Gardening of the Martian Regolith by Diurnal CO2 Frost and the Formation of
Slope Streaks
L.Lange, Laboratoire de Météorologie Dynamique, Institut Pierre-Simon Laplace (LMD/IPSL), Sorbonne Université, Centre National de la Recherche Scientifique (CNRS), École Polytechnique, École Normale Supérieure
(ENS), Paris, France (lucas.lange@lmd.ipsl.fr), S.Piqueux, Jet Propulsion Laboratory, California Institute of
Technology, Pasadena, CA 91109, USA, C.S.Edwards, Northern Arizona University, Department of Astronomy
and Planetary Science PO BOX 6010 Flagstaff, AZ 86011, USA.

µm

µm
µm

z = 10 µm

5
~

~

32

µm

18

µm

Fig 1. Schematic model of
the CO2 frost (blue squares)
and regolith dust grains (orange circles) relationship in
dusty low thermal inertia regions, and its evolution
throughout a night and sunrise. In this model, CO2 ice
forms at depth, within the
pores of the dusty surfaces. At
sunrise, CO2 ice sublimates,
creating an upward sublimation-driven wind that could
lead to grain displacement.

D

12

.5
z = 5 µm

D

Surface

5

~

10

Regolith Grain Movement

m

CO2 ice sublimation

Approach:
We use coincident visible and temperature observations acquired by THEMIS [10], a multispectral
visible and thermal infrared wavelength imager. We
only analyze data acquired near sunrise (i.e., 6 A.M.
- 8 A.M), when adequate lighting allows the acquisition of multiband visible wavelength imagery of the
ground. We only study infrared images coupled with
visible wavelength images acquired simultaneously.
Surface temperatures are derived from THEMIS
band 9 centered at 12.57 µm as they offer the best
signal on cold surfaces [10]. We eliminate image
pairs clearly impacted by calibration issues (e.g.,
[11]) and poor observation conditions (mainly the
identifiable presence of clouds in visible wavelength
imagery). We identify the presence of ice on an image if the surface temperature retrieved is at the CO2
condensation temperature TCO2, within a margin of 5
K to account for instrumental noise.
Once an image is flagged for CO2 ice, we inspect
the associated visible wavelength image. The
THEMIS visible camera has a resolution of 18 m/
pixel and has five filters with band centers located at
425 (band 1), 540 (band 2), 654 (band 3), 749 (band
4), and 860 nm (band 5) [10]. When available, we
use « R2B » or « RGB » products available with the
THEMIS public viewer: viewer.mars.asu.edu/viewer/
themis.

7µ

1 µm CO2 crystal
1 µm dust grain

We proposed here to characterize the regolith/ice
relationship in low thermal inertia terrains near equator through visible and infrared wavelength imagery
analysis, surface features mapping, and numerical
modeling.

D

Introduction:
The seasonal transfer of CO2 between the atmosphere and the surface at high latitudes is associated with a wide range of exotic processes shaping
the surface morphology [1 and references therein].
In contrast, the existence of a diurnal CO2 cycle at
mid and low-latitudes has only recently been exposed [2,3]. [2,4] proposed that this diurnal CO2 cycle could impact the physical state of the regolith,
preventing dust induration/duricrust formation,
maintaining large mobile dust reservoirs, or the formation of geophysical features known as slopes
streaks [5].
[2] suggests that diurnal frost in low thermal inertia terrains (i.e., dusty surfaces) should be present
in the pores of the regolith, not exclusively on top of
it (Fig. 1). With this regolith/ice interaction, the recurring diurnal growth and sublimation of CO2 crystals could indeed cryoturb the very surficial regolith.
One consequence of this dirty frost model is that
diurnal CO2 frost in low thermal inertia terrains
should be excessively difficult to detect in visible
wavelength imagery, in contrast with diurnal or seasonal CO2 frost present elsewhere on Mars forming
on the top of the regolith grains.
In addition, one potential implication for this
recurring diurnal growth and sublimation of CO2
frost crystals within dusty terrain may include regular overnight surficial mechanical disruption of the
soil, and possible fluffing by vertical sublimationdriven winds at sunrise. This process could maintain
high regolith porosity, and prevent compaction or
inter-grain induration/cementation that seems to be
ubiquitous elsewhere on Mars [6,7] except in low
thermal inertia terrains [8,9].

