SURFACE DUST REDISTIBUTION ON MARS FROM INTERANNUAL
DIFFERENCES IN TEMPERATURES AND ALBEDO
J. Bapst (jonathan.bapst@jpl.nasa.gov), S. Piqueux, D.M. Kass, A. Kleinböhl, Jet Propulsion Laboratory,
California Institute of Technology, Pasadena, USA, C.S. Edwards, C. Wolfe, Northern Arizona University,
Flagstaff, USA, P.O. Hayne, University of Colorado, Boulder, USA.

Introduction: The focus of this work is to
advance our understanding of the impact of global
dust storms (GDS) on the state of dust reservoirs
because of their propensity for dust lifting and
redistribution [1-4]. Mapping the global distribution
of surface dust, and identifying dust reservoirs has
been carried out previously using derived albedo
[6,1], visible imagery [7,8,3,9], and dust cover index
derived from infrared spectra [10].
Outstanding questions pertain to dust fluxes and
include whether replenishing of dust reservoirs is
necessary for GDS initiation (i.e., “are dust source
regions exhausted during GDSs?”). General
Circulation Models (GCM) initialized with a finite
supply of surface dust show that source regions will
be quickly depleted and inhibit dust storm activity
[11]. This is not consistent with observations [12].
Thus, [11] concluded that source regions are large
reservoirs, providing dust for lifting on at least multidecadal timescales, and that the interannual
variability in dust storms is not due to availability of
surface dust. Another hypothesis is that dust must be
deposited or removed in certain geographic locations
to explain interannual variability of GDS [13].
Approach: We identify the location of dust
sources and sinks and study their changes before and
after GDS. We leverage the nearly continuous and
global surface temperature record available at Mars.
Orbital measurements of surface temperature
spanning 10 Mars Years (MY), along with lesscomplete albedo datasets, are used as proxies for
surface dust distribution, allowing us to track changes
across Mars through time. The high sensitivity of
those measurements provides a different and
complimentary perspective into dust redistribution
that is not possible using visible datasets alone.
Our approach primarily involves the analysis and
identification of temperature changes from one Mars
year to the next. Martian dust is characterized by its
relatively high albedo (>0.25; [6,14]) and low thermal
inertia (20–100 J m-2 K-1 s-1/2), in contrast to the rest
of the surface which, besides ices that exhibit high
albedo value, is characterized by lower albedo (<0.25)
and has thermal inertia values between 100–400
J m-2 K-1 s-1/2 [15,16]. Owing to differences between
dust and other surface types, the redistribution of
surface dust can result in changes in albedo and/or

thermal insulation, which affects surface temperature
via surface-energy balance [17,6,1]. Therefore,
surface temperatures changes can help us understand
the spatial and temporal nature of dust redistribution.
To study dust redistribution globally, we consider
two instrument datasets: Thermal Emission
Spectrometer (TES [14]) aboard Mars Global
Surveyor and Mars Climate Sounder (MCS [18])
aboard Mars Reconnaissance Orbiter. TES and MCS
data are binned in an identical manner. Data are
separated by local time into AM (<12 PM) and PM
(>12 PM). We divide the data by season into 15° LS
windows, which is necessary when carrying out
difference operations as seasonal differences will
dominate for larger LS windows. An LS window of
15° provided the best spatial coverage while limiting
the seasonal influence on temperature at 1°×1°
latitude by longitude spatial resolution. To achieve a
more-complete map concerning data density, we
average (median) a full season (90° of LS or six 15°
maps) of data after calculating interannual
differences. Analysis is limited below 60° latitude in
both hemispheres. Temperatures at higher latitudes
should be interpreted with caution, as they are subject
to variable presence of seasonal surface frosts [19,20],
subsurface frost with the potential for gardening [21],
and enhanced aerosol and condensate loading in the
atmosphere (i.e., polar hoods).
Following binning, interannual difference maps of
temperature and albedo are generated and averaged
over a larger seasonal baseline (90° of LS) for
analysis. These changes in surface temperature and
albedo are interpreted in terms of dust redistribution;
i.e., AM temperatures would decrease with dust
deposition as it would increase albedo and lower the
effective surface thermal inertia. Geographical
naming convention follows [8].
MY25 GDS: Interannual changes in albedo before
and after the MY25 GDS exhibit the largest and most
extensive changes throughout the TES observations
included in our analysis. Extensive areas of change
are observed south of the Tharsis province (10°N, 120°E) in Daedalia Planum (-25°N, -110°E), south of
Valles Marinaris (-10°N, -70°E) in Solis Planum (25°N, -75°E), in and around Hellas Planitia (-45°N,
70°E), and in Syrtis Major (10°N, 70°E)—all
consistent with findings from previous studies [8,1,3].

