Monitoring of Dust Devil Tracks Around the InSight Landing Site, Mars, and Comparison With In Situ Atmospheric Data

Abstract

The NASA InSight mission on Mars is a unique opportunity to study atmospheric processes both from orbit and in situ observations. We use post-landing high-resolution satellite images to monitor dust devil activity during the first eight months of the mission. We perform mapping and semi-automatic detection of newly formed dust devil tracks and analyze their characteristics (sizes, azimuths, distances, and directions of motion). We find a large number of tracks appearing shortly after landing, followed by a significant decrease of activity during late winter, then a progressive increase during early spring. New tracks are characterized by dark linear, to slightly curvilinear, traces ranging from a few to more than ten meters wide. Tracks are oriented in the ambient wind direction, according to measurements made by InSight's meteorological sensors. The systematic analysis of dust devil tracks is useful to have a better understanding of atmospheric and aeolian activity around InSight.

Full text

manuscript submitted to Geophysical Research Letters
Monitoring of Dust Devil Tracks Around the InSight1
Landing Site, Mars, and Comparison with in-situ2
Atmospheric Data3
Perrin, C.1, Rodriguez, S.1,14, Jacob, A.1, Lucas, A.1, Spiga, A.2,14 , Murdoch,4
N.3, Lorenz, R.4, Daubar, I. J.5, Pan, L.6, Kawamura, T.1, Lognonn´e, P.1,14,5
Banfield, D.7, Banks, M. E.8, Garcia, R. F.4, Newman, C. E.9, Ohja, L.10,6
Widmer-Schnidrig, R.11, McEwen, A. S.12, Banerdt, W. B.13
7
1Universit´e de Paris, Institut de Physique du Globe de Paris, CNRS F-75005 Paris, France8
2Laboratoire de M´et´eorologie Dynamique / Institut Pierre Simon Laplace (LMD/IPSL), Sorbonne9
Universit´e, Centre National de la Recherche Scientifique (CNRS), ´
Ecole Polytechnique, ´
Ecole Normale10
Sup´erieure (ENS), Campus Pierre et Marie Curie BP99, 4 place Jussieu, 75005 Paris, France11
3Institut Sup´erieur de l’A´eronautique et de l’Espace SUPAERO, 10 Avenue Edouard Belin, 3140012
Toulouse, France13
4Johns Hopkins University Applied Physics Laboratory, Laurel, Maryland 20723, USA14
5Department of Earth, Environmental, and Planetary Sciences, Brown University, Providence, USA15
6Universit´e de Lyon, Universit´e Claude Bernard Lyon 1, ENS de Lyon, CNRS, UMR 5276 Laboratoire de16
G´eologie de Lyon -Terre, Plan`etes, Environnement, 69622 Villeurbanne, France17
7Cornell Center for Astrophysics and Planetary Science, Cornell University, Ithaca, USA18
8NASA Goddard Space Flight Center, Greenbelt, MD 20771, USA19
9Aeolis Research, 333 N Dobson Road, Unit 5, Chandler AZ 85224-4412, United States20
10Department of Earth and Planetary Sciences, Johns Hopkins University, Baltimore, MD, USA21
11Black Forest Observatory, Stuttgart University, Heubach 206, D-77709 Wolfach, Germany22
12Lunar and Planetary Laboratory, University of Arizona, Tucson, Arizona, USA23
13Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California, 911098099 USA24
14Institut Universitaire de France, 1 rue Descartes, 75005 Paris, France25
Key Points:26
•Many fresh linear and dark tracks caused by dust devils are detected from orbit27
near the InSight landing site.28
•Dust devil track formation rate decreases in late northern winter then increases29
in early spring.30
•Dust devil tracks’ azimuths are in good agreement with InSight wind direction mea-31
surements.32
Corresponding author: Cl´ement Perrin, perrin@ipgp.fr
–1–

manuscript submitted to Geophysical Research Letters
Abstract33
The NASA InSight mission on Mars is a unique opportunity to study atmospheric34
processes both from orbit and in situ observations. We use post-landing high-resolution35
satellite images to monitor dust devil activity during the first eight months of the mis-36
sion. We perform mapping and semi-automatic detection of newly formed dust devil tracks37
and analyze their characteristics (sizes, azimuths, distances, and directions of motion).38
We find a large number of tracks appearing shortly after landing, followed by a signif-39
icant decrease of activity during late winter, then a progressive increase during early spring.40
New tracks are characterized by dark linear, to slightly curvilinear, traces ranging from41
a few to more than ten meters wide. Tracks are oriented in the ambient wind direction,42
according to measurements made by InSight’s meteorological sensors. The systematic43
analysis of dust devil tracks is useful to have a better understanding of atmospheric and44
aeolian activity around InSight.45
Plain Language Summary46
The NASA InSight mission landed on Mars in November 2018. It carries weather47
and seismic stations that are now working continuously. We are also able to observe the48
InSight region from orbit using high-resolution satellite images that have been acquired49
regularly over the first year of the InSight mission. They show a lot of dark traces on50
the surface, which are caused by whirlwinds called dust devils raising dust into the air.51
This phenomenon is not observed at the same rate over the entire year, as it depends52
on atmospheric conditions which vary with season. Our study with satellite images al-53
lows us to understand the characteristics of dust devil tracks and compare them with54
related measurements from the weather station on board InSight. These two sets of ob-55
servations are well correlated to each other and provide significant constraints to bet-56
ter characterize the atmospheric activity around InSight and in the region of Elysium57
Planitia, Mars.58
1 Introduction59
Dust devils form on Earth and Mars when atmospheric convective vortices, result-60
ing from daytime turbulence in the Planetary Boundary Layer, are able to lift dust from61
the surface. Behind their passage, they usually leave albedo markings, characterized by62
dark or bright lineaments, called Dust Devil Tracks (DDTs). Even when convective vor-63
tices do not carry enough dust particles to make them visible as dust devils, their pas-64
sage over the surface may lead to tracks, also named DDTs for simplicity.65
DDTs have been well documented on Earth and Mars during the last 50 years (e.g.,66
Balme & Greeley, 2006; Reiss et al., 2016, and references therein). On Mars, successive67
orbital missions have been able to study dust devils and their associated tracks with ever68
greater imaging resolution, from Mariner 9 and Viking (e.g., Veverka, 1976; Thomas &69
Gierasch, 1985; Grant & Schultz, 1987) to Mars Global Surveyor images (e.g., Fisher et70
al., 2005; Cantor et al., 2006), Mars Express (e.g., Stanzel et al., 2006, 2008) and most71
recently MRO (Mars Reconnaissance Orbiter) images from the CTX (Context) and HiRISE72
(High Resolution Imaging Science Experiment) cameras (e.g., Verba et al., 2010; Reiss73
et al., 2014). In their review, Reiss et al. (2016) classified DDTs into three main fam-74
ilies, based on their morphology and albedo contrast with the surrounding surface: dark75
linear DDTs, bright linear DDTs, and dark cycloidal DDTs. The distinct appearances76
of tracks result from differing conditions of the atmosphere or properties of the surface77
dust layer.78
The NASA InSight (Interior Exploration using Seismic Investigations, Geodesy and79
Heat Transport) mission successfully landed in Elysium Planitia (Banerdt et al., 2020;80
–2–

