Mars may look like a frozen, airless desert from the ground, but high above its surface the planet’s atmosphere is a constantly changing laboratory. In the thin region between roughly 100 and 180 kilometers above Mars, carbon dioxide molecules absorb extreme-ultraviolet sunlight, temperatures rise and fall, dust storms send disturbances upward, and the remaining gas gradually merges into space. Now, researchers have carried out the most systematic comparison yet of several independent measurements of this region from NASA’s Mars Atmosphere and Volatile EvolutioN, or MAVEN, spacecraft. Their study shows how observations made by different instruments can be placed on a common scale, potentially turning a collection of partially overlapping datasets into a much more powerful record of the Martian thermosphere.

The challenge is not simply that MAVEN has many instruments. It is that those instruments observe the atmosphere in fundamentally different ways. The Neutral Gas and Ion Mass Spectrometer, known as NGIMS, samples gas directly at the spacecraft as MAVEN sweeps through the upper atmosphere. The Imaging Ultraviolet Spectrograph, or IUVS, views ultraviolet light emitted or absorbed along long paths through the atmosphere. The Extreme Ultraviolet Monitor, EUVM, measures how sunlight is weakened during solar occultations as the spacecraft moves into or out of eclipse. Each technique provides information about carbon dioxide density and atmospheric temperature, but each also has its own viewing geometry, calibration history, altitude range, assumptions and sources of uncertainty. Comparing them is essential before combining them into a single climate record.

The new analysis focuses on five MAVEN datasets: EUVM solar occultations, two IUVS airglow products based on carbon dioxide ion ultraviolet-doublet emissions and atomic oxygen emission at 297.2 nanometers, IUVS stellar occultations, and NGIMS in-situ measurements. Together, they cover different combinations of latitude, local time, solar zenith angle and altitude. The airglow observations are confined to the dayside because they depend on sunlight-driven emissions. EUVM solar occultations are restricted largely to the dawn and dusk terminators, where the spacecraft’s line of sight crosses the Sun through the atmosphere. Stellar occultations and NGIMS can sample both day and night, although their coverage is controlled by MAVEN’s orbit and its changing orientation around Mars.

For the comparison, the researchers used observations collected from Mars Year 33 through the end of Mars Year 36, spanning June 2015 to December 2022. Four Martian years allowed the team to include seasonal changes, different levels of atmospheric dust and variations in solar activity, from the declining phase of solar cycle 24 into the rising phase of solar cycle 25. The analysis concentrated on the overlapping altitude range of 100 to 180 kilometers, which extends from near the Martian mesopause toward the variable exobase. The exobase is the altitude where collisions become infrequent enough for particles to travel long distances without interacting with one another; depending on atmospheric conditions, it can lie between about 160 and 200 kilometers. This overlap region is especially important because it contains the thermosphere, where solar energy strongly controls atmospheric expansion and escape.

A central feature of the work is its attempt to compare measurements under genuinely similar conditions. The researchers sorted observations into bins according to altitude, latitude, solar zenith angle, solar Lyman-alpha irradiance and a global dust activity index. Solar zenith angle—the angle between the direction to the Sun and the local vertical—was used instead of local solar time because Mars’s terminator shifts in local time during the year. Lyman-alpha irradiance served as a tracer of solar ultraviolet forcing from above, while the dust index represented changes originating lower in the atmosphere. Each comparison bin had to contain at least 20 measurements from every dataset, and at least 10 valid bins were required for an inter-comparison. Within each bin, measurements were combined using their reported uncertainties, while the spread of the observations was retained as a practical measure of variability.

The technical differences among the retrieval methods help explain why agreement cannot be assumed. During an EUVM solar occultation, the measured irradiance is interpreted using the Beer–Lambert law, in which transmission decreases exponentially with the column abundance of absorbing gas and its absorption cross section. The observed column density is then converted into a local density profile using an Abel transform, a mathematical inversion suited to measurements made along curved lines of sight. Hydrostatic equilibrium and the ideal gas law are subsequently used to derive pressure and temperature. The EUVM retrieval treats carbon dioxide as the absorbing constituent, an approximation that can overestimate carbon dioxide at high altitudes where atomic oxygen becomes increasingly important.