Figure 1: Global map of MCS AM surface temperature change, LS=0°–90°. Note the lack of data coverage near Gale Crater
in MY31 and after.

We see excellent agreement between the location
of albedo changes and inversely corresponding
changes in surface temperature. We would expect the
strongest agreement with albedo change in PM data,
due to the local time dependency. Compared to
albedo, changes observed in PM temperatures are not
as clearly defined and are more susceptible to orbitto-orbit variability than albedo changes.
AM surface temperature changes display a similar
pattern, and the geographic extent of the changes are
in good agreement. One major difference is that AM
temperature changes are the same sign but are less
than half the magnitude of those changes observed in
PM. Because the magnitude of AM changes is less
than PM changes (i.e., smaller signal), the AM maps
are more susceptible to small changes that do not
represent surface dust redistribution. The similarities
between interannual PM and AM surface
temperatures changes bracketing the MY25 GDS are
noteworthy because both AM and PM surface
temperature changes inversely correlate with albedo.
This implies that changes in temperature are driven
primarily by changing surface albedo opposed to the
insulating effects of dust.
Overall, the interannual changes for the pair of
years beyond the MY25 GDS show albedo and
temperature changes smaller in magnitude compared
to the years surrounding the GDS. Changes in albedo
supports the idea that the surface may require more
than one Mars year to return to a MY24-like or “prestorm” state. Surface temperatures show the same
behavior and do not invert over the course of one MY.
MY28 GDS: No TES or MCS surface observations are available before and after the MY28 GDS,
but substantial changes in AM temperature are
observed in MY29 and MY30, Fig. 1. Persistent
temperature changes in the region east of Valles
Marineris and north of Tharsis could be due to
atmospheric activity such as water ice clouds [22].
MY29 to MY30, show substantial changes in
temperature, most notably north and northwest of
Tharsis in Amazonis Planitia. [3] identify a spatially-

coincident change in reflectance in this region
between global mosaics between MY29 and MY27.
This is consistent with dust deposition in this region
about the MY28 GDS. Because our analysis brackets
the years directly following the MY28 GDS (i.e.,
MY29 to MY30) we would expect a decrease in
albedo and thus a positive temperature change. This
was observed in the years following the MY25 GDS,
in regions where dust was initially deposited during
the storm (e.g., Hellas Planitia).
MY34 GDS: A positive temperature change
identified in the AM temperature map bracketing the
MY34 GDS suggests a decrease in albedo and/or an
increase in the effective thermal inertia, both
consistent with dust removal. Although we do not
have simultaneous, coincident observations of
broadband albedo, observed changes in temperature
guide us towards visible (or other) datasets that may
corroborate the temperature changes (Fig. 2). This
region, along the border of Arabia and Acidalia, is a
well-documented storm track [23,24] where
substantial changes in albedo have been observed
over decades of monitoring [3,9].
Two MRO Context Camera (CTX [25]) images
bracket the MY34 GDS in this region of observed
change and share almost exactly the same viewing
geometry (Fig. 2)—The latter restriction is employed
to rule out explanations, besides dust redistribution,
for the observed changes between images. A pair of
images captured by the High-Resolution Imaging
Science Experiment (HiRISE;[26]), with similar
viewing geometries, are also identified (Fig. 2).
Visible images capture clear evidence for dust
removal about the MY34 GDS, located in the broad
region of AM temperature change identified in MCS
data within NW Arabia Terra. The “before” scenes
show brightness is primarily a function of incidence
angle (i.e., slope). West-facing slopes receive more
solar flux (~3 PM local time), but have similar
reflectance properties, and are therefore brighter in
the image. The “after” scenes are markedly different,
where brightness corresponds to local topographic
highs and visible surface roughness. The eroded rim

