◦) and at 55◦ [6], [7].
allowing the bands to be selected anywhere in the instrument’s
In the context of the PROBA/CHRIS mission, the view zenith
spectral range. In this mode, three standard spectral band sets
angle of the satellite at the fly-by position is known as the Min-
have been defined. These are intended for studies of water
imum Zenith Angle (MZA). The convention adopted is that this
bodies (Mode 2; Fig. 4(a)), the land surface and atmospheric
angle is reported as a negative number when the target lies to
aerosols (Mode 3; Fig. 4(b)), and chlorophyll (Mode 4). Where
the east of the sub-satellite track, and positive when it lies to the
this can also be satisfied, the standard bands sets have been con-
west. The scan direction in which the five images are acquired
figured to permit inter-comparisons with other satellite sensors,
alternates, to minimise the slewing required between images,
such as MODIS and MERIS. In any case, the convergence of
and each image is assigned a unique tag number (Table I).
interest in a similar set of science problem lead us to locate the
CHRIS bands at similar wavelengths to those chosen for other
A.3 Temporal Sampling
satellites.
It is possible to take still further advantage of the across-track
A.1.c Mode 5: Half Swath, Full Spatial Resolution.
CHRIS
pointing of PROBA-1, to acquire up to three separate sets of
can also be used to acquire image data across half of the nominal
multiple-view-angle CHRIS images with respect to a given tar-
swath at the highest nominal spatial resolution. In this mode, the
get area, each from a different orbital overpass, over the period
instrument is able to record image data in 37 spectral channels.
of several days. The opportunities for repeated imaging of a
target are determined by its latitude and a factor related to the
A.2 Angular Sampling
immediate properties of the orbit. At the equator, any point can
As was noted earlier, the PROBA-1 satellite can be manoeu-
be imaged by CHRIS once every seven to eight days. If the
vred in all three axes to view any target of interest, which is
overpass has a small MZA, this will be the only chance to im-
very important for an instrument with a small image swath, such
age the site but, if the MZA is larger, the site may be visible on
as this, and only sites at the very highest latitudes are inacces-
two successive days. At higher latitudes, the opportunities for
sible to the mission. More significantly, by tilting PROBA-1
imaging on two successive days increases and, if the satellite
in both the along-track and across-track directions during orbit,
is close to apogee, it is possible that the site will be visible to
it is possible to acquire a set of up to five CHRIS images of
PROBA/CHRIS on three successive days — one view will be
a given target area, each at a different view zenith angle, in a
from the east, one from the west, and one from an overpass with
single data acquisition sequence; that is, during a single orbital
a relatively small (<5◦) MZA. The different sensor viewing an-
overpass (see, for example, Fig. 5). It is important to realise
gles and, to a lesser extent, the small changes in solar zenith an-
that the CHRIS instrument is only occasionally able to image a
gle at which these images are acquired enable PROBA/CHRIS
target area from directly overhead (i.e. in nadir-viewing mode)
to sample the BRDF of the target surface. For example, on the
because of its relatively narrow field-of-view. More generally,
12th, 13th and 14th of July 2003, the community site at Bar-
PROBA-1 must be tilted at some angle in the across-track direc-
rax, Spain (latitude 39◦N) was visible to PROBA/CHRIS with
tion so that the target area falls within the sensor’s field-of-view.
MZAs of 20◦, −4◦ and −27◦, respectively. The pattern with
Each target site therefore has an associated ‘fly-by position’ on
which any site is accessible to the platform varies at roughly
any given day; this is the point on the sub-satellite track that is
eight-day intervals, but with some shifting in MZA, because the
closest to the target. The platform acquires images of the target
orbit does not repeat exactly. For example, Barrax was visible
when the zenith angle of the platform, with respect to the fly-
for just two successive days a week later (MZA 15◦ and −9◦ on
by position, is one of the following: ±55◦, ±36◦ and 0◦. The
the 20th and 21st, respectively).
zenith angles with respect to the target will typically be some-
B. Mission Science
what larger than these values, especially for the ‘nadir’ images.
