Seasonal Variations in Atmospheric Composition as Measured in Gale Crater, Mars

Abstract

The Sample Analysis at Mars (SAM) instrument onboard the Mars Science Laboratory Curiosity rover measures the chemical composition of major atmospheric species (CO2, N2, ⁴⁰Ar, O2, and CO) through a dedicated atmospheric inlet. We report here measurements of volume mixing ratios in Gale Crater using the SAM quadrupole mass spectrometer, obtained over a period of nearly 5 years (3 Mars years) from landing. The observation period spans the northern summer of MY 31 and solar longitude (LS) of 175° through spring of MY 34, LS = 12°. This work expands upon prior reports of the mixing ratios measured by SAM QMS in the first 105 sols of the mission. The SAM QMS atmospheric measurements were taken periodically, with a cumulative coverage of four or five experiments per season on Mars. Major observations include the seasonal cycle of CO2, N2, and Ar, which lags approximately 20–40° of LS behind the pressure cycle driven by CO2 condensation and sublimation from the winter poles. This seasonal cycle indicates that transport occurs on faster timescales than mixing. The mixing ratio of O2 shows significant seasonal and interannual variability, suggesting an unknown atmospheric or surface process at work. The O2 measurements are compared to several parameters, including dust optical depth and trace CH4 measurements by Curiosity. We derive annual mean volume mixing ratios for the atmosphere in Gale Crater: CO2 = 0.951 (±0.003), N2 = 0.0259 (±0.0006), ⁴⁰Ar = 0.0194 (±0.0004), O2 = 1.61 (±0.09) x 10‐3, and CO = 5.8 (±0.8) x 10‐4.

Full text

Seasonal Variations in Atmospheric Composition as
Measured in Gale Crater, Mars
Melissa G. Trainer
1
, Michael H. Wong
2
, Timothy H. McConnochie
3
, Heather B. Franz
1
,
Sushil K. Atreya
2
, Pamela G. Conrad
4
, Franck Lefèvre
5
, Paul R. Mahaffy
1
, Charles A. Malespin
1
,
Heidi L.K. Manning
6
, Javier Martín‐Torres
7,8
, Germán M. Martínez
9,2
, Christopher P. McKay
10
,
Rafael Navarro‐González
11
, Álvaro Vicente‐Retortillo
2
, Christopher R. Webster
12
, and
María‐Paz Zorzano
13,7
1
NASA Goddard Space Flight Center, Greenbelt, MD, USA,
2
Climate and Space Sciences and Engineering, University of
Michigan, Ann Arbor, MI, USA,
3
University of Maryland, College Park, MD, USA,
4
Geophysical Laboratory, Carnegie
Institution of Washington, Washington, DC, USA,
5
LATMOS, CNRS, Sorbonne Université, UVSQ, Paris, France,
6
College
of Arts and Sciences, Misericordia University, Dallas, PA, USA,
7
Department of Computer Science, Electrical and Space
Engineering, Luleå University of Technology, Luleå, Sweden,
8
Instituto Andaluz de Ciencias de la Tierra (CSIC‐UGR),
Granada, Spain,
9
Lunar and Planetary Institute, Universities Space Research Association, Houston, TX, USA,
10
NASA
Ames Research Center, Moffett Field, CA, USA,
11
Instituto de Ciencias Nucleares, Universidad Nacional Autónoma de
México, Ciudad de México, Mexico,
12
NASA Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA,
USA,
13
Centro de Astrobiología (INTA‐CSIC), Torrejón de Ardoz, Madrid, Spain
Abstract The Sample Analysis at Mars (SAM) instrument onboard the Mars Science Laboratory Curiosity
rover measures the chemical composition of majoratmospheric species (CO
2
,N
2
,
40
Ar, O
2
, and CO) through a
dedicated atmospheric inlet. We report here measurements of volume mixing ratios in Gale Crater using
the SAM quadrupole mass spectrometer, obtained over a period of nearly 5 years (3 Mars years)from landing.
The observation period spans the northern summer of MY 31 and solar longitude (L
S
) of 175° through spring of
MY 34, L
S
= 12°. This work expands upon prior reports of the mixing ratios measured by SAM QMS in the
first 105 sols of the mission. The SAM QMS atmospheric measurements were taken periodically, with a
cumulative coverage of four or five experiments per season on Mars. Major observations include the seasonal
cycle of CO
2
,N
2
, and Ar, which lags approximately 20–40° of L
S
behind the pressure cycle driven by CO
2
condensation and sublimation from the winter poles. This seasonal cycle indicates that transport occurs on
faster timescales than mixing. The mixing ratio of O
2
shows significant seasonal and interannual variability,
suggesting an unknown atmospheric or surface process at work. The O
2
measurements are compared to
several parameters, including dust optical depth and trace CH
4
measurements by Curiosity. We derive annual
mean volume mixing ratios for the atmosphere in Gale Crater: CO
2
= 0.951 (±0.003), N
2
= 0.0259 (±0.0006),
40
Ar = 0.0194 (±0.0004), O
2
= 1.61 (±0.09) x 10
‐3
, and CO = 5.8 (±0.8) x 10
‐4
.
Plain Language Summary The atmosphere of Mars is made up of primarily carbon dioxide,
and during the Martian year, the barometric pressure is known to cycle up and down substantially as this
carbon dioxide freezes out and then is rereleased from polar caps. The Mars Science Laboratory Curiosity
rover has now acquired atmospheric composition measurements at the ground over multiple years,
capturing the variations in the major gases over several seasonal cycles for the first time. With the Sample
Analysis at Mars instrument, the annual average composition in Gale Crater was measured as 95.1% carbon
dioxide, 2.59% nitrogen, 1.94% argon, 0.161% oxygen, and 0.058% carbon monoxide. However, the
abundances of some of these gases were observed to vary up to 40% throughout the year due to the seasonal
cycle. Nitrogen and argon follow the pressure changes but with a delay, indicating that transport of the
atmosphere from pole to pole occurs on faster timescales than mixing of the components. Oxygen has been
observed to show significant seasonal and year‐to‐year variability, suggesting an unknown atmospheric or
surface process at work. These data can be used to better understand how the surface and atmosphere
interact as we search for signs of habitability.
1. Introduction
The present‐day 25.2° axial tilt or obliquity of Mars is similar to the Earth's 23.5°, resulting in seasonal
changes as on Earth. However, each season on Mars is nearly twice as long as on Earth due to Mars'
RESEARCH ARTICLE
10.1029/2019JE006175
Key Points:
•First multiyear in situ
measurements of the major
components of the Mars atmosphere
have been obtained by the
MSL/SAM investigation
•Seasonal variation of CO
2
,N
2
,and
Ar reveals differences in
atmospheric transport and mixing
timescales
•Oxygen varies seasonally and
interannually, independently from
Ar and N
2
, on timescales too fast to
be explained by known chemistry
Supporting Information:
•Supporting Information S1
•Table S1
•Table S2
TRAINER ET AL. 3000
Correspondence to:
M. G. Trainer,
melissa.trainer@nasa.gov
Citation:
Trainer, M. G., Wong, M. H.,
McConnochie, T. H., Franz, H. B.,
Atreya, S. K., Conrad, P. G., et al. (2019).
Seasonal Variations in Atmospheric
Composition as Measured in Gale
Crater, Mars. Journal of Geophysical
Research: Planets, 124, 3000–3024.
https://doi.org/10.1029/2019JE006175
Received 17 AUG 2019
Accepted 22 OCT 2019
Accepted article online 12 NOV 2019
Published online 21 NOV 2019
Corrected 2 MAY 2023
This article was corrected on 2 MAY
2023. See the end of the full text for
details.
©2019. The Authors.
This is an open access article under the
terms of the Creative Commons
Attribution‐NonCommercial License,
which permits use, distribution and
reproduction in any medium, provided
the original work is properly cited and
is not used for commercial purposes.

longer orbital period of 687 days compared to the Earth's 365 days, and the more elliptical orbit of Mars
strongly affects the seasonal variation as compared to Earth. The seasonal change in Mars' meteorological
parameters has been monitored by a number of spacecraft, beginning with the Viking landers (VL1 and
VL2) in 1976 and continuing to this day on the Curiosity rover. As a result, a long time record of the surface
pressure and temperature has become available, largely from VL1 and VL2 from MY 12 to MY 15 at their
respective landing sites of 22.4°N and 47.9°N, MER‐A (Spirit) and MER‐B (Opportunity) from MY 26 to
MY 33 at 14.6°S and 1.9°S, and now Curiosity at Gale Crater from MY 31 to present at 4.6°S. Other environ-
mental properties, including aerosol opacity, UV flux, relative humidity, and water vapor content, are also
being measured on the Curiosity rover (Martínez et al., 2017). These parameters are valuable for interpreting
the observations of seasonal, temporal, or sporadic changes in other atmospheric properties such as the
abundances of atmospheric volatiles, which are the focus of this paper.
Prior to Curiosity, little information was available on long‐term trends in the behavior of the atmospheric
constituents. While the GCMS on the Viking landers measured the bulk volume mixing ratios (VMR) of
the main atmospheric constituents CO
2
,N
2
, and
40
Ar (95.3%, 2.7%, and 1.6%, respectively), O
2
and CO
had large uncertainties (Owen, 1992; Owen et al., 1977; Oyama & Berdahl, 1977). No useful information
about their temporal or seasonal change could be derived from those data; however, on the other hand,
the seasonal change in surface pressure measured by Viking was instrumental in understanding the unique
annual cycle of condensation and sublimation of atmospheric carbon dioxide to and from the poles of Mars
(Hess et al., 1980; James et al., 1992). As CO
2
is the principal component of the atmosphere, and the other
two gases N
2
and Ar are not condensable at Martian temperatures and pressures, the observed change in
the surface pressure can be attributed to seasonal change in the atmospheric CO
2
content. Those data also
revealed a time lag between the onset of CO
2
deposition/sublimation and the resulting surface pressure
change, which is related to the dynamics of CO
2
migration to and from the poles.
Seasonal variation of the second most abundant gas on Mars, N
2
, could not be studied in situ or by remote
sensing before Curiosity. However, the next most abundant constituent, radiogenic argon (
40
Ar, referred to
generally as Ar), was measured over several years in situ by the APXS instrument on MER (VanBommel
et al., 2018) and by remote sensing using the gamma subsystem of the gamma ray spectrometer on Mars
Odyssey orbiter (Sprague et al., 2012). The MER data correspond to the equatorial region, where the VMR
of Ar were found to have a relatively small seasonal variation of 10%. The Mars Odyssey data lacked the pre-
cision for such small changes in Ar in the equatorial region but revealed a dramatic change in Ar over the
poles. The Ar mixing ratio was found to increase by a factor of 6 over the southern pole in winter and by
a factor of 3 over the northern pole in winter. As argon is a noncondensable gas, its total atmospheric content
remains unchanged through the Martian seasons. The observed seasonal variation in its mixing ratio is
therefore due to the dynamical processes induced by the deposition and sublimation of the principal atmo-
spheric constituent, CO
2
, at the poles.
Like N
2
and Ar, O
2
and CO are long‐lived constituents on Mars, with lifetimes greater than the year on Mars
(Atreya & Gu, 1995; Krasnopolsky, 1993; Wong et al., 2003). Besides aforementioned Viking measurements,
only a few sporadic ground‐based measurements of O
2
were available pre‐Curiosity. However, an extensive
set of data on CO has been collected between MY 28 and MY 33 by CRISM on MRO (Smith et al., 2017). The
CO VMRs from those orbital observations are found to be generally slightly larger than the values obtained
from the Earth or from Mars Express observations for corresponding regions (e.g., Billebaud et al., 2009;
Encrenaz et al., 2006; Krasnopolsky, 2015).
The SAM results discussed in this paper provide the first simultaneous measurements in the equatorial
region of all key constituents. Including CO
2
,N
2
, Ar, O
2
, and CO, these measurements were made over nearly
6 years using the same mass spectrometer of the SAM instrument suite. Additionally, methane has been mea-
sured over the same period with the tunable laser spectrometer (TLS) of the SAM suite. Mars is found to have
a persistent low background level of methane (CH
4
) with a mean value of 0.41 ± 0.16 ppbv, but it undergoes
an unexpected seasonal variation of a factor of ~3 from 0.24 to 0.65 ppbv. The observed variation is unrelated
to any known environmental factors, which can account for only about ±20% seasonal change in CH
4
. The
magnitude of the seasonal change in the CH
4
background is much greater than in other long‐lived atmo-
spheric volatiles (N
2
, Ar, O
2
, and CO) discussed in this paper. The TLS results on the methane background
and occasional spikes have been published elsewhere (Webster et al., 2015; Webster et al., 2018).
10.1029/2019JE006175
Journal of Geophysical Research: Planets
TRAINER ET AL. 3001

