APXS Raster Localization Using MAHLI

46th Lunar and Planetary Science Conference (2015)
Gellert1, L.M. Thompson2, J.A. Berger3, N.I. Boyd1, J.L. Campbell1, K.S. Edgett4, M.E. Minitti5, M.E. Schmidt6,
A.S. Yen7, and the MSL Team, 1University of Guelph, Guelph ON, Canada, 2University of New Brunswick, Fredericton NB, Canada, 3University of Western Ontario, London ON, Canada, 4Malin Space Science Systems, San Diego
CA, USA, 5Planetary Science Institute, Tucson AZ, USA, 6Brock University, St. Catharines ON, Canada, 7California
Institute of Technology, Pasadena CA, USA.
Introduction and Rationale: The Alpha Particle
X-ray Spectrometer (APXS) on the Mars Science Laboratory (MSL) rover, Curiosity, is an arm-mounted
bulk chemistry instrument that performs high-precision
in-situ measurements of rocks and soils [1]. Compared
to the Mars Exploration Rover APXS, the MSL APXS
is 5x more sensitive and has a Peltier cooler that improves the nominal ambient operational temperature
from -40o C to -5o C [1, 2]. The increase in sensitivity
and improved integration during warmer temperatures
allows the APXS to characterize a larger area than its
nominal field of view (FOV) by utilizing a raster technique. Morning or evening rasters can be used to investigate sample heterogeneity (compositional and/or textural) on the order of the APXS FOV.
Arm-mounted instruments, including the APXS,
have placement uncertainty (accuracy <15 mm; precision <10 mm) [3]. Arm placement adds uncertainty to
the interpretation of APXS data given an in-contact
APXS FOV of ~15 mm [1]. Here, we combine APXS
data with images from the Mars Hand Lens Imager
(MAHLI) to constrain APXS positioning on the Martian targets Sayunei, Stephen and Morrison (Fig. 1).
Figure 1: MAHLI images of APXS raster targets
Sayunei (top left, sol 165), Stephen (top right, sols
627-629) and Morrison (bottom, sols 767-779).
Method: MAHLI images were analyzed by classifying pixels into orthogonal bins based on the assumption that visually distinct phases have a distinct APXS
composition. The orthogonality correlates directly to
elemental and/or spectral variation observed in the
APXS data. The APXS raster FOVs are overlaid on the
MAHLI image. The positional assumption, given iden-
tical MAHLI and APXS targets, is that the central
APXS FOV is aligned with the center of the MAHLI
image. The pixels within each APXS FOV are analyzed and tabulated based on their radially dependent
contribution resulting in a relative abundance of each
phase for each APXS FOV. The test raster is translated
and/or rotated to assess a different orientation. The
linear relationship between phase abundances and the
appropriate APXS spectral variation is compared using
a bivariate weighted least squares regression [4] to
infer a more-likely APXS raster position and orientation.
Sayunei is a mudstone from the Yellowknife Bay
area (John Klein class [5]) with a cross-cutting vein.
APXS data collected from a 2x2+1 raster showed elevated CaO and SO3 in three of the five raster positions
and was complimented by MAHLI images of the central APXS spot only [6]. This light-toned vein material
is visually distinct, resulted in CaO and SO3 variation
in the APXS spectra and is interpreted to be CaSO4. As
the amount of vein material in the APXS FOV increases, the SO3 measured by the APXS will increase. This
linear relationship between vein material (light-toned
pixels) and SO3 (wt%) was used to infer an APXS raster position that not only agrees with proportionality
but also with upper- and lower-limit bounds (Fig. 2).
Stephen is a thin fracture fill/layer within
silt/sandstone at the Kimberley waypoint. APXS data
collected from a 2x2+1 raster showed variation in the
APXS norm (proxy for instrument standoff) as well as
dust-dependent signals. The APXS raster was complimented by 5 cm MAHLI images on the all raster spots.