In practice, this means that the angles at which images are ac-
Reflectance from Earth’s surface is almost always anisotropic.
quired vary from site-to-site, depending on their positions with
As a result, a measurement of surface reflectance acquired at a
respect to the orbital track. By comparison, the view zenith an-
single sensor view angle has only limited usefulness in terms of
gles at which the MISR cameras acquire image data are, to the
estimating the surface albedo, which is a critically important pa-
nearest degree, ±70◦, ±60◦, ±46◦, ±26◦ and nadir [5], while
rameter of the surface radiation budget and, hence, the surface
the Along Track Scanning Radiometer series (ATSR, ATSR-2
energy budget [8], [9]. The directional reflectance properties

---

7
of Earth’s surface also mean that a comparison of reflectance
measurements made at two different times (e.g. for change de-
tection purposes) is likely to be compromised, if variations in
reflectance with viewing geometry are not taken into account.
Thus, the proper interpretation of a measurement made at a sin-
gle view angle often requires an understanding of the directional
reflectance properties of the surface; simple spectral discrimina-
tion of different cover types is perhaps the only exception, al-
though even here a set of directional-spectral measurements can
help to improve the separability of different vegetation types,
when compared to a single multispectral image [10].
At the risk of over-simplification, one might argue that spec-
tral variation in reflectance is largely determined by the chem-
(a) −55◦
ical composition of the surface materials, while angular varia-
tion is determined their structural properties [11]. One of the
main contributions of the PROBA/CHRIS mission, therefore,
will be the validation of surface reflectance models, at spatial
scales that can be tied to simultaneous ground measurements of
canopy variables [12].
The atmosphere also reflects light non-isotropically. Scatter-
ing by permanent gases is well characterised by theory and it
is relatively simple to account for this contribution to the top-
of-atmosphere signal. Scattering by optically active aerosols is
more problematic. This has a strong forward-scattering compo-
nent, but the precise phase function depends on the size distri-
bution of the particles present. Again, as for the land, the di-
(b) −36◦
rectional quality of the atmospheric reflectance distribution car-
ries information on the causes of the anisotropy, and multi-view
data have been used to estimate concentrations of optically ac-
tive aerosols [7], [13].
The initial plan for the data from CHRIS was designed to ex-
ploit the multiple-view-angle capabilities of the instrument to re-
trieve either atmospheric aerosol properties or land-surface geo-
physical properties related to the BRDF. In each case, the inten-
tion was, initially at least, to collect data over a limited num-
ber of well-characterised and instrumented sites (e.g. Aeronet
sites in the case of aerosols and selected MODIS MODLAND
Cal/Val sites for the surface properties). Coastal studies were
later added, since the CHRIS instrument shares some heritage
(c)
in design terms with the MERIS sensor, and a decision was
+36◦
therefore made to broaden the scientific user base. It was subse-
quently decided to open up the mission to a still wider scientific
audience by means of an Announcement of Opportunity, which
was issued by ESA in 1999. As a result, there are now some
60 locations on the list of scientific sites for which CHRIS ac-
quires images, and these have been partly prioritised. The mis-
sion now includes a number of sites for the study of coastal and
inland waters, and a wider range of land applications including
forestry, agriculture and surface energy budgets. Some of these
exploit the programmability of the CHRIS to select different sets
of spectral channels for specific applications.
B.1 Land-Surface Studies
(d) +55◦
Models of the BRDF for vegetation canopies show that the
Fig. 5.
Four single-band images from a multiple-view-angle set acquired by
varying intensity in reflected radiance is determined principally
PROBA/CHRIS over Lana’i, one of the Hawaiian islands, on December 6,
by structural features, such as gaps and the arrangement, orienta-
2002.
tion and spacing of the surface-scattering elements (e.g. leaves,
stems and branches) [14]. The more parsimonious models of

---

8
BRDF allow most of the surface reflection, aside from the ‘hot-
AATSR and MISR pixels, and should make possible a more
spot’, to be captured in terms of a few parameters. Character-
sensible validation of the latter with the former. At some, non-
isation of the surface in terms of an appropriate BRDF model
Aeronet sites, ground-based lidars will be deployed as part of the
enables robust, physically based relationships between albedo
PROBA/CHRIS programme to obtain vertical profiles of aerosol
and directional radiance to be developed, but the validation of
distributions.