2. Methods
2.1. SAM QMS Measurements
We report here the measurements of atmospheric volume mixing ratios on Mars using the SAM QMS
onboard the Mars Science Laboratory Curiosity rover, taken in Gale Crater (4.5°S, 137°E) over a period of
almost 3 Mars years (>5 Earth years) from landing (Vasavada et al., 2014). Although Gale Crater is located
just south of the equator, in this paper, the seasons will be referenced with respect to the northern
hemisphere. The observation period spans the northern summer of MY 31, solar longitude (L
S
) of 175°,
and through northern spring of MY 34, L
S
= 85°. During the measurement timeframe, Curiosity traversed
a distance of over 16 km in Gale Crater, spanning elevations from –20 to +329 m relative to the landing site
(–4501 m). The SAM QMS atmospheric measurements were taken periodically during this time, interspersed
between solid sample measurements and other rover activities, with a cumulative coverage of four or five
experiments per season (Figure 1).
Atmospheric mixing ratios measured by the SAM QMS were reported for the first 105 sols of the mission
(Mahaffy et al., 2013). These were recently updated to account for newly developed calibration factors
following in situ calibration experiments using SAM's onboard calibration cell (Franz et al., 2017).
Periodic sampling has continued throughout the mission to explore variations related to the seasonal CO
2
cycle, as indicated by the annual pressure curves (Figure 1). The SAM atmospheric measurements taken
through MY 34 L
S
85° overlap seasonally with the first reported measurements in MY 31, testing whether
there is a repeatable annual cycle in the atmospheric composition. With the exception of three daytime runs,
atmospheric ingestions were taken near local midnight (Table 1).
Following the acquisition of the atmospheric data set presented in this manuscript, the SAM instrument
completed its first full derivatization experiment on the Ogunquit Beach (OG) dune sample (Malespin
et al., 2018). High signal levels in the wet chemistry experiment on sol 1909 (December, 2017) caused a shift
in the sensitivity of the instrument detectors, requiring a change in the QMS electron multiplier gain setting.
The gain state change restored QMS count rates to their previous levels, but detailed analysis is ongoing to
ensure that deadtime corrections and calibration constants for all species are adjusted. Thus, this report is
restricted in scope to atmospheric measurements prior to sol 1909.
2.1.1. Experimental Details and Calibration
The SAM instrument suite consists of three instruments supported by a gas separation and processing
subsystem and a solid sample manipulation system (Mahaffy et al., 2012). Results presented here were
obtained with the QMS through a dedicated heated (50 °C) atmospheric inlet. The QMS employs
Figure 1. The timing of SAM atmospheric QMS experiments is plotted as the atmospheric pressure at the time of ingestion
(left axis) against the MSL mission sol (bottom axis) and solar longitude (top axis). The majority of the atmospheric ingests
for mixing ratio derivation was conducted during local night (closed symbols), with three daytime experiments (open
symbols). The figure background is shaded by northern season, and the REMS daily pressure maximum and minimum
values are given by the dotted lines. Seasonal trends are tracked through the direct atmospheric runs, with attention paid
to possible diurnal variations. Mars year (MY) and Earth year (EY) are indicated across the top of the figure.
10.1029/2019JE006175
Journal of Geophysical Research: Planets
TRAINER ET AL. 3002

hyperbolic rods, redundant 70‐eV electron beam energy ion sources, and redundant pulse counting
Channeltron detectors. Separate miniaturized turbomolecular pumps (compression ratio ~5x10
8
) evacuate
the small QMS sensor volume and the much larger inlet manifold volume prior to Martian atmosphere
ingestion. Experiments include two background scans: One with the QMS sensor volume isolated from the
manifold and one with the QMS open to the evacuated manifold. The QMS is operated in a dynamic
sampling mode with continuous pumping by one of the turbomolecular pumps. An atmospheric sample is
acquired by opening a valve on the sample inlet tube for ~30 sec to introduce gas to a portion of the
manifold. A small fraction of this gas is then leaked into the QMS and scanned over a specified mass range
(1.5–149.9 Da) for several minutes. This process is repeated for most runs to provide two ingestions per
experiment (Table 1). Detailed descriptions of the experimental design and flight instrument calibration
for the atmospheric investigation can be found in Franz et al. (2014).
Direct atmospheric QMS data were acquired in both a fractional scan mode with 0.1‐Da step size and a
unit scan mode with 1.0‐Da step size. A typical fractional scan spectrum acquired during an
atmospheric experiment is shown in Figure 2. Fractional scan mode reduces uncertainties related to
the tuning and shape of spectral peaks, while unit scan mode reduces uncertainties due to time varia-
tion in the signal (as both background and sample are pumped out). Repeated experiments have shown
that sample and background signals are well characterized by exponential functions of time, so best
results are determined through processing of data acquired in fractional scan mode (Franz et al.,
2014; Franz et al., 2015). The method involves correcting mass spectra for detecting deadtime effects
at high count rates and integrating peak areas (±0.4 Da) for the m/z values of the major atmospheric
components, CO
2
, Ar, N
2
,O
2,
and CO.
Table 1
SAM Atmospheric Ingest Conditions with Test Identification Numbers (TIDs).
Mars year Test ID (TID) MSL sol L
s
(°) Ingest start LMST REMS air temp (°C) REMS ground temp (°C) REMS pressure (mbar)
31 25012 45.95 175.59 22:43 –59 ±3 –71 ±2 7.71 ±0.07
25027 77.88 194.34 21:07 –54 ±1 –65 ±2 7.94 ±0.03
77.95 194.38 22:42 –57 ±1 –69 ±3 8.06 ±0.03
25084
a
278.54 320.77 13:02 –22 ±1 –8 ±2 8.36 ±0.04
278.61 320.81 14:36 –16 ±2 –7 ±1 8.15 ±0.04
25088 284.96 324.48 22:55 –49 ±1 –58 ±2 8.61 ±0.04
285.02 324.51 00:30 –58 ±1 –62 ±2 8.68 ±0.04
25095 292.10 328.56 2:25 –64 ±1 –66 ±2 8.75 ±0.04
292.17 328.60 4:00 –66 ±1 –69 ±2 8.79 ±0.04
25106
a
321.67 344.92 16:06 –18 ±0 –8 ±1 8.02 ±0.05
321.74 344.95 17:41 –24 ±0 –24 ±1 7.96 ±0.05
32 25150 434.88 40.89 21:07 –65 ±1 –67 ±2 8.90 ±0.04
434.95 40.92 22:41 –70 ±7 –71 ±3 8.97 ±0.06
25172 538.94 87.87 22:30 –74 ±7 –74 ±5 8.70 ±0.06
539.00 87.90 0:05 –74 ±1 –75 ±3 8.81 ±0.04
25195 638.05 134.67 1:05 –73 ±1 –75 ±3 7.70 ±0.04
638.11 134.70 2:39 –78 ±1 –80 ±3 7.71 ±0.04
25217 753.93 198.85 22:24 –57 ±4 –65 ±3 8.12 ±0.06
754.00 198.89 23:58 –62 ±1 –67 ±2 8.15 ±0.04
25232 830.88 247.83 21:05 –47 ±1 –53 ±2 9.27 ±0.04
830.94 247.87 22:40 –51 ±1 –57 ±3 9.40 ±0.04
33 25301 1145.96 60.21 22:57 –68 ±3 –71 ±3 9.09 ±0.04
1146.02 60.24 0:32 –68 ±1 –75 ±3 9.24 ±0.04
25337 1251.98 108.28 23:37 –75 ±1 –79 ±3 8.19 ±0.05
1252.05 108.32 1:12 –79 ±1 –81 ±3 8.25 ±0.05
25343
a
1319.68 141.28 16:18 –26 ±1 –25 ±2 7.05 ±0.05
25346 1357.07 161.02 01:46 –69 ±5 –72 ±3 7.57 ±0.05
25372 1457.07 220.39 01:35 –60 ±1 –64 ±2 8.80 ±0.05
25395 1600.07 311.55 01:46 –62 ±1 –69 ±3 8.75 ±0.05
34 25409 1711.06 11.86 01:26 –67 ±1 –73 ±2 8.65 ±0.05
25440 1869.07 84.70 01:42 –73 ±2 –82 ±5 8.68 ±0.07
a
Daytime ingest.
10.1029/2019JE006175
Journal of Geophysical Research: Planets
TRAINER ET AL. 3003

