Testing the efficacy of a Mars submillimeter (submm) sounder for atmospheric measurements
L. K. Tamppari, Jet Propulsion Laboratory/California Institute of Technology, Pasadena, CA, USA
(leslie.tamppari@jpl.nasa.gov), N. J. Livesey, W. Read Jet Propulsion Laboratory/California Institute of
Technology, Pasadena, CA, USA, D. Banfield, NASA Ames Research Center, Moffett Field CA, USA, B. Ward,
Flatiron Institute, New York, NY, USA, F. Forget, E. Millour, Laboratoire de Météorologie Dynamique, Paris,
France, L. Steele, European Centre for Medium-Range Weather Forecasts, Reading, UK, M. Kahre, NASA
Ames Research Center, Moffett Field CA, USA, R. Haberle, Capitola, CA, USA, G. Chattopadyay, D. Hayton,
Jet Propulsion Laboratory/California Institute of Technology, Pasadena, CA, USA, P. Hartogh, Max Planck
Institute for Solar System Research, Göttingen, Germany.
Introduction: The study of the Martian atmosphere has advanced significantly over the last 30
years, however, daily, global measurements of winds
and water vapor have not been obtained, yet are crucial for understanding vapor transport, illuminating
the role of water-regolith interaction, and quantifying
the large-scale circulation and wave activity that are
present in the current climate on Mars. Furthermore,
understanding the current climate and being able to
constrain global circulation models (GCMs) are necessary for accurately evaluating scenarios for the past
Martian climate, during a time when it was vastly
different and likely had flowing liquid water and
possibly microbial life.
Here we describe our efforts to construct the key
components of an end-to-end Observing System Simulation Experiment (OSSE) suitable for quantifying
the potential scientific value of vertically resolved
wind and water vapor observations of the Martian
atmosphere from an orbital vantage point. OSSEs,
originally developed for quantifying the benefits of
potential new Earth meteorological observations, are
increasingly used to measure the potential utility of
observation concepts across a wide range of disciplines. An end-to-end OSSE comprises a truth data
set (such as GCM output) often known as the “nature
run”, a synthetic instrument data set sampled from
the truth data set, the capability to infer atmospheric
phenomena from both, and evaluation of the ability
of the putative measurements to answer scientific
questions. We recognize that, as for many others, our
approach stops short of a full OSSE which, in its
original form, includes work to assimilate observations, or in some cases radiances, into a model. Accordingly, we refer to our process as a pseudo-OSSE.
In our pseudo-OSSE, we assess the efficacy of a
submm sounder [1] capable of retrieving vertical
profiles of water vapor, winds, and temperature to
reproduce the “true” atmospheric fields provided by
the LMD GCM. Our objective is to determine how
well the simulated measurements are able to match
the input data, giving insight into the ability of the
submm instrument to answer outstanding scientific
questions and to identify potential improvements to
the instrument and/or its measurement strategy. Hav-

ing a system to generate and compare simulated
measurements to “truth” input allows for future assessments of various sampling strategies, orbital patterns, or instrument configurations and identification
of those optimal for addressing a variety of science
questions.
Methods: We assumed that the submm instrument would be flown on a spacecraft orbiting Mars at
~300 km altitude, with a 92 inclined, sunsynchronous, near-circular orbit, with an ascending
nodal time of 3 p.m. We chose this orbit because it
is similar to orbits that have been used for MRO and
MGS, and would allow for qualitative comparisons
to published data acquired from those spacecraft
(e.g., water vapor column abundances from TES).
We assumed the instrument would have two identical instrument assemblies, each consisting of a 2axis scanning telescope, a submm receiver, spectrometer, and associated electronics, and that each
telescope would have both vertical and azimuthal
scanning capability. We initially chose to examine
an azimuthal sampling pattern in which one antenna
was continuously pointed such that the line of sight is
oriented south-to-north at the measurement location,
and the other was continuously steered such that the
line of sight is oriented west-to-east at the measurement location. This pointing was chosen to simplify
the evaluation of the zonal and meridional winds.
This sampling pattern, using the current sampling
frequency, produced some unsampled latitude bands
near the southern pole. We assume a vertical scan
from the surface to 150 km over 76 sec. This provides a sampling frequency of one sample roughly
every 4 on a great circle.
We evaluated the submm single-profile precisions and vertical for wind as a function of varying
amounts of water vapor (Fig. 1). Because all available emission lines are used in combination to retrieve
all quantities of interest (temperature, winds, water
vapor), the precision due to the varying water vapor
can affect precisions of the other retrieved quantities
as well, though to a far smaller extent.
Results:
Wind and Temperature evaluation. We started
with an evaluation of seasonally varying zonal-mean

zonal wind cross-sections and stationary wave phases
and amplitudes. The ability of the simulated measurements to yield accurate and detailed information
about these atmospheric dynamical processes was
evaluated against similar analysis of the complete,
regular latitude/longitude gridded, full timeresolution, un-averaged and noise-free data from the
GCM to quantify the simulated instrument’s ability
to capture these scientifically important phenomena.
These analyses follow from the work of Banfield et
al. [e.g., 2-4].

