Figure 17. Ratio between the
measured IOF and the model IOF values of (a) the primary white clean
spot, (b) the light gray clean spot, (c) the white grayscale ring and
(d) the secondary horizontal white patch. The colors of the points are
the same for the four plots and represent the narrow-band filters. The
vertical dashed line is the martian aphelion.
For the primary white patch (Figure 17a), the points were systematically
just below 1 for all filters spanning the range between 677 nm and 1022
nm, while filters L4 (605 nm), L5 (528 nm) and L6 (442 nm) displayed
progressively lower values. These three filters followed a similar
trend: a first interval, from landing to sol 100, characterized by a
linear and faster decrease, followed by a flatter decline which lasted
at least until the solar conjunction of sols 217-235. After conjunction,
they tended to be overall more stable (L6 seemed to increase slightly)
up to the period of the major dust event of sol 315. There was not a
clear relationship with the solar aphelion. For reference, Figure 17b
shows unchanging ratios for the light gray primary patch. This material
did not exhibit any significant deviation from the unit ratio before the
dust event. Although manufactured from the same material as the other
white patches, the white primary ring was not or only very slightly
influenced by the yellowing effect (Figure 17c).
Discussion
In this section we discuss the performance of the Mastcam-Z radiometric
calibration targets in light of the methods presented in section 3 and
the results presented in section 4. Principally, we want to evaluate the
performance of the cal-targets over the first 350 sols on Mars and the
model used to generate the reflectance-calibrated products. In addition,
we display a basic assessment of the dust on the cal-targets and the
effect on the diffuse light. Eventually, we describe the tests that were
aimed at understanding the visual and spectral deterioration of the
AluWhite98 material.
The Cal-Target Performance and the Linear Fit Model for IOF
Calibration
The performance of the Mastcam-Z calibration targets can be assessed
from the results reported in sections 3.2 and 4.2, and the visual
inspection of the regions involved in the IOF calibration process within
the color images (section 4.1). The plots of Figure 5 show a limited
dispersion of the data from the fit lines (except for the white spot),
where this dispersion was quantified as a relative error on the slopes
of less than 3.5% on average. This small deviation indicates that the
cal-targets were successful in achieving their main goal, especially
thanks to their design and the inclusion of the permanent magnets. The
number of clean spots, higher than in the previous missions,
considerably reduced the impact of the exclusion of one of those regions
(the AluWhite98) from calibration, thanks to the other seven regions.
The presence of the hollow magnets below the primary patches but not in
the secondary target was a useful way to visually infer the accumulation
of airfall dust on the surfaces. The magnet rings, as expected, were the
first regions to be covered in dust, as shown in and Figure 11. We
should also consider the action of the wind, which contributed to a
frequent and efficient deposition and cleaning from dust on other parts
of the cal-targets (e.g., on sols 327 and 349, as shown in the movie
S2). These effects on the clean spots were critical for the radiometric
calibration, because they ensured minimal dust disturbance of the clean
spots’ radiances when the linear fits were made with the corresponding
laboratory reflectances (Figure 5). As a consequence, the data points in
those plots were well described by the one-term fit model, which was
expected by the theory (equation (2 )). The ‘cleanliness’ of the
clean spots due to the magnets and the wind was adequately maintained in
time, as suggested by the solar irradiance time series (Figure 8a).
Indeed, the irradiance \(F\) followed a quite stable and smooth trend in
all filters up to the dust event of sols 314-316 that caused a
significant unsettlement in images and radiance values, but that did not
compromise the cal-targets or their surfaces. Further proof of the
effectiveness of the clean spot-magnet ring system was the solar
irradiance spectrum at the aphelion (Figure 8b), which is consistent
with a solar black body model allowing for some atmospheric absorption,
particularly at short wavelengths.
As mentioned in section 3.2 and fully treated in section 4.4, though the
one-term linear fits characterize well the radiance-reflectance data
points, we noticed a small offset in the distribution of the data points
with respect to the straight line passing through the origin. The
testing of a two-term model for the linear fits gave better statistical
outcomes. Over the whole sets of radiance-reflectance data, the one-term
model yielded a reduced \(\chi^{2}\) between 10.7 and 34.4, the two-term
yielded values between 1.43 and 4.71. The exact nature of this offset is
not yet known. It might be due to some computational source, such as
residuals from the radiance-calibration process when the corrections
(e.g., bias frames, flat fields, shutter frame subtraction) are applied
to the raw images, or slight discrepancies in the reflectance model that
is employed to give an estimate of the expected reflectance of each
clean spot at any illumination geometry. The presence of dust on the
surfaces would also tend – to first order – to result in a straight
line with an offset. The slowly increasing trend of the offset in all
the narrow-band filters might indicate a dust-related origin. This may
imply that a small fraction of weakly or even non-magnetic airfall dust
(not attracted or repelled by the magnets) gradually deposited within
the clean spots and adhered electrostatically, such that it could not be
swept by wind interaction. The appearance of the offset shortly after
landing might suggest that dust and sand were raised by the rockets of
the skycrane during landing. Deeper investigation is required to fully
understand the origin and the nature of this offset.
