ASSESSING
BRIGHTNESS
TEMPERATURE
SENSITIVITY
TO
AEROSOLS IN THE MARTIAN ATMOSPHERE USING THE ENSEMBLE
MARS ATMOSPHERE REANALYSIS SYSTEM (EMARS)
R. P. McMichael, Department of Meteorology and Atmospheric Science, The Pennsylvania State University,
University Park, PA, USA (rpm5826@psu.edu), S. J. Greybush, Department of Meteorology and Atmospheric
Science, The Pennsylvania State University, University Park, PA, USA, R. J. Wilson, Space Science and Astrobiology Division, NASA Ames Research Center, Moffett Field, CA, USA.
Introduction:
The Ensemble Mars Atmosphere Reanalysis System (EMARS; Greybush et al., 2019a) assimilates retrievals from the Thermal Emission Spectrometer
(TES; Smith et al., 2001) and Mars Climate Sounder
(MCS; Kleinböhl et al., 2009) instruments into the
Geophysical Fluid Dynamics Laboratory Mars Global
Climate Model (GFDL/NASA MGCM) using the Local Ensemble Transform Kalman Filter (LETKF;
Hunt et al., 2007). One of the primary science objectives of EMARS is to better model and understand
global dust storms, in particular characterizing which
atmospheric states lead to their initiation and how
these storms evolve. Of particular importance to this
objective is determining what modeling and assimilation techniques and observing system design is the
most helpful for constraining aerosols, estimating
dust lifting, and improving dust storm forecasting.
Hinson (2006) suggests that dust storm genesis is
linked to baroclinic wave activity in the northern midlatitudes. For EMARS to investigate dust storm genesis, conditions in the lower atmosphere (pressures exceeding 350 Pa, the altitudes associated with peak
baroclinic wave activity) need to be well-characterized. Unfortunately, at the times of year, latitudes, and
altitudes associated with baroclinic wave activity, the
success rate of MCS retrievals falls off precipitously
(Hinson and Wilson, 2021). When reanalysis uses
these MCS temperature retrievals, the 16-member ensemble does not always converge to a unique synoptic
state (Greybush et al., 2019b; Battallio, 2022), compromising the integrity of model output. This indicates that the information contained in MCS retrievals
needs to be used in conjunction with additional data
to support a comprehensive investigation of near-surface wave activity, and in turn dust storm initiation.
EMARS currently assimilates atmospheric temperature retrievals to update atmospheric temperatures,
winds, and surface pressures, and updates dust via the
Montabone et al. (2015) dust scenarios, which incorporate MCS dust retrievals. To meet the EMARS science objectives, we turn to MCS brightness temperature observations, which are likely impacted by surface temperatures, lower atmospheric temperatures,
and aerosols (Hinson and Wilson, 2021). Therefore,
the next version of EMARS plans to use brightness
temperature observations to update the atmospheric
temperature and dust fields. This work picks up where
Wilson et al. (2011) left off. Before full assimilation

is implemented, we need to quantify the correlation
between brightness temperatures and model fields
(e.g., temperature, dust, water ice, surface properties).
To do this, experiments have been performed with
model output 23-μm and 32-μm brightness temperatures (T23 and T32), corresponding to the TES and B1
MCS observation channels sensitive to dust (Christensen et al., 2001; McCleese et al., 2007). The primary question these experiments are hoping to answer
is whether the transient eddies observed in brightness
temperatures (Hinson and Wilson, 2021) are the result
of surface temperature fluctuations, near-surface air
temperature fluctuations, or dust being carried by
transient eddies.
Multiple Linear Regression Model:
Experimental Setup: A multiple linear regression
model was built to recreate T23 time series from various physical fields within EMARS. The primary goal
of this portion of the project was not to create a datadriven forward model that accurately predicts brightness temperatures but rather to understand what
model variables the brightness temperatures are most
sensitive to at this wavelength. The version of the
model shown here utilized surface temperature, column dust opacity, column water ice, and surface carbon dioxide and water ices as predictors, such that T23
was the predictand. The statistical model was trained
using predictor and predictand data from EMARS reanalysis output between MY 28 LS 111º and MY 29
LS 125º. This model was then verified against
EMARS reanalysis predictand data spanning MY 2933. The model was tested at tropical (2.57º S) and
midlatitude (43.7º N) locations, all hours of the day,
and four times of year – 30º areocentric longitude (Ls)
chunks starting at the solstices and equinoxes. During
model training, regression coefficients constraining
the relationships between T23 and the physical predictors were calculated.
Results: The relationship between T23 and column
dust opacity is illustrated in Figure 1. During the day,
elevated dust levels correspond to decreased T23,
whereas at night, elevated dust levels correspond to
increased T23. This suggests that airborne dust is absorbing and re-emitting upwelling longwave radiation
at night and reflecting incident shortwave radiation

