Global circulation of Mars’ upper atmosphere

See allHide authors and affiliations

Science  13 Dec 2019:
Vol. 366, Issue 6471, pp. 1363-1366
DOI: 10.1126/science.aax1553

Mapping winds in Mars' upper atmosphere

The atmospheric loss processes that stripped Mars of most of its ancient atmosphere are poorly understood. Benna et al. analyzed atmospheric measurements collected by the Mars Atmosphere and Volatile Evolution (MAVEN) spacecraft as it repeatedly dipped into the red planet's upper atmosphere. Combining multiple observing modes allowed the authors to derive wind speeds and map the global circulation of the atmosphere at altitudes of ∼150 kilometers. In some locations, winds followed the slope of the surface topography far below. Such insights into the upper levels of planetary atmospheres are limited, even for Earth.

Science, this issue p. 1363


The thermosphere of Mars is the interface through which the planet is continuously losing its reservoir of atmospheric volatiles to space. The structure and dynamics of the thermosphere is driven by a global circulation that redistributes the incident energy from the Sun. We report mapping of the global circulation in the thermosphere of Mars with the Mars Atmosphere and Volatile Evolution (MAVEN) spacecraft. The measured neutral winds reveal circulation patterns simpler than those of Earth that persist over changing seasons. The winds exhibit pronounced correlation with the underlying topography owing to orographic gravity waves.

Exploration of the atmospheres of the Solar System terrestrial planets (Venus, Earth, and Mars) has shown that global circulation of neutral and ionized gases in the high-altitude regions of the atmosphere (including the thermosphere and ionosphere) dominates the dynamical state and evolution of these planetary environments [e.g., (1)]. Thermospheric circulation locally and globally distributes the deposited energy and momentum imparted both from below, by lower atmosphere waves, and from above, by the solar wind, and subsequently determines how the lower atmosphere and magnetosphere combine to interact with the thermosphere [e.g., (2)]. Although Earth’s thermospheric circulation has been the target of extensive studies [e.g., (3, 4)], the nature of Mars’ thermospheric transport and its variability remain largely uncharacterized owing to the scarcity of direct observations. The Mars Global Surveyor (MGS), Mars Odyssey, and Mars Reconnaissance Orbiter spacecraft collected data on the dynamics of Mars’ global circulation and its seasonal variability [e.g., (58)]. However, those observations provided very limited and incomplete constraints for global circulation models [e.g., (912)]. We report observations from the Neutral Gas and Ion Mass Spectrometer (NGIMS) (13) on the Mars Atmosphere and Volatile Evolution (MAVEN) spacecraft to investigate Mars’ planet-wide thermospheric circulation.

Since entering Mars’ orbit in 2015, NGIMS has collected in situ measurements of the composition of the thermosphere and ionosphere of Mars. In April 2016, the instrument’s operation mode was modified to measure both speed and direction of neutral flows along the spacecraft path through the atmosphere (14). These observations took advantage of MAVEN’s articulated payload platform (APP), on which the instrument is mounted, to rapidly and periodically swing NGIMS’ boresight direction by ±8° in the spacecraft’s local horizontal plane. This technique allows the instrument to measure neutral winds at altitudes of ~140 to 240 km with typical uncertainty of 20 m s−1 for the along-track component and 6 m s−1 for the cross-track component, with a measurement frequency of ~0.03 Hz. Flow directions and speeds are reconstructed during data processing of the modulated CO2 abundance in the thermosphere using spacecraft pointing and trajectory information (14). In a typical orbit, neutral wind observations are conducted along a spacecraft track that extends over 2000 km and an altitude change of up to 80 km. Two successive wind measurements are separated in time by the 30 s it takes for the APP to complete a single swing and in space by ~130 km horizontally and ~10 km vertically as a result of the spacecraft motion.

To mitigate the effect of short- and long-term variabilities of the neutral flow patterns (on scales of minutes and hours, respectively), most wind observations were collected over monthly campaigns. During most campaigns, NGIMS conducted a set of 5 to 10 successive neutral wind observations at nearly the same local time–latitude (LT-LAT) combination. These observations were separated by the 4.5-hour orbital period of MAVEN. Between April 2016 and December 2018, a period of 1.5 Mars seasons, 33 campaigns (designated C#1 to 33) were completed, spanning latitudes of ±70° and nearly all local times. The list of the individual orbits that make up these campaigns is provided in Data S1. For each campaign, the dominant wind direction and magnitude were assessed by averaging the wind vectors at a fixed LT-LAT location and altitude. The magnitude of the orbit-to-orbit variability of the winds was parametrized by the generalized coefficient of variation (GCV), a consolidated measure of changes in both flow direction and magnitude (14). The GCV metric is a multivariate extension to the classical single-variable coefficient of variation, which provides a consolidated measure of wind variability across all campaigns. Dominant neutral winds (average wind vectors assessed for each campaign) measured over the 33 campaigns are depicted in Fig. 1 and provided in numerical form in Data S2.