Preliminary dust assessment
We observed the martian dust both directly (by deposition on the
surfaces from which we measured the radiance and their surroundings) and
indirectly (by analyzing its effect on the light that illuminated the
cal-targets). As shown in the irradiance time series in Figure 8a, which
only refer to the clean spots (expected to be the cleanest regions of
the cal-targets), in absence of larger events dust never had a
significant impact on the cal-targets or their ability to correctly
perform the IOF calibration. On the other hand, as mentioned above, the
frames of the movie S2 express a frequent change in the material that
accumulated on the deck of the rover. We could recognize patterns of
fine dust layers, which were more evident on the deck and on the bright
materials due to the higher light/dark-toned contrast, and single larger
grains of sand or regolith raised from the surface, which in some cases
moved a few mm or even cm across two consecutive sols. Whereas the
peripheral regions of the primary target were likely ruled by the
magnetic field of the permanent magnets (in particular the round patches
and the outermost grayscale rings), which produced an optically thick
buildup of dark reddish dust on the magnet rings, the variations in all
the other regions were probably controlled by wind (e.g., several dust
devils were observed by Mastcam-Z in the rover site; Newman et
al. , 2022) and possibly aided by vibrations caused by the motion of the
rover, which traveled almost 4 km in the first 350 sols. This fine dust
not only deposited on horizontal surfaces, but also adhered to vertical
sides, such as the lower part of the golden base of the primary target
where the magnetic field from the permanent magnets is prominent and on
the cylindrical structure of the gnomon (the reddish dust coating is
just perceivable on its boundaries in Figure 7).
The model of the direct fraction of sunlight on the illuminated and
shaded grayscale rings was a powerful method to follow the presence of
dust suspended within the martian atmosphere. The two time series shown
in Figure 12 display a roughly inverse correlation in the sense that\(F_{d}\) rises when \(\tau_{I}\) falls and vice-versa. This can be
interpreted in a way that a growth in the density of dust in the
atmosphere leads to an increase of the optical depth measured from
Mastcam-Z images and increased diffuse scattering of sunlight, leading
to a decrease in \(F_{d}\). In addition, \(F_{d}\) and \(\tau_{I}\) show
a shallow local maximum and minimum, respectively, around the martian
aphelion. The correlation can be recognized after the solar conjunction,
when stronger perturbations caused a higher variability in \(F_{d}\) and\(\tau_{I}\), and upon the major dust event from sol 314.
The spectra of dust accumulated on the magnet rings (Figure 11d) are
consistent with previous observations of martian dust attracted to
permanent magnets (e.g., Madsen et al. , 2009). The dust is
brownish-red with very low reflectance factor at ultraviolet and blue
wavelength, a characteristic rise from green to red consistent with the
presence of ferric iron and higher reflectance factors in the red and
infrared. Overall, though, the spectrum is darker than a typical
spectrum of martian bright dust, again consistent with expectations for
magnetically attracted material that can be expected to be richer in
magnetite and with a larger average grain size (Kinch et al. ,
2006).
The AluWhite98
We have not yet been able to identify a root cause of the observed
yellowing of the white patches of the cal-targets. The white material is
different (AluWhite98 manufactured by Avian Technologies) from the seven
other materials (glazed ceramic material manufactured by Lucideon)
(Kinch et al ., 2020). This explains why this effect was only
observed in the white patches but it remains enigmatic why the white
ring was not or almost not affected. One hypothesis suggested that a
possible cause was the different type of epoxy adhesive employed to fix
on one hand the primary and secondary patches and tiles (Henkel/Loctite
EA-9309NA) and on the other hand the white ring (3M-2216B/A Gray) to
their supports.
Several tests of martian environment simulation were carried out on four
spare AluWhite98 samples. Three of these samples were fixed onto an
Aluminum support using the two types of epoxies mentioned above, whereas
the fourth was put in place without any adhesive. The samples underwent
the same baking process of preparation as those currently within the
cal-targets on Mars, and were treated with UV irradiation at the
University of Winnipeg. However, the tests did not reproduce the
visible, radiometric, and spectral outcomes of the in-flight materials.
This issue therefore remains unsolved.
Conclusions and future work
In this work we assessed the performance of the Mastcam-Z radiometric
Calibration Targets (or cal-targets; Kinch et al. , 2020),
regularly employed to convert Mastcam-Z images (Bell et al. ,
2021) from units of radiance to reflectance, over the first 350 sols on
Mars.
The cal-targets proved to be efficient not only for calibration, but
also to retrieve information on the environment and the dust dynamics
within Jezero crater. The design of the regions of interest for
calibration, surrounded by strongly magnetized hollow cylindrical
magnets, allowed accurate measurements of the local radiances (Hayeset al. , 2021) involving low disturbance due to dust, which was
mostly attracted or repelled by the magnets, or deposited and cleaned
off by the wind. Linear fits between model reflectances and observed
radiances are of good quality with only limited dispersion of data
points around the fit line. A continuous monitoring of the linear fits,
as well as their slopes (equal to the instantaneous local solar
irradiance), will ensure a correct application of the reflectance
calibration procedure in the future, as significant natural events (such
as e.g., the major dust event of sols 314-316) can directly affect the
atmospheric optical thickness and the cal-target materials, hence
perturbing the stability of the calibration process. This includes the
implementation of a dust model, which was already performed in the MER
mission (Kinch et al. , 2007). Due to the satisfying results from
Mastcam-Z cal-targets within the first 350 sols, we do not contemplate
the urgency of a dust model.
We did observe that the linear fits could consistently be improved by
the inclusion of a slight offset term. The origin of this offset is not
yet understood but plausible hypotheses include residuals of the
radiance calibration, imperfections of the reflectance model, or dust,
or some combination of the three.
Finally, the yellowing effect of the AluWhite98 patches of the primary
and secondary cal-targets could not yet be reproduced by experiments and
therefore for now it remains an unexplained phenomenon. Determining the
physical trigger of this effect could help establish the starting point
and provide a useful reference in the design of the new calibration
targets for the cameras of coming planetary missions.