these models over large areas is difficult. In this context, the
directional data from PROBA/CHRIS will be used both for the
B.3 Coastal and Inland Waters
validation and refinement of top-of-canopy BRDF models, and
The remote sensing of coastal and inland waters requires sen-
for the retrieval of geophysical parameters, such as leaf area
sors with a high spatial resolution and the ability to provide
index and canopy chlorophyll content, by inversion of simpli-
multi-temporal imagery. Cracknell has argued that Earth obser-
fied versions of those models, or through the use of look-up
vation has been far less successful in the shallow coastal zone
table (LUT) approaches applied to more sophisticated models
compared to the deep ocean and atmosphere, in part because
[12]. This original theme is now joined by a range of land
limited scanner resolution and image frequency have prevented
surface studies, including the assimilation of CHRIS data into
observation of the relevant scales in shelf seas and estuaries [26].
models of plant growth, as well as more conventional red-edge
PROBA/CHRIS provides a higher spatial resolution than exist-
position (REP) investigations and land cover studies. In many
ing ocean colour sensors (MERIS currently has the highest spa-
cases, simultaneous measurements of surface reflectance and at-
tial resolution, which is 300m). The critical question will be
mospheric optical properties are needed to validate the CHRIS
whether PROBA/CHRIS has the required radiometric sensitiv-
data and the derived products, so that dedicated fieldwork is re-
ity to map not just suspended particulate matter, but also phyto-
quired, although automatically instrumented sites have also been
plankton, since it was not originally designed to act a sensor for
included.
the marine environment.
CHRIS is being used over a small number of coastal and in-
B.2 Atmospheric Aerosols
land water sites in Europe. The core test sites are well charac-
Aerosols (liquid droplets and atmospheric particulate matter)
terised and represent a range of Case II waters (dominated by
are an important radiative constituent of the atmosphere, with
sediment or coloured dissolved organic matter), although some
sulphate aerosols in particular able to reflect significant amounts
of which on occasion are more similar to Case I (phytoplankton
of energy back to space. Apart from their direct radiative forcing
dominated) waters. Wherever possible, the sites and acquisi-
effect, aerosols have an indirect effect through their role as cloud
tions are chosen to augment existing, pre-planned, campaigns
condensation nuclei, although the interaction between clouds
with research ship support. Most sites have in situ sampling
and aerosols is a complex story with many uncertainties. Lit-
(both optical and bio-geochemical) during the CHRIS acquisi-
tle is known about global aerosol distribution, and how it varies
tion periods, sun photometer (atmospheric) measurements and
in time, although this is being rectified with the expansion of
other forms of remote sensing (additional airborne/satellite sys-
the Aeronet system [15] and the continuous delivery of global
tems). Accurate atmospheric correction is very important to
aerosol products from instrument such as MISR [5].
these studies, as the water-leaving signal represents a small con-
MVA data can be used to estimate aerosol properties under
tribution to the top-of-the-atmosphere value. The main contri-
appropriate directional sampling conditions [16], [17], which
bution of the directional component of the data, here, is to make
was the rationale behind the nine-view MISR instrument [18],
possible a more accurate atmospheric correction.
[19], [5]. Nine viewing directions may not always be needed;
the ATSR2 instrument [20] combines just two views of the
C. Acquisition Planning/Schedule : Ground Segment
ground in four solar channels, but this has been found ade-
In keeping with the experimental nature of the overall mis-
quate for good estimates of aerosol optical depth [21], [22], [7],
sion, the ground segment for the scientific component is rela-
[23]. In comparison with the existing MVA instruments, which
tively modest. There is a steering committee for the CHRIS
have just a few spectral channels, the CHRIS allows an almost
instrument, which considers requests for scientific acquisitions
continuous spectrum to be obtained at high spectral resolution
and which has developed a simple scheme for prioritising data
(∼ 10nm), and it is hoped the extra spectral information will al-
acquisitions among the different sites and users, with the aim
low inferences to be made about the particle size distribution of
of maximising the acquisition of useful data. Highest priority
the aerosols.