manuscript submitted to Geophysical Research Letters
M. Golombek et al., 2020), carrying the seismometer SEIS (Seismic Experiment for In-81
terior Structure; Lognonn´e et al., 2019) and meteorological instruments through the APSS82
experiment (Auxiliary Payload Sensor Suite; Banfield et al., 2019; Spiga et al., 2018).83
Results from APSS have shown intense atmospheric activity at the landing site. In par-84
ticular, a great number of rapid, short-lived pressure drops occurring between late morn-85
ing and late afternoon (Banfield et al., 2020), which are diagnostic of a vortex passing86
over or close to the lander and indicate intense convective vortex activity. Seismic sig-87
natures corresponding to convective vortices have also been detected by the SEIS instru-88
ment (Lognonn´e et al., 2020; Kenda et al., 2020; Murdoch et al., 2020). Surprisingly, to89
date neither of Insight’s cameras (the Instrument Context Camera, ICC, and Instrument90
Deployment Camera, IDC, located respectively below the deck and on the robotic arm91
of the InSight lander, Maki et al. (2018)) have directly imaged any dust devil, unlike pre-92
vious missions such as Pathfinder (Ferri et al., 2003), Spirit (Greeley et al., 2006) and93
Curiosity (Lemmon et al., 2017). Nevertheless, preliminary analysis of ICC images re-94
vealed DDTs less than 20 m from the InSight lander (Banerdt et al., 2020). The com-95
bination of orbital and InSight images associated with meteorological and seismic data96
enables the first-ever joint orbital and in situ interpretation of the impact of a convec-97
tive vortex on a planetary surface. It also enables the improvement of seismic velocity98
models of the sub-surface near InSight (Banerdt et al., 2020; Lognonn´e et al., 2020).99
Reiss and Lorenz (2016) analyzed eight pre-landing HiRISE images (25 cm/pixel,100
McEwen et al., 2007) acquired between March 2010 and February 2014, covering par-101
tially a wide area of ∼20,000 km2, corresponding to the four candidate landing sites in102
Elysium Planitia (M. P. Golombek et al., 2014). They did not observe active dust dev-103
ils directly but identified more than 500 newly-formed DDTs in 8 study areas, mostly104
characterized by narrow linear dark tracks (diameter <10 m). In this paper, we exam-105
ine the evolution of DDTs around the InSight lander over the first 220 sols of the mis-106
sion (i.e. from December 2018 to July 2019). Thanks to the efforts of the MRO and HiRISE107
teams, multiple orbital images of the InSight landing site have been acquired repeatedly108
over time during the first months of the mission. This is a unique opportunity to study109
the activity of convective vortices, both from orbit and in situ, while seismic, pressure110
and wind data are acquired continuously at high frequency and accuracy.111
2 Regional observation of dust devil tracks in the vicinity of InSight112
2.1 Data and methods113
Starting soon after InSight’s landing on November 26th , 2018, HiRISE images have114
been acquired periodically to monitor surface changes around the lander due to active115
aeolian processes, which are potentially detectable by InSight’s seismic and meteorolog-116
ical sensors. In this study we analyze seven HiRISE images (Table S1) taken on Decem-117
ber 6th and 11th (respectively InSight mission Sol 9 and Sol 14; ESP 057939 1845 and118
ESP 058005 1845), February 4th and 21st (respectively Sol 68 and 84; ESP 058717 1845119
and ESP 058928 1845), March 20th (Sol 111; ESP 059284 1845), April 6th (Sol 127; ESP 059495 1845)120
and July 8th (Sol 218; ESP 060695 1845). All were taken at the highest HiRISE reso-121
lution, i.e. a ground sampling of 25 cm/pixel. Red filter (RED) Reduced Data Records122
(RDRs) images were geo-referenced and co-registered relatively to the first image (ESP 057939 1845123
shown in Fig. 1a) and projected on the Mars 2000 geographic coordinate system (Seidelmann124
et al., 2007).125
We analyze consecutive images in order to detect new DDTs formed in between126
image acquisitions. Direct comparison using blinking techniques (i.e. alternate viewing127
between two scenes) can be used to detect wide and clear tracks but is not effective for128
detecting the majority of DDTs, especially small tracks with a low albedo contrast. To129
overcome those limitations, we additionally performed ratios between images (i.e. im-130
age(t+1) / image(t)) using the raster calculator in the QGIS (Quantum Geographical131
–3–