Data corrections are needed to account for isobaric interference (multiple species contributing to count rates
at a single m/z value) and for pressure‐related variation in CO
2
‐splitting fractions (ratios between yields of
products such as CO
2+
,CO
2++
, and O
2+
upon ionization of atmospheric CO
2
molecules). As described in
Franz et al. (2015), calibration experiments on the SAM flight instrument and laboratory QMS test bed have
demonstrated an increase with pressure in the ratio of doubly to singly ionized CO
2
. The high abundance of
CO
2
in atmospheric samples leads to saturation at m/z 44 (Figure 2), so we rely on the CO
2++
signal at m/z
22 as the reference marker for mixing ratio measurements. We adjust for the pressure dependence of the CO
2
++
‐/CO
2+
‐splitting fraction using empirical corrections from Franz et al. (2015):
m22corr ¼F22 ×m22obs;(1)
where the correction factor is a linear function of the uncorrected count ratio at m/z 22:
F22 ¼a×m22obs þb;(2)
with coefficients a=–2.321 (±0.1094) x 10
‐7
cps
‐1
and b= 1.000 ± 0.003. To isolate the signal at m/z 32
due to atmospheric O
2
(m32
corr
) from the observed signal at m/z 32 (m32
obs
), we apply an assumption that
the O
2+
/CO
2+
splitting fraction is pressure invariant. For the atmospheric data in which the m44 peak is
saturated, the correction is applied as follows:
m32corr ¼m32obs –c×m44=m22ðÞ
CO2 ×m22corr;(3)
where c= 4.558 (±0.07104) x 10
‐4
and (m44/m22)
CO2
= 145.88 ± 1.17 as in Tables 1 and 2, respectively, in
Franz et al. (2015).
Overall sources of error in the mixing ratio calculation include (i) measurement noise (detector noise follow-
ing Poisson counting statistics), (ii) errors in the data corrections (detector deadtime, background
subtractions, and isobaric interferences), and (iii) uncertainties in the calibration constants. Some of these
errors can be reduced by averaging data points within individual experiments and by averaging multiple
experiments (i.e., measurement noise and background subtractions), while other sources are systematic
and are not reduced by averaging multiple experiments together (i.e., calibration constants and corrections
for deadtime and isobaric interference). Discussions of these sources of error and their estimation are
included in Franz et al. (2014) and in the supplemental material of Wong, Atreya, et al. (2013). For the
Figure 2. A typical atmospheric mass spectrum taken in fractional scan mode with the residual background signal
subtracted. TID 25217 is shown here; see Table 1 for details. The error bars show the standard deviation of the average
counts per second (cps) from multiple scans over a ~7‐minute period. The gray bars indicate regions of the spectrum
which are not used for VMR calculations, either due to decreasing residual water signal or high counts from the primary
CO
2
ion (m/z 44) that saturate the detector. Integrated peak areas used for quantitative determination of the volume
mixing ratios are determined from a +/‐0.4 Da window around each nominal m/z.
10.1029/2019JE006175
Journal of Geophysical Research: Planets
TRAINER ET AL. 3004

major atmospheric gases with multiple ion fragments measured by the
QMS, the m/z values used for calculation of mixing ratios were selected
to have large enough count rates to minimize detector noise but low
enough count rates to minimize saturation and deadtime effects.
Although contributions of residual gas molecules to the background are
minimal for most species, the background signal for the primary peak
for O
2
(m/z 32) is high, thus a large proportion of the total signal. This is
due to O
2
permeation from the high conductance valve seat after opening,
and the background signal decreases exponentially once the QMS and
manifold are actively evacuated. This background correction is therefore
challenging and is a major source of uncertainty in the O
2
mixing ratios
measured by the SAM QMS (see Figure S3 of the supporting information).
Figure 3 shows a typical experiment timeline with backgrounds and
atmospheric samples for the primary atmospheric species of interest. The
approach used for the background correction is to fit an exponential signal
as a function of time to the data within the background interval (specifi-
cally, the background scan with the QMS exposed to the evacuated mani-
fold) and extrapolating to the sample interval to determine the
time‐dependent background signal. This correction is a small percentage
of the total signal for four of the species under study and thus introduces
a minor uncertainty to the final mixing ratio. For the O
2
peak at m/z 32,
the correction isa large proportion ofthe signal (≥50%) and varies between
different experiments and different ingested samples within
each experiment. To accurately capture the inherent uncertainty in this
correction in the O
2
mixing ratio, the 1‐σconfidence intervals for the expo-
nential fitatt
a
are included as partof the error estimation. Our uncertainty
estimates include variations in the functional formof the background fit, in
specific cases where the form of the fitsignificantly influenced the results.
As seen in Figure 3, the magnitude of the QMS signal for the atmospheric
sample is also a time variable, as the sample trapped in the manifold is
gradually pumped through the QMS. Ratios for the major ions of each
gas are taken at each time point (i.e., from each mass spectrum) and then
averaged to remove the time variability from the mixing ratio determina-
tions. Resulting ratios are constant across a sampling interval. Mixing
ratios are then computed at each time point and averaged for each inges-
tion. For experiments with two atmospheric ingests, the resulting mixing ratios from each ingestion are com-
bined into a weighted mean for the sol, reducing systematic errors due to the background corrections. The
values reported in Table S1 are these weighted means.
Finally, for CO
2
, the calculated uncertainties in the reported mixing ratios in this manuscript differ from those
in the previous publications (Franz et al., 2014; Franz et al., 2015; Franz et al., 2017; Mahaffy et al., 2013). For
this work, we are mostinterested in identifying seasonal trends and therefore relative behavior of the measured
mixing ratios throughout the mission. For CO
2
in particular, the errors previously reported are largely intro-
duced by the uncertainty in the calibration constant derived during the prelaunch calibration. This constant
introduces an uncertainty on the CO
2
VMR on the order of 3%. This uncertainty would apply to the absolute
amount of CO
2
(such as a calculated partial pressure or number density) but is not optimal for characterizing
the VMR. The uncertainty on the relative abundance of CO
2
is best modeled by assuming that the total atmo-
spheric composition must be equal to 1 and thus the uncertainty on the CO
2
is equal to the total uncertainties
on the trace gases (Ar, N
2
,O
2
, and CO, added in quadrature) that comprise the balance of the atmosphere.
2.2. Supporting Curiosity Measurements
2.2.1. Rover Environmental Monitoring Station
Throughout the mission, meteorological conditions along the traverse in Gale Crater have been measured by
the Rover Environmental Monitoring Station (REMS) sensor suite (Gómez‐Elvira et al., 2012). The REMS
Figure 3. Integrated peak areas (cps * 0.8 Da‐width peak) for the preferred
ions of CO
2
(m/z 22), Ar (m/z 40), N
2
(m/z 14), O
2
(m/z 32), and CO (m/z 12,
with contributions from CO
2
) are plotted against time for a typical SAM
atmospheric experiment. TID 25217 is shown here; see Table 1 for details.
The background measurement (no atmospheric gas ingested into the QMS)
is accomplished in the experiment time between 6,278 and 7,468 s. Mars
atmosphere is then ingested and measured by the mass spectrometer
between 8,211 and 10,000 s. To perform the background correction, the
exponential decay in the residual gas background signal shown in the
bracketed region is fit and extrapolated to the time of the first fractional
atmospheric scan at time t
a
. The fit value at t
a
is then used to estimate the
contribution of the background to the signal during the sample ingest. As
can be seen in this example, only the m/z 32 background is >10% of the
atmospheric signal. Confidence bands show the fit uncertainties used to
estimate the error in the background correction.
10.1029/2019JE006175
Journal of Geophysical Research: Planets
TRAINER ET AL. 3005

suite performs measurements of atmospheric pressure, ground and atmospheric temperatures, atmospheric
relative humidity, UV radiation fluxes, and horizontal wind speeds (e.g., Martínez et al., 2017). To provide
context for the SAM VMR measurements and the seasonal behavior of the major gases, we focus
primarily on the analysis of REMS pressure (P) and atmospheric temperature (T) measurements.
Comparisons to other environmental conditions, including those measured by REMS, are discussed below in
section 4.2, with more detail in the supporting information Figure S7.
The pressure and temperature conditions at the time of sample ingestion for each QMS experiment are
provided in Table 1. Conditions were calculated as described in Wong, LeFavor, et al. (2013). Uncertainties
include estimated extrapolation uncertainties at times when simultaneous REMS measurements are
not available, variation over the duration of the ingest event, and instrumental uncertainties. Updated
instrumental uncertainty values correspond to the 9th PDS data release. Ingest start times are rounded to
the nearest minute, in local mean solar time (LMST). As indicated in the table, the majority of the atmospheric
samples was acquired in near midnight. Three experiments were conducted in the midafternoon to late
afternoon, two of these in close proximity to nighttime experiments (within ten sols). Figure 4 shows the daily
temperature and pressure curves for two of the sols in which the QMS sampled, 292 and 278, acquired during
northern winter/southern summer. The vertical gray bars indicate the times of the QMS sample ingestions,
showing a typical nighttime experiment and one of the daytime experiment. The search for possible diurnal
trends was of interest in part because of the large diurnal pressure variations observed in Gale Crater, driven
by a combination of thermal tides and topographical effects. Haberle et al. (2014) discussed this in their
interpretation of the pressure cycles over the first 100 sols on Mars, concluding that in addition to the global
thermal tides in the atmosphere from solar heating, the diurnal cycle of upslope/downslope flows driving
crater circulation likely have a significant effect on the pressure amplitudes, with implications for the mixing
of air in the bottom of the crater with air on the surrounding plateau [see also Rafkin et al., 2014; Tyler &
Barnes, 2013]. The planetary boundary layer (PBL) in Gale Crater may be particularly suppressed in compar-
ison with other locations on Mars, with significant impact on the observations of trace gases made by
MSL/SAM (Moores et al., 2019; Newman et al., 2017; Rafkin et al., 2016).
2.2.2. ChemCam Passive Sky Observations
Although the MSL ChemCam spectrometer was designed primarily for laser‐induced breakdown
spectroscopy (LIBS) of Martian surface materials (Maurice et al., 2012; Wiens et al., 2012), it can also operate
in “passive”mode to observe solar radiation scattered by the surface and atmosphere. The ChemCam passive
mode was initially used only for reflectance spectroscopy of surface materials (Johnson et al., 2015).
However, routine ChemCam passive sky spectroscopy started on sol 230 and McConnochie et al. (2017a)
has used these observations to derive aerosol properties and water vapor column abundances using the
instrument's visible and near‐infrared (VNIR) spectral band. ChemCam also observes O
2
absorption near
Figure 4. The QMS sample ingestion times are shown in the context of the daily pressure and temperature cycles for
(a) a typical nighttime ingestion, sol 292, and (b) one of the daytime ingestions, sol 278. The temperature at the
SAM QMS inlet (at approximately 1 m) is between the values measured for the ground (0 m) and by the atmospheric
sensor on the rover mast (1.6 m).
10.1029/2019JE006175
Journal of Geophysical Research: Planets
TRAINER ET AL. 3006