Probability Density of
Photons Path Length
Low

Early Night

Mid-Night

Late Night

Pre-Dawn

Post-Sunrise

High

b

a

N

1

c

N

2.5

f

g

Ls = 142.1° LTST = 7.43

1
42.8°S, 204.1°E

146

150 155 159

Tsurf [K]

km

163

N

2

km

km
Ls = 132.1° LTST = 7.37

d

148

156 164 172
Tsurf [K]

km

180

N

2

h

Ls = 85.6° LTST = 7.03

i

Ls = 126.1° LTST = 7.29

2
km

28.8°S, 43.3°E

150

158 165 173

Tsurf [K]

180

N

2.5

km

km

2.5
41.2°S, 111.2°E

e

km

j

Ls = 13.9° LTST = 6.07

2
32.3°S, 133°E

147

160 173 186

Tsurf [K]

km

199

2.5
26.5°N, 266°E

144

151 159 166

Tsurf [K]

km

173

Fig. 2 Examples of THEMIS visible wavelength (a-e) and corresponding thermal infrared (f-j) images acquired simultaneously near
dawn. The blue/white surface patches (a-d) and low surface temperatures within 5 K of TCO2 in infrared images (f-j). Coordinates, solar
longitude (Ls), and local true solar time (LTST) are given in the different panels. Red arrows emphasize hard-to-distinguish blue/white
patches. White arrow point to the position of the sun in the sky. a: Class 1, V71796004; b: Class 2, V63705007; c: Class 3, V78900003;
d: Class 4, V63305007; e: Class 5, V76958011; f: I71796003 associated with a; g: I63705006 associated with b; h: I78900002 associated with c; i: I63305006 associated with d. j: I76958010 associated with e. f-j underlain with a THEMIS daytime IR mosaic to enhance
topography [11]. Some terrains appear black in the thermal infrared images because of the background mosaic (not because of an absence of measurement)
Surface frosts are identified based on their blue/
white hues, in stark contrast with the orange/brown/
grey surrounding terrains (Fig. 2). The vast majority
of these blue/white units are confidently attributed to
the surface (as opposed to the atmosphere) based on
their sharp boundaries following morphometric or
color units, topography, or preferential slope orientation.
[2] demonstrated that CO2 diurnal frost should
form thin layers (i.e., a few microns to tens of microns in thickness) of micrometer size ice crystals
(as opposed to translucent slab-like ice) in dusty
surfaces. Micrometer size CO2 ice crystals are associated with high albedo values at visible wavelengths
[12], but dust contamination can drastically reduce
it. [13] showed however that micrometer size ice
crystals forming the top of dusty surfaces can be
detected in these dusty surfaces. Hence, the albedo
(or color) of frosted terrains provides an excellent
diagnostic tool of the relationship between the ice
and the regolith grains. Furthermore, the surface
emissivity is highly sensitive to the presence of frost
on the surface at λ ~ 12.57 µm (THEMIS band 9,
[10]), even for thin frost layer. We therefore also use
the surface emissivity to further constrain the frost/

regolith relationship in the low thermal inertia regions of Mars.
Results:
THEMIS surface temperatures acquired at sunrise are consistent with the presence of CO2 frost at
virtually all latitudes (black and white dots in Fig. 3),
confirming results presented by others with MCS [2]
and THEMIS [3]. We find that 848 image pairs (out
of 2,761, i.e. ,~30%, that have been flagged for CO2
ice) that do not show a signature of frost in visible
wavelength imagery, while the ground is at CO2 ice
temperature. Noticeably, no identification in visible
wavelength imagery is located in the 45°N-15°S
latitude band in the low thermal inertia terrains (i.e.,
purple in Fig.3,), where widespread diurnal CO2
frost forms during a significant fraction of the year
[2]. In contrast, at other latitudes, 87.5% of the image pairs at the CO2 frost point temperature are associated with bright patches on the ground (whether
contaminated by water ice or not). Most of the missing frost signatures in visible wavelength imagery
when the surface temperatures at TCO2 at high latitudes are generally linked to poor image quality, with
challenging illumination conditions that prevent us
from clearly assessing the color of the surface. This