manuscript submitted to Geophysical Research Letters
Information System) software. The ratio images, as is exemplified in figure 1b and 1c,132
highlight the surface changes that occurred between the two scenes. The unchanged ar-133
eas are characterized by a grey faded background where sparse shadowing effects of lo-134
cal relief can be seen (e.g. craters, Fig. 1b), due to the different illumination conditions135
between images. In contrast, surface changes clearly appear with darker or brighter tones136
in the ratioed images (fig. 1b, 1c and Supp. Fig. 1). In particular, new DDTs appear137
dark in our ratios since vortices remove bright-toned dust to reveal the underlying darker138
surface. We do not observe new bright DDTs in RDRs images, as were previously re-139
ported on Earth and Mars (e.g., Reiss et al., 2011, and references therein), often found140
in regions with a thick layer of dust. However, we do observe fading DDTs that appear141
bright in the ratio images since bright-toned dust depositions occur and the track slowly142
fades (Supp. Fig. S1d). In this study, we did not perform measurements of fading rates143
of DDTs but previous studies in Gusev Crater have shown that the minimum fading time144
of a dust devil track can be up to 211 sols (or 217 terrestrial days; Daubar et al., 2018).145
2.2 Manual mapping and characteristics of DDTs146
Figure 1a presents the regional mapping of the new DDTs identified from both RDRs147
and ratio images. Detailed analysis of each image has been performed to identify new148
DDTs in a wide area, up to 100 km2(Table S1). Systematic comparison of resulting DDT149
maps was done in order to ensure that the same DDT was not mapped twice. New DDTs150
were detected during the following time periods: December 6th -December 11th 2018 (5151
sol period, HiRISE interval 1), December 11th 2018-February 4th 2019 (54 sol period;152
HiRISE interval 2) and April 6th -July 8th 2019 (91 sol period; HiRISE interval 6). Con-153
versely, no tracks were detected during the following time periods: February 4th-February154
21th (16 sol period, HiRISE interval 3), February 21st-March 20st (27 sol period, HiRISE155
interval 4) and March 20th-April 6th (16 sol period, HiRISE interval 5). This shows a156
significant seasonal variability of convective vortex activity over almost a terrestrial year,157
as already pointed out by Verba et al. (2010) based on DDTs detected by HiRISE.158
Figure 1a also shows that the highest density of new tracks was observed in the first159
weeks of the InSight mission, denoting an exceptionally high level of atmospheric activ-160
ity. Indeed, up to 339 new tracks were mapped in the HiRISE interval 1, which was only161
five sols in duration (see blue DDTs in Fig. 1a), representing a formation rate as high162
as 0.68 DDT/sol/km2. The following longer HiRISE interval 2 (54 sols) covered by the163
images was much less active: only 125 new tracks were mapped (see orange DDTs in Fig.1a),164
corresponding to a formation rate of 0.03 DDT/sol/km2, a factor of 20 drop from the165
first weeks of the InSight mission. Then, after a couple of months in which no tracks were166
detected (i.e., HiRISE intervals 3, 4 and 5), we mapped 213 new tracks (see green DDTs167
in Fig.1a) that formed during HiRISE interval 6 (91 sols), corresponding to a formation168
rate of 0.04 DDT/sol/km2. These formation rates are gathered in Table S1. We explain169
in section 4 the caveats of such estimations.170
In RDRs and ratio images, new DDTs are characterized by dark linear traces rang-171
ing from a few meters to more than ten meters wide (Fig. 1 and Supp. Fig. S1). The172
along-tracks’ lengths range from 40 m to 6770 m (Supp. Fig. S2), with an average track173
length of 658 m for complete tracks (i.e. start and end of the DDT visible on the HiRISE174
image) and >1347 m for incomplete tracks (i.e. DDTs continue past the edges of the HiRISE175
footprint) over the three HiRISE intervals considered in Figure 1. These estimates are176
in good agreement with pre-landing measurements made by Reiss and Lorenz (2016).177
The mean direction of DDTs (180◦ambiguity) is persistent over the 8 months, trend-178
ing mainly NW-SE (Fig. 1a; mean direction of all tracks is N138±22◦E, i.e. measured179
clockwise from North, see also Supp. Fig. S3). We note, however, a slight shift in az-180
imuth for the most recent time period (∼N135◦E, HiRISE interval 6) compared to the181
beginning of the mission (∼N144◦E, HiRISE interval 1; ∼N147◦E, HiRISE interval 2).182
–4–

manuscript submitted to Geophysical Research Letters
We also distinguish locally wider tracks consisting of feathered structures (Fig. 1b,183
1c and Supp. Fig. S1). They correspond generally to ‘asymmetric’ convective vortices184
(Reiss et al., 2013) which remove dust mainly on one side of their path, creating incom-185
plete cycloidal tracks each made of a wall track and a section of the circular trace of the186
convective vortex (see Fig. 1d). This often occurs when the ratio between ambient wind187
speed and vortex-induced wind speed is large, which is the case for the time considered188
here (see section 4). Feathered structures are useful to infer the sense of rotation and189
direction of motion of the dust devils (Greeley, 2004; Reiss et al., 2016). These feath-190
ered tracks are associated with the path of a dust devil, where dust is removed and pro-191
jected toward the rear of the vortex (solid black lines in Fig. 1d). Feathered structures192
in the front of the dust devil can be distinguished, but they are rapidly faded by the pas-193
sage of the vortex (dotted black lines in Fig. 1d). Hence, the feathered tracks, only clearly194
observed in the HiRISE interval 1 images, indicate both clockwise and counter-clockwise195
rotations of DDTs moving from the NW to the SE (Fig. 1b, 1c and 1d and Supp. Fig.196
S1a, S1b and S1c). We also observe a single complete cycloidal track a few kilometers197
north of the lander (Supp. Fig. S1b) strikingly showing the perfectly circular shape of198
a convective vortex. We detect only seven new feathered and cycloidal tracks in ratio199
images, and only in HiRISE interval 1. Based on those feathered tracks, we estimate that200
the radii r(Fig. 1d) of the corresponding vortices range here from 22 to 70 m.201
3 Semi-automatic detection of dust devil tracks202
In this section we compare the manual mapping presented in the previous section203
with an alternate approach based on a semi-automatic detection of new DDTs in the HiRISE204
images. This allows us to have an objective analysis of the images and assess the robust-205
ness of our mapping. Besides, by focusing on a smaller area around the landing site (square206
of 12.5 km2, Fig. 1b and 2), the semi-automated detection algorithm is able to detect207
new tracks closer to the lander, and thus is more representative of the local atmospheric208
conditions recorded by APSS sensors and potentially detectable by the SEIS instrument.209
The semi-automated algorithm to detect DDTs in the ratio images is based on the210
use of the Radon transform (Toft, 1996), an efficient technique to detect linear or curvi-211
linear features in 2D images. The main strength of the Radon transform is the ability212
to extract lines (curves in general) from very noisy images (see method in Supp. Fig. S4).213
Figure 2 presents the ratio images (top) and results of the semi-automatic detec-214
tion of new DDTs (black lines; bottom) in a ∼12.5 km2area centered on InSight. 27 new215
DDTs have been detected in HiRISE interval 1, while 22 new DDTs have been detected216
in HiRISE interval 2, corresponding to a formation rate of ∼0.43 DDTs/sols/km2and217
∼0.03 DDTs/sols/km2, respectively. No new DDTs were found in images in the HiRISE218
intervals 3, 4 and 5. Part of the semi-automatic analysis of new DDTs in the HiRISE219
interval 6 can be found in Banerdt et al. (2020).220
Our method allows us to extract the trajectory of the DDTs relative to the lan-221
der (direction, distance; see Supp. Fig. S5). As seen in the manual mapping results, the222
DDTs show a persistent azimuth, ranging from N120◦E to N140◦E. Most of the DDTs223
are 8 to 10 m wide (Supp. Fig. S5), but they can reach tens of meters for larger tracks,224
averaging their width by considering both the wall track and feathered structures (for225
example ∼26 m wide for the feathered track on Fig. 1c and Supp. Fig. S5). In the lat-226
ter case of a large convective vortex (i.e. diameter >10 m), the DDT’s width represents227
only a lower limit on their size. However, for small convective vortices (i.e. diameter <228
10 m), the DDT’s width can actually be a good estimate of their size. Together with in-229
formation about DDTs’ trajectories, they can be used to calculate expected minimum230
pressure drops and seismic signals in comparison with InSight measurements (Banerdt231
et al., 2020; Murdoch et al., 2020).232
–5–

