Figure 4 evaluates the potential effects of varying illumination angle on the azimuth measurements, comparing average azimuth for each sol (through Sol 238) and local true solar time of image acquisition. The Mastcam-Z mosaics that were analyzed typically required a few minutes to acquire, so the local true solar time of the midpoint of each mosaic acquisition was used for this analysis.  While there is more scatter in the measurements when the sun was lower, there is no apparent systematic effect of illumination differences on the results. Potential systematic errors from viewing azimuth (measured clockwise from N, centered on Mastcam-Z) were also evaluated in ventifact orientations measured through Sol 238. Figure 5 shows no apparent dependence of feature azimuth on viewing geometry either. We therefore conclude that neither illumination nor viewing geometry significantly affects our results.
Figure 4. Bearing (azimuth clockwise from north) of ventifact features on the floor of Jezero crater, averaged over measurements made in each Mastcam-Z panorama (acquisition sol indicated in boxes) through Sol 238. Vertical bars show the standard deviation of each range of measurements. Three panoramas were acquired between 12:04 and 12:41 on Sol 180; the midpoint time of the second one is plotted here. The two bearing measurements of features seen in the Sol 112 panorama differ by only 0.1° so standard deviation is not visible for the Sol 112 data point. There is no significant correlation between time of day and ventifact orientation, indicating that illumination conditions do not affect these three-dimensional measurements.
Figure 5 . Bearing (azimuth clockwise from north) of ventifact features on the floor of Jezero crater vs. viewing azimuth for data acquired through Sol 238.
Overall, variations in ventifact orientations along the first 400 sols of the rover traverse are minor, as shown in Figure 6 , perhaps because the area explored by the Mars 2020 rover over that period is only about 1 km across. Some of the variability in linear feature orientations is due to variations in the 3D geometry of the rock faces relative to the wind azimuth, as expected. It is also expected that local topography will deflect sand-transporting winds and that some of the observed variability may be caused by nearby topographic obstacles. To determine the significance of such effects on our measurements, individual ventifact azimuth measurements that differ from the global average by more than two standard deviations were examined.
Many of the >2-sigma variations in azimuth appear to be caused by the deflection of winds on more steeply inclined rock faces (e.g., on Sols 53, 254, 353, and 362). In these cases, differences between the normal to the abraded rock face and the direction of strong winds may cause the winds to be deflected horizontally. Such deflections have been observed on terrestrial ventifacts. To understand this, consider a hemispherical rock that is subjected to abrasion by saltating sand blowing from a single direction. The orientation of aeolian flutes will depend on the local surface normal at each grain impact site, resulting in variable flute orientations that radiate away from the upstream point on the rock. Lower-resolution topographic data on more distant features also contributed to uncertainty and more scatter in measured azimuths on some features. Evidence for large-scale wind deflections by upstream topography is limited, with possible cases observed on Sols 53 and 180. As described above, the Rochette float block may have moved slightly after the ventifact features were formed, contributing to differences in their orientations relative to other ventifacts.
Figure 6 . Locations of measurements of aeolian feature orientations (arrows) plotted on HiRISE color image (north up, illumination from left). White dots show rover locations, connected by grey traverse path with white arrows showing drive direction. Sols and average orientations of ventifact features shown by yellow arrows, average orientations of wind tails shown as green arrows. Wind tails were measured at multiple locations on Sols 340, 341, and 360; suffixes “a” and “b” were added to the sol numbers annotating these locations. Wind tail orientations measured later in the mission are not included, as the traverse extended to the north after Sol 379, beyond the area mapped.
4 Comparison with indicators of recent aeolian transport
Wind tail orientations, averaged over observations made at each rover location, are summarized in Table 2. These 101 downwind orientations were measured using Navcam stereo data in all cases except on Sols 34 and 52, when Mastcam-Z stereo images were analyzed. Sometimes images obtained at mid-drive pauses showed measurable features (i.e., in addition to features at the final, end-of-drive position). These instances are distinguished in Table 2 with letters appended to the sol number. Downwind azimuths that differ from the mean by more than two standard deviations were measured in images acquired on Sols 130, 137, 173, 199, 202, and 355. In all cases, topographic features near the measured wind tails are too small to have affected wind flow significantly, so the cause of the orientation differences is unclear. There is no evidence that differences in grain or feature size influenced susceptibility to recent wind modification.
Table 2. Sol number of image acquisition, number of measurements (N), average wind tail azimuth (clockwise from north, in degrees) and standard deviation. Measurements were made at multiple rover locations on Sols 32, 47, 340, 341, 355, and 360.