tends to be given to experiments where large ground-campaigns
Existing MVA satellite-borne instruments tend to have large
are taking place simultaneous to the CHRIS overpasses, particu-
footprints — 2–4km in the case of ATSR2, about 14km in the
larly those involving, for example, ship-time or the deployment
case of POLDER [24], [25], and roughly 20km in terms of the
of ground-based lidars, or similar equipment, away from their
global aerosol product delivered by MISR. The dangers of scal-
usual base. Investigations where angular sampling is an impor-
ing errors in such retrievals may be real, given that aerosol con-
tant element of the science are preferred to those that are inter-
centration can vary significantly over smaller lengths scales, es-
ested merely in hyperspectral data; similarly, instrumented sites
pecially in the vicinity of aerosol sources. The high spatial res-
are preferred to non-instrumented ones. Perhaps the strongest
olution of the CHRIS may allow any such scaling effect to be
determinant of all, however, is weather. A prioritised short-list
characterised. Perhaps more importantly, the retrievals from
of candidate sites is compiled for each day, a week or so in ad-
CHRIS are at a scale between the point measurements of the
vance, but the choice of which image to take on any given day
Aeronet network, and the multiple-kilometre areas covered by
rests on the expected cloud cover for each site, based on predic-

---

9
tions obtained from the UK MetOffice. For major campaigns,
Verstraete, “Multi-angle imaging spectroradiometer (misr): instrument de-
the image is taken even if the likelihood of overcast conditions
scription and experiment overview,” IEEE Transactions on Geoscience
and Remote Sensing, vol. 36, no. 4, pp. 1500–1530, 1998.
is greater than 50%, but for remotely-instrumented sites, such as
[6]
C. Godsalve, “Bidirectional reflectance sampling by atsr-2 - a combined
Aeronet locations, a threshold (10%) of acceptable cloud cover
orbit and scan model,” International Journal of Remote Sensing, vol. 16,
is imposed. This experience of scheduling is expected to be
no. 2, pp. 269–300, 1995.
[7]
P. North, S. Briggs, S. Plummer, and J. Settle, “Retrieval of land surface
very useful material in terms of planning SPECTRA and other
bidirectional reflectance and aerosol opacity from atsr-2 multi-angle im-
follow-on missions to PROBA/CHRIS.
agery,” IEEE Transactions on Geoscience and Remote Sensing, vol. 37,
no. 1, pp. 526–537, 1999.
[8]
D. S. Kimes, P. J. Sellers, and D. J. Diner, “Extraction of spectral hemi-
V. SUMMARY
spherical reflectance (albedo) of surfaces from nadir and directional re-
flectance data,” International Journal of Remote Sensing, vol. 8, pp. 1727–
The PROBA platform is one of the first products arising from
1746, 1987.
a new and increasingly influential philosophy within the Euro-
[9]
M. J. Barnsley, P. Lewis, M. Sutherland, and J. Muller, “Estimating land
pean Space Agency and the British National Space Centre, fol-
surface albedo in the HAPEX-Sahel southern super- site: inversion of two
BRDF models against multiple angle ASAS images,” Journal of Hydrol-
lowing a NASA lead. To ensure that satellite-based Earth ob-
ogy, vol. 189, no. 1-4, pp. 749–778, 1997.
servation missions can be delivered in a time-frame that makes
[10] A. H. Hyman and M. J. Barnsley, “On the potential for land cover mapping
their data useful to scientists, missions need to become more
from multiple-view-angle (mva) remotely-sensed images,” International
Journal of Remote Sensing, vol. 18, no. 11, pp. 2471–2475, 1997.
tightly focused, and the time taken from the original concept to
[11] M. J. Barnsley, D. Allison, and P. Lewis, “On the information content
launch and operation must be greatly reduced, if the necessary
of multiple view angle (mva) images,” International Journal of Remote
responsiveness is to be realised. The mission considered here
Sensing, vol. 18, no. 9, pp. 1937–1960, 1997.