90

T12 ~ TCO2 No blue surfaces
T12 ~ TCO2 With blue surfaces

60

25

125

325

225

400+

Latitude

30

0

-30

-60

-90
0

45

90

135

180

225

270

315

360

Fig. 3 Distribution of dawn THEMIS
visible and infrared wavelength image pairs within 5 K of the local CO2
frost point. Black dots indicate no
signature of frost in visible wavelength images (848 images). White
dots indicate the presence of image
pairs where frost is identified at visible wavelengths (1,931 cases). Colorized background is a thermal inertia map overlaid on with a MOLA
shaded relief only shown outside the
maximum extent of the continuous
seasonal caps. During the winter, the
Northern high/mid latitudes are subject to much fewer observations than
the Southern high/mid latitudes, partially explaining the hemispheric
asymmetry

Longitude (East)

difference suggests that on low thermal inertia terrains, the visible wavelength optical properties of
diurnal frost are uniquely dominated by those of the
surface dust, i.e., ice forms within the pores the
dusty layer.

The observations are thus consistent with dirty
diurnal CO2 ice in these terrains, following the hypothesis formulated by [2] , and illustrated in Fig.1.

Discussions:
Based on Earth analogies [14,15], we assume
We acknowledge that a regolith/ice relationship
that the recurring growth and removal of ice may
diagnosis solely based on terrain color can be comcreate a stress cycle in the regolith leading to internal
plicated by the lowering of the ice's albedo by dust
grain displacement and microscopic weathering. At
(e.g., [12]). Fig. 4 shows that most of THEMIS imsunrise, the rapid sublimation of the diurnal ice creages in dusty grounds display a high surface emissivates a short-lived wind field within the regolith that
ity, close to the emissivity of ice-free dusty grounds
exerts a drag on individual grains, that could pro(0.96 [2]). We find that more than 70% of the immote the displacement of the dust grains. [13] comages show a CO2 frosted surface emissivity between
putations show that frost sublimation-driven wind at
0.96 ± 0.04 (uncertainty defined as the 3-σ spread
sunrise reaches up to 3.2 cm s-1, with a mean value
from our 5 K tolerance). Such high emissivities at
between 1.5 and 2.0 cm s-1 (standard deviation of 0.8
12.57 µm suggest the absence of CO2 frost over
cm s-1 as a function of season, latitude, and elevation.
dusty surfaces, or over micrometer size thickness
These values are upper limits and are generally larg(Fig. 11 in [2]). But in the latter case the ice thicker than (but comparable to) other values reported in
nesses would be inconsistent with mass/energy balthe literature for seasonal ice sublimation and Knudance results [2]. Thus, the high surface emissivity
sen pumping [16,17]. Using a simple 1-D balance
values are most consistent with an absence of frost
model, we show that [13] the vertical drag exerted
on the surface.
on individual grains can be larger and opposite in
direction to cohesion plus gravity forces, suggesting
Thermal Inertia [J m K s ]
that motion can be initiated by this wind. Local
b 70
slope, and more importantly grain packing angle as
THEMIS emissivity with T ⩽ T
well as cohesion forces between grains are important
Dust Emissivity
factors controlling whether grain motion can occur.
60
Surface Emissivity of 10 µm CO Frost + Dust
Surface Emissivity of 100 µm CO Frost + Dust
Winds faster than 2.5 cm s-1 are occasionally encountered and should be able to disrupt the ground on
50
most sloped and poorly cohesive low thermal inertia
terrains;
Consequently, this grain-displacement induced
40
by CO2 sublimation could promote other dynamic
phenomena as slope streaks. Slopes streaks are dark
30
wedge-shaped surface features on sloped dusty terrains, associated with mass movement downslope
[18]. The formation process for slope streaks is still
20
debated, and generally falls into two categories: wet
vs. dry mechanisms, although both types can be associated with dust avalanches. [19]. These slope
10
streaks are observed where diurnal dirty frost is expected most of the year [13]. We therefore assume
0
that at sunrise, this diurnal frost sublimates, initiat0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Number of image pairs
ing a vertical drag on individual grains able to set
Mean Emissivity ε
those uncohesive in motion . In this model, steep
Fig. 4 CO2 emissivity derived at λ = 12.57 µm. Dust emissivity dusty terrains already close to the angle of repose
at 12.57 µm (red line) from [2]. Modeled emissivity of the
dusty surface overlaid by 10 µm (solid line) and 100 µm
(dashed line) CO2 ice layer indicated in blue [2].
Number of image pairs

