Introduction

Strong geographical and mineralogical evidence on Mars suggests that liquid water once flowed across its surface1,2. Over the past decade, multiple studies analyzing the deuterium-to-hydrogen (D/H) ratio in Martian water reservoirs have revealed an enrichment of deuterium relative to Earth. Since hydrogen (H) atoms escape more easily than deuterium (D) atoms, the D/H ratio serves as a key indicator of water loss to space over the planet’s history. Recent estimates place the D/H ratio at  ~ 0.001, five to eight times higher than on Earth, suggesting that Mars has lost the equivalent of hundreds of meters of global equivalent water layer (GEL) over the past 4 billion years3,4,5,6,7. This substantial water loss has played a major role in shaping Mars’ current dry environment.

Like Earth, Mars experiences four seasons due to a similar axial tilt. Superimposed on that, its elliptical orbit creates notable seasonal differences. During northern summers, Mars is farther from the Sun, resulting in colder temperatures and a relatively dust-free atmosphere. In contrast, southern summers are warmer, as they coincide with Mars’ closest approach to the Sun, allowing a larger solar flux to reach the planet. The warmer atmosphere causes stronger winds and circulation, and more lifting of dust from the surface, which in turn heats the atmosphere further. Additionally, superimposed to this orbital effect, the Martian climate is strongly affected by the topographic asymmetry of the planet with higher elevations in the southern hemisphere controlling the differences between northern and southern summers8. This unique combination of the Martian orbit and its surface dichotomy lead to the formation of regional dust storms each southern summer, which in some cases, can escalate to global dust storms9,10,11.

Recent observations from Mars Express and ExoMars Trace Gas Orbiter (TGO) have revealed a striking contrast in water vapor vertical distribution between northern and southern summers12,13,14,15. During northern summer, water vapor remains confined below 20 km, as it condenses into ice clouds at these altitudes16,17. In contrast, the warmer atmosphere of southern summer prevents condensation at low altitudes, allowing the meridional circulation to transport water vapor above 100 km14,18,19. Large dust storms, when present, further enhance this upward transport18,20,21, which after the photolysis of water vapor, ultimately drives hydrogen escape at the exobase22,23.

This pattern, characterized by higher hydrogen escape fluxes during southern summer, on the order of  ~ 109 cm−2 s−1, repeats annually. As a result, the second half of the Martian year is thought to be the primary driver of H escape, in contrast to the much lower flux of  ~ 107 cm−2 s−1 typically observed during northern summer7.

Results

Unexpected water vapor enhancement

Here, we report a newly observed increase in water vapor in the middle atmosphere of Mars during the northern summer of Mars Year (MY) 37 (Fig. 1). The Nadir and Occultation for Mars Discovery (NOMAD) instrument onboard TGO has been continuously monitoring water vapor abundance from the lower troposphere up to the mesopause (0–120 km) since 2018 through solar occultation measurements. These data are unique in revealing the vertical distribution of water vapor in detail14,18,20,24,25,26. We analyzed this dataset (see Methods) and compared vertical profiles of water vapor during the northern summer across different MYs.

Fig. 1: Latitudinal and seasonal vertical distribution of water vapor.
Fig. 1: Latitudinal and seasonal vertical distribution of water vapor.
Full size image

Vertical distribution of water vapor volume mixing ratio (VMR) as a function of latitude (a1, a2) and solar longitude for the northern (b1,b2) and southern (c1,c2) hemispheres during MYs 35 (left) and MY 37 (right). Dots over each panel indicate the solar longitude, latitude and local time of the NOMAD observations. Blue and red dots indicate morning and evening observations, respectively.