Figure 2: (left) Regional view of the positive temperature anomaly (See Fig. 1). Temperature changes <+/-2 K
are omitted from view. CTX view of a region in NW Arabia Terra (a) before and (b) after the MY34 GDS.
Stretched HiRISE color images (red outline) show stark differences in the color information (c) before and (d)
after the MY34 GDS. CTX IDs from left to right are J20_052788_2109_XI_30N016W and K22_061795_
2095_XI_29N015W, and for HiRISE, ESP_053500_ 2095 and ESP_059329_2095, respectively. North is up.
of the larger crater is bright as well as the rim of the
smaller crater in the scene. This difference suggests
that dust in smooth or flat-lying regions of the scene
was removed in the time between images (i.e., MY34
GDS), resulting in a net decrease in albedo (and
potentially net-increase in thermal inertia) and
therefore a net-increase in temperature, consistent
with our MCS observations. CTX images bracketing
the MY34 GDS were also identified in the region that
did not overlap with MCS temperature changes and
did not show clear evidence for surface changes.
Implications for the dust cycle: One outstanding
question is whether surface dust reservoirs are finite
in volume and need to be resupplied for GDS to occur
[15] or have a virtually unlimited dust volume (i.e.,
deep reservoir). Our results for MY25 GDS are
partially ambiguous due to the start of observations of
MGS and the lack of data before the GDS. The first
useful interannual changes are after the global event
has occurred. The changes in years following the
MY25 GDS mostly support a return to “pre-storm”
conditions, supporting a finite reservoir that needs
resupply before another GDS can initiate. However,
the magnitude of change (in albedo and temperature)
is smaller than those across the years directly
following the storm. Some regions appear not to
revert or are doing so slowly (e.g., Daedalia and
Tharsis), likely requiring a GDS-like event, or
multiple years—or both—to return to a prior state, if
at all [7,1]. These observations support the deep
reservoir model, where dust is available in quasiunlimited supply.
MCS observations support the deep reservoir
model as no large changes or systematic changes over
time are observed in the leadup to the MY34 GDS. In
particular, the region showing the strongest evidence
for dust removal, NW Arabia Terra, showed no

significant interannual changes in AM temperature
for the previous five Mars Years. Thus, results
suggest replenishment of well identified discrete dust
source regions is not necessary for initiating GDS.
Using a thermal model to quantify the amount of
dust necessary to fit the observed temperature
changes, the MY25 interannual changes in AM and
PM temperature are consistent with dust layer
thickness changes <100–1000 microns as an upper
limit, and down to 5–10 microns of dust displacement
as a lower limit based on the interpretation of albedo
changes. Results support the notion that dust can
significantly impact albedo, requiring only addition or
removal on the order of monolayers, while having a
negligible impact on surface thermal properties [5].
These observations support a deep reservoir for dust
sources, insofar regions of dust are known to be at
least decimeters thick and potential many 10s of
meters thick [27]. Annual or interannual removal of
surface layers on the upper end of our estimates (~100
microns) implies dust reservoir lifetimes of at least
103 years and potentially much longer.
Laboratory work by [5] linking albedo changes to
dust deposition thickness can provide a lower limit of
the total mass of dust transported globally about
GDSs. This quantity, summed across the entire
surface about the MY25 GDS equates to 8.0x1010 kg
of dust removed and 5.8x1010 kg dust deposited
calculated over one Mars year. The values imply at
least 8.0x1010 kg of dust involved in transport about
the storm. Comparing this to historical estimates of
dust flux supports this value as a lower limit. For
example, [28] estimate from Viking Orbiter data a
yearly dust flux of 2.9x1012 kg about the MY13 GDS
(1977). [29] estimates 4.3x1011 kg of dust as peak
atmospheric loading of the same storm with Viking
Orbiter data. [30] report 5.0x1011 kg of surface dust

annually transported. The estimated flux of dust lifted
by dust devils are of similar magnitude as other
reported estimates, where between 2.3–5.8x1011 kg of
dust is lifted annually (limits reported by [31] and [8],
respectively).
Although we have not derived quantitative albedo
from MCS visible channel observations to support
calculations of dust flux directly, as was performed
with TES data, we can calculate the local dust flux
from the region in NW Arabia where we identified
dust removal and estimated albedo changes (Fig. 3).
Following the same approach as performed with TES
albedo, we sum the dust thickness change across the
region of interest. This equates to 5.4x108 kg of dust,
roughly 2–3 orders of magnitude less than the global
transported mass during other GDS, and notably less
than the flux estimated for a regional storm in [29]—
1.3x1010 kg. Again, comparing our results to those in
the literature suggests our model may be too sensitive
(i.e., the required dust mass to effect albedo change is
larger than what we assume) and/or the assumption of
uniform layers being added or removed at the spatial
resolution of our analysis is unrealistic (e.g., [32,6]).
Allowing for heterogeneous dust addition/removal
(i.e., non-uniformity or sub-pixel variability), and