[12] M. Barnsley, P. Lewis, S. O’Dwyer, M. Disney, P. Hobson, M. Cutter, and
has managed to put an imaging spectrometer (CHRIS) into Earth
D. Lobb, “On the potential of chris/proba for estimating vegetation canopy
orbit in a little under three years, notwithstanding a long delayed
properties from space,” Remote Sensing Reviews, vol. 19, pp. 171–189,
launch, for a cost of about 10 million US dollars, or Euros. The
2000.
[13] P. R. J. North, “Estimation of aerosol opacity and land surface bidirec-
mission is about more than showing how quickly something can
tional reflectance from atsr-2 dual-angle imagery: operational method and
be put into space, however. Innovative design in both the plat-
validation,” Journal of Geophysical Research, vol. 107 (D12), no. 4149,
form and spectrometer have also been pursued successfully and,
2002.
[14] N. S. Goel, “Models of vegetation canopy reflectance and their use in the
although the data from the CHRIS will not, strictly, answer fo-
estimation of biophysical parameters from reflectance data,” Remote Sens-
cused science questions that can only be addressed through the
ing Reviews, vol. 3, pp. 1–212, 1987.
use of a pointable imaging spectrometer, the highly resolved,
[15] B. N. Holben, T. F. Eck, I. Slutsker, D. Tanré, J. P. Buis, A. Setzer, E. Ver-
mote, J. A. Reagan, Y. J. Kaufman, T. Nakajima, T. Lavenu, I. Jankowiak,
multi-view, hyperspectral datasets that it generates are provid-
and A. Smirnov, “Aeronet — a federated instrument network and data
ing a rich source of data that will take several years to analyse,
archive for aerosol characterization,” Remote Sensing of Environment,
vol. 66, no. 1, pp. 1–16, 1998.
and that is already helping to shape the related missions that are
[16] J. V. Martonchik and D. J. Diner, “Retrieval of aerosol and land surface
to follow.
optical properties from multi-angle satellite imagery,” IEEE Transactions
on Geoscience and Remote Sensing, vol. 30, pp. 223–230, 1992.
A
[17] J. V. Martonchik, D. J. Diner, R. A. Kahn, T. P. Ackerman, M. M. Ver-
CKNOWLEDGEMENTS
straete, B. Pinty, and H. R. Gordon, “Techniques for the retrieval of aerosol
properties over land and ocean using multiangle imaging,” IEEE Geo-
MJB would like to acknowledge the support of the UK
science and Remote Sensing, vol. 36, pp. 1212–1227, 1998.
Natural Environment Research Council, through grant number
[18] D. J. Diner, C. J. Bruegge, J. V. Martonchik, T. P. Ackerman, R. Davies,
NER/Z/S/1999/00130. Thanks are also due to Dr. P. Lewis,
S. A. W. Gerstl, H. R. Gordon, P. J. Sellers, J. Clark, J. A. Daniels, E. D.
Danielson, V. G. Duval, K. P. Klaasen, G. W. Lilienthal, D. I. Nakamoto,
Dr. T. Quaife and Dr. G. Thackrah at University College London
R. J. Pagano, and T. H. Reilly, “Misr: A multiangle imaging spectro-
for their help and advice in connection with the PROBA/CHRIS
radiometer for geophysical and climatological research from eos,” IEEE
project. The image data presented in this paper are derived from
Transactions on Geoscience and Remote Sensing, vol. 27, pp. 200–214,
1989.
the CHRIS instrument, developed by Sira Technology Ltd (for-
[19] D. J. Diner, C. J. Bruegge, J. V. Martonchik, G. W. Bothwell, E. D. Daniel-
merly Sira Electro-Optics Ltd), with support from the British
son, V. G. Ford, L. E. Hovland, K. L. Jones, and M. L. White, “A multi-
National Space Centre, mounted on board the European Space
angle image spectroradiometer for terrestrial remote sensing with the earth
observing system,” International Journal of Imaging Systems and Technol-
Agency’s PROBA-1 platform.
ogy, vol. 3, pp. 92–107, 1991.