Figure 1 -a1 and a2 illustrate the latitudinal variation in the vertical distribution of water vapor volume mixing ratios at the evening terminator, for Solar Longitudes (LS) between 105° and 130° (late northern summer period) in MYs 35 and 37, respectively. Due to the TGO’s orbital geometry, only specific latitudes are sampled at particular local times and seasons. To ensure a meaningful comparison between different MYs, we focused on comparing MY 37 with MY 35, as in both years TGO covers similar sampling patterns in terms of latitude, LS, and local time. A notably higher abundance of water vapor is observed in the middle atmosphere (up to 80 km) at latitudes poleward of 45°N in MY 37 compared to MY 35. Low water vapor abundances comparable to MY 35 were also seen for MY 3614,26, and are expected by global climate model (GCM) simulations27,28, currently able to reproduce the seasonal variability of the observed water vapor vertical distribution fairly well29,30.

The temporal variation in the vertical distribution of water vapor is shown in Fig. 1-b1 and c1 for the northern and southern hemispheres in MY 35, respectively, while Fig. 1-b2 and c2 illustrate the same for MY 37. These panels highlight the stark contrast between the two MYs, particularly in the northern hemisphere in MY 37 (Fig. 1b2), where a significant water vapor injection of approximately 70 parts per million (ppm) at 60 km is observed at LS = 111°. In contrast, during the same period in MY 35, water vapor abundances in this hemisphere and at these altitudes do not exceed  ~ 4 ppm. This water vapor enhancement persists, with relatively high abundances of around 10 ppm remaining at 60-80 km altitude for several sols until LS  ~ 117°. All individual profiles of water vapor retrieved above 50°N during LS=110°-130° are shown in Fig. S1 from the Supplementary Information (SI). Moreover, at this time, the aforementioned MY 37 water increase is observed systematically at all longitudes (Fig. S1, see also Fig. S2 and S3 bottom panels), suggesting a quick zonal transport distributing water vapor globally at latitudes poleward of 50°N. After this period beyond LS=120°, the water vapor distribution in MY 37 returns to patterns similar to those observed in previous years, with the majority of water vapor confined to altitudes below 30 km as shown in Fig. 1b2.

Moreover, the Emirates Mars Ultraviolet Spectrometer (EMUS) onboard the Emirates Mars Mission (EMM) detected a significant increase in hydrogen density at the exobase of about 2 × 105 cm−3 (Fig. 2). This value is comparable to the densities observed during late-season regional dust storms (C-storms) towards the end of the dusty season9, indicated in Fig. 2 with the green shaded area. This density increase is accompanied by a larger hydrogen escape flux of about 5 × 108 cm−2s−1, representing an increase of about 2.5 times relative to the escape observed in the previous year during the same season. The values for the H escape during the aphelion of MY 36 are consistent with those reported in previous analyses31,32,33 and close to the value expected from diffusion-limited escape processes34,35. However, discrepancies with36 arise, likely due to assumptions in the retrieval and the locations where the data were gathered.36 use climatological temperatures in combination with an assumed response of local H density to local temperature to correct their nadir measurements to the global escape rate. In contrast, our EMM/EMUS retrieval approach simultaneously constrains H density and temperature from Ly-α and Ly-β brightness (with Ly-β providing reduced sensitivity to radiative transfer effects), and derives a global escape rate from full phase-angle coverage.

Fig. 2: Seasonal variation of H density and H escape at the exobase.
Fig. 2: Seasonal variation of H density and H escape at the exobase.
Full size image

Seasonal variation of (first row) the average Mars Ly β brightness, (second row) hydrogen density, (third row) derived hydrogen escape rate as observed by EMM/EMUS and (bottom) water vapor VMR retrieved by NOMAD at 60 km at latitudes above 50°N during MYs 36 and 37. Both the hydrogen density and loss rate are the ones at the exobase. The Mars Ly β emission has been corrected for the solar illumination as inferred from MAVEN/EUVM so that time periods with illumination differences due to solar rotation and the solar cycle can be directly compared. The increase in sun-corrected Ly β brightness at LS ~ 100° in MY 37 following the dust event is a model-independent indicator of an increase in the abundance/temperature of hydrogen in the Mars upper atmosphere, indicating enhanced H escape at this time. The solid lines show the smoothing of the data. The shaded areas indicate (red) the period where the water vapor increase is observed in MY 37 at LS=105°-130°, and (green) the end of the year C-storm season. For each magnitude, the right hand side panels show the ratio of the values within the shaded areas between MYs 36 and 37. The red line indicates the ratio during the same period in both MYs (red shaded areas), whereas the green line indicates the ratio between the peak within the red shaded area in M 37 and the mean value within the green area.

