Plain Language Summary
Strong winds mobilize sand grains that can abrade rock surfaces to form
lineated bedrock features. These wind abrasion textures record the
direction of the sand-driving winds that created them. Wind abrasion
textures were observed on rocks on the floor of Jezero crater traversed
during the early part of the Mars 2020 mission. Orientations of these
abrasion textures were measured using rover-acquired three-dimensional
stereo image data. Results indicate that the sand-driving wind
directions that abraded these rocks are very different—nearly
opposite—from the current strongest wind directions indicated by
orientations of sandy wind tails extending from behind obstacles,
measured wind velocities, and climate model predictions for the area.
Collectively, these results provide a record of changing atmospheric
circulation patterns in the Jezero region.
1 Introduction
Ventifacts are rocks that have been abraded and shaped by wind-blown,
sand-size particles. The abrasion process forms flutes, pits, and
grooves that typically align along the flow direction of strong,
sand-driving winds on both Earth and Mars (Bridges et al., 2004; Laity
& Bridges, 2009). There is abundant evidence that aeolian abrasion
modifies exposed rock surfaces on Mars, based on observations of
ventifacts at previous landing sites (Bridges et al., 1999, 2014;
Greeley et al., 2006, 2008; Thomson et al., 2008). Ventifacts also are
present on the floor of Jezero crater, as observed by thePerseverance rover during the first 400 sols (martian days) of
the Mars 2020 mission (Farley et al., 2020) and are thought to record
the direction of the winds that formed them (Herkenhoff et al., 2021).
This study reports orientation measurements of ventifact textures on the
floor of Jezero and compares these with orientation measurements of
recent wind direction features (regolith wind tails, i.e., sandy drifts
behind obstacles). Similar comparisons of wind direction inferences at
previous landing sites have shown significant differences in recent and
paleo-wind indicators, suggesting significant changes in wind
directions. The implications of our observations in Jezero crater for
paleoclimatic variations are discussed along with similar results at
other Martian landing sites. Atmospheric circulation modeling results
are then presented and compared with the measured orientations.
2 Methods
Ventifacts and other wind-formed features were identified and measured
in stereo data derived from images acquired by the Navcam (Maki et al.,
2020; Maki, 2021) and Mastcam-Z (Bell et al., 2021; Hayes et al., 2021;
Bell & Maki, 2021) cameras on the Perseverance rover. Navcam has
a focal length of 19.1 mm and instantaneous field of view (IFOV) of 330
µrad/pixel, while Mastcam-Z’s focal length ranges from 26 mm (IFOV of
283 µrad/pixel) to 110 mm (IFOV of 67.4 µrad/pixel) (Maki et al., 2020;
Bell et al., 2021). At a typical range from the cameras of 3 m, Navcam
and Mastcam-Z 110 mm focal length images have pixel scales of 1 and 0.2
mm, respectively. Mastcam-Z stereo mosaics acquired at 110 mm focal
length allowed more detailed views supporting more accurate assessments
of ventifact texture and orientations compared with shorter focal
lengths and other onboard camera systems, so these were used heavily for
ventifact orientation measurements in this study. All measured features
were within 30 meters of the rover cameras when images were acquired.
Care was taken to measure only aeolian abrasion textures with clearly
expressed orientations; features that appeared to be aligned along
bedding or other primary structures were not analyzed further.
Once linear abrasion features of interest were identified, their
orientations were measured in digital ordered point cloud (OPC) datasets
using the PRo3D software package (Barnes et al., 2018). The OPCs were
produced by 3D vision processing of Mastcam-Z stereo pairs (Paar et al.,
2022), georeferenced by the 20 to 50-cm-level-accurate rover
localization (relative to HiRISE DTMs) provided by the Mapping
Specialists on the Mars 2020 science team. At the maximum distance of
measured features (30 m), the Mastcam-Z stereo range error is 7 cm. Each
end of linear features, such as flutes, was selected manually and the
bearing and slope of the line between them were recorded; an example is
shown in Figure 1. Even at the maximum range of 30 m, feature
measurements span multiple OPC grid points, so that the features are
well resolved in the OPCs. Upwind points were picked first, followed by
downwind points, to maintain a consistent dataset. Typically, the upwind
ends of flutes are at a lower elevation than the downwind ends (Laity &
Bridges, 2009), but an exception was found on the rock named
“Rochette.” As shown in Figure 2, Rochette does not appear to be
in-place bedrock, and therefore may have tilted since the aeolian
abrasion features were formed. In this one case, the downwind direction
was reversed for comparison with other measurements, assuming that after
most ventifact textures were formed on Rochette, the block tilted
(counterclockwise about a ~horizontal axis in the mosaic
view) slightly to place the downwind ends of the aeolian flutes at a
lower elevation than the upwind ends. This assumption is supported by
observations of linear features on larger rocks in the same Mastcam-Z
mosaic, which show more typical flute orientations (e.g., Figure 1).