(a)
(a)

(b)
(b)

Figure 1: Diurnal evolution of column dust opacity regression coefficient at (a) 2.5714 ºS and (b) 43.7143
ºN for the northern vernal equinox (dashed blue), summer solstice (red), autumnal equinox (green), and
winter solstice (dashed cyan) seasons, where each season is represented by a 30º areocentric longitude period starting at the respective date. Asterisks denote seasons with higher dust opacity (LS 180-210º, 270300º). Positive values indicate column dust opacity is positively correlated with T23, where negative values
indicate negative correlation.
during the day. The impact of this greenhouse effect
varies by latitude and time of year, for instance becoming significantly weaker during the northern midlatitude winter. There is also a significant positive correlation between surface temperature and brightness
temperature (not pictured), expected since the brightness temperature mirrors the surface temperature in
clear-sky conditions. Since T23 is sensitive to both
dust and surface emission, it stands to reason that the
height at which dust is present (and corresponding atmospheric temperature) could also have a significant
impact on the observed brightness temperature. Constraining brightness temperature behavior by column
variables alone is a good starting point, but a comprehensive picture requires altitude-dependent temperature and dust fields as well.
Ensemble Sensitivity Analysis:
Experimental Setup: One of the significant impacts on the diurnal variability of brightness temperature is airborne dust, which serves to reduce the diurnal temperature range (as shown in Figure 1). To further quantify the impact of airborne dust, we wanted
to assess whether airborne dust drives a greenhouse or

anti-greenhouse effect. To examine the impact of
dust, we examined the relative impacts between ensemble members with different dust forcing. The 16
ensemble members in EMARS vary as a function of
both dust and water ice forcing, enabling the model to
account for some of the uncertainties associated with
aerosol particle size distributions and composition
(Greybush et al., 2019a). Using individual ensemble
members enables us to visualize different potential atmospheric scenarios based on varying physical assumptions and improve our best estimate from this
more detailed examination. The dust forcing tuning
parameter in EMARS is a scaling factor applied when
converting the dust mass mixing ratio to opacity,
which is adjusted such that the dust opacity increases
uniformly from 0.7 to 1.3 times the amount of the average ensemble member. This ensemble comparison
currently examines the members with the strongest
and weakest dust radiative forcing but also identical
water ice forcing to isolate the aerosol impacts of dust.
T32 was taken from these members from EMARS free
run data, reflecting the behavior of the MGCM without the influence of assimilated temperatures to better
determine the model relationship between atmospheric dust, surface temperature, and brightness

(b)

(a)

(c)

(d)

Figure 2: The seasonal evolution of the difference between zonally-averaged (a) column dust opacity, (b) T32
standard deviation, (c) T32, and (d) surface temperature between the EMARS ensemble members with highest
and lowest dust radiative forcing. The areas with little to no shading at the highest latitudes correspond to the
polar ice caps, and areas with bright shading near these edges correspond to differences driven by varying locations of the ice cap edge between ensemble members.
temperature. This not only provides a clearer picture
of how brightness temperature is currently impacted
by other model variables, but also provides a framework for how dust and temperature could be updated
by brightness temperature during ensemble data assimilation. From this point, brightness temperatures
were evaluated across a latitude circle at a given time
from MY 30 LS 101.42º to MY 31 LS 105.75º. Zonally-averaged brightness temperature was taken to reflect the mean and used to assess whether dust was
heating or cooling the atmosphere at a given latitude.
The standard deviation of brightness temperature was
taken across the same latitude circle. Since all this
data was taken from the same standard time, and not
local time, these brightness temperatures span the full
diurnal cycle. The standard deviation is used here as a
metric to evaluate changes to magnitude of the diurnal
cycle. In comparing the relative impacts of dust forcing, we compared the dustiest conditions (member 16)
to the clearest conditions (member 1). This includes
comparison of the column dust opacity, as well as the
mean and standard deviation (as defined above) of T32
and surface temperature.
Results: The horizontal and temporal dust distribution within EMARS is derived from the Mars Climate Database version 5 gridded dust scenarios (Montabone et al., 2015). As expected, the dustiest times of
year are shortly before and after the winter solstice.
(a) column
Since this is the time of year with the highest
dust opacities, this is also the time of year where the