IUVS uses different physical signals. The carbon dioxide ion ultraviolet doublet near 288–289 nanometers is produced mainly when carbon dioxide is photoionized or struck by energetic electrons. Because this wavelength region experiences little pure absorption, emission modeling can be used to infer carbon dioxide density. The atomic oxygen 297.2-nanometer airglow is generated primarily through the photodissociation of carbon dioxide, and its altitude profile can contain an upper peak near 120 kilometers and a lower peak near 85 kilometers. In both cases, the AURIC first-principles radiative model is used in an optimal-estimation retrieval. Temperatures are then inferred by applying hydrostatic integration to the retrieved density, with an exospheric temperature supplied by fitting a generalized Chapman function to the emission profile.

Stellar occultations provide yet another route. As a bright star rises or sets behind Mars, IUVS records how the stellar spectrum is attenuated by gases along the line of sight. In the far-ultraviolet range from 115 to 170 nanometers, carbon dioxide and molecular oxygen absorb at different wavelengths. The transmission is modeled as an exponential function of the absorption cross sections and line-of-sight column abundances. A Levenberg–Marquardt fitting algorithm retrieves the columns of both species, after which vertical inversion—often described as onion peeling—converts the integrated measurements into local densities. Temperature and pressure are then calculated by imposing hydrostatic equilibrium. NGIMS, by contrast, ionizes atmospheric particles inside the instrument, separates the resulting ions according to their mass-to-charge ratios and counts them. Carbon dioxide is primarily measured through the mass-44 channel, with isotope and fragment channels used when detector saturation occurs.

To quantify differences between datasets, the team applied reduced major axis regression, a method designed for situations in which both variables contain measurement uncertainty. For carbon dioxide density, the fit was performed on the base-10 logarithm of density, producing a relationship of the form that one dataset’s density equals a constant multiplied by the other dataset’s density raised to a fitted power. This approach captures both multiplicative offsets and changes in the scaling between instruments. Temperature comparisons used either a linear regression or, where the datasets were weakly correlated, the median difference between them. Bins lying more than three standard deviations from the initial relationship were identified as outliers before the calculations were repeated. Regression uncertainties were estimated through 10,000 bootstrap resamplings, and the resulting slopes, intercepts and covariances were propagated into the adjusted measurements.

The outcome is a set of adjustment factors intended to allow the five MAVEN products to be used together rather than treated as isolated records. The comparisons cannot connect every dataset to every other one: EUVM’s terminator-only observations, for example, do not overlap sufficiently with the dayside-only airglow measurements, leaving EUVM directly comparable in this analysis only with NGIMS. That limitation illustrates both the value and the difficulty of multi-instrument planetary science. No single MAVEN instrument observes every location, altitude, time of day or season, but their complementary coverage can reveal patterns that an individual dataset would miss. By correcting systematic offsets and carrying their uncertainties into combined analyses, the researchers provide the Mars aeronomy community with a practical framework for expanding coverage without mistaking instrumental differences for real atmospheric behavior.

The work matters because the Martian upper atmosphere is where several major planetary processes meet. Solar extreme-ultraviolet radiation deposits energy at altitudes where carbon dioxide absorbs it, heating and expanding the thermosphere. Atmospheric dust can alter circulation and temperature far above the storms visible from the surface. Changes in density affect the drag experienced by spacecraft, while the distribution of light and heavy species helps determine how Mars loses atmospheric material to space. A common measurement framework could therefore improve models of present-day weather in near-space, clarify how Mars responds to solar storms and dust events, and strengthen attempts to reconstruct how the planet evolved from a warmer, wetter world into the cold and dry environment observed today. The researchers also intend this comparison to support future links with atmospheric datasets from the European Space Agency’s Trace Gas Orbiter, creating an even broader observational network around Mars.