DETECTION OF GRAVITY WAVES ON MARS USING THEMIS BAND 10.
J. M. Battalio, Dept. of Earth and Planetary Sciences, Yale University, New Haven, CT (joseph.battalio@yale.edu),
N. Heavens, A. Pankine, Space Science Institute, Boulder, CO, J. Cowart, Dept. of Geosciences, Stony Brook University, Stony Brook, NY.

Introduction

Dataset Calibration

Atmospheric gravity waves (GW) emerge from disturbances in a stably stratified air mass and are restored
by buoyancy 1 . GWs control many processes found
throughout all atmospheric layers so are extremely important for understanding the dynamics occurring in
each layer, for capturing the coupling between the nearsurface and the highest levels of planetary atmospheres,
and for closing energy and momentum budgets of the
general circulation 2 . GWs are important for understanding atmospheric dynamics on Mars because they impact
atmospheric escape processes, and as shown by Curiosity, they may instigate cloud activity 3 .
Recently, Heavens et al., (2020, 2022) 4;5 developed
a climatology of GWs from the Mars Climate Sounder
(MCS). However, MCS can only look for GWs in the
meridional direction; it cannot detect GWs in the zonal
direction. To look for GW in the zonal direction, an
imager can be used, like the Thermal Emission Imaging
System (THEMIS). Specifically, we use Band-10, which
has a weighting function that peaks at about 25 km above
the surface, to search for GW activity in both the meridional and zonal directions. THEMIS has been in orbit
on Mars Odyssey for two decades, yielding hundreds
of thousands of image swaths. We focus on a season
that overlaps with the brief time that MCS took nadir
observations, Ls =120◦ –150◦ . Here, we summarize an
analysis of GW activity from THEMIS observations over
Mars years (MY) 26–33 6 .

Bands 1–9 of each THEMIS image swath undergo a
multi-step calibration pipeline before GW activity is
calculated. However, the calibration for bands 1–9 is
performed assuming that Band 10 is thermally homogeneous across the imaging scene, and Band 10 is left
unaltered so that the calibration is reversible for other
bands. GW activity invalidates this assumption, so the
calibration must be performed for Band 10 separately
from data on the PDS.
After converting the THEMIS RDR files into ISIS3
format (Fig. 1, Raw Swath), each image cube has banddependent row and line correlated noise removed via
the DaVinci “deplaid” and “destreak” functions 9 (Fig.
1, Destreak & Deplaid). Time-dependent focal plane
variations due to long-period thermal variation are removed 10 , similar to the “uddw” (UnDrift-DeWobble)
DaVinci function but implemented for THEMIS Band
10. The Band 10 implementation necessitates a longer
1x400 boxcar filter (Fig. 1, Undrift & Dewobble). Temperature variations across the calibration flag are removed, similar to the “rtilt” DaVinci function 9 . An additional deplaid is applied to remove the band-independent
Band 10 noise (Fig. 1, Rtilt & Deplaid). An additional set of destreak filtering is applied to remove several columns that are particularly oversaturated in Band
10. Each swath is sinusoidally map projected. Finally,
swaths are binned at 1 km to improve the signal to noise.

Calculation of Gravity Wave Spectra
THEMIS Band 10
The THEMIS instrument consists of a set of two cameras
that image Mars at thermal-infrared and visible wavelengths 7 . The IR camera images across ten spectral
filters between 6.5 and 14.88 µm, and the visible images
between 0.45 and 0.85 µm. We diagnose GW activity on Mars using only THEMIS Band 10 observations.
We focus on THEMIS Band 10 because it is centered
on the CO2 absorption band at 14.88 µm, so emission
predominantly comes from the atmosphere.
THEMIS images are 320 pixels across-track and
have a spatial resolution of ∼100 m/pixel, with a width
of 32 km. Images are taken in the along-track direction due to spacecraft motion and vary in length from
∼30 km to hundreds of km, depending on the scientific
motivation for a specific observation.
As Mars Odyssey maintains a Sun-synchronous orbit, images are captured twice a sol between 3:00 and
7:30 AM/PM depending on year and time of year. For
Ls =120◦ –150◦ , observations during MYs 26–28 occurred around 5:00, migrated to 4:00 for 29–31, and
then shifted to 7:00 for MY 33 8 .

