WATER VAPOR ON MARS: A REFINED CLIMATOLOGY AND NEARSURFACE CONCENTRATION ENABLED BY SYNERGISTIC
RETRIEVALS
E W. Knutsen, LATMOS/IPSL, UVSQ Université Paris-Saclay, Sorbonne Université, CNRS, Guyancourt,
France (elise-wright.knutsen@latmos.ipsl.fr), F. Montmessin, LATMOS/IPSL, UVSQ Université Paris-Saclay,
Sorbonne Université, CNRS, Guyancourt, France, L. Verdier, LATMOS/IPSL, UVSQ Université Paris-Saclay,
Sorbonne Université, CNRS, Guyancourt, France, G. Lacombe, LATMOS/IPSL, UVSQ Université ParisSaclay, Sorbonne Université, CNRS, Guyancourt, France, F. Lefèvre, LATMOS/IPSL, UVSQ Université ParisSaclay, Sorbonne Université, CNRS, Guyancourt, France, S. Ferron, ACRI-ST, boulevard des Garennes,
Guyancourt, France, M. Giuranna, IAPS-INAF, Rome, Italy, P. Wolkenberg, IAPS-INAF, Rome, Italy, A.
Fedorova, Space Research Institute (IKI), Moscow, Russia, A. Trokhimovskiy, Space Research Institute (IKI),
Moscow, Russia, O. Korablev, Space Research Institute (IKI), Moscow, Russia.

Introduction:
Until recently, knowledge of the near-surface
H2O profile on Mars mostly relied on general
circulation models. The vertical distribution of water
vapor has been inferred from nadir measurements
(e.g.[1], [2]) and measured directly by solar
occultation viewing geometry with SPICAM on
Mars Express since 2004, and with the ExoMars
Trace gas Orbiter and its infrared spectrometers
NOMAD and ACS since 2018 (e.g. [3], [4]).
With solar occultation measurements, a very fine
vertical resolution can be obtained, yet
measurements below 10 km are sparse as aerosols at
low altitudes lead to high opacities which
significantly reduce the transmittance. The lower
limit for observation is typically 5-10 km for dustfree conditions, and as high as 20-30 km during the
dusty perihelion season (e.g., Aoki et al., [3]). Only
under very clear conditions are solar occultation
observations able to probe below 10 km, however
such conditions mostly occur at high latitudes. As a
result, information about the low-atmosphere water
vapor profile remains exceptional.
Below 10 km, surface-atmosphere interactions
such as convection, frost sublimation and deposition
are expected to be the main forcers on the vertical
distribution, along with adsorption and desorption.
Above 10 km, water ice clouds are thought to be
dominant [5], [6]. Below the saturation level,
controversy exists regarding whether water vapor is
well mixed with CO2, or distributed in a more
complex manner.
The goal of this study is two-part: firstly, to
distinguish the near-surface water content from the
rest of the column, and secondly, by having
constrained the near-surface concentration, a refined
and highly accurate water vapor climatology is
obtained. The intention of this study is to be largely
descriptive, as this is the first time the spectral
synergy has been applied to a large dataset.

Method:
Two infrared spectrometers onboard the Mars
Express satellite were utilized for this work. Water
vapor
column abundances
were retrieved
simultaneously with PFS (sensing the thermal
infrared range) and SPICAM (sensing the nearinfrared range), yielding distinct yet complementary
sensitivity to different parts of the atmospheric
column.
TIR measurements are mostly sensitive to the
middle atmosphere (at the origin of the photon
emission) where the temperature contrast of the
atmosphere with respect to the surface is high. NIR
measurements on the other hand are sensitive to any
molecule present along the column as the technique
relies on solar photons traversing the entire
atmosphere back and forth.
When observing an atmosphere in nadir viewing
geometry, the outcome is normally a column
abundance value of the target species. However, it is
possible to obtain information about the vertical
distribution of the species by combining multiple
spectral domains in the retrieval process.
By combining two spectral domains, the degree
of freedom of the signal (DOF) is increased
compared to single-domain retrievals. The DOF
gives an estimate of the number of independent bits
of information in an atmospheric measurement [7],
and a DOF higher than 1 indicates the presence of
some amount of profile shape information.
Dataset and retrieval. The synergistic approach
requires a set of co-located PFS and SPICAM
observations on which to apply the retrieval method.
The dataset presented here consists of co-located
observations taken over 1379 individual orbits
distributed across seven Mars years from Ls 334° of
MY 26 to Ls 297° of MY 34, with no measurements
from MY 30-31.
To obtain a satisfying PFS SNR for the fitting of
multiple parameters, nine consecutive spectra are
averaged together. The SPICAM observation closest
in time to the central PFS spectrum is then selected

and averaged with the seven observations prior to it
and the seven after it, resulting in a combined FOV
similar to that of the nine combined PFS
observations.
A priori water vapor and temperature profiles are
extracted from the Mars Climate Database (MCD)
version 5.3, based on the general circulation model
developed at the Laboratoire de Météorologie
Dynamique (LMD GCM), [8], [9].
Temperature and aerosol parameters are retrieved
individually from the PFS average spectra, which are
then injected into the synergistic routine. The overall
spectral fitting procedure uses the HITRAN 2012
spectroscopic database [10] as a baseline for the
computation of absorption coefficients of H2O and
CO2, and then relies on a Bayesian approach that
consists in maximizing the probability that a given
retrieval satisfies both the observed averaged spectra
and falls within a range of plausible a priori values
specified by assumptions on the value and its
dispersion. The weight of the a priori assumption in
the retrieval is dictated by its a priori uncertainty,
which is set equal to the integrated a priori water
vapor profile.
Water vapor is inferred from the set of combined
NIR and TIR spectra, by a simultaneous inversion
from both spectral domains. In practice, the
algorithm adjusts the water vapor abundance along
the vertical profile at nine altitude points separated