Fig. 1 Dominant neutral winds observed using NGIMS during 33 monthly campaigns from April 2016 to December 2018.

The average neutral winds are shown as whiskers. The magnitude of their orbit-to-orbit variability are reflected in the whisker colors. The orbit-to-orbit variability of the winds for a given location is captured by the GCV (color bar). No variability is assessed for campaigns that involved a single orbit (black whiskers of C#1, 13, 16, and 23). The MAVEN ground tracks are shown as black traces, with periapsis locations indicated by black dots. The season of each campaign is indicated by the solar longitude (LS). A neutral wind flow map generated by the M-GITM model (9) is shown in gray arrows for comparison. This map represents the expected winds at 150 km for LS = 180° season and under moderate solar EUV input. More detailed and season-specific comparisons between winds produced by the M-GITM model and a select number of neutral wind campaigns are provided elsewhere (16).

These average neutral wind observations provide a global view of thermospheric circulation on Mars. On the day side, winds originate from the equatorial postnoon region, corresponding to the location of the expected high–atmospheric density and high-temperature region [e.g., (9)]. The stagnant winds of this region were captured during campaigns C#8 and 9. The observations show that this thermospheric mass is transported from day to night either through transpolar eastward and westward flows (e.g., C#1 and 2, 11 and 12, 21 to 23, and 30 and 31) or through an equatorial westward path (e.g., C#17 and 18 and 25 to 27). The high-latitude transpolar flows on the dusk side (C#1 and 2, 21 and 22, and 30 and 31) exhibit small orbit-to-orbit variations and have a mean speed of 211 m s−1. Those on the dawn side are slower (an average of 97 m s−1) and more variable. In the postdusk area (LT = 19 to 23 hours), the observed rotation of the flow westward (C#1, 13, 5, 21, and 30) is consistent with the presence of a convergent zonal wind region where dayside eastward flowing and nightside westward flowing winds meet. The location of the convergence region was captured at LT = 19.5 hours as the very slow but highly variable winds of C#7 during solar longitude (LS) = 332°. The location of this convergence region seems to vary with seasons, because it was not present when the same LT-LAT was mapped 4.5 months earlier during C#5 (LS = 227°). In the predawn, the observed winds (C#3 and 4, 19 and 20, 25 to 29, and 33) are consistent with the presence of a second region of flow convergence near LT = 2 to 6 hours and LAT = ±30° where the thermospheric gas mass is most likely descending (15). The center of this convergence region was most likely captured as the slow and variable winds of C#28.

Figure 1 also provides a qualitative comparison between the average measured winds and the neutral wind flow map predicted by the Mars Global Ionosphere-Thermosphere Model (M-GITM) (9) at 150 km for LS = 180° season and under moderate solar extreme ultraviolet (EUV) input. The northern hemisphere autumn equinox and moderate solar EUV settings correspond to the average conditions encountered by NGIMS over the course of the 33 campaigns. This comparison is notional because the NGIMS observations span multiple seasons and solar activities and are taken over a wide range of altitudes (~140 to 240 km). A quantitative comparison between the measured winds and those modeled by M-GITM has been performed by Roeten et al. for five selected campaigns (16). They found that, in some cases, the magnitudes and/or directions of the M-GITM–simulated winds match those of the observed winds, whereas in others, large differences occur. The role of normal solar forcing [i.e., net radiative heating largely owing to the difference between the absorbed solar EUV and ultraviolet (UV) irradiance and which radiated back to space as CO2 15-μm emission] in driving thermospheric winds at Mars seems to be variable. In cases with large differences between the observed and simulated winds, processes other than normal solar forcing may have altered the expected circulation pattern.

All NGIMS wind observations were taken under conditions of moderate or minimum solar EUV activity. The global-scale thermospheric wind pattern is predicted to remain much the same during time periods when higher solar fluxes occur [e.g., (9)]. However, the magnitude of the winds should increase substantially as solar maximum conditions are approached and in situ EUV and UV heating is enhanced. Simulations with the Mars Thermosphere General Circulation Model predicted that horizontal wind magnitudes at equinox should increase by nearly ~125 m s−1 at ~200 km as solar fluxes rise from solar minimum to maximum conditions [10.7-cm solar radio flux index (F10.7) = 70 to 200 solar flux units] (17). Likewise, M-GITM predicts terminator horizontal winds to vary by ~100 m s−1 (equinoxes) and ~140 to 200 m s−1 (solstices) for the same increase of solar activity (9). Thus, the combination of solar activity (primary) and seasonal variations is expected to contribute substantially to the change in magnitude of the martian thermospheric winds when in situ EUV-UV heating is the primary driver.