762 nm (McConnochie et al., 2017a), and preliminary results (McConnochie et al., 2017b) indicate that this
can be used to derive quantitative O
2
column abundances, pending further work to better characterize
measurement uncertainties. Because the ChemCam passive sky data are taken at a higher frequency
and lower precision than the SAM QMS mixing ratios, they may be able to provide a complementary
measurement to track the seasonal behavior of molecular oxygen in and above Gale Crater. Future
reports from the ChemCam passive sky measurements may prove valuable for the interpretation of the
SAM O
2
mixing ratios.
3. Results: In situ Volume Mixing Ratios
3.1. Carbon Dioxide (CO
2
)
Carbon dioxide is the primary component of the Martian atmosphere, with large seasonal variations in the
global surface density due to the condensation and sublimation of CO
2
in the polar regions during winter
and spring, respectively. This was first observed by the Viking landers through global fluctuations of about
30% in surface pressure (e.g., Tillman et al., 1993). Curiosity pressure measurements to date have shown this
cycle as well (Haberle et al., 2014; Harri et al., 2014). The Curiosity landing and science investigations
commenced shortly after the annual pressure minimum, which corresponds with the formation of the larger
southern polar cap (Figure 1). During the first 200 sols of the mission, the average surface pressure rose
steadily to the annual maximum during northern fall, shortly after perihelion. The pressure then decreased
to a local minimum near MSL sol 470 during the formation of the north polar cap, which begins to sublime
again during the northern spring. The pressure was observed to decrease again as the southern seasonal
polar cap was formed, repeating the annual cycle as observed in Gale Crater. SAM QMS measurements after
MSL sol 680 have been timed to check for interannual repeatability and to achieve reasonable coverage over
the pressure curve.
The SAM measurements are the first comprehensive compositional measurements of the atmosphere taken
at intervals throughout Mars' CO
2
cycle. Figure 5 shows the local VMR values determined through MSL sol
1869 (MY 34, L
S
85°) as a function of solar longitude, with the daily average pressure included for
comparison to the global cycle. Tabulated data are given in Table S1 and are publicly available (Trainer,
2019a, 2019b). Although there are large fluctuations in CO
2
VMR driven by seasonal and diurnal cycles,
CO
2
is so dominant that the volume mixing ratio only varies by 1% about a computed average mixing ratio
of 0.951 ± 0.003.
Figure 5. SAM measurements of the CO
2
volume mixing ratio (symbols, left axis). The color scale is matched to Mars
year, with tones going from lighter to darker as the points move from Mars year 31 to 34. The error bars on individual
points are derived from the uncertainties on the trace gas measurements as described in the text. The daily mean pressure
at the surface pressure (gray dashes, right axis), indicating the general pressure cycle on Mars. The vertical axes are set so
that the minimum and maximum points in each data set are aligned allowing for visual comparison of the trends. An
average CO
2
mixing ratio of 95.1% ± 0.3% is derived from these data.
10.1029/2019JE006175
Journal of Geophysical Research: Planets
TRAINER ET AL. 3007

The small seasonal perturbation (~1%) in the measured VMR is related to,
but does not strictly follow, the surface pressure cycle. The decrease of the
relative amount of CO
2
in the atmosphere shows a lag behind the decrease
of the average surface pressure by roughly 20–40° of L
S
during the north-
ern summer as the pressure approaches the annual minimum. The data
are sparser during the northern winter, but a similar lag is also indicated
as the pressure approaches the seasonal minimum. The hemisphere‐to‐
hemisphere redistribution of mass that occurs during these periods of
cap sublimation and condensation pulls atmospheric components toward
the winter pole. Transport acts very rapidly to maintain pressure
equilibrium, but changes in VMRs lag behind because physical mixing
of air masses is a slower process.
To better characterize the mixing process, we introduce an annual‐mean
VMR (VMR') that differs from the instantaneously measured VMR by a
simple correction factor. The simplest scenario (although inaccurate
because the mixing timescale is not as fast as the transport timescale) is
that composition and total pressure at a specific location on Mars adjust
simultaneously, as CO
2
condenses/sublimes in polar regions, while other
species remain in the gas phase. If this simple scenario held, then compo-
sition and total pressure could be related by the equation:
Patm tðÞ¼∑PiþPCO2 tðÞ¼∑PiþPCO2 avgðÞ
þdPCO2 tðÞ;(4)
where P
atm
(t) is the total atmospheric pressure as a function of time of
year; P
i
is the partial pressures of the long‐lived, noncondensable species;
P
CO2(avg)
is the average annual partial pressure of CO
2
; and dP
CO2
(t) is the variation in the partial pressure of
CO
2
as a function of the time of year. The total average annual pressure of the atmosphere (P
atm(avg)
) is then
the sum of the partial pressures of the total noncondensable species and P
CO2(avg)
. The value of P
atm(avg)
for
Gale Crater was determined by fitting multiyear daily mean pressure measurements from the REMS data,
binned by integer L
S
values to provide a seasonally averaged pressure. The average pressure computed in this
way is 8.46 mbar. The average CO
2
VMR is used to determine the values for P
CO2(avg)
= 8.05 mbar and ∑P
i
=
0.41 mbar. The P
atm
(t) term was fit to a polynomial and a correction factor (F
P
) developed to allow the mea-
sured VMR values for the noncondensable species to be adjusted to the annual average volume mixing ratio,
VMR':
VMR0¼FP•VMR:(5)
The details of the pressure fit, this calculation, and the derived F
p
values are provided in Figure S2.
Similarly, in the (unrealistic) fast‐mixing case, P
CO2
(t) and P
atm
(t) should be linearly related. This scenario is
plotted as VMR versus P
atm
in Figure 6, where the expected relationship is represented by the black curve,
and measured values are shown with symbols. The expected value for the CO
2
mixing ratio was calculated
by subtracting ∑P
i
= 0.41 mbar from the daily mean pressure at the same L
S
of the measurement and dividing
the P
atm
(t) to get the mixing ratio (equation (4)). The SAM CO
2
VMR data show a deviation from this relation,
in particular the seasons containing the pressure minima. (N. summer and winter) show consistent
enhancements above the fast‐mixing relation. In the periods during which the seasonal caps are subliming,
the measured CO
2
mixing ratio is more likely to match or run below the fast‐mixing relation, as can also be
seen in Figure 5. By comparing the actual data with the fast‐mixing relation in Figure 6, we calculate a
correlation coefficient of R
2
= 0.36. This poor correlation indicates that the fast‐mixing model does not
adequately match the observations.
The seasonal fluctuations of the atmospheric pressure, and the influences of dynamics on the atmospheric
composition at this location on Mars, are more easily identified and understood by studying the behavior
of the trace, noncondensable components, which are more sensitive to the changes in global pressure.
These are discussed in the next section.
Figure 6. The SAM‐measured CO
2
volume mixing ratios (filled symbols)
are plotted against the average pressure for the L
s
on which the data point
was taken, based on approximately 2 years of data taken by the REMS
pressure sensor. The colors indicate the time of year: N. spring (Ls 0°–90°),
N. summer (90°–180°), N. fall (180°–270°), and N. winter (270°–360°). The
expected value for the CO
2
mixing ratio, based on constant composition, is
shown as the solid black line.
10.1029/2019JE006175
Journal of Geophysical Research: Planets
TRAINER ET AL. 3008

3.2. Noncondensable Atmospheric Components
The mixing ratios of argon (
40
Ar), nitrogen (N
2
), and oxygen (O
2
) are shown in Figure 7, and carbon mon-
oxide (CO) is shown in Figure 8. These next four most abundant species are not condensable at Mars surface
and atmospheric temperatures and pressures and thus are not expected to deposit or sublime from the polar
caps as does CO
2
. However, there are seasonal trends in the VMR of these molecules that are a response to
mixing of air masses during the seasonal CO
2
cycle and unexplained additional processes affecting O
2
and CO.
For reference, Table 2 provides instantaneous volume mixing ratios for these gases measured by MSL/SAM
at approximately the same time of the Mars year as the Viking landers. The previous in situ measurements
were not operational for the full Mars seasonal cycle.
3.2.1. Argon and Nitrogen
Argon and nitrogen serve as excellent tracers of global transport, since they are chemically inert in addition
to remain solely in the gas phase. Figure 7b shows that, as expected, these two species track each other
consistently through the Mars year, with good year‐to‐year repeatability. Instantaneous composition is
plotted (i.e., VMR as opposed to VMR'). Pressure maxima occur near L
S
60° and 250°, after the seasonal
Figure 7. Seasonal trends in the instantaneous volume mixing ratios of the three most abundant noncondensable gases
in the Mars atmosphere show a general inverse relationship with pressure due to the condensation and sublimation of
CO
2
from the polar deposits. (a) Daily mean atmospheric pressure (gray dashes, left axis) and air temperature (red
dashes, right axis) show the environmental conditions in Gale Crater as influenced by the seasonal cycle on Mars.
Darkest points are the most recent. (b) Nitrogen (triangles), argon (circles), and oxygen (diamonds) are plotted versus.
L
s
. Point shading corresponds with Mars year (light to dark) as indicated in the upper left corner of each plot. Error
bars shown are 1‐σ, computed as described in section 2.1.1. All plotted data are publicly available (Trainer, 2019a,
2019b).
10.1029/2019JE006175
Journal of Geophysical Research: Planets
TRAINER ET AL. 3009