manuscript submitted to Geophysical Research Letters
DDT formation rates deduced from the semi-automatic detection are similar to those233
obtained from manual mapping: during HiRISE interval 1 and 2, 96% and 81% of the234
DDTs detected by the semi-automatic method are observed in our manual mapping, re-235
spectively. Also, the mean azimuth of each track deduced from both methods is in ex-236
cellent agreement (Supp. Fig. S6) with a coefficient of determination R2= 0.98 and 0.99237
for HiRISE interval 1 and 2, respectively. Since the semi-automatic detection method238
is more efficient at detecting linear or slightly curvilinear features, more highly curved239
DDTs are not always detected or are cut into small linear traces, compared to the man-240
ual mapping (Supp. Fig. S7). On the contrary, small low contrast DDTs are detected241
by the semi-automatic method but can be hardly visible by eye and thus not considered242
in the manual mapping (Supp. Fig. S7). Furthermore, on one hand, the semi-automatic243
method removes the possible bias of manual mapping, which might consider an inter-244
rupted linear track as several individual tracks. On the other hand, the manual map-245
ping can be used as a support to set the threshold of the semi-automatic detection. Both246
methods are thus complementary and allow us to perform a thorough cross-checking of247
the information coming from the two analyses to fully characterize the DDTs around In-248
Sight.249
4 Comparison with InSight meteorological measurements and impli-250
cations for DDT formation rates251
Figures 3a, b and c present atmospheric measurements from InSight (Temperature252
and Wind for InSight ”TWINS”, and pressure, sensors, part of the APSS instrumental253
suite; Spiga et al., 2018; Banfield et al., 2019) since the beginning of the mission up to254
sol 220 (July 10th, 2019), hence covering the range of the analyzed HiRISE images. The255
mean ambient wind speed and direction are averaged between 8 am and 5 pm Local Mar-256
tian Solar Time, the period during which convective vortices are active according to pres-257
sure drop analysis (Banfield et al., 2020). We distinguish three main ’APSS stages’ in258
the data:259
i The first ∼50 InSight sols (Solar longitude Ls 295-326◦) are characterized by a very260
active atmosphere, with a fairly constant mean daytime wind direction ranging261
from -40◦to -80◦(i.e. wind blowing from the NW to the SE according to atmo-262
spheric science convention, written asc. below), and high variations of mean wind263
speed (from 5 to 11 m/s) associated with a large number of pressure drops denot-264
ing convective vortices (up to 40 per sol).265
ii This is followed by an abrupt change of meteorological conditions at the InSight266
landing site, caused by a large regional dust storm (Banfield et al., 2020). This267
causes the daytime wind direction (Fig. 3a) to vary significantly, from -90◦(wind268
blowing from the W) to +120◦(wind blowing from the SE), and the mean wind269
speed to decrease significantly to reach a steady value of about 5 m/s until the270
decay of the dust storm between sol 130 to 140 (Ls 8-13◦). Convective activity is271
closely linked to surface sensible heat flux (e.g., Renn´o et al., 1998; Newman et272
al., 2019), which increases with both the wind stress at the surface and the surface-273
to-air temperature difference. Thus, following the decrease of both ambient wind274
speed and daytime surface temperature in those dustier conditions, the number275
of pressure drops caused by convective vortices also decreases during this stage,276
ranging from 1 to 12 pressure drops per sol.277
iii From sol 130 to sol 220 (Ls 8-51◦), a significant change appears again in the at-278
mospheric activity: the mean wind direction rapidly switched from -40◦to +130◦,279
denoting a reversal of the wind direction from ”NW to SE” to ”SE to NW”. This280
change in Elysium Planitia was predicted by pre-landing climate modeling (Spiga281
et al., 2018). It occurs at the seasonal transition between northern winter and north-282
ern spring. The switch of wind direction is accompanied by a progressively larger283
increase in mean wind speed and numbers of large pressure drops per sol. Between284
–6–