Owing to planetary rotation, the MAVEN spacecraft overflies different martian terrains on successive orbits. Analysis of orbit-to-orbit variations in the measured winds during campaigns C#17 and 18 revealed a pronounced correlation between the observed wind variability and the underlying topography of the planet. These two campaigns are unlike the others in two aspects. First, they covered the latitudinal band of ±30° over which Mars exhibits its topographic transition between the southern highlands and the northern lowlands. This band also encompasses the pronounced elevated terrains of the Tharsis plateau and its shield volcanoes. Second, these two campaigns captured the strong and steady equatorial westward flows that advect thermospheric gases from day to night. These steady flows exhibit the lowest measured orbit-to-orbit GCV of our equatorial campaigns. The inherent stability of the neutral flows in this region allows us to detect the faint disturbances that are induced by the varying martian topography. By contrast, campaigns that occurred in the equatorial eastward regions (C#7 to 9) exhibited a high intrinsic variability owing to the presence of the stagnant wind region (LT = 15 to 17 hours) and the convergence region at LT = 19.5 hours where flows are inherently slow and unorganized. Campaigns C#25 and 26, although also equatorial and with strong winds, took place during the planet-encircling dust event of June 2018 (18) that disturbed the thermosphere.

The correlation between the measured winds and the topography was assessed for campaigns C#17 and 18 by comparing the measured wind vectors for a given orbit with the mean vectors derived by averaging measurements taken at a given altitude from all orbits of the campaign (as shown in Fig. 1). The average wind provides a close approximation of the thermospheric flow over a featureless terrain (devoid of any topographic relief). The deviation from the average of the wind vectors that were observed on a given orbit were compared to topographic slopes of the terrain reconstructed from a 1°-resolution MGS elevation map (19).

Figure 2 shows that the wind disturbances of C#17 and 18 tend to preferentially align with the downslope direction of the underlying terrain. This alignment occurs for topographic slopes higher than 0.122° and is more pronounced over regions that exhibit large topographic gradients. However, over locations where the terrain has no pronounced sloping, the directions of the wind disturbances are nearly randomly distributed about the direction of the downslope. The disturbances that are shown in Fig. 2 have an average magnitude of 75 m s−1 compared with average wind speeds of 195 m s−1. A statistical hypothesis test (14) shows that this correlation has a probability of ~1.8 × 10−8 of resulting from a combination of statistical randomness and the undersampling of observations over terrains with high topographic elevation. We therefore conclude that the correlation is statistically significant.

Fig. 2 Topographic correlation of the deviations from the mean wind for campaigns C#17 and 18.

Wind disturbances tend to follow the downslope of the local terrain (low angular deviations) when its topographic slope is high. This preferential alignment creates the exclusion region depicted in gray. The bounds of the exclusion region (dashed red line) are used as a measure of the degree of alignment of the wind disturbances. This degree of alignment increases linearly with the sloping of the terrain, independently of the disturbance magnitude. For flat terrains (slopes <0.15°), the disturbance ceases to follow a preferential direction. The error bars reflect the propagated 1σ uncertainties of the measured disturbances.

An overlay of the wind disturbance vectors on a surface elevation map are shown in Fig. 3 for three orbits of campaigns C#17 and 18. The wind disturbances measured during MAVEN orbits 6047 and 6199 flow away from the elevated terrains of the Tharsis plateau. However, the same wind pattern is disturbed from its average direction during MAVEN orbits 6045 and 6202, flowing down toward the low elevations of Arabia Terra and the Hellas basin. These winds flow downslope toward the Isidis basin and Hellas Planitia on orbits 6044 and 6201.

Fig. 3 Relationship between surface topography, mean flows, and wind disturbances observed during three orbits of campaigns C#17 and 18.

The mean flow (black arrows) is derived by averaging measurements taken at the same altitude from all orbits of the campaign. The shown wind disturbances (red arrows) flow away from high elevation following the terrain local gradient.

The apparent alignment of the observed wind disturbances with the underlying terrain is most likely the result of the winds flowing around and past standing features in the thermosphere. We interpret these standing thermospheric features as perturbations in both density and temperature that are linked to the underlying topography. The ionosphere can be ruled out as a possible source of these topographically linked thermospheric features for two reasons. First, the wind disturbances were observed at altitudes well above the dynamo region (120 to 160 km) where ion-neutral collisions dominate (20) and where ion drag can exert some influence on the thermosphere. Second, the crustal magnetic fields that can generate ionospheric structures with strong connection to the surface exhibit a topology that is weakly linked to the topography of the planet. Because the topographic relief that is responsible for these thermospheric standing features lies 150 to 200 km below the locations where NGIMS is sensing the winds, they most likely exert their influence in the lower thermosphere by means of upward propagating gravity waves.