Measurements of features that were questionable or poorly resolved in
the OPC were not included in statistical analyses. Flutes on nearly
vertical rock faces showed higher variability in azimuth, as expected
because wind-driven particles are deflected laterally when impacting
such rock faces (Laity & Bridges, 2009).
Figure 1. Example
measurements of flute orientations on ventifact. View toward south,
approximate true color. (left) Part of Mastcam-Z image acquired on Sol
180 by sequence ZCAM08195. Three-dimensional (3D) downwind orientations
of flutes indicated by yellow arrows were measured. (right) PRo3D
visualization of ordered point cloud data derived from Mastcam-Z 110-mm
stereo images acquired by same sequence. Endpoints of flutes were
selected manually and are shown connected by blue lines. The highlighted
line with red endpoints is 2 cm long. Context is shown in Movie S1.
Figure 2. Rochette
ventifact. (top) Measurements of linear abrasion features on Rochette
rock using PRo3D, shown as blue lines. Data at lower right for line
highlighted with red dots at endpoints (length is 9.6 cm). (bottom) Part
of enhanced-color Mastcam-Z mosaic ZCAM08195 acquired on Sol 180,
showing Rochette rock and nearby terrain. View toward south-southeast.
The orientations of regolith wind tails (drifts or sand shadows)
extending from behind rocks were measured using Navcam and Mastcam-Z
stereo data acquired through Sol 400. The three-dimensional (3D)
locations of the upwind and downwind ends of each wind tail were
measured using Mars 2020 flight operations ASTTRO software (Advanced
Science Targeting Toolkit for Robotic Operations; Abercrombie et al.,
2019), allowing slope and azimuth (clockwise from north) of each feature
to be derived. An example downwind endpoint is shown in Figure
3 . Similar to ventifact measurements, azimuths of regolith wind tails
were calculated in the inferred downwind direction. An important
qualitative criterion for feature selection was relative isolation in
open ground away from adjacent large obstacles like outcrops, boulder
clusters, and other features that might have blocked winds and/or
wind-blown material from some azimuth ranges, or locally altered the
wind direction.
Figure 3. Screen
capture of Mars 2020 ASTTRO visualization of part of Navcam image
acquired on Sol 355, showing location of end of wind tail (length = 17.3
cm) at white cross. Other wind tails are visible near right side of
frame and 3-D (XYZ) and other data shown at far right.
3 Ventifact Orientation Results
Measurements of flutes using Mastcam-Z mosaics of the rock “Rochette”
acquired on Sols 180 and 197 (after sample acquisition) differ by 7
degrees, reflecting the uncertainties inherent in our approach including
manual point selection and variations in viewing geometry. Standard
deviations of each sol’s measurements are similar to this value
(Table 1 ). Early in the mission, OPCs were generated in the
rover coordinate frame, so that the rover heading had to be subtracted
from measurements of features (corrections for tilt were not performed
due to our focus on the azimuth of ventifact features). More recently,
OPCs have been generated in the Mars local level coordinate frame.
Measurements made using the new OPCs agree with measurements of the same
features made using the old OPCs to within 4 degrees, reflecting errors
in coordinate transformations and uncertainties in picking endpoints of
linear features. The orientations of linear abrasion features seen in
Navcam and Mastcam-Z stereo data acquired through Sol 57 were measured
independently, and agree very well with the OPC measurements: the
average downwind azimuth of features measured in terrain meshes
generated at the Jet Propulsion Laboratory (JPL) using Navcam and
Mastcam-Z images is 97° ± 6° while the average downwind azimuth of
features measured in Mastcam-Z OPCs is 95° ± 4° (in both cases, the
standard deviation in the measured azimuths is given rather than the
uncertainty in the individual measurements). Similarly, the average
downwind azimuth of all features measured in OPCs derived from Mastcam-Z
stereo images acquired through Sol 400 is 94° ± 7°. Based on the
comparisons summarized above, true variations in the orientations of
linear features exceed the uncertainties in the measurements.
Table 1 . Sol number of
image acquisitions for each set of aeolian abrasion features, number of
orientation measurements (N), average downwind feature azimuth
(clockwise from north, in degrees) and standard deviation for each sol.