manuscript submitted to Geophysical Research Letters
sols 180 and sols 220, they both reach similar values to the beginning of the mis-285
sion, but with lower standard deviations.286
The regular acquisition of HiRISE images since the beginning of the InSight mis-287
sion (grey dashed lines in Fig. 3) covers the three APSS stages described above. The DDT288
formation rates deduced from HiRISE images are quite consistent with InSight atmo-289
spheric measurements (see also Table S1). The first HiRISE interval (in blue) corresponds290
to the highest DDT formation rate we measure, and also shows the highest atmospheric291
activity during this time (APSS stage i). The second HiRISE interval (in orange) cov-292
ers the transition from high to low atmospheric activity (from APSS stages i to ii), as-293
sociated with the regional dust storm during northern winter on Mars. We do observe294
DDTs during this period, which might have formed in the first part of this interval (i.e.295
before the transition to APSS stage ii). Also, the regional dust storm probably affected296
the surface signature of previously formed DDTs, covering them via dust deposition (Supp.297
Fig. 1d). The following third, fourth and fifth HiRISE intervals (in grey, corresponding298
to late northern winter season at this location) were quiet in comparison to the other299
time periods discussed, with respect to both atmospheric conditions (APSS stage ii) and300
formation rate of new DDTs. Indeed we did not detect any new DDTs in HiRISE im-301
ages during the third, fourth, and fifth intervals. The sixth HiRISE interval (in green)302
falls within the increase of atmospheric activity during the northern spring (from APSS303
stages ii to iii), and also corresponds to the time period during which we do detect new304
DDTs. This correlation possibly allows us to identify a threshold in the InSight data above305
which DDTs are formed. The number of pressure drops per sol and wind speed did not306
exceed 12 and 6 m/s, respectively, during the time periods for which we observe no new307
DDTs.308
The seasonal occurrence of dust devils and their associated tracks was already ob-309
served on Mars in other sites than InSight, suggesting also a higher frequency of dust310
devil formation during local late spring through fall in both the northern and southern311
hemispheres (Greeley et al., 2006; Whelley & Greeley, 2006, 2008; Reiss et al., 2016). De-312
spite the fact that the HiRISE intervals 1 and 6 include quiet periods with a low num-313
ber of large pressure drops (<12 per sol), we do observe new DDTs, implying that they314
formed mainly during a shorter amount of time when the atmospheric activity was higher315
(i.e., first half of APSS stage i and last half of APSS stage iii, respectively). The DDT316
formation rates are minimum rates since they strongly depend on the acquisition date317
of HiRISE images. For an equivalent number of DDTs, the longer the time period be-318
tween two images (large number of sols), the lower will be the formation rate of DDTs.319
Moreover, long time periods will favor fading processes of tracks through cumulative dust320
deposition. This process might hide some tracks formed in the earlier part of the period,321
artificially lowering the DDTs formation rate. However, we consider that this effect is322
negligible in our cases, since the time periods between two scenes are all shorter than323
the fading time (≤91 sols, Table S1; Reiss & Lorenz, 2016; Daubar et al., 2018), and324
most of the DDTs are still observed in the following images. We compare our DDT for-325
mation rates with those found by Reiss and Lorenz (2016) in Elysium Planitia (Fig. 3d).326
We find complementary estimates from northern mid-winter to mid-spring (Ls 301◦to327
50◦). Figure 3d completes the overall seasonal variation of the DDT formation rate in328
Elysium Planitia: it is higher during northern late autumn-early winter (from 0.02 to329
0.68 DDTs/sol/km2) and mid-spring (∼0.05 DDTs/sol/km2), while a major decrease oc-330
curs during both northern late winter and late spring.331
Despite the large range of DDT directions measured close to InSight (area in Fig.332
2) from both methods, their mean directions are in good agreement with the ambient333
wind directions measured by InSight during the HiRISE intervals where we detected new334
DDTs (see also Supp. Fig. S8). The HiRISE intervals 1, 2 and 6 present a mean azimuth335
of about N140±5◦E (or -40◦asc.) close to the measured mean wind direction during336
each period and despite the dust storm perturbation (∼N120◦E, or -60◦asc.). Feath-337
–7–

manuscript submitted to Geophysical Research Letters
ered and complete cycloidal tracks were only observed in the first HiRISE interval, and338
confirm that the wind blew from the NW to the SE (Fig. 1c and 1d). However, we were339
not able to deduce such information from DDTs during HiRISE intervals 2 and 6, due340
to a lack of feathered and cycloidal tracks. Generally, the mean wind direction value is341
lower than the mean DDTs direction (Supp. Fig. S8). HiRISE intervals include large342
time periods where the wind activity varies, hence affecting our calculation of the mean343
direction. For example, during the HiRISE interval 6 (and APSS stage iii), InSight recorded344
an inversion in the wind direction (from SE to the NW). Between sol 160 and sol 220,345
it reaches a steady value of ∼N130◦E, in better agreement with DDTs directions. This346
suggests again that DDTs were likely formed during this shorter time period when the347
atmospheric activity was higher (i.e. last half of APSS stage iii).348
The short time period during the first HiRISE interval might was a chance to com-349
pare in more detail with the variation of InSight wind measurements on a timescale of350
several sols. Unfortunately, at this early stage of the mission, the wind sensors were turned351
on only for a few hours on December 11th (Supp. Fig. S9). The measured wind direc-352
tion at that time shows a fairly constant value of ∼-55◦asc., which is slightly different353
from the direction deduced from the DDTs map (N144◦E, or -36◦asc.). However, there354
is an apparent absence of winds coming from -40◦asc. (or ∼N140◦E; Supp. Fig. S9),355
surrounded by most of the wind data. This may be due to the wind approaching the TWINS356
wind sensors from exactly that direction being blocked by the presence of other instru-357
ments (the Rotation and Interior Structure Experiment (RISE) and Ultra-High Frequency358
(UHF) antennas; Banfield et al., 2019). The difference between our DDTs measurements359
and InSight might come from this unfortunate configuration of instruments on the lan-360
der deck.361
Previous studies on Earth found a relationship between the standard deviation of362
DDT migration direction and the ambient wind speed (Supp. Fig. 10a; Balme et al.,363
2012; Lorenz, 2016). Under low ambient wind speeds, the DDT is sinuous and random364
as its path is mainly governed by the internal advection speed of the vortex. As the am-365
bient wind speed increases, the dust devil tends to form a more linear track (Reiss et al.,366
2016). We can not assess that this relationship can be linearly applied on Mars (e.g. dif-367
ferent wind characteristics, atmospheric pressure and density, etc.). Nevertheless, we do368
see variations in standard deviation of DDT azimuths and mean ambient wind speed be-369
tween the different APSS stages. Despite our small data set and the uncertainties, we370
note a similar trend on Mars as on Earth, showing a decrease of the standard deviation371
of DDT azimuths as the ambient wind speed increases (Supp. Fig. 10b). However, we372
note that all the DDTs mapped in this study are near-linear. The regional flat topog-373
raphy of Elysium Planitia might contribute to the formation of linear tracks, even un-374
der lower ambient wind speeds. Upcoming acquisition of new orbital images above In-375
Sight will allow us to explore this relationship further for Mars.376
Studying the characteristics of DDTs is a good indicator of the wind regime mea-377
sured locally by InSight (Fig. 3 and Supp. Fig. S8 and S10). Apart from short-term tur-378
bulent fluctuations, the wind is dominated by a combination of global-scale circulations379
(Hadley cells, thermal tides) and regional contributions (i.e., western boundary currents,380
moderate slope flows) (Spiga et al., 2018). The wind direction measured by InSight ap-381
pears to be reasonably steady in the daytime hours (Banfield et al., 2020). Daily vari-382
ability is also very moderate in the equatorial region where InSight landed; changes of383
wind direction are mostly seasonal, or driven by rare regional dust storms as occurred384
around sol 50 (Banfield et al., 2020).385
5 Conclusions386
We monitor the activity of convective vortices around InSight on Mars as evinced387
by their tracks visible in orbital HiRISE images. Manual and semi-automatic detection388
–8–