Figure 1. Wind retrieval single-profile precisions (left)
and vertical resolutions (right) for varying water column
abundances.
In Figure 2, the panels show the comparison between the LMD GCM and the synthetic data for the
zonal means of temperature, zonal wind, and
meridional wind for a Mars month of data in late
northern winter. The overall temperature structure
agrees quite well, but there are locations where the
synthetic observations are off by about 5K. The
zonal-mean zonal wind structure also agrees quite
well, except that the chosen cadence of observations
causes a latitudinal data gap in the polar vortex location at that season. Consequently, the position of this
jet stream is well-estimated (from the temperatures
and using the thermal wind equation), but the zonal
winds are not directly measured in the jet itself.
This insight is valuable and argues in favor of choosing other steering/scanning strategies, such as a more
rapid scan to achieve denser latitudinal coverage.
The zonal-mean meridional wind shows a relatively
poor match, not only in magnitudes, but also in overall meridional structure. The cause of this discrepancy may be due to the sampling pattern and further
investigation would assess this by examining a variety of patterns.
The wavenumber 2 stationary wave results (Figure 3) exhibit the same latitudinal gaps as the zonal

mean results, but in general have overall better
agreement with the LMD GCM. In all cases, a gap
was apparent in the wind results for Z<1H (i.e., the
bottom scale height of the atmosphere). This is a
result of the poorer quality of the wind information in
this region as shown in Fig. 1.

Figure 2. Comparison between the LMD GCM (left
column, labeled “Truth”) and the L2 synthetic data (right
column) for the zonal means of temperature (a), zonal
wind (b), and meridional wind (c). Note that gaps in the
data result from the directly east-west and directly northsouth views that were assumed in generating the synthetic
data set (b and c, respectively; for a both are used).

Figure 3. Comparison between the LMD GCM (left column, labeled “Truth”) and the L2 synthetic data (right column) for the zonal mean wavenumber 2 stationary wave
as reflected in amplitude (contours) and East longitude of
maximum (hue; scale bar at bottom) for temperature (a),
zonal wind (b), and meridional wind (c).
Water vapor evaluation: Our first comparison
was to create a representation of the annual water
cycle (Fig. 4) in both the LMD GCM output and the
synthetic data. Our goal was to determine if the
submm sounder could adequately reproduce the water vapor column abundances throughout the annual
water cycle.

Figure 4. Seasonal zonally-averaged water vapor
column abundances from the synthetic L2 data (top).
(Bottom) Absolute difference in water vapor abundance, in
pr m, between the LMD GCM and synthetic data. The
differences are generally <5%, even with the absolute
differences are largest.
We also examined how well we could reproduce
the spatial distribution of the water vapor as a function of season (Figures 5, 6).

Figure 5. Seasonal snapshots of the spatial distribution of water vapor column abundances. The left column
shows LMD GCM output and the right column shows the
synthetic data. The top row shows the average column
abundances in 3.75 x 5.625 lat/lon bins (the LMD output
resolution) for a 30 seasonal window, centered at Ls =
30, which represents the driest season planet-wide. The
middle row shows similar plots for Ls = 120, which repre-

sents the annual maximum in atmospheric water vapor
column abundances. The bottom row shows similar plots
for Ls = 270, which represents the annual maximum atmospheric water vapor column abundances in the south
polar region. Note the vapor abundance color scale bar
changes for each row. The water vapor abundances are
not scaled for surface pressure, in order to highlight the
changing amounts with season and location.