relative difference between ensemble members 16
and 1 is most noticeable (Figure 2a). As expected
from the results of the regression model, the times of
year and latitudes associated with the largest departure in dust opacity is aligned with the times of year
and latitudes associated with the largest reduction in
diurnal brightness temperature standard deviation
(Figure 2b). The mean brightness temperature is not
uniformly increased or decreased, but rather varies
with respect to the location of the subsolar point (Figure 2c). One of the physical mechanisms likely driving this T32 behavior is as follows. In the summertime
hemisphere, enhanced dust reflects the ample incident
shortwave radiation, reducing the total energy in the
system. In the wintertime hemisphere, enhanced dust
absorbs and re-emits upwelling longwave radiation,
preventing energy from leaving the system. Additional support for this physical mechanism driving the
model brightness temperature behavior can be in the
seasonal evolution of surface temperature (Figure 2d),
where the wintertime surface temperatures are elevated in dusty conditions. There are significant departures in the surface temperature and T32 behavior in
the summertime hemisphere, indicating the importance of additional variables to T32 behavior.
As indicated earlier, brightness temperatures are
highly dependent on the vertical temperature and dust
profiles. For the vertical dust profile, the model equations for the lowest model level include a source and
sink term for dust that relaxes the model column opacity towards the observations. Otherwise, three dust
tracers are advected by model winds such that the

vertical profile is driven by advection and sedimentation (Greybush et al., 2019a). While the dust forcing
tuning knob neatly increases the column dust opacity
value between ensemble members, dust tracer mass
mixing ratios do not uniformly increase in turn, likely
due to the impact of dust radiative forcing on the
model winds that scatter these tracers. More detailed
analysis relating the vertical profiles of dust and air
temperature to brightness temperature using EMARS
free run data is forthcoming. . In particular, ensemble
sensitivity analysis (especially across the full ensemble of 16 members) can inform how brightness temperature is related to atmospheric temperatures and
dust at particular vertical levels. These ensemble correlations are then used by the data assimilation system
to update the atmospheric state based on these observations. This study paves the way for use of actual
MCS atmospheric brightness temperatures, along
with MCS retrievals, for the next version of EMARS.
Acknowledgments:
We thank Hartzel Gillespie and Noah Polek-Davis for
facilitating access to data from EMARS and consultation regarding its use and interpretation. We are
grateful to Dave Hinson for insights on MCS brightness temperatures, and to the MCS team for their
guidance on the dataset. This research was supported
by the NASA Mars Data Analysis Program Grant
80NSSC20K1054. EMARS data utilized in this paper
are available from the Penn State Data Commons at
https://doi.org/10.18113/d3w375 (Greybush et al.,
2019).
References:
Battalio, J. M., Transient Eddy Kinetic Energetics on
Mars in Three Reanalysis Datasets, J. Atmos.
Sci., 79(2), 361-382, 2022.
Christensen, P. R. et al., Mars Global Surveyor Thermal Emission Spectrometer experiment: Investigation description and surface science results, J.
Geophys. Res., 106(E10), 23823, 2001.
Greybush, S. J., et al., The Ensemble Mars Atmosphere Reanalysis System (EMARS) Version 1.0,
Geosci. Data J., 6(2), 137–150, 2019.
Greybush, S. J., H. E. Gillespie, and R. J. Wilson,
Transient eddies in the TES/MCS Ensemble Mars
Atmosphere Reanalysis System (EMARS), Icarus, 217, 158-181, 2019.
Hinson, D. P., Radio occultation measurements of
transient eddies in the northern hemisphere of
Mars, J. Geophys. Res. Planets, 111(E5), 2006.
Hinson, D. P. and R. J. Wilson, Baroclinic waves in
the northern hemisphere of Mars as observed by
the MRO Mars Climate Sounder and the MGS
Thermal Emission Spectrometer, Icarus, 357,
114-152, 2021.
Hunt, B. R., et al., Efficient data assimilation for Spatiotemporal Chaos: A local ensemble transform

Kalman filter, Phys. D: Nonlinear Phenom.,
230(1-2), 112-126, 2007.
Kleinböhl, A. et al., Mars Climate Sounder limb profile retrieval of atmospheric temperature, pressure, and dust and water ice opacity, J. Geophys.
Res. Planets, 114(E10), 2009.
McCleese, D. J. et al., Mars Climate Sounder: An investigation of thermal and water vapor structure,
dust and condensate distributions in the atmosphere, and energy balance of the polar regions, J.
Geophys. Res., 112(E5), 2007.
Montabone, L., et al., Eight-year climatology of dust
optical depth on Mars, Icarus, 251, 65-95, 2015.
Smith, M. D., et al., One Martian year of atmospheric
observations by the thermal emission spectrometer, Geophys. Res. Lett., 28(22), 4263-4266, 2001.
Wilson, R. J., J. Noble, and S. J. Greybush, The Derivation of Atmospheric Opacity from Surface
Temperature Observations, Mars Atmosphere
Modeling and Observations, 4th, 2011.