Pixels from the MAHLI image(s) were classified as
either dust, rock or negligible based on pixel saturation
level and sample topography. A combination of elemental and topographic linear relationships was used to
infer a more-representative APXS raster position (Fig.
Morrison is a brushed, raised, diagenetic feature at
the Pahrump waypoint [7, 8]. APXS data obtained via
a 2x2+1 raster showed elevated MgO, SO3, Ni and, to
a lesser extent, Cl and Br in the diagenetic feature relative to the immediately adjacent bedrock [7, 8].
MAHLI images of the target were taken on earlier sols
of the central APXS raster spot only. As with Stephen,
a ternary pixel classification system was used on Morrison, binning pixels orthogonally as dust, bare rock or
46th Lunar and Planetary Science Conference (2015)
rosette, with various linear correlations to spectral heterogeneity observed in the APXS data.
Results and Observations: Improved positioning
of APXS rasters through MAHLI pixel analysis resulted in significant shifting (4-8 mm) of the APXS raster
center from the center of the MAHLI image considering the size and sensitivity of the APXS’s FOV. APXS
placement deviation from the center of MAHLI images
is comparable to observed offsets of a Dust Removal
Tool brushed center and a post-brush MAHLI image
center (i.e. ~3 mm offset on sol 755 Maturango). The
APXS raster edges are not parallel to MAHLI image
edges as confirmed when MAHLI images are acquired
on each APXS raster spot.
APXS raster positioning for Sayunei exists within
arm placement uncertainty such that the elevated CaO
and SO3 observed is attributed to the vein material and
is not elevated in the host rock (Fig. 2). Combining
target topography and visually distinct phases improved the Stephen raster location (Fig. 3, rotation was
minimal as initial orientation was dictated by MAHLI
images of each APXS raster location). The bestagreement raster offsets from MAHLI image center for
Sayunei and Stephen are in the same direction, despite
the flipped arm orientation between the two locations,
hinting perhaps at a systematic placement error (Fig.4).
A preliminary analysis of the Morrison raster indicates
strongly that the APXS raster is offset from the
MAHLI image center as well.
Conclusions: APXS rasters are a powerful tool for
quickly analyzing heterogeneous samples. MAHLI
images with visually distinct phases can reduce APXS
FOV uncertainty, ultimately improving interpretation
of the target and its geological context.
Figure 2: Early raster position (top, raster parallel to
MAHLI image edges and centered on 5 cm MAHLI
image) and improved raster position (bottom) yielding
a strong correlation between vein material in the
APXS FOV and APXS SO3 concentration (wt%). For
scale, the central APXS FOV diameter is ~15 mm.
Figure 3: Raster position yielding an improved correlation between radially weighted APXS FOV pixel fill
and measured APXS norm. The updated position also
improved the correlation between dust-related signals
and the amount of dust in the APXS FOV.
Figure 4: Correlation plots of Sayunei (top) and Stephen (bottom) illustrating a raster center offset of ~8
mm and ~4 mm respectively from the appropriate
MAHLI image centers (marked by X) to the region of
best agreement (dark red).
References: [1] Gellert et al. (2014), LPSC XLV,
#1876. [2] Gellert et al. (2006), JGR, 111-E2. [3] Robinson et al. (2013), IEEE SoSE, 184-189. [4] York et
al. (2004), AJP, 72-3. [5] Schmidt et al. (2014), LPSC
XLV, #1504. [6] McLennan et al. (2013), Science,
1244734. [7] Thompson et al. (2015), LPSC XLVI
(this meeting). [8] Gellert et al. (2015), LPSC XLVI
(this meeting).
Acknowledgements: The MSL APXS is financed
and managed by the Canadian Space Agency (CSA)
with MacDonald Dettwiler and Associates as the primary contractor. Funding is provided by the CSA and
NASA. Much appreciation goes to JPL for their support, dedication and invaluable expertise.