northern and southern polar caps (respectively) have sublimated. Minima in Ar and N
2
VMRs lag behind
these pressure maxima by ~30° L
S
, suggesting a slower mixing timescale (although the timing of the
northern winter minima in Ar and N
2
VMRs are poorly constrained by the SAM QMS data, we assume
the minimum is at L
S
280°). The lag between pressure minima and VMR maxima (~15° L
S
) is much
smaller than the lag between pressure maxima and VMR minima (~30° L
S
), but note that VMR' is better
suited to compare global transport and mixing timescales. The VMR data are provided in Figure 11 with a
detailed discussion in section 4.1.
The four right‐most points in Figure 7b for both Ar and N
2
show the time period in MY 31 during which
daytime and nighttime measurements were taken within close proximity. From left to right, this group of
points represents: day–night–night–day (Table 1). Although there appears to be a change in the mixing ratio
among these points, there does not appear an obvious correlation with time of day.
N
2
and
40
Ar have mean mixing ratios of approximately 2.6% and 1.9%, respectively, with a seasonal variation
of ±10% of these values throughout the year (Table 3). Note that the temporal average reported here is
affected by limited number of samples (Figure 1); we use a strict numerical average without regard to
temporal coverage. The nitrogen mixing ratio has been increased substantially (Franz et al., 2017) from
the originally published value in Mahaffy et al. (2013) and is now consistent with the Viking values within
the uncertainty of those previous measurements. A detailed discussion of the updated calibration constants
for the SAM QMS and the adjusted mixing ratios for the first sols of the SAM QMS measurements is given in
Franz et al. (2017), which included corrections based on a calibration cell experiment on Mars.
Figure 8. The CO measurements for the first 830 sol (shaded by Mars year) appear to follow the general trend of Ar,
particularly in the spring through summer. Data after mission sol 1,000 show substantially elevated signal at m/z 12,
and the derived mixing ratios are consistently high with no apparent seasonal modulation. These data are not reported in
this publication.
Table 2
Viking and SAM Mixing Ratio Values for Similar Seasonal Periods
Viking GCMS
a
MSL SAM Viking GEX
b
MSL SAM
L
s
(°) 100 108 121 –142 134,141
c
CO
2
95.32% 95.2% ± 0.1% 96.2% ± 5% 95.0% ± 0.2%
N
2
2.7% ± 0.5% 2.55% ± 0.06% 2.3% ± 0.3% 2.70% ± 0.11%
40
Ar 1.6% ± 0.3% 1.95% ± 0.03% 1.5% ± 0.3% 2.03% ± 0.06%
O
2
1300 ± 260 ppmv 1940 ± 40 ppmv <1500 ppmv 1950 ± 290 ppmv
d
a
Owen et al., 1977 (±20% uncertainty on values were prefinal calibration estimates, as stated in their text, and O
2
might
be based on earlier ground‐based observations (see discussion in section 4.2 here)).
b
Oyama & Berdahl, 1977.
c
Average of two measurements taken at these L
s
values. Local “instantaneous”mixing ratio is increasing during
this time.
d
Oxygen exhibits large interannual variability in the MSL/SAM data set, reflected in the uncertainty
on this average.
10.1029/2019JE006175
Journal of Geophysical Research: Planets
TRAINER ET AL. 3010

The
40
Ar/
14
N ratio of the Mars atmosphere, in combination with
14
N/
15
N, has been used as a diagnostic tool
for verifying the inclusion of trapped atmosphere in Martian meteorites, as it has a unique signature as
compared to Earth (Becker & Pepin, 1984). This has been discussed previously in the context of the SAM
14
N/
15
N measurements (Wong, Atreya, et al., 2013), but the value has been updated with the new SAM
calibration (Franz et al., 2017). The
40
Ar/
14
N ratio also serves as a useful metric by which to evaluate the
robustness of the mixing ratio measurements of these trace gases over time. The relative abundance of these
two gases should remain constant, since neither is expected to react or condense, and their similar
abundances ensure near‐identical transport during the seasonal cycle on Mars. The
40
Ar/
14
N measured by
SAM throughout the mission is shown in Figure 9 as a function of MSL sol and indeed shows multiyear con-
sistency within measurement uncertainty, with an average value of 0.376 ± 0.008.
3.2.2. Oxygen
The measured mixing ratio of O
2
has varied from approximately 1,300 to 2,200 ppmv during MSL's first 1,900
sols at Gale Crater (Figure 7b). Two unexpected features of the seasonal behavior of O
2
are immediately
apparent when comparing to the other major inert species (Figure 7b). First, O
2
does not follow the same
general pattern as the Ar and N
2
, particularly through the beginning of the year. Second, the O
2
mixing ratio
shows substantial interannual variability. Both of these features are surprising, because the chemical
lifetime of Martian atmospheric O
2
is estimated to be ≥10 Earth years (Krasnopolsky, 2017). Like Ar and
N
2
,O
2
does not condense under Mars atmospheric conditions, so the O
2
/
40
Ar ratio is expected to be
constant, as for the Ar and N
2
. Despite the larger uncertainties on the derived O
2
mixing ratios (section 2.1.1),
two prominent seasonal features are apparent in the O
2
VMR data (Figure 7b) and the O
2
/
40
Ar ratio
(Figure 10): There is a gradual northern spring/summer increase in O
2
, followed by a potentially rapid reset
to a constant level over much of northern summer and fall (L
S
160°–315°).
The spring/summer increase in O
2
can only be characterized in a broad sense, due to the large error bars and
coarse sampling of the time series. If we make the simplest assumption, a linear change in VMR as a function
of L
S
in the L
S
0°–150° period, there is almost a factor of 3 variation in rate of change in MY 32, MY 33, and
MY 34 (from about 1.3 to 3.6 ppm/°L
S
; Figure 7b). The VMR values plotted in the figure include the effects of
both seasonal CO
2
condensation/sublimation and global mixing. By instead considering the O
2
/
40
Ar ratio
(Figure 10), we can eliminate all changes due to condensation/sublimation and global mixing, and it
becomes more reasonable to apply a constant rate of change to all 3 years of O
2
observations. Any remaining
changes in the O
2
/
40
Ar ratio indicate other factors controlling the local mixing ratio of oxygen in Gale Crater
besides the large‐scale global dynamics controlling transport during the seasonal cycle. We will discuss this
result in more detail below in section 4.2. A significant change is clear over the L
S
0°–150° period, with a
consistent rate of 0.012%–0.015% /°L
S
. The exact onset and end of the O
2
increase season are not well
defined, and the data are not finely sampled enough to determine whether the increase is truly linear (or
whether there is variation year to year).
The second O
2
seasonal feature is that O
2
/
40
Ar mixing ratios (Figure 10) are more or less constant and
identical in all Mars years, over the period of L
S
160°–315°. During this part of northern summer and fall,
Table 3
Annual Mean Volume Mixing Ratios for Mars Atmosphere
a
Atmospheric
component
Annual
mean
mixing ratio
Uncertainty
on mean
b
Seasonal
variation
from mean
Approximate
measurement
error
c
CO
2
0.951 ±0.003 1% 2.9%
N
2
0.0259 ±0.0006 10% 3.2%
Ar 0.0194 ±0.0004 9.7% 2.0%
O
2
0.00161 ±9 × 10
‐5
13% 18%
CO
d
0.00058 ±8 × 10
‐5
36% 6.1%
a
Corrected for annual mean pressure (× P/P
0
; using equation (5)), except for CO
2
. P is the annual mean pressure in Gale
Crater, 8.46 mbar.
b
Uncertainty on the annual mean mixing ratio is reported as 2× the standard deviation of the
mean.
c
Typical uncertainty on the individual instantaneous measurements for each atmospheric component.
d
Reflects measurements from the first 830 sol only; see discussion in section 3.2.3. This likely skews the mean value
toward northern winter (36% of measurements; Figure 8).
10.1029/2019JE006175
Journal of Geophysical Research: Planets
TRAINER ET AL. 3011

the O
2
/
40
Ar mixing ratio is identical to the annual mean of 0.083 (within uncertainties). It is important to
note that the sparse temporal sampling of the data is not sufficient to precisely characterize the bounds of
this constant period. However, MY 33 data on L
S
141° and 161° show a rapid decrease of 20%–25% in
O
2
/
40
Ar. This is a remarkably rapid change, potentially giving insight into the processes modifying
atmospheric O
2
abundances, though we note that the rapid decrease was only observed during MY 33.
Finally, interannual variation is clearly apparent from offsets between measurements in the spring/summer
season of increasing O
2
/
40
Ar mixing ratio. Full characterization of the interannual variation is not possible
given the sparse temporal sampling, but the onset of the variable period seems to be sometime later than L
S
270° and sometime earlier than L
S
30°.
We investigated the possibility of instrument effects or other measure-
ment artifacts affecting the retrieval of the O
2
mixing ratio and thus con-
tributing to the apparent variability. As noted in section 2.1.1, the O
2
mixing ratio is computed based on the signal at m/z 32, which has both
a high background signal and a contribution from the fragmentation of
CO
2+
. The backgrounds and corrections have been tracked through the
course of the mission, and we have found no correlation with the
increases in both absolute and relative O
2
mixing ratios. Further, the O
2
measurements were checked against the solid sample analyses to deter-
mine whether there could be contamination from a pyrolysis experiment.
There was no unique correspondence between solid samples with O
2
release and large increases in the atmospheric O
2
mixing ratio or back-
ground measurement (Figure S4).
3.2.3. Carbon Monoxide
Carbon monoxide (CO) has been detected and is quantified as reported in
Franz et al. (2017, 2015). As discussed in those publications and in
Mahaffy et al. (2013), the quantification of CO relies on a marginal detec-
tion above the dominant CO
2
component, which generates CO
+
and other
interfering fragments in the QMS. Even when the CO abundance was
measured at a relatively high value (i.e., near L
S
180°, TID 25012), the sig-
nal at m/z 12 is estimated to be comprised of 15% directly ionized CO and
Figure 10. The O
2
/
40
Ar ratio, plotted against L
s
and shaded by Mars year,
highlights the variability in O
2
that exists outside global seasonal trans-
port. The light blue‐shaded region shows the overlapping mean and median
of the data set, an O
2
/
40
Ar of 0.083 (±5%). Dashed lines show fits to the data
in the L
S
0°–150° range, under a simple linear model that fits a single slope
to all the data, but separate intercepts for each Mars year.
Figure 9. The
40
Ar/
14
N ratio is consistent over the course of the mission within the measurement uncertainties, verifying
the robustness of the mixing ratio measurements during nearly 5 years on Mars. The SAM
40
Ar/
14
N is greater than the
values reported by Viking instruments (GCMS, red line; GEX, blue line) but is within the range of uncertainties, as
indicated by the red‐and blue‐shaded regions. SAM has measured a larger proportion of Ar in the atmosphere than the
previous landed instruments. Small perturbations in this ratio correlated to local air temperature are discussed in
Figure S9.
10.1029/2019JE006175
Journal of Geophysical Research: Planets
TRAINER ET AL. 3012