a

-2

CO

2

2

Mean Emissivity

Number of image pairs

2

-1

-1/2

(i.e., 30-35° [20]) are destabilized and initiate an
avalanche of dust. Our results also show that the
velocities reached during sublimation can be high
enough to exceed the velocity threshold computed in
[21] required for fluidization of avalanching material. These fluidization of the avalanching material is
necessary to explain the length and width of the
slope streaks given.
We hypothesize that diurnal CO2 ice sublimation
may be triggering and sustaining the growth of slope
streaks, but other factors such as local winds may be
involved in this process. Our proposed model overcomes several limitations of competing dry mechanisms presented in the literature, but it is not unequivocally validated by observations constraining
the seasonality or orientation of slopes highlighted
by others.
Conclusions:
We have conducted an analysis of THEMIS visible and thermal infrared data acquired at dawn. This
work constrains the relationship between diurnal
frost and the surficial regolith on Mars. It unveils the
potential geomorphological impact of the diurnal
CO2 cycle. Specifically:
• The distribution of THEMIS thermal infrared
data acquired at dawn confirms the widespread
nature of CO2 frost on Mars previously reported
[2,3];
• Multiband THEMIS visible wavelength images acquired at dawn frequently show blue/white
hues interpreted as clean surface frost over a significant fraction of the surface. However, In the
mid-to-low latitude low thermal inertia terrains
(45°N-15°S), surface temperatures consistent with
the presence of CO2 frost on the ground do not
show any frost signature in visible wavelength
imagery, suggesting the formation of dirty frost.
• This conclusion is also supported by the high
emissivity at 12.57 µm of these frosted surfaces;
• Recurring regolith grain movement on Mars
instigated by overnight CO2 crystal growth and
rapid sublimation at sunrise could prevent the induration of the regolith observed elsewhere, maintaining large surface dust reservoirs available for
lifting in the atmosphere;
• Within the regolith, wind generated by the sublimation of CO2 ice could exert a drag on the dust
grains, promoting the displacement of these grains;
• This displacement induced by the sublimation
of diurnal frost could trigger the formation of slope
streaks, especially since wind velocities required
for fluidization are met. The sublimation of CO2
frost after dawn may not be the only necessary
factor required for their formation
• This potential CO2 sublimation-driven geomorphological activity show that CO2 is a geomorphological agent on all of the surface of Mars, not only
in the polar latitudes.
Acknowledgements: Part of this work was performed at the Jet Propulsion Laboratory, California
Institute of Technology, under a contract with

NASA. US government acknowledged.
References:
[1] Diniega et al. (2021), 10.1016/j.geomorph.
2021.107627
[2] Piqueux et al. (2016), 10.1002/2016je005034
[3] Khuller et al. (2021), 10.1029/2020JE006577
[4] Mischna & Piqueux (2020), 10.1016/j.icarus.
2019.113496
[5] Schorgoffer et al. (2007), 10.1016/j.icarus.
2007.04.026

[6] Jakosky & Christensen (1986), 10.1029/
jb091ib03p03547
[7] Mellon et al. (2000), 10.1006/icar.2000.6503
[8] Piqueux & Christensen (2009),
10.1029/2008je003308, 10.1029/2008je003309
[9] Putzig et al. (2005) 10.1016/j.icarus.
2004.08.017
[10] Christensen et al. (2004),
10.1007/978-0-306-48600-5_3
[11] Edwards et al. (2011), 10.1029/2010je003755
[12] Singh, & Flanner (2016), 10.1002/2016je005040
[13] Lange et al. (2022), 10.1029/2021JE006988
[14] Ruedrich & Siegesmund (2006), 10.1007/
s00254-006-0585-6
[15] Woronko & Pisarska-Jamroży (2015),
10.1002/ppp.1855
[16] Kieffer (2007), 10.1029/2006je002816
[17] Schmidt et al. (2017), 10.1038/ngeo2917
[18] Chuang et al. (2007), 10.1029/2007gl031111
[19] Bhardwaj et al. (2019),
10.1029/2018RG000617
[20] Atwood & McEwen (2013), 10.1002/grl.
50586
[21] Cedillo et al. (2011),
10.1029/2011GL049403