The C storm in MY 36 occurred during LS=308°–320°37, which is close to the perihelion season (Ls = 251°), when the solar flux reaching the planet is significantly higher than during the aphelion season, when the MY 37 event occurred. As a result, the temperature at the exobase, which drives the thermal escape of hydrogen7, is expected to be lower during the aphelion of MY 37 than during the perihelion of MY 36. Even though the hydrogen densities are comparable between these two periods, this could explain the observed two times lower escape rates at Ls = 110° in MY 37 compared to those at Ls = 300° in MY 36 in the third row of Fig. 2.

In addition to the dust storm-driven increase in H escape in MY 37 at LS ~ 110°, the total H loss during the beginning of this year (LS = 30° − 100°) was larger than that estimated for the aphelion of MY 36. The magnitude of escape fluxes is expected to be controlled by the replenishment of hydrogen from the lower atmosphere7,22. However, when estimating the H loss rate, parameters other than H density play a key role, such as exospheric temperatures. Since both the water vapor in the middle atmosphere and the H density at the exobase are similar in the beginning of both MYs 36 and 37 (Fig. 2), the observed interannual variation in the H loss rate during this period is likely due to a progressive increase in solar flux during MY 37, as the Solar Cycle reached its maximum in 2023-2024, leading to higher atmospheric temperatures. H escape is more sensitive to changes in temperature than to changes in H density (ref. 7, see their Materials and Methods section), hence, although there is a notable increase in H density due to the dust storm at LS ~ 110° compared to MY 36, its effect on the H loss rate is moderate relative to the elevated background levels during MY 37.

Although being only a factor 4–6 smaller than the fluxes observed during the perihelion season, the main factor that could lessen the effect of this event on the annual budget of hydrogen loss is its short duration compared to the escape during the second half of the year. The aforementioned escape increase occurs several degrees in LS after the observed water vapor enhancement in the middle atmosphere (Fig. 2). The observed time lag of approximately one week aligns with the expected hydrogen transport timescale via photochemical processes22,38,39.

Martian year 37 dust event

What triggered this unusual vertical transport of water vapor? A combination of observations from different space instrumentation provides the answer. We identified a strong, localized dust storm in the northern hemisphere that coincided in time with the water vapor enhancement. The Emirates Mars Infrared Spectrometer (EMIRS) onboard EMM detected a rapid increase in dust opacity between 30°–60° latitude in the northern hemisphere and 0°–60° east longitude at LS = 108°–111° (Fig. 3a2). This observation is corroborated by nadir measurements of dust opacity from the NOMAD UV-Visible channel (UVIS) (top panel of Fig. 4). Moreover, visible images taken by the MARs Color Imager (MARCI) on Mars Reconnaissance Orbiter (MRO) show how this dust event arose from or around the Antoniadi crater north-west of Syrtis Major (bottom panels of Fig. 4). Atmospheric temperatures retrieved by EMM/EMIRS (see Methods) showed a rapid global increase of 15 K at an altitude of 4 scale heights ( ~ 40 km) (Fig. 3b2). The onset of this increase in temperature coincides in time with the dust event and is observed close to the region where the storm is initially detected. This global increase acts as an indicator of warming the lower-to-middle atmosphere and suggests that it is caused by an increase in atmospheric dust, originating from the localized regional dust storm and consequently transported across the planet. The increased temperatures at 4 scale heights ( ~ 40 km) observed by EMM/EMIRS imply transport of fine dust across the planet at these altitudes, enough to significantly warm the middle atmosphere without substantially increasing the total dust column outside the immediate region of the dust event.