85% from CO
2
fragmenting into CO
+
. The derived CO mixing ratios for the first part of the mission are given
in Figure 8. It can be seen that these species approximately track the seasonal trend in argon, behaving much
like a passive tracer species through the atmosphere. This is generally what is assumed in photochemical
models, although there has been some observational variability that has challenged this assumption [see
discussion in Krasnopolsky, 2015]. Smith et al. (2017) recently published CRISM column observations of
CO for MY 28 through MY 33, therefore overlapping in the time period of the reported MSL
measurements. The CO derived from CRISM data for the latitude band at 0°–20°S shows a similar trend,
with a minimum at L
S
90° and a subsequent rise to a peak value just before L
S
180°. The absolute value of
the mixing ratio measured by SAM ranges from 400 ppm to 800 ppm during this period, which is slightly
lower than the 700–1000 ppm reported by CRISM for this region on Mars. Other orbital and ground‐based
measurements also report greater CO mixing ratios than the in situ values (Billebaud et al., 2009; Encrenaz
et al., 2006; Hartogh, Błęcka, et al., 2010; Krasnopolsky, 2003, 2015; Sindoni et al., 2011; Smith et al., 2009).
There are several possible explanations for any discrepancy between the surface and the orbital and ground‐
based observations, including the difference in spatial sampling, the effect of any local depletions in Gale
Crater, errors in the SAM CO measurement, and the distinction between a point and column‐
integrated measurement.
Although we have measured the CO in each atmospheric experiment through MSL sol 1,711, measurements
after sol 1,000 show significantly elevated signal at m/z 12 and therefore very high CO mixing ratios. The
VMR for CO obtained in MY 32, near L
S
250° (MSL sol 830, TID 25232, Tables 1 and S1) shows the onset
of this trend, in which the CO begins to diverge from the repeated seasonal trend in Ar (Figure 8). In addition
to the derived CO mixing ratio, more than doubling from MY 32 to MY 33, the elevated measurements have
not decreased or shown any seasonal modulation in MY 33 and MY 34, in contrast to the O
2
measurement.
This behavior is suspect, and at this time, a possible contamination or instrument effect cannot be ruled out.
Especially because we know the m/z 12 signal to be highly sensitive to such effects (Franz et al., 2015), we are
cautiously omitting the questionable observations from this paper. Those CO measurements thus require
further investigation and will be reported at a later time.
4. Discussion
4.1. Seasonal Transport
Argon and nitrogen in Mars atmosphere have extremely long lifetimes (~Gyr) against losses from
photochemistry, sputtering, and escape, and they do not condense under any of the conditions reached
during the current seasonal cycle. Condensation points at Mars ambient pressures are ~63 and 53 K for
Ar and N
2
, respectively. Thus, these gases serve as excellent tracers of the complicated dynamics induced
by the cycling of CO
2
into and out of the polar caps on yearly timescales. Measurements of Ar from orbit
using gamma subsystem (GS) of the gamma ray spectrometer (GRS) on the Mars Odyssey spacecraft first
identified a “freeze distillation”effect, in which Ar (and other noncondensables) are carried to the polar
regions by advection of the bulk atmosphere during autumn and winter when CO
2
freezes out on the polar
cap (Sprague et al., 2007). The GRS measurements provide a column‐averaged mean mixing ratio (mmr) of
Ar relative to the total atmospheric gas and show an enrichment of up to 6 times at the southern winter pole
and 3 times at the northern winter pole (Lian et al., 2012; Sprague et al., 2012). However, the orbital
measurements of Ar at the lower latitudes have too much scatter to detect what could be seasonal variations
of a smaller magnitude (Sprague et al., 2012).
Recently published work from the alpha particle X‐ray spectrometer (APXS) on the Mars Exploration Rover
Opportunity (MER‐B) has reported a relative atmospheric argon density over 6 years of observation from
~2°S, a similar latitude as MSL (VanBommel et al., 2018). The MER‐B measurements, from Mars years
28–33, provide normalized argon mixing ratios that show a seasonal variation of up to 15% from L
S
0°, which
is comparable in magnitude to the observed variation in absolute mixing ratios measured by SAM, with
broadly similar seasonal trends. More detailed analysis, beyond the scope of this paper, will be needed to
assess the statistical and physical significance of the modest differences between SAM and APXS. Note
however that the spatial separation of MER‐B and MSL and the much larger volume of atmosphere sampled
by the APXS detection method could contribute to such differences due to the significance of atmospheric
dynamics in mixing ratio trends.
10.1029/2019JE006175
Journal of Geophysical Research: Planets
TRAINER ET AL. 3013

With SAM QMS measurements, we have the first data set of compositional measurements near the
equator of all the major species, thus painting a more complete picture of the behavior of Mars' atmosphere
throughout the seasonal cycle. The instantaneous mixing ratio data presented in Figure 7 and Table S1 were
subsequently adjusted for the annual global pressure cycle, using the method described in section 3.1. We
used equation (5) to correct the in situ values to produce global mean annual mixing ratios (VMR',
Figure S5). If the atmospheric gases were perfectly globally mixed and transport effects on trace species were
negligible at this low latitude, the amount of Ar and N
2
would show a constant value once the pressure
variation was removed. In other words, the Ar and N
2
VMR' values in Gale Crater would remain constant
with time, if atmospheric transport and mixing timescales were identical. The corrected mixing values
(Figure 11) show that there are clear deviations to such behavior, because the transport (traced by
adjustment of atmospheric pressure due to polar CO
2
condensation/sublimation) is faster than mixing
(traced by adjustment of composition). Global mean values were calculated from the Ar and N
2
VMR'
(0.0194 and 0.0259, respectively), about which the individual mixing ratios are plotted. In this manner,
the two atmospheric components can be plotted on the same axis, highlighting deviations from the average
value as a function of L
S
. The data indicate that the minimum in Ar and N
2
VMR' in N. spring/summer lags
behind the pressure maximum by approximately 78° of L
S
, and similarly the maximum in Ar and N
2
in N.
summer/fall follows the pressure minimum by 67° of L
S
, due to delay in mixing.
The corrected mixing ratios indicate mixing of air masses of distinct composition, moving through the
equator during the seasonal cycle. At the south pole, maximum enrichment of Ar observed at the southern
pole by the GRS (Sprague et al., 2012) occurs near L
S
120°, an effect of “freeze distillation”of trace gases
occurring at the pole during the local winter (Lian et al., 2012). Although pressure at all latitudes adjusts
relatively and rapidly to polar sublimation/condensation, compositional differences can be maintained
longer by the southern polar vortex until it mixes with higher latitude air in southern spring, pushed by
the large amounts of CO
2
subliming off the cap. Repeating cycles of enrichment and depletion of Ar and
N
2
are observed at Gale Crater as the composition mixes between low latitude air masses and seasonally
varying polar air masses. It should be noted that the orbital GRS measurements of Ar at low latitudes (where
both Opportunity and Curiosity sit), and even in north polar regions, are dominated by scatter and the var-
iations described here are not resolvable.
Differences in the timescales for atmospheric transport and mixing are emphasized by plotting global mean
values (VMR'; Figure 11) using the pressure correction described above, since VMR' should not change if
Figure 11. The global annual mean values of Ar and N
2
(Figure S5) were subtracted from the pressure‐corrected
mixing ratios of Ar and N
2
(VMR', equation (5)) at each seasonal point and plotted as a percent deviation from the
average. In this way, the observed seasonal cycle of Ar and N
2
can be compared directly to a flat annual mean that
removes the mixing ratio changes introduced by fluctuating CO
2
pressures. If there were no influence of transport, the
corrected VMR' values would appear on the dashed zero line. The polynomial fit to the daily mean pressure data is
shown as a gray dashed line for comparison to the annual pressure cycle.
10.1029/2019JE006175
Journal of Geophysical Research: Planets
TRAINER ET AL. 3014

transport and mixing timescales are identical. We use pressure‐corrected VMR' data to compute mean
annual mixing ratios (Table 3), but as indicated by the residuals plotted in Figure S5, uncertainties in the
global average are affected more by seasonal variation than by instrumental uncertainties. In Figure 12,
we plot the instantaneous VMR of Ar and N
2
on top of a modeled Ar mixing ratio in Gale Crater as
predicted by the Laboratoire de Météorologie Dynamique (LMD) Mars GCM (MGCM) with coupled
photochemistry (Lefevre et al., 2004). All are normalized to L
S
345°, selected as one of the SAM
measurement points closest to L
S
0° (360°). The MGCM prediction shows generally good agreement with
the measured mixing ratios for Ar and N
2
, with some discrepancy between L
S
90° and180° when a lower
mixing ratio is observed by SAM. The discrepancy between the model and the data may reflect an
inaccurate estimation of the local isolation in Gale Crater that limits mixing between the MSL
measurement sites and the surrounding atmosphere (Pla‐Garcia et al., 2016; Rafkin et al., 2016). Thus, the
SAM data set provides unprecedented ground truth measurements of these tracers to aid in improving the
representation of various processes within the Martian climate system in global circulation models. This
has significant implications for proper understanding of the water cycle, CH
4
abundance and distribution
(Webster et al., 2018), and possibly also the dust cycle on Mars (Lian et al., 2012).
MER‐B used APXS data to measure seasonal trends in Ar mixing ratio at Meridiani Planum. General trends
are very consistent with SAM results at Gale Crater (Figure 7), but the cadence of the SAM data preclude a
definitive observation of the argon “pulse”near L
S
150° observed in the APXS data [see Figure 9 in
VanBommel et al., 2018]. The SAM data are not inconsistent with a pulse; qualitatively, in particular with
the pressure‐corrected data (Figure 11), there appears to be a small increase in Ar around L
S
150° that
deviates from an apparent seasonal curve. Interestingly, L
S
150° also corresponds to the timing of the
transition between seasonal periods of increasing O
2
and constant O
2
in the SAM data set (section 3.2.2).
4.1.1. Nitrogen Cycle
Recent detections of nitrate in Martian sediments have indicated the presence of nitrogen fixation cycles on
Mars at least at one time in its history (Navarro‐Gonzalez et al., 2019; Stern et al., 2015; Stern et al., 2017).
Because nitrogen is an essential element for life, nitrogen fixation is a critical process required to support
habitable environments as we know them. If currently active, a large nitrogen flux into the regolith could
potentially affect the abundance of atmospheric N
2
over seasonal or long‐term timescales. We explored
Figure 12. (a) The instantaneous mixing ratios of N
2
, Ar, and O
2
(Figure 7, Table S1) are normalized to their values at
MY 31, Ls 344.9°, and plotted on top of the MGCM output for relative Ar mixing ratio within Gale Crater. Similar
models have produced seasonal variations in Ar at low latitudes that are on the same scale (±10%) (Lian et al., 2012).
(b) The background seasonal CH
4
abundance reported in Webster et al. (2018) is shown here as the instantaneous mixing
ratio (right axis), normalized to L
S
330° (MY 32), which allows for comparison to the O
2
measurements (left axis, replotted
as in the top panel). Correlations between CH
4
and O
2
are addressed in section 4.2.2.
10.1029/2019JE006175
Journal of Geophysical Research: Planets
TRAINER ET AL. 3015