Flow over terrain with a stably stratified atmosphere is known to produce vertically propagating gravity waves (21). Such gravity waves are often referred to as orographic gravity waves (or mountain waves) to distinguish them from convectively generated gravity waves. Gravity waves transport their energy and momentum upward to the middle and upper atmosphere, where they dissipate into the mean flow through wave breaking, diffusion, nonlinear interactions, and viscosity (22). This localized energy and momentum deposition is the most likely source of the observed correlation. However, gravity wave–driven transport was not forecast (23) to reach the high altitudes of the MAVEN observations (up to 240 km).

Depending on the background winds and temperature, orographic waves are predicted to rise nearly vertically to the thermosphere. Their signature can also reach indirectly higher altitudes through nonlinear interaction (24, 25) with upward propagating nonorographic gravity waves. A combination of the two processes may be responsible for the observed correlation. The ability of gravity waves to propagate upward is controlled to a large extent by their intrinsic phase speed—specifically, horizontal phase speed relative to the background atmospheric flow. Harmonics with larger intrinsic phase speeds have a greater potential to penetrate deeper into the thermosphere (26). Terrain-generated gravity waves have zero phase speeds with respect to surface features and appear fixed in space for a stationary observer. However, relative to the background circulation, these waves are propagating upwind at the same speed as the background wind, which allows them to reach higher altitudes. The spectrum of the waves generated in response to the irregular topographical features is expected to be a function of the flow conditions and topographical properties.

Overall, the global circulation pattern detected by MAVEN appears simpler and more stable over martian seasons than predicted by models of Earth [e.g., (27)]. The persistence of this global pattern over changing seasons indicates the stability of the martian climate on seasonal time scales. By contrast, the observed high variability of the winds on a time scale of hours is consistent with the observed short-term fluctuations and extensive structures of thermospheric densities and temperatures (2830). This short-term variability of the neutral flows reflects the high sensitivity of the thermosphere to small changes in the underlying or overlying sources of forcing (e.g., gravity waves, pressure changes in the lower atmosphere, solar and magnetospheric forcing, etc.). The mapped flow patterns contain no circulation cells at high latitudes (>60°) like those observed at Earth (31). This is most likely due to the lack of a strong dipolar intrinsic magnetic field at Mars. Polar circulation cells are the signatures of strong momentum transfer from convecting ions along magnetic field lines with the neutral gas of the upper atmosphere (32). At Earth, this process provides a pathway for energy transfer from the magnetosphere to the thermosphere, especially during geomagnetic disturbances (33). The absence of this pathway at Mars most likely prevents the thermosphere from experiencing heating and upwelling at high latitudes from particle precipitation and collisions between ions and neutrals, limiting atmospheric escape.

Our observations constrain the current state of Mars’ upper-atmosphere dynamics. The observed global thermospheric circulation must be consistent with the seasonal patterns of the underlying density and temperature structure as well as the short-term (time scale of hours) variations of this underlying structure. The latter may be driven by surface topography and the resulting lower- to upper-atmosphere coupling by means of waves.

Supplementary Materials

Materials and Methods

Figs. S1 to S7

References (3438)

Data S1 and S2

References and Notes

  1. Materials and methods are available as supplementary materials.
Acknowledgments: We are grateful for support from all components of the MAVEN project, including the spacecraft, operations, instrument, science, and project teams. Tests and calibrations of the NGIMS instrument were completed at the Planetary Environment Laboratory of NASA’s Goddard Space Flight Center. Funding: The MAVEN mission is supported by NASA. Author contributions: M.B. conceived the measurement technique. M.B., P.R.M., and B.M.J. planned the observations and collected the data. M.B. and Y.L. processed the data. M.B., S.W.B., K.J.R., and Y.L. analyzed and interpreted the global circulation map. M.B., S.W.B., K.J.R., and E.Y. analyzed and interpreted the gravity wave signatures. M.B. wrote the manuscript with input from all co-authors. Competing interests: The authors declare no competing interests. Data and materials availability: The NGIMS data are available at the Planetary Data System, in the NGIMS bundle at The orbits and times we analyzed are listed in Data S1. The numerical values of the wind vectors plotted in Fig. 1 are provided in Data S2. The 10 wind data files that went into plotting Figs. 2 and 3 are located in the NGIMS bundle directories /l3/2017/11/ and /l3/2017/12/, respectively.

Stay Connected to Science

Navigate This Article