Fig. 3: Column integrated dust optical depth and temperature.
Fig. 3: Column integrated dust optical depth and temperature.
Full size image

Geographical distribution of column integrated dust optical depth (top panels) and temperature (bottom panels) as observed by EMM/EMIRS from LS= 105∘ to 117∘ in MY 37 (averaged over 3∘ of LS). Dust optical depth shows the column measured in the vertical region from the surface up to 4 scale heights (from 0 to  ~ 40 km altitude). Temperature corresponds to measurements at roughly 40 km altitude.

Fig. 4: UV and visible images of the MY 37 aphelion local dust storm.
Fig. 4: UV and visible images of the MY 37 aphelion local dust storm.
Full size image

Geographical distribution of the dust optical depth (300 nm) retrieved from NOMAD/UVIS nadir measurements between LS=108∘-111∘ of MY37 (using a 2∘  × 2∘ averaging grid). (Bottom) MRO-MARCI daily global map images of the early growth of a once in a Martian decade regional dust storm in northwest Syrtis Major (red box in top panel) observed on (left) 21 August 2023 at LS = 107.6∘, (center) August 2023 at LS = 108.0∘, reaching an areal extent of 1.2 × 106 km2. The regional storm persisted for 5-sols. A more common local storm followed, (right) originating in western Syrtis Major on 1 September 2023 at LS = 112.7∘. White features are water ice clouds, dust storms appear as yellowish-orange clouds. Note the water ice clouds above the eastern side of the two storms on the first sol of each event. Color composite created from (437, 546, and 604 nm) filter images and have been cylindrically projected from 0∘-40∘N, 45∘-75∘E, at 1 km pixel−1.

Limb observations from the Mars Climate Sounder (MCS) onboard MRO provided global vertical distributions of dust and water ice cloud extinction, and temperature during this event (Fig. 5; Fig. S2; Fig. S3; Fig. S4). Between LS=108°-111°, dust concentrations increased significantly in the lower-to-middle atmosphere (within 50° latitude) at altitudes of 20–40 km, causing a temperature rise of approximately 10 K. A decrease in water ice concentration is also observed, which can be explained by the higher saturation vapor pressure caused by the higher temperatures. From then until LS = 117°, the dust continued to accumulate between 20–40 km at high northern latitudes, as well as below 20 km at mid-low southern latitudes. The radiative heating induced by the dust further increased temperatures by more than 15 K at altitudes of 30–60 km, across a broad latitudinal range from 50°S to 80°N (Fig. 5), consistent with the EMIRS measurements. This temperature increase is also confirmed by CO2 density analysis from NOMAD measurements (Fig. S5). These temperature changes led to a decrease in water ice cloud content at altitudes of 20–40 km and the formation of a water ice cloud layer above 40 km, similar to the patterns observed in the southern summer season during MY 34’s Global Dust Storm (GDS,40,41,42). This is corroborated by MARCI VIS blue channel images at the time of the storm, sensitive to water ice in the atmosphere, that reveal themselves surprisingly bright, indicating water ice was mixed in with the dust at the top of the storm, or above it (Fig. 4).

Fig. 5: Aerosol extinction and temperature during MY 37’s local dust storm.
Fig. 5: Aerosol extinction and temperature during MY 37’s local dust storm.
Full size image

Latitudinal distribution of dust extinction [1/km] (top), temperature [K] (second row), water ice extinction [1/km] (third row) as measured by MCS, and water vapor volume mixing ratio [ppmv] (bottom) as measured by NOMAD, from LS=105∘ to LS=117∘. Each panel shows the distribution within LS periods of 3∘. For MCS results (three top rows), the first period LS=105∘-108∘ (first column) shows absolute values, whereas the rest of the periods (second, third and fourth columns) show differences with respect to the first one. Bottom panels show absolute water vapor abundances in all columns. NOMAD and MCS vertical profiles have been averaged within bins of 5∘ latitude. The reference wavelength for the dust and water ice extinctions are 461 cm−1 and 843 cm−1, respectively.