whether the available SAM atmospheric data set could shed any light on whether there is a currently active
nitrogen cycle. As on Earth, the primary reservoir of nitrogen is in the form of N
2
in the atmosphere. On
Earth, nitrogen fixation proceeds abiotically and biologically, with the biological rate (10
14
gNyr
‐1
)
occurring about 100 times higher than the abiotic rate triggered by lightning (Menge et al., 2013;
Schumann & Huntrieser, 2007; Zehr et al., 2001). Nitrogen‐containing end products of biological processes
are rapidly recycled back into the atmosphere at a rate such that the nitrogen budget on Earth appears to be
in an approximate equilibrium. On Mars, nitrogen fixation could potentially occur both biologically
(Klingler et al., 1989) and abiotically (Navarro‐Gonzalez et al., 1998; Navarro‐Gonzalez et al., 2001; Segura
& Navarro‐Gonzalez, 2005). Currently there exist no data capable of providing evidence of biological activity
on Mars. If reported, CH
4
detections (Formisano et al., 2004; Mumma et al., 2009; Webster et al., 2013;
Webster et al., 2015; Webster et al., 2018) were assumed to be solely a result of active biological activity;
an estimated biomass of ~10
9
gyr
‐1
would be required to maintain a CH
4
concentration of 10 ppbv
(Krasnopolsky et al., 2004). A methanogenic population of this size would be capable of a hypothetical
biological nitrogen fixation rate of ~10
8
gNyr
‐1
(Frigstad et al., 2011; Leigh, 2000). Conversely, photochemical
models predict an abiotic nitrogen fixation rate of ~10
9
gNyr
‐1
in the form of nitrates (Yung et al., 1977).
Therefore, the abiotic rate in the current atmosphere would be expected to exceed the hypothetical biological
rate by at least an order of magnitude. The current annual flow of nitrogen from the atmosphere to the surface
is negligible considering a repository in the atmosphere on the order of 10
18
gN
2
.
Further, the abiotic denitrification of nitrates is a viable mechanism to recycle back surface nitrogen to the
atmosphere, but the rate of this process is currently unknown. Therefore, it is concluded that current
nitrogen cycle on Mars has no measurable impact on the atmospheric N
2
mixing ratio and consequently
implies nitrogen seasonal stability.
4.2. Oxygen
4.2.1. Interannual Variability
The 1300–2200 ppmv abundances of oxygen measured by SAM are generally in the same range as prior
measurements that span different sets of in situ and remote sensing observations of O
2
in the upper
atmosphere and at the surface of Mars. The repeated measurements of SAM provide the most robust
measurements to date, as the previously reported values have been made with substantial uncertainty and
limited frequency, besides being in different regions of the atmosphere and the surface. From five sets of mass
spectral measurements obtained within days of landing of Viking lander 1 (VL1) on 20 July 1976, Owen and
Biemann (1976) reported an O
2
mixing ratio in the 1000–4000 ppmv range. Owen et al. (1977) subsequently
concluded that the oxygen measurements had considerable scatter of a factor of 2, which they attributed to
instrumental causes. It seems likely, as England and Hrubes (2004) suggest, that Owen et al. (1977) based
the 1300 ppmv value for O
2
mixing ratio given in their Table 1 on measurements reported by
Barker (1972) and Carleton and Traub (1972), which were obtained with high‐resolution 762 nm O
2
‐band
spectroscopy from terrestrial observatories (see also Trauger & Lunine, 1983 who report a slightly lower value
from ground‐based spectroscopy: ~1200 ppm when scaled to 6 mbar surface pressure). The Viking GEx
experiment reported an upper limit of 1500 ppmv (Oyama & Berdahl, 1977). England and Hrubes (2004)
calculated a seasonal variation in O
2
between 2,500 and 3,300 ppmv, by inverse scaling of O
2
to atmospheric
pressure measured by the Viking. It is important to note that the reference O
2
value they used in their scaling
calculation (3,000 ppmv) is the value VL1 measured at 125–300 km, in the upper atmosphere above the
homopause (Nier & McElroy, 1977), which is not representative of O
2
at the surface. Mars Express
SPICAM observed 4,000 ppmv O
2
averaged over 90–130 km and six observations (Montmessin et al., 2017;
Sandel et al., 2015). Any seasonal or temporal variations in O
2
at the surface are not expected to propagate
to the upper atmosphere. Recent measurements made through disk‐averaged observations of Mars with
the Herschel Space Observatory's HIFI instrument retrieved a value of 1400 ± 120 ppmv, though they caution
the reader the value may not be vertically uniform (Hartogh, Jarchow, et al., 2010). This measurement was
taken during MY 30, L
S
77°. SAM measurements from surface from similar times of year are slightly higher
but with overlapping uncertainty (Table S1).
The SAM measurements of O
2
in Gale Crater do not show the annual stability or seasonal patterns that
would be predicted based on the known sources and sinks in the atmosphere. As mentioned in
section 3.2.2, based on known sources and sinks, O
2
should show the same seasonal patterns and annual
10.1029/2019JE006175
Journal of Geophysical Research: Planets
TRAINER ET AL. 3016

repeatability as Ar. Given the known chemical cycles, the formation of atmospheric O
2
is controlled
primarily by photochemistry of H
2
O and CO
2
(e.g., Atreya & Gu, 1994):
CO2þhν→CO þO;(R1)
H2Oþhν→HþOH;(R2)
OH þO→HþO2;(R3)
and
OþOþM→O2þM;(R4)
where M is the background gas (CO
2
). The abundance is then controlled by the balance between formation
and loss through photolysis and formation of H
2
O
2
,HO
2
, and NO
2
(Krasnopolsky, 1993).
To quantify the enrichment of the observed O
2
compared to what would be predicted, we account for the
changes in total number density, which are mainly due to pressure changes (Figure 7a) caused by the
global‐scale CO
2
condensation‐sublimation cycle, as well as the expected changes in the mixing ratios of
all noncondensable gases, which are caused by the interaction of the global circulation with that
condensation‐sublimation cycle and are readily visible to us via Ar results. After accounting for both of
these, it can be estimated that approximately 10
14
O
2
molecules cm
‐3
must be added to the atmosphere
sampled by Curiosity in order to explain the observed 1,700 to 2,200 ppmv increase in O
2
between L
S
60°
and 140° in MY 33. In other words, the number density of O
2
molecules at L
S
140° in MY 33 is ~10
14
molecules cm
‐3
larger than it would have been had the ratio of O
2
to Ar remained constant as expected.
For reference, given the 1,700 ppm starting value at L
S
60°, the O
2
mixing ratio would have been ~1830
ppm at L
S
140°, had the O
2
to Ar ratio remained constant over that time span.
Using the O
2
/
40
Ar ratio as an indicator for O
2
variability outside of the known and observed seasonal
dynamics, a simple linear model can be used to fit a single slope to all the data (Figure 10). This model
highlights a consistent seasonal increase in O
2
/
40
Ar ratio of 0.014/100° L
S
during the L
S
0–150° period, with
an interannual variation in the mean O
2
/
40
Ar ratio in this period. For L
S
>150°, O
2
/
40
Ar seems to be more or
less constant, with no significant interannual variation. The low values at L
S
>310° in MY 31 suggest a
possibility of additional interannual variation late in the Mars year, and potentially the onset of the increases
is observed through the spring of the following years, but future observations would be needed to confirm
this possibility.
Within the uncertainties caused by limited sampling and measurement error, this magnitude appears typical
of the unexpected seasonal increase, and so going forward, we will adopt ~400 ppm and ~0
14
molecules cm
‐3
as the amount that needs to be resolved. Assuming that the unexpected O
2
is uniformly mixed in the lower
atmosphere, as seems likely for perturbations of this timescale given current assumptions about the eddy
diffusion coefficient, the 10
14
molecules cm
‐3
become 10
20
molecules cm
‐2
in the atmospheric column
(see, e.g., Krasnopolsky, 2010 who adopts 10
7
cm
2
s
‐1
for the eddy diffusion coefficient, which gives a ~2‐day
characteristic timescale for the bottom scale height of the atmosphere).
Given photochemical schemes above, this 400 ppm of extra O
2
would require a corresponding destruction of
CO
2
and H
2
O molecules in approximately 170 sol. Considering H
2
O alone, ~800 ppm of H
2
O would need to
be destroyed, which is more than five times larger than the maximum abundance of H
2
O measured in and
around Gale Crater by REMS (Martínez et al., 2016) and ChemCam (Fig. 111 in McConnochie et al., 2018).
Thus, it appears unlikely that the needed O
2
could be produced from the available atmospheric water for any
plausible H
2
O photolysis or dissociation mechanism. Furthermore, H
2
O abundance shows an increase
during this time period and no strong correlation with O
2
(Figure S6). Estimates for the production of O from
CO
2
, using CO
2
photolysis rates for the lower atmosphere of Mars (Table 1 in Wong et al., 2003), indicate
that this process is much too slow to generate the observed rise over the short (~1/2 yr) time period. For
completeness, note that photolysis or other dissociation of CO is negligible, and in any case a seasonal
removal of ~800 ppm of CO is clearly ruled out by observations (e.g., Smith et al., 2009).
10.1029/2019JE006175
Journal of Geophysical Research: Planets
TRAINER ET AL. 3017