[20] A. J. Prata, R. P. Cechet, I. J. Barton, and D. T. Llewellyn-Jones, “The
REFERENCES
along-track scanning radiometer for ers-1 : Scan geometry and data sim-
ulation,” IEEE Transactions on Geoscience and Remote Sensing, vol. 28,
[1]
M. J. Barnsley, “Environmental monitoring using multiple-view-angle
pp. 3–13, 1990.
(mva) remotely-sensed data,” in Environmental remote sensing from global
[21] R. J. Flowedew and J. D. Haigh, “Retrieval of aerosol optical thickness
to regional scales (G. M. Foody and P. J. Curran, eds.), pp. 181–201,
over land using the ATSR-2 dual look satellite radiometer,” Geophysical
Chichester: John Wiley and Sons, 1994.
Research Letters, vol. 23, pp. 351–354, 1996.
[2]
H. E. McCurdy, Faster, Better, Cheaper. Boston, USA: Johns Hopkins
[22] G. Mackay, M. D. Steven, and J. A. Clark, “An atmospheric correction
University Press, 2001.
procedure for the ATSR-2 visible and near-infrared land surface data,” In-
[3]
M. J. Barnsley, K. P. Morris, A. H. Strahler, and J.-P. Muller, “Sampling
ternational Journal of Remote Sensing, vol. 19, pp. 2949–2968, 1998.
the surface bidirectional reflectance distribution function (brdf): Evalu-
[23] P. North, “Estimation of fapar, lai, and vegetation fractional cover from
ation of current and future satellite sensors,” Remote Sensing Reviews,
atsr-2 imagery,” Remote sensing of Environment, vol. 80, pp. 114–121,
vol. 8, pp. 893–916, 1994.
2002.
[4]
P. van den Braembussche, L. Dayers, D. Rosseneu, and J. Bermyn, “Proba
[24] P.-Y. Deschamps, F. M. Bréon, M. Leroy, A. Podaire, A. Bricaud, J. C.
mission analysis,” Tech. Rep. PROBA-TN-080-VE, ESA, Noordwijk,
Buriez, and G. Seze, “The polder mission — instrument characteristics
Netherlands, 1999.
and scientific objectives,” IEEE Transactions on Geoscience and Remote
[5]
D. J. Diner, J. C. Beckert, T. H. Reilly, T. P. Ackerman, C. J. Bruegge,
Sensing, vol. 32, no. 3, pp. 598–615, 1994.
J. E. Conel, R. Davies, S. A. W. Gerstl, H. R. Gordon, R. A. Kahn, J. V.
[25] M. Leroy, J. L. Deuze, F. M. Breon, O. Hautecoeur, M. Herman, J. C.
Martonchik, J.-P. Muller, R. Myneni, B. Pinty, P. J. Sellers, and M. M.
Buriez, D. Tanre, S. Bouffies, P. Chazette, and J. L. Roujean, “Retrieval

---

10
of atmospheric properties and surface bidirectional reflectances over
land from polder/adeos,” Journal of Geophysical Research-Atmospheres,
vol. 102, no. D14, pp. 17023–17037, 1997.
[26] A. P. Cracknell, “Remote sensing techniques in estuaries and coastal zones
— an update,” International Journal of Remote Sensing, vol. 20, pp. 485–
496, 1999.
Michael J. Barnsley is Research Professor of Remote
Sensing and Head of Department in the Department of
Geography, University of Wales Swansea, Singleton
Park, Swansea SA2 8PP, UK. He is Director of the
NERC Climate and Land-Surface Systems Interaction
Centre (CLASSIC), based at Swansea.
Jeff Settle is Senior Research Fellow in the NERC
Environmental Systems Science Centre, University of
Reading, UK. He holds a MA from the University
of Cambridge, UK and a PhD from the University of
London, UK.
Mike Cutter is Business Director for Space at Sira
Technology Ltd, part of the Sira Group of companies,
based in Chislehurst, Kent, UK.
Dan Lobb is the Senior Systems Engineer at Sira
Technology Ltd, part of the Sira Group of companies,
based in Chislehurst, Kent, UK.
Frederic Teston is the PROBA Project Manager at
ESA. He is currently based at ESA/ESTEC, Noord-
wijk, The Netherlands.

---