manuscript submitted to Geophysical Research Letters
of DDTs are performed and present similar and complementary results. We observe a389
large number of dark and linear DDTs in the first days after the InSight landing, followed390
by a significant decrease and then a progressive increase of their numbers. These obser-391
vations match the measurements of atmospheric activity obtained by APSS onboard the392
lander. This behavior correlates well with the transition between seasons on Mars from393
northern fall, to northern winter and then northern spring. The mean azimuth of DDTs394
is very persistent (∼N138◦E), in good agreement with APSS measurements of the mean395
wind direction. InSight’s meteorological dataset, with its high frequency acquisition rates,396
provides a unique opportunity to conduct DD activity monitoring from both Mars’s or-397
bit and surface. The link with APSS data is important because InSight has detected a398
large number of convective vortices in the pressure and seismic data but surprisingly has399
imaged no dust devils yet.400
Characterization of DDTs is important to better understand seismic, pressure and401
wind data signals of passing vortices and to investigate the vortex and ground proper-402
ties around the InSight lander (Banerdt et al., 2020; M. Golombek et al., 2020; Murdoch403
et al., 2020, this issue). Future HiRISE images are thus needed to monitor DDTs for-404
mation over the summer period around InSight, and have a complete overview of atmo-405
spheric processes throughout an entire martian year, in concert with in-situ measurements.406
Acknowledgments407
We thank the HiRISE team for their effort on acquiring new orbital images. All HiRISE408
images credit NASA/JPL/University of Arizona. Data from APSS and TWINS sensors409
referenced in this paper are available from the PDS (https://atmos.nmsu.edu/ data and services/atmospheres data/INSIGHT/insight.html).410
We acknowledge NASA, CNES and its partners (UKSA, SSO, DLR, JPL, IPGP-CNRS,411
ETHZ, IC, MPS-MPG) and the flight operations team at JPL, CAB, SISMOC, MSDS,412
IRIS-DMC and PDS for providing InSight data. Funding support was provided by Agence413
Nationale de la Recherche (ANR-14-CE36-0012-02 SIMARS and ANR-19-CE31-0008-414
08 MAGIS). We acknowledge the support of NVIDIA Corporation with the donation of415
GPU used for this research. This research benefits from the QGIS, an Open Source Geospa-416
tial Foundation Project (http://qgis.org). Statistical analysis of DDTs maps have been417
done using the FracPaQ matlab toolbox (Healy et al., 2017). This is InSight Contribu-418
tion Number 124 and IPGP 4130.419
References420
Balme, M., & Greeley, R. (2006). Dust devils on Earth and Mars. Reviews of421
Geophysics,44 (3), RG3003. Retrieved from http://doi.wiley.com/10.1029/422
2005RG000188 doi: 10.1029/2005RG000188423
Balme, M., Pathare, A., Metzger, S., Towner, M., Lewis, S., Spiga, A., . . . Verdasca,424
J. (2012, nov). Field measurements of horizontal forward motion velocities425
of terrestrial dust devils: Towards a proxy for ambient winds on Mars and426
Earth. Icarus,221 (2), 632–645. Retrieved from http://dx.doi.org/10.1016/427
j.icarus.2012.08.021https://linkinghub.elsevier.com/retrieve/pii/428
S0019103512003405 doi: 10.1016/j.icarus.2012.08.021429
Banerdt, W. B., Smrekar, S. E., Banfield, D., Giardini, D., Golombek, M., John-430
son, C. L., . . . Wieczorek, M. (2020, mar). Initial results from the In-431
Sight mission on Mars. Nature Geoscience,13 (3), 183–189. Retrieved432
from http://www.nature.com/articles/s41561-020-0544-y doi:433
10.1038/s41561-020-0544-y434
Banfield, D., Rodriguez-Manfredi, J. A., Russell, C. T., Rowe, K. M., Leneman,435
D., Lai, H. R., . . . Banerdt, W. B. (2019, feb). InSight Auxiliary Pay-436
load Sensor Suite (APSS). Space Science Reviews,215 (1), 4. Retrieved437
from http://link.springer.com/10.1007/s11214-018-0570-x doi:438
10.1007/s11214-018-0570-x439
–9–

manuscript submitted to Geophysical Research Letters
Banfield, D., Spiga, A., Newman, C., Forget, F., Lemmon, M., Lorenz, R., . . .440
Banerdt, W. B. (2020, mar). The atmosphere of Mars as observed by InSight.441
Nature Geoscience,13 (3), 190–198. Retrieved from http://www.nature.com/442
articles/s41561-020-0534-0 doi: 10.1038/s41561-020-0534-0443
Cantor, B. A., Kanak, K. M., & Edgett, K. S. (2006). Mars Orbiter Camera ob-444
servations of Martian dust devils and their tracks (September 1997 to January445
2006) and evaluation of theoretical vortex models. Journal of Geophysical446
Research E: Planets,111 (12), 1–49. doi: 10.1029/2006JE002700447
Daubar, I. J., Ojha, L., Chojnacki, M., Golombek, M. P., Lorenz, R. D., Wray, J.,448
& Lewis, K. (2018). Lifetime of a dust devil track and dust deposition rate in449
Gusev Crater. In 49th lunar and planetary science conference.450
Ferri, F., Smith, P. H., Lemmon, M., & Renn´o, N. O. (2003). Dust devils as ob-451
served by Mars Pathfinder. Journal of Geophysical Research (Planets),108 ,452
5133-+. doi: 10.1029/2000JE001421453
Fisher, J. A., Richardson, M. I., Newman, C. E., Szwast, M. A., Graf, C., Basu,454
S., . . . Wilson, R. J. (2005). A survey of Martian dust devil activity using455
Mars Global Surveyor Mars Orbiter Camera images. Journal of Geophysical456
Research E: Planets,110 (3), 1–18. doi: 10.1029/2003JE002165457
Golombek, M., Warner, N. H., Grant, J. A., Hauber, E., Ansan, V., Weitz,458
C. M., . . . Banerdt, W. B. (2020, dec). Geology of the InSight land-459
ing site on Mars. Nature Communications,11 (1), 1014. Retrieved460
from http://www.nature.com/articles/s41467-020-14679-1 doi:461
10.1038/s41467-020-14679-1462
Golombek, M. P., Wigton, N., Bloom, C., Schwartz, C., Kannan, S., Kipp, D., . . .463
Banerdt, B. (2014). Final Four Landing Sites for the InSight Geophysical464
Lander. In 45th lunar and planetary science conference,held 17-21 march, 2014465
at the woodlands, texas. lpi contribution no. 1777, p.1499.466
Grant, J. A., & Schultz, P. H. (1987, aug). Possible Tornado-Like Tracks on Mars.467
Science,237 (4817), 883–885. Retrieved from http://www.sciencemag.org/468
cgi/doi/10.1126/science.237.4817.883 doi: 10.1126/science.237.4817469
.883470
Greeley, R. (2004). Martian dust devils: Directions of movement inferred from their471
tracks. Geophysical Research Letters,31 (24), L24702. Retrieved from http://472
doi.wiley.com/10.1029/2004GL021599 doi: 10.1029/2004GL021599473
Greeley, R., Whelley, P. L., Arvidson, R. E., Cabrol, N. A., Foley, D. J., Franklin,474
B. J., . . . Thompson, S. D. (2006). Active dust devils in Gusev crater, Mars:475
Observations from the Mars Exploration Rover Spirit. Journal of Geophysical476
Research E: Planets,111 (12), 1–16. doi: 10.1029/2006JE002743477
Healy, D., Rizzo, R. E., Cornwell, D. G., Farrell, N. J., Watkins, H., Timms, N. E.,478
. . . Smith, M. (2017). FracPaQ: A MATLAB toolbox for the quantifica-479
tion of fracture patterns. Journal of Structural Geology,95 , 1–16. doi:480
10.1016/j.jsg.2016.12.003481
Kenda, B., Drilleau, M., Garcia, R., Kawamura, T., Murdoch, N., Compaire, N.,482
. . . Spohn, T. (2020). Subsurface structure at the InSight landing site from483
compliance measurements by seismic and meteorological experiments. Journal484
of Geophysical Research E: Planets,this issue .485
Lemmon, M. T., Newman, C. E., Renno, N., Mason, E., Battalio, M., Richardson,486
M. I., & Kahanp¨a¨a, H. (2017, mar). Dust Devil Activity at the Curiosity Mars487
Rover Field Site. In Lunar and planetary science conference (p. 2952).488
Lognonn´e, P., Banerdt, W. B., Giardini, D., Pike, W. T., Christensen, U., Laudet,489
P., . . . Wookey, J. (2019, feb). SEIS: Insight’s Seismic Experiment for Inter-490
nal Structure of Mars. Space Science Reviews,215 (1), 12. Retrieved from491
http://dx.doi.org/10.1007/s11214-018-0574-6http://link.springer492
.com/10.1007/s11214-018-0574-6 doi: 10.1007/s11214-018-0574-6493
Lognonn´e, P., Banerdt, W. B., Pike, W. T., Giardini, D., Christensen, U., Gar-494
–10–