The primary destruction pathways for O
2
are through direct photolysis in the upper atmosphere and reaction
with photolysis products of H
2
O, HO
2
, and CO
2
deeper (Atreya et al., 2006; Lefèvre & Krasnopolsky, 2017;
Wong et al., 2003). Even factoring in the effects of dust devils and large dust storms (though no major
dust events occurred during the measurement period) (Atreya et al., 2006), the lifetime of O
2
against
photochemical destruction in the Mars atmosphere is expected to be at least 10 years, possibly longer
(Krasnopolsky, 2010; Lefèvre & Krasnopolsky, 2017). Again using the O
2
/
40
Ar ratio as an indicator for O
2
variability (Figure 10), the SAM measurements in MY 33 show a relative decrease of 23% in a period of 39
days (38 sol). (This is a ~500 ppm absolute O
2
decrease from what would be expected from a constant
O
2
/
40
Ar ratio given the starting O2 abundance at L
S
140°.) There is a similar decrease observed from fall
to winter of MY 31, although the measurement frequency is such that the period of change appears longer
(–20% in 201 sol). In particular, the rapid drop in MY 33 corresponds to a lifetime of 150 days, several orders
of magnitude less than the photochemical equilibrium lifetime of ≥10 years. The drop in MY 31 corresponds
to a lifetime of ~1000 days, which is still relatively short.
The lack of a known atmospheric source or sink that could explain the apparent behavior of the O
2
in
Gale Crater suggests the possibility of a temporary surface reservoir. Previously, Herschel WIFI observa-
tions found that the O
2
vertical profile above the surface is not constant with altitude (Hartogh, Jarchow,
et al., 2010) and preliminary analysis of the data shows that a surface flux of O
2
may be required to
explain the observations (Paul Hartogh, personal communication, 2016). A surface sink has been
previously invoked to balance the current redox budget (Zahnle et al., 2008), and the surface is known
to harbor a variety of oxidant species (e.g., Lasne et al., 2016 review). In fact, the Viking Gas Exchange
experiments found that a significant quantity of O
2
was released whenever soil samples were humidified
(Klein, 1978; Oyama & Berdahl, 1977) although these experiments were all done at ~10 °C rather than
Mars ambient temperatures.
Deposition of oxygen could occur in the form of more reactive oxidized species, superoxides, hydrogen
peroxide (H
2
O
2
), ozone (O
3
), or perchlorates, all of which are assumed to have a higher surface reaction
probability (γ) with surface materials than molecular oxygen. It is possible to conceive of an oxygen cycle
with the appropriate seasonal and interannual variability if oxygen were effectively converted to these species,
deposited into the regolith, and then rereleased due to thermal, chemical, or radiation perturbations.
Perchlorates, found to be prevalent in the surface materials in Gale Crater at 0.03–1 wt% level (Sutter et al.,
2017), are very stable. They have also been detected at 0.4–0.6 wt% level in the polar landing site of the
Phoenix Lander (Hecht et al., 2009). To put perchlorates in context, a 1‐cm depth of soil containing 1% by
weight of calcium perchlorate has slightly more than enough oxygen to contribute the apparent ~10
20
molecules cm
‐2
of unexpected column O
2
variation, and O
2
has been shown to be a high yield product of
radiolysis of surface perchlorate salts (Quinn et al., 2013). However these results point to a long‐term
accumulation of O
2
in the Martian soil; the production rate of O
2
from perchlorate radiolysis is insufficient
to produce the observed recurring unexplained signal. More specifically, based on the Quinn et al. (2013)
experimental radiation dose and yield, and on their estimated Martian dose rates and estimated 2‐m cosmic
ray penetration depth, it would take on order of 1 Myr to accumulate enough trapped O
2
for one season worth
of 10
20
molecules cm
‐2
of column O
2
variation. Similarly, the proposed “superoxide”O
2
‐
ions, suggested to
explain the results of the Viking soil reactivity experiments, could form from ultraviolet radiation on surface
minerals and lead to the observed release of O
2
with humidification (Yen et al., 2000), but the reported rate
of superoxide generation is too small to be consistent with the inferred column O
2
signal yet again by
afactorof~10
6
.
The Viking Gas Exchange experiments released up to 770 nanomoles of O
2
from a 1 cm
3
sample over a
period of less than 11 days upon “humidification”at ~10° C (Klein, 1978; Oyama & Berdahl, 1977). If
the same abundance of rapidly releasable O
2
was present across 2 m of depth (i.e., 200 cm
3
), this would
yield the 10
20
molecules cm
‐2
that would explain the atmospheric measurements. This indicates that
sufficient rapidly releasable O
2
is present in the Martian soil, although it is not clear that such a rapid
release of O
2
could have occurred at Mars ambient temperatures. More importantly, this serves to illus-
trate that explaining a one‐time release of O
2
is not the main problem. The primary difficulty is that the
slow rates of accumulation in the processes considered so far cannot explain the seasonal recurrence of
excess O
2
.
10.1029/2019JE006175
Journal of Geophysical Research: Planets
TRAINER ET AL. 3018

Hydrogen peroxide is worth considering as a solution to this problem, because it is less stable than
perchlorates, and is therefore more likely to provide a rapidly exchangeable reservoir of O
2
.H
2
O
2
has been
detected in the atmosphere (Clancy et al., 2004; Encrenaz et al., 2004) and exhibits seasonal and interannual
variability (Encrenaz et al., 2019), and it has been suggested that diffusion of atmospheric H
2
O
2
into the
regolith and/or mineral/water interactions could supply H
2
O
2
in the near subsurface (Bullock et al., 1994;
Lasne et al., 2016).
Current knowledge of H
2
O
2
physics and chemistry in Martian soil is very limited, but based on a coupled
soil‐atmosphere model by Bullock et al. (1994), it appears that the magnitude and thermal sensitivity of
H
2
O
2
soil adsorption is potentially close to the right order of magnitude to supply the unexpected 10
20
molecules cm
‐2
of O
2
. However, this conclusion only follows from assuming a 10
7
‐year chemical lifetime
for adsorbed H
2
O
2
, which is the longest that Bullock et al. (1994) considered plausible, and it depends on
their adopted absorption isotherm, which was based on the Fanale and Cannon (1971) empirically derived
expression for H
2
O since no data for H
2
O
2
were available. Furthermore, once the depth penetration of the
annual temperature wave (e.g., Grott et al., 2007) is considered, the amount of H
2
O
2
potentially cycled in
and out of the soil is estimated at an order of magnitude less than needed here, at 10
19
molecules cm
‐2
.
Further, the timescale for diffusion from meter‐scale depths may be far too long. Finally, note that although
seasonal trends are in fact observed for atmospheric H
2
O
2
, this mechanism would require a rapid conversion
to O
2
immediately at the surface as the observed amount of H
2
O
2
is only on the order of ppb (Encrenaz
et al., 2015).
Another potential mechanism involving H
2
O
2
was proposed by Quinn and Zent (1999). They showed that
H
2
O
2
‐TiO
2
complexes rapidly released O
2
upon humidification at warm (for Mars) temperatures of 10 °C;
a more extensive study of regolith analogs and environmental conditions for this type of H
2
O
2
O
2
conversion
in the regolith is needed.
Finally, any release of O
2
to the atmosphere from surface/subsurface reservoirs of H
2
O
2
and perchlorates
(ClO
4
‐
) would be associated with concomitant flux of hydrogen and chlorine, respectively, which would
impact the chemistry of the atmosphere in ways not seen as well as require their removal from the
atmosphere by processes that are poorly understood.
4.2.2. Correlations With CH
4
and Environmental Parameters
The in situ CH
4
abundances measured by SAM/TLS (Webster et al., 2018) are plotted with the relative
instantaneous O
2
mixing ratios in Figure 12b. Here, the annual global mean pressure correction is omitted,
and the CH
4
data are normalized to the value at L
S
330°, whereas the O
2
data are normalized to L
S
344.9°
(Figure 12a). Although not displayed on identical vertical scales, it can be seen that both trace gases exhibit
seasonal variations with much greater amplitudes than Ar and N
2
(Figure 12a). The available CH
4
data
indicate a smoother seasonal trend than the O
2
within the sampling frequency. The observed behavior of
either molecule is not currently understood, and a strong relationship between the two might inform the
root cause of observed changes in both O
2
and CH
4
, such as the potential seepage or release mechanisms
hypothesized for CH
4
(Moores et al., 2019). However, it appears that the O
2
and CH
4
show a similar trend
for only part of the year. In particular, the relative decrease during northern fall into winter (as the smaller
of the two polar caps forms) is similar for both species. This is a period when O
2
also more closely follows
the Ar and N
2
measurements as well as the modeled seasonal trends. Yet in the northern spring/summer,
O
2
shows an earlier increase and much more interannual variability than the CH
4
measurements (see
Figure 13a for O
2
and CH
4
vs. sol). Thus it seems that at least some component of the variability in the O
2
cycle is unique or not directly affected by the processes that regulate CH
4
. Seemingly, with respect to O
2
and
CH
4
on Mars, the observations to date are inconclusive as to whether there is a definitive correlation between
the them.
Finally, we tested correlations between O
2
abundance variations with a wide range of environmental
parameters in an attempt to explain the compositional variability. For most parameters, no correlations
were observed (discussed in more detail in Figure S6). Figure 13 shows the correlations identified between
O
2
/
40
Ar and the dust opacity and UV absorption in the atmosphere. An overlay of O
2
/
40
Ar ratios and
dust opacity as a function of MSL sol number (Figure 13a and Figure S6) suggests some type of inverse
relationship between dust opacity and oxygen abundance. A similar relationship is apparent in comparisons
with atmospheric UV absorption (Figure 13b and Figure S7). Note that dust opacity and atmospheric UV
10.1029/2019JE006175
Journal of Geophysical Research: Planets
TRAINER ET AL. 3019

absorption are closely related, because atmospheric UV absorption is more sensitive to variation in dust
loading than to insolation.
The observed correlations are not very strong, with reduced χ
ν
2
values significantly greater than 1
(Figures S6c and S7c). This suggests that if a real relationship between atmospheric dust loading (and/or
atmospheric UV absorption) is present, that relationship is more complex than a simple linear relationship.
Agreement between these dust‐related environmental parameters and the O
2
abundance seems weaker in
MY 32, compared to the other Mars years. Within the limitations of this analysis, a real relationship between
O
2
abundance and atmospheric dust loading (and/or atmospheric UV absorption) could suggest either
unknown photochemical or surface chemistry controls on the O
2
abundance.
Thus, the observed O
2
variability remains a mystery until further measurements, models, or experiments are
able to identify likely mechanisms through which the O
2
can vary on short timescales. It is hoped that
hypotheses that may be testable with further in situ measurements by Curiosity arise, while the mission is
still operating in Gale Crater.
5. Summary
The atmospheric compositional data obtained by the SAM instrument in Gale Crater offer unprecedented
seasonal and multiyear coverage of the Mars atmosphere at the surface. The abundances of the major
atmospheric gases CO
2
,N
2
,
40
Ar, O
2
, and CO have been measured from MY 31–MY 34, through three full
Martian seasonal cycles. Measurements were sufficiently distributed throughout the mission to provide
insight on the behavior of the major atmospheric components in response to the seasonal variations driven
by Mars' obliquity and its large orbital eccentricity.
Figure 13. Ratios of O
2
/
40
Ar (black squares with sol numbers) compared with (a) dust optical depth (colored circles)
and (b) UV atmospheric absorption, as a function of MSL mission sol number. Symbol color (and vertical lines)
delineate Mars years. The dust opacity axis is reversed, because O
2
/
40
Ar and dust opacity appear to be inversely related.
SAM/TLS high‐precision enrichment measurements of CH
4
volume mixing ratio are shown as gray circles in the top
panel, as described in Webster et al. (2018). CH
4
volume mixing ratio values are shown on a different vertical scale,
indicated by horizontal ticks near sol 900. Dust opacity data have been previously described (Martínez et al., 2017; Smith
et al., 2016; Vicente‐Retortillo et al., 2017; Vicente‐Retortillo et al., 2018). Absorption energies are integrated over each sol,
and derived from the model of Vicente‐Retortillo et al. (2017, 2015).
10.1029/2019JE006175
Journal of Geophysical Research: Planets
TRAINER ET AL. 3020

The SAM measurements of volume mixing ratios reveal repeatable cycles in which CO
2
is modulated by the
formation and sublimation of the polar caps, but with a lag of approximately 20–40° of L
S
behind the
maxima and minima of the total pressure cycle. Similarly, the inert tracer species N
2
and Ar show a similar
delayed response to the pressure cycle. These data demonstrate that transport acts very rapidly to maintain
pressure equilibrium, but changes in VMRs lag behind because physical mixing of air masses is a slower