GW activity is calculated along multiple, unique baselines across each THEMIS Band 10 swath. To select the
baselines, points are taken every 16 km in the cross-track
direction and every 32 km in the along-track direction
on swath edges. This establishes a 16 × 32 km grid
of points. A line is swept 180◦ through each point, and
the unique intersections generate the baselines on which
GW activity is calculated. The 180◦ sweep constitutes
the various orientations that GW activity can take, because the direction of wave propagation cannot be determined from a single swath. This results from spacecraft
motion  wave phase speed, so each swath is considered instantaneous. The baseline length varies from 32
km to ∼2,000 km, constrained by the swath dimensions.
For each baseline, GW activity is estimated following Heavens et al., (2020) 4 . The variance of the GW
is estimated from the brightness temperature measurement 11;12 , which is calculated from the retrieved Band 10
radiance. To account for differences in average temperature, the GW variance in an observation is normalized
by the mean of the squared temperature in each measurement within the observation. This is denoted as Ω̂GW ,

THEMIS Gravity Waves

Figure 1: Sequence of the pre-processing steps for THEMIS band 10 swath (I55019007) at Ls =128.8◦ , MY 32 which occurs at
297.7◦ E, 40.2–43.3◦ S. To show each step, the swaths are each sinusoidally projected, but the preprocessing is all performed before
map projection. Gray shading indicates areas outside of the swath. Adapted from Battalio et al., (submitted) 6 .

with units K 2 K −2 . The associated GW activity variance is binned for length-orientation on each baseline
every 1 km × 2◦ . All swaths with length > 70 km are
subset into multiple trapezoids of size ∼32 × 32 km, and
the GW activity procedure is repeated on them. These
are called “symmetric” swaths.

-70◦ orientation (black, dashed lines), a 30 km baseline
wave with -50◦ orientation (red, dashed lines), and an
80 km baseline wave with 50◦ orientation (red, dashed
lines). The average spectrum (Fig. 2b) exhibits a peak
for each of these modes.

Each THEMIS swath has an associated unique GW
spectrum, and the above method robustly identifies the
main GW modes. The swath calibrated in Fig. 1 has
several GW modes superimposed, and the average spectrum captures each (Fig. 2a). The spectrum is binned
by 1 km of baseline and by 2◦ of orientation. The spectral peaks above the mean are indicated and numbered
in decreasing amplitude. For sufficiently long swaths, a
particular baseline (length, orientation) bin may contain
multiple baselines; these are averaged together. By inspection, there are at least three GW peaks denoted by
the dashed lines in Fig. 2a: a 20 km baseline GW with

Average Gravity Wave Activity
GW activity exhibits several overarching patterns (Fig.
3a,d). Activity maximizes near the southern polar cap,
poleward of 60◦ S both during the day (Fig. 3a) and
night (Fig. 3d), greater than log10 (Ω̂GW )=−5 K 2 K −2 .
Within this zonal band of greatest activity, there is some
longitudinal localization at 150◦ –100◦ W and 30◦ –70◦ E,
to the east of the large Argyre and Hellas basins. Outside of this zonal band, GW activity is reduced in the
equatorial regions, particularly at night, and there is low
activity throughout the northern hemisphere.

THEMIS Gravity Waves

Figure 2: The THEMIS band 10 swath processed in Fig. 1 binned at 1 km resolution (a) and the associated average GW spectrum
binned every 1 km of baseline by 2◦ or orientation (b). Blue crosses indicate locations defining baseline sweeps, and dashed lines
denote GW fronts in (a), and numbers indicate the spectral peaks in (b). Gray shading indicates a lack of data. Adapted from
Battalio et al., (submitted) 6 .