Figure 6. Seasonal snapshots of the spatial distribution of water vapor column abundances shown in Fig. 5.
The left column shows absolute difference between the
LMD GCM and the synthetic data, and the right column
shows the percent differences between them. Note that
the absolute difference scale bar (left column) changes by
row. The top row shows the results for the Ls = 30 window, the middle row shows the results for the Ls = 120
window, and the bottom row shows the results for the Ls =
270 window. White contours show the column abundances.
It is also of interest to evaluate whether or not the
submm instrument would be able to determine the
timing of the seasonal global as well as hemispheric
maximum and minimum water vapor in the system to
within 5 Ls, to identify any interannual changes that
may occur, and to reproduce both the magnitude of
the total vapor as a function of season as well as its
hemispheric distribution. [5] showed that there was
interannual variability in the yearly peak of the water
vapor abundances of about 0.1 x 1015 g and the seasonal date of that peak varied by 30-40 Ls centered
around Ls = 120. The LMD GCM shows the annual
water cycle peak closer to Ls = 100. Figure 7 shows
that the synthetic L2 data do a very good job representing the LMD GCM. The curves are within 4%
in all cases, and outside of the Ls = 120-150 season
they are within ~1-2%, so determining the seasonal
total and hemispheric abundance maxima and minima and timing is achievable.
We examined the LMD GCM and synthetic data
vertical water vapor magnitude and distribution for
the same three 30 seasonal periods as before: Ls =
30, 120, and 270 (Fig. 8). The zonally-averaged
synthetic data for each seasonal period examined are

very similar to the LMD GCM, with latitudes and
altitudes of water vapor mass mixing ratios matching
well. The maximum absolute difference between the
LMD and the synthetic data is 8x10-5 kg/kg, which
occurs for annual maximum water vapor abundance
of 0.001 kg/kg in the high northern latitudes, at low
altitude, during northern summer. This is equivalent
to an 8% difference. When vapor mass mixing ratios
are extremely small, which occurs in the polar regions during dry conditions and at high altitudes,
differences can be up to 100%.

Figure 7. (Top) Total atmospheric water vapor abundance (solid line) as a function of season and separated
by northern (dashed line) and southern (dotted line) hemispheres as a function of season for the LMD GCM (blue)
and synthetic L2 data (“simulated” shown in red). Cf.
Smith (2002), Figure 8. (Bottom) Percent difference between LMD GCM and synthetic L2 total global atmospheric
water vapor abundances.

Figure 8. Zonally-averaged vertical distribution of water vapor mass mixing ratio for three 30 seasonal periods:
(top row) Ls = 30, planet-wide dry season, (middle row) Ls
= 120, northern hemisphere vapor maximum, and (bottom row) Ls = 270, S hemisphere vapor maximum. The
left column shows the LMD GCM output and the right column shows the synthetic L2 data.
We also examined the ability of the submm to reproduce the zonal-mean vapor mass stream function.
Figure 9 shows this for the LMD and the synthetic
data sets for the early northern summer time period,
near the water vapor column maximum for the year.
The mass stream function contours compare well,

with a slightly higher estimate in the L2 synthetic
data. This may result from the finer vertical gridding
used in the L2 synthetic data set, the lat/lon gridding,
and/or the fact that there are fewer time steps in the
L2 data used (2deg LS, whereas the LMD is averaged
out to once-per-day).

Figure 9. The zonal-mean vapor mass stream functions (contours) for the LMD data (top) and the synthetic
data (bottom) in units of 108 kgs-1 (white, dashed contours). In each case, the water vapor mass mixing ratios
are shown as well, in color.
Conclusions and Future Work: We have constructed key components of an end-to-end Mars
OSSE and have demonstrated the efficacy of a twoantenna submm instrument for measuring vertical
profiles of temperature, winds, and water vapor from
an orbital platform. The pseudo-OSSE we have developed (process ends prior to full assimilation)
demonstrates the framework on which future instrument evaluation/optimization could be conducted.
Our results show that a submm instrument, operating in a near-sun-synchronous orbit, would provide
temperature, wind, and water vapor profiles with
accuracy sufficient to measure key aspects of the
martian atmosphere, many of which have not been
measured or have been insufficiently measured, such
as the strength of the zonal jets and a seasonal &
global picture of water vapor vertical distribution.
This initial analysis is valuable to direct future efforts for improvement. Ensuring adequate coverage
of the high latitudes is among the top priorities and
could be improved by scanning a smaller altitude
range. This could be evaluated in combination with
a slower (faster) scan speed to determine if relatively
higher (lower) precision is required balanced against
latitudinal sampling (a faster scan speed for a given
altitude range would yield more frequent latitudinal

sampling). In addition, specialized scanning patterns
could be evaluated.
Bibliography: [1] Read et al., 2018. [2-4]
Banfield et al., 2000, 2004, 2019. [5] Smith 2002.