manuscript submitted to Geophysical Research Letters
cia, R. F., . . . Zweifel, P. (2020, mar). Constraints on the shallow elastic495
and anelastic structure of Mars from InSight seismic data. Nature Geo-496
science,13 (3), 213–220. Retrieved from http://www.nature.com/articles/497
s41561-020-0536-y doi: 10.1038/s41561-020-0536-y498
Lorenz, R. D. (2016, jun). Heuristic estimation of dust devil vortex parameters and499
trajectories from single-station meteorological observations: Application to In-500
Sight at Mars. Icarus,271 , 326–337. Retrieved from http://dx.doi.org/501
10.1016/j.icarus.2016.02.001https://linkinghub.elsevier.com/502
retrieve/pii/S0019103516000580 doi: 10.1016/j.icarus.2016.02.001503
Maki, J. N., Golombek, M. P., Deen, R., Abarca, H., Sorice, C., Goodsall, T., . . .504
Banerdt, W. B. (2018, sep). The Color Cameras on the InSight Lander. Space505
Science Reviews,214 (6), 105. Retrieved from http://link.springer.com/506
10.1007/s11214-018-0536-z doi: 10.1007/s11214-018-0536-z507
McEwen, A. S., Eliason, E. M., Bergstrom, J. W., Bridges, N. T., Hansen, C. J., De-508
lamere, W. A., . . . Weitz, C. M. (2007, may). Mars Reconnaissance Orbiter’s509
High Resolution Imaging Science Experiment (HiRISE). Journal of Geophys-510
ical Research,112 (E5), E05S02. Retrieved from http://doi.wiley.com/511
10.1029/2005JE002605 doi: 10.1029/2005JE002605512
Murdoch, N., Spiga, A., Lorenz, R. D., Garcia, R. F., Perrin, C., Widmer-Schnidrig,513
R., . . . Bruce Banerdt, W. (2020). Constraining Martian dust devil vortex and514
regolith parameters from combined seismic and meteorological measurements.515
Journal of Geophysical Research E: Planets,submitted to this issue.516
Newman, C. E., Kahanp¨a¨a, H., Richardson, M. I., Mart´ınez, G. M., VicenteRetor-517
tillo, A., & Lemmon, M. T. (2019, dec). MarsWRF Convective Vortex and518
Dust Devil Predictions for Gale Crater Over 3 Mars Years and Comparison519
With MSLREMS Observations. Journal of Geophysical Research: Planets,520
124 (12), 3442–3468. Retrieved from https://onlinelibrary.wiley.com/521
doi/abs/10.1029/2019JE006082 doi: 10.1029/2019JE006082522
Reiss, D., Fenton, L., Neakrase, L., Zimmerman, M., Statella, T., Whelley, P.,523
. . . Balme, M. (2016, nov). Dust Devil Tracks. Space Science Reviews,524
203 (1-4), 143–181. Retrieved from http://link.springer.com/10.1007/525
s11214-016-0308-6 doi: 10.1007/s11214-016-0308-6526
Reiss, D., & Lorenz, R. D. (2016). Dust devil track survey at Elysium Planitia,527
Mars: Implications for the InSight landing sites. Icarus,266 , 315–330. Re-528
trieved from http://dx.doi.org/10.1016/j.icarus.2015.11.012 doi: 10529
.1016/j.icarus.2015.11.012530
Reiss, D., Raack, J., & Hiesinger, H. (2011). Bright dust devil tracks on Earth: Im-531
plications for their formation on Mars. Icarus,211 (1), 917–920. Retrieved532
from http://dx.doi.org/10.1016/j.icarus.2010.09.009 doi: 10.1016/j533
.icarus.2010.09.009534
Reiss, D., Spiga, A., & Erkeling, G. (2014). The horizontal motion of dust devils on535
Mars derived from CRISM and CTX/HiRISE observations. Icarus,227 , 8–20.536
Retrieved from http://dx.doi.org/10.1016/j.icarus.2013.08.028 doi: 10537
.1016/j.icarus.2013.08.028538
Reiss, D., Zimmerman, M. I., & Lewellen, D. C. (2013). Formation of cycloidal539
dust devil tracks by redeposition of coarse sands in southern Peru: Impli-540
cations for Mars. Earth and Planetary Science Letters,383 , 7-15. doi:541
10.1016/j.epsl.2013.09.033542
Renn´o, N. O., Burkett, M. L., & Larkin, M. P. (1998, nov). A Simple Thermo-543
dynamical Theory for Dust Devils. Journal of the Atmospheric Sciences,544
55 (21), 3244–3252. Retrieved from http://journals.ametsoc.org/doi/abs/545
10.1175/1520-0469{\%}281998{\%}29055{\%}3C3244{\%}3AASTTFD{\%}3E2546
.0.CO{\%}3B2 doi: 10.1175/1520-0469(1998)055h3244:ASTTFDi2.0.CO;2547
Seidelmann, P. K., Archinal, B. A., A’hearn, M. F., Conrad, A., Consolmagno,548
G. J., Hestroffer, D., . . . Williams, I. P. (2007, jul). Report of the IAU/IAG549
–11–