by 2.5 km from ground to 10 km, and by 5 km from
10 to 30 km. All points are correlated with a
Gaussian kernel, such that the points are less
strongly correlated when the distance between them
is increasing. Some example spectra are shown in the
top row of Figure 1, where the selected NIR and TIR
spectral intervals include strong diagnostic features
of water vapor. The co-located observations shown
here are from early summer of MY 27 at high
latitudes. The corresponding vertical profile obtained
from the synergistic retrieval is shown in the bottom
left, and is compared to the MCD a priori profile. In
the bottom right panel, the synergy is compared to
the single spectral domain retrievals by computing
ratios of a posteriori-to-a priori (post-to-prior for
short) error profiles. MCD a priori profiles are used
for all retrieval techniques (synergy, PFS-only,
SPICAM-only). The post-to-prior profile indicates
the amount of added information at each altitude,
and shows that the synergy is more sensitive to the
lower atmosphere than both PFS and SPICAM.

Results:
After imposing a series of selection criteria to
ensure retrievals with satisfactory quality and
sufficient vertical information to justify further
analysis, a global composite climatology is
assembled, as shown in Figure 2.

Figure 1: Example spectra from SPICAM and PFS, with the corresponding retrieved
vertical profile and the a priori profile, along with the a posteriori-to-a priori ratios for the
synergy and the single domain retrievals.

Figure 2: Seasonal and geographical variations in normalized total column abundances of water vapor. Panel A:
Seasonal distribution of all qualified retrievals averaged in bins of 2°x2° Ls and latitude. Panels B-E: Geographical
distribution of qualified retrievals for each seasonal interval, averaged in bins of 2°x4° latitude and longitude.

Figure 3: Composite maps of normalized water vapor column-integrated abundance and partitioning index. For the
partitioning index, values higher than 0.5 (in green) indicate that more than 50% of the water vapor column is
confined below 5 km, while lower values (in brown) indicate that water vapor is more evenly dispersed with altitude.
Data from all available years fulfilling all requirements have been averaged in bins of 2° latitude, 2° Ls and 4°
longitude.

Even with a DOF higher than with regular singledomain retrievals, the synergy still cannot resolve a
vertical profile with several completely independent
points. Because of this, we have defined a parameter
called the Partitioning index (PI), a dimensionless
value defined as the ratio of the column integrated
water content from the surface up to 5 km, divided
by the full column abundance. The PI is thus a
measure of the near-surface water concentration,
ranging from 0 (no water in the first 5 km) to 1 (all
water contained below 5 km).
Figure 3 focuses on the Northern hemisphere
spring and summer season, and shows the average
trends in the total column abundance and PI at
latitudes between 45°S and the North Pole, covering
the polar ice sublimation season.
References:
[1] T. Fouchet et al., “Martian water vapor: Mars
Express PFS/LW observations,” Icarus, vol.
190, no. 1, pp. 32–49, Sep. 2007, doi:
10.1016/j.icarus.2007.03.003.
[2] A. A. Pankine and L. K. Tamppari, “Constraints
on water vapor vertical distribution at the
Phoenix landing site during summer from
MGS TES day and night observations,”
Icarus, vol. 252, pp. 107–120, May 2015, doi:
10.1016/j.icarus.2015.01.008.
[3] S. Aoki et al., “Water Vapor Vertical Profiles
on Mars in Dust Storms Observed by
TGO/NOMAD,” J. Geophys. Res. Planets,
vol. 124, no. 12, pp. 3482–3497, Dec. 2019,
doi: 10.1029/2019JE006109.
[4] A. Fedorova et al., “Stormy water on Mars: The
distribution and saturation of atmospheric
water during the dusty season,” Science, vol.
367, no. 6475, pp. 297–300, Jan. 2020, doi:
10.1126/science.aay9522.
[5] F. Montmessin, F. Forget, P. Rannou, M.
Cabane, and R. M. Haberle, “Origin and role
of water ice clouds in the Martian water cycle
as inferred from a general circulation model,”
J. Geophys. Res., vol. 109, no. E10, p.
E10004, 2004, doi: 10.1029/2004JE002284.
[6] M. I. Richardson, “Investigation of the nature
and stability of the Martian seasonal water
cycle with a general circulation model,” J.
Geophys. Res., vol. 107, no. E5, p. 5031, 2002,
doi: 10.1029/2001JE001536.
[7] C. D. Rodgers, Inverse Methods for
Atmospheric Sounding: Theory and Practice,
vol. 2. WORLD SCIENTIFIC, 2000. doi:
10.1142/3171.
[8] F. Forget et al., “Improved general circulation
models of the Martian atmosphere from the
surface to above 80 km,” J. Geophys. Res.,
vol. 104, no. E10, pp. 24155–24175, Oct.
1999, doi: 10.1029/1999JE001025.
[9] E. Millour, F. Forget, A. Spiga, M. Vals, V.
Zakharov, and L. Montabone, “The Mars

climate database (version 5.3),” Paper presented at the Mars Science Workshop “From
Mars Express to ExoMars”, held 27-28
February 2018 at ESAC, Spain, id.68., Feb.
2018.
[10] L. S. Rothman et al., “The HITRAN2012
molecular spectroscopic database,” Journal of
Quantitative Spectroscopy and Radiative
Transfer, vol. 130, pp. 4–50, Nov. 2013, doi:
10.1016/j.jqsrt.2013.07.002.