thus thicker quantities of dust than what we estimate,
is one potential explanation, however there is no
evidence these thicknesses are significant in terms of
affecting insulating properties of the surface.
Differences between observations about the
MY25 GDS and MY34 GDS are substantial. Dust
redistribution across the MY25 event results in clear
changes in both PM and AM temperatures while AM
temperature changes are less pronounced across the
MY34 GDS. One potential cause is the areal extent of
dust removed between the two GDSs. Interannual
temperature changes about the MY25 GDS showed
broad areas of substantial dust removal across the
Tharsis Plateau and Daedalia Planum whereas in
MY34, only a localized region of substantial dust
lifting was identified in northwest Arabia Terra.
Annual storm activity has been reported along the
border of Chryse Planitia and Arabia Terra (i.e.,
Acidalia storm track; [8,33] but was likely more
impactful in the MY34 GDS than the MY25 GDS.
These observations support spatial variability
between GDSs and the inclusion or exclusion of dust
reservoirs that play a major role in the global event.

Figure 3: AM MCS temperature
change from MY34 to MY35 (0 <
Ls < 90) interpreted in terms of surface albedo variation (left) and
corresponding dust thickness using
[5], right. Assuming a bulk dust
density of 1,000 kg m-3, a dust
thickness change of 5x10-3 kg m-2
is equivalent to a surficial dust
layer thickness of ~ 5 µm.

Acknowledgements: Part of this work was
performed at the Jet Propulsion Laboratory,
California Institute of Technology, under a contract
with NASA. US government support acknowledged.
References: [1] M. A. Szwast et al., Journal of
Geophysical Research 111 (2006). [2] B. A. Cantor et al.,
Journal of Geophysical Research 111 (2006). [3] P. E.
Geissler et al., Icarus 278, 279 (2016). [4] T. Bertrand et al.,
Journal of Geophysical Research-Planets 125, 36,
e2019JE006122 (2020). [5] E. N. Wells et al., Icarus 58,
331 (1984). [6] P. R. Christensen, Journal of Geophysical
Research 93, 7611 (1988). [7] P. E. Geissler, Journal of
Geophysical Research 110 (2005). [8] B. A. Cantor, Icarus
186, 60 (2007). [9] D. F. Wellington and J. F. Bell, Icarus
349, 25, 113766 (2020). [10] S. W. Ruff and P. R.
Christensen, Journal of Geophysical Research 107 (2002).
[11] M. A. Kahre et al., Geophysical Research Letters 32,
L20204 (2005). [12] R. W. Zurek and L. J. Martin, (Amer
Geophysical Union, 1993), pp. 3247. [13] D. P. Mulholland
et al., Icarus 223, 344 (2013). [14] P. R. Christensen et al.,
Journal of Geophysical Research 106, 23 (2001). [15] F. D.
Palluconi and H. H. Kieffer, Icarus 45, 415 (1981).

[16] N. E. Putzig and M. T. Mellon, Icarus 191, 68 (2007).
[17] H. H. Kieffer et al., Journal of Geophysical Research
82, 4249 (1977). [18] D. J. McCleese et al., Journal of
Geophysical Research 112 (2007). [19] J. Bapst et al.,
Icarus 260, 396 (2015). [20] S. Piqueux et al., Icarus 251,
164 (2015). [21] C. Pilorget and F. Forget, Nature
Geoscience 9, 65 (2016). [22] M. D. Smith, Icarus 202, 444
(2009). [23] H. Q. Wang and M. I. Richardson, Icarus 251,
112 (2015). [24] M. Battalio and H. Q. Wang, Icarus 354,
20, 114059 (2021). [25] M. C. Malin et al., Journal of
Geophysical Research 112, S504 (2007). [26] A. S.
McEwen et al., Journal of Geophysical Research 112,
E05S02 (2007). [27] P. R. Christensen, Journal of
Geophysical Research 91, 3533 (1986). [28] J. B. Pollack et
al., Journal of Geophysical Research 84, 2929 (1979). [29]
T. Z. Martin, Journal of Geophysical Research 100, 7509
(1995). [30] B. A. Cantor et al., Journal of Geophysical
Research 106, 23653 (2001). [31] P. L. Whelley and R.
Greeley, Journal of Geophysical Research 113 (2008). [32]
P. Thomas et al., Icarus 60, 161 (1984). [33] D. M. Kass et
al., Geophysical Research Letters 43, 6111 (2016).