manuscript submitted to Geophysical Research Letters
Working Group on cartographic coordinates and rotational elements: 2006.550
Celestial Mechanics and Dynamical Astronomy,98 (3), 155–180. Retrieved551
from http://link.springer.com/10.1007/s10569-007-9072-y doi:552
10.1007/s10569-007-9072-y553
Spiga, A., Banfield, D., Teanby, N. A., Forget, F., Lucas, A., Kenda, B., .. .554
Banerdt, W. B. (2018, oct). Atmospheric Science with InSight. Space Sci-555
ence Reviews,214 (7), 109. Retrieved from http://link.springer.com/556
10.1007/s11214-018-0543-0 doi: 10.1007/s11214-018-0543-0557
Stanzel, C., P¨atzold, M., Greeley, R., Hauber, E., & Neukum, G. (2006). Dust558
devils on Mars observed by the High Resolution Stereo Camera. Geophysical559
Research Letters,33 (11), L11202. Retrieved from http://doi.wiley.com/10560
.1029/2006GL025816 doi: 10.1029/2006GL025816561
Stanzel, C., Patzold, M., Williams, D., Whelley, P., Greeley, R., Neukum, G.,562
& And, T. H. C.-I. T. (2008, sep). Dust devil speeds, directions of mo-563
tion and general characteristics observed by the Mars Express High Reso-564
lution Stereo Camera. Icarus,197 (1), 39–51. Retrieved from https://565
linkinghub.elsevier.com/retrieve/pii/S0019103508001875 doi:566
10.1016/j.icarus.2008.04.017567
Thomas, P., & Gierasch, P. J. (1985, oct). Dust Devils on Mars. Science,230 (4722),568
175–177. Retrieved from http://www.sciencemag.org/cgi/doi/10.1126/569
science.230.4722.175 doi: 10.1126/science.230.4722.175570
Toft, P. (1996). The Radon Transform - Theory and Implementation. (Doctoral dis-571
sertation, Kgs. Lyngby, Denmark, Technical University of Denmark.) Retrieved572
from https://backend.orbit.dtu.dk/ws/portalfiles/portal/5529668/573
Binder1.pdf574
Verba, C. A., Geissler, P. E., Titus, T. N., & Waller, D. (2010). Observations from575
the High Resolution Imaging Science Experiment (HiRISE): Martian dust dev-576
ils in Gusev and Russell craters. Journal of Geophysical Research E: Planets,577
115 (9), 1–11. doi: 10.1029/2009JE003498578
Veverka, J. (1976). Variable features on Mars. VII. Dark filamentary markings on579
Mars. Icarus,27 (4), 495–502. doi: 10.1016/0019-1035(76)90165-2580
Whelley, P. L., & Greeley, R. (2006). Latitudinal dependency in dust devil activity581
on Mars. Journal of Geophysical Research E: Planets,111 (10). doi: 10.1029/582
2006JE002677583
Whelley, P. L., & Greeley, R. (2008, jul). The distribution of dust devil ac-584
tivity on Mars. Journal of Geophysical Research,113 (E7), E07002. Re-585
trieved from http://doi.wiley.com/10.1029/2007JE002966 doi:586
10.1029/2007JE002966587
–12–

manuscript submitted to Geophysical Research Letters
Figure 1. Observation and mapping of dust devil tracks (DDTs) on HiRISE images at the
InSight landing site. a) HiRISE image acquired on December 6th, 2018 (ESP 057939 1845 RED
RDR) and map of new DDTs manually identified for different periods of time: from December 6th
to 11th 2018 (5 sol period, blue lines, HiRISE interval 1), from December 11th 2018 to February
4th 2019 (54 sol period, orange lines, HiRISE interval 2) and from April 6th to July 8th 2019 (91
sol period, green lines, HiRISE interval 6). Grey lines are DDTs identified in the previous time
periods. Dashed lines are uncertain DDTs barely seen in ratio images. Black frames surround-
ing DDTs maps show the overlapping footprint of the successive ratio between HiRISE images.
The red dots indicate the InSight lander. Rose diagrams showing the range of DDTs azimuth
are presented below each HiRISE interval. Red lines on rose diagrams are mean azimuths. b)
Example ratio image (Dec. 11th/Dec. 6th 2018) presenting a closer view of new DDTs (dark
traces) around the InSight lander; c) and d) Zoom and sketch of the biggest new track observed
in December (see white arrows in 1b). Its feathered structures allow an estimate of the radius (r)
of the vortex and its sense of rotation (clockwise) and direction of motion (from the Northwest to
the Southeast). All HiRISE images credit NASA/JPL/University of Arizona.
Figure 2. Original ratios (a, b) and DDTs detections using the semi-automated method (c,
d) for ratio images during HiRISE intervals 1 (a, c) and 2 (b, d) (see Supp. Table S1). Width,
azimuth and distance of each track relative to the lander are shown in Supp. Fig. S5
Figure 3. (a) Mean wind direction (shown according to atmospheric science convention, see
text for details), (b) mean wind speed and (c) number of rapid, short-lived pressure drops exceed-
ing 0.3 Pa measured by APSS between sols 0 and 220. Wind measurements are averaged values
between 8 am and 5 pm LMST (Local Martian Solar Time), corresponding to the time period
where convective vortices are active during the martian day (Banfield et al., 2020). The three
different APSS stages are indicated above (see text for details). Colored rectangles correspond to
the six HiRISE intervals described in the text (colors similar to fig. 1). The HiRISE acquisition
dates (UTC) are indicated by vertical grey dashed lines; (d) DDT formation rates in Elysium
Planitia as function of the solar longitude (Ls) from this study (colored dots and grey rectangle)
and from Reiss and Lorenz (2016) (grey dots). The red box bounds the 220 sols covered in this
study. Seasons in the northern hemisphere are indicated.
–13–

Figure 1.

Figure 2.

Figure 3.