Water vapor retrievals from NOMAD data show that the sudden, strong increase in water vapor at altitudes above 15 km immediately followed this dust event (bottom panel of Fig. 5 and Fig. S2). NOMAD water vapor retrievals from previous years do not show such an increase at this time of year (Fig. 1 for MY 35 and14,26 for MY 36). This provides evidence that the exceptional increase in water vapor in the middle atmosphere in northern summer is a direct result of the dust storm during MY 37 and its subsequent atmospheric heating, reducing cloud formation at low altitudes. A schematic representation of this event is illustrated in Fig. 6.

Fig. 6: Effects of the MY 37 aphelion regional dust storm.
Fig. 6: Effects of the MY 37 aphelion regional dust storm.
Full size image

Schematic illustrating the atmospheric response to a localized dust storm in the northern hemisphere during the local summer season. High dust loadings significantly increase the absorption of solar radiation, leading to enhanced atmospheric heating, particularly in the middle atmosphere. This thermal response in turn affects the water ice cloud deck, which becomes more vertically extended and less opaque due to reduced water vapor condensation. Additionally, the enhanced atmospheric circulation associated with the dust storm strengthens vertical transport of water vapor from the lower atmosphere, favoring the injection of water to higher altitudes and enhancing the hydrogen escape at the exobase.

Although the major effects of the storm on water vapor are observed in the northern hemisphere, its impact is also noticeable in other regions of the planet. As shown in Fig. 1c1 and -c2, another enhancement in water vapor abundances between 20–50 km altitude is observed in the southern hemisphere, coinciding in time with the northern hemisphere’s dust injection. This southern hemisphere increase of approximately 10–20 ppm was captured by NOMAD during morning terminator observations, as opposed to the evening observations that recorded the northern event. Despite these observational limitations, the increase in water vapor in the southern hemisphere is likely due to the enhanced global circulation driven by the temperature increase and the additional water vapor injected into the northern hemisphere near the upwelling branches of the Hadley cells (Fig. S6).

The behavior of temperature, dust and water ice observed by MCS in MY 37 early summer (LS=90°-120°), was never before observed at this season since the start of the MCS observations in MY 28. Although some dust storm events can occur during the northern spring, including a recent dust storm during LS ~ 40° in MY 3543, the northern summer season is particularly calm in terms of dust events44 and no comparable dust activity was seen at this time (after LS=90°) and location in images by the Mars Orbiter Camera (MOC) on Mars Global Surveyor (MGS) taken between MY 24 and 28. Additionally, the amount of water vapor available in the atmosphere at this moment of the year is significantly larger than that during the spring season45,46. This makes the MY 37 regional dust storm unique in 14 MY (about 24.4 Earth years) of orbital remote sensing.

Discussion

Although during northern summer the Mars’ atmosphere reaches a minimum dust opacity level of about 0.2-0.5 optical depth47,48,49, exceptions occur around the north polar cap and later in the season around the north polar hood50,51,52, or at lower latitudes when localized, intense storms form50,53,54. As suggested by the simulations of55, these storms can rapidly inject dust into the atmosphere, reaching altitudes of 30–50 km, coining the name “rocket dust storms”. The process is somewhat similar to deep convective cumulonimbus clouds on Earth. However, unlike Earth’s storms, which are driven by latent heat release, Mars’ rocket dust storms are fueled by solar absorption by dust particles56.

Such rocket dust storms are expected to be particularly intense during the northern summer season (from late northern winter to late northern summer) in cases when the contrast between the background dust opacity and the storm’s opacity is largest. This leads to stronger temperature gradients and more intense vertical winds that sustain the storm55. In MY 27, a rocket dust storm was observed in the equatorial region (around 3°S, 24.7°E) at LS = 135°54. Given the location, season, and duration of the explosive dust event in MY 37, we attribute it to being similar to the rocket dust storm observed in MY 27. Further detailed modeling simulations should be done to better understand how this particular storm developed.

How do the effects of this storm compare with global and regional dust