Figure 3: Peak GW activity (log10 Ω̂GW K 2 K −2 ) (a,d), orientation (deg from north-south) (b,e), and baseline length (km) (c,f)
for all THEMIS GW observations during Ls =120◦ –150◦ , MY 26–33 binned at the median at 4◦ resolution. Daytime observations
(a–c) and nighttime observations (d–f). Gray shading indicates lack of observations. Adapted from Battalio et al., (submitted) 6 .

THEMIS Gravity Waves

Figure 4: Average GW activity maximum at a particular orientation for the day (top) and night (bottom) for all complete THEMIS
swaths (blue) and for all swath subsets (orange). Shading indicates standard deviation. Adapted from Battalio et al., (submitted) 6 .

There is considerable variance in the approximate
wavelength of the GWs, measured as the length of the
baseline. The shortest observed wavelengths of less than
40 km are near the southern pole, both at day (Fig. 3c)
and night (Fig. 3f); however, daytime wavelengths are
shorter than nighttime. The longest lengths of greater
than 100 km are found just to the north, colocated with
the highest latitudinal gradients of GW activity along
60◦ S. Outside of these areas, there is little organization
of the baseline lengths at night.
The average GW orientation is generally north-south,
northward of the greatest activity along 60◦ S, both during the day (Fig. 3b) and night (Fig. 3e), though the
activity is somewhat west of due north at night and somewhat east of due north at day by approximately 10◦ . This
effect is an observational bias in the data caused by the
trajectory of Odyssey in its sun-centric orbit, which is
south to north during the day and north to south at night.
Despite this bias, GWs during Ls =120◦ –150◦ are
isotropic, meaning that no particular orientation is favored when considering observations at all spatial locations both during the day (Fig. 4, top) and night (Fig.
4, bottom). For both symmetrically subdivided swaths
(Fig. 4, orange) and even for the full swath lengths (Fig.
4, blue), activity is not distinguishable from a constant
value across all orientations. This shows that despite
biases from spacecraft motion, the method of evaluating
GW activity by rotating through all baselines is robust.
Summary
Gravity waves on Mars contribute to atmospheric escape, impact cloud formation, and provide an impor-

tant mechanism for energy and momentum transport.
With increasing complexity of gravity wave parameterizations used by Mars global circulation models, the
development of complete gravity wave datasets is crucial for model verification. Until now, many gravity
wave climatologies have used instrumentation sensitive
to waves oriented only in the meridional direction, missing activity oriented zonally. To remedy this coverage
gap, gravity waves are analyzed from THEMIS Band 10
over approximately one Mars month, totaling >30,000
spectra. Gravity wave activity exhibits an isotropic distribution among wavelengths detectible in all directions
(<20 km). In directions sensitive to longer wavelengths,
activity at shorter baselines remains the most prominent.
Activity peaks near the southern pole, at appreciably
longer wavelengths (∼100–200 km).

Acknowledgments
This work was funded by NASA’s Mars Data Analysis
Program (80NSSC19K1215).
R EFERENCES : [1] Fritts et al. 2003, Rev. of Geophys., 41, 1. [2] Medvedev et al. 2019, Atmosphere, 10,
531. [3] Guzewich et al. 2021, JGR: P, 126. [4] Heavens et al. 2020, Icarus, 341, 113630. [5] Heavens
et al. 2022, PSJ, 3. [6] Battalio et al. submitted, JGR:
P. [7] Christensen et al. 2004, Space Science Reviews,
110, 85. [8] Smith, M. D. 2019, Icarus, 333, 273. [9] Edwards et al. 2011, JGR: P, 116, 1. [10] Banfield et al.
2004, Icarus, 170, 365. [11] Wu et al. 1996, GRL, 23,
3289. [12] Wu et al. 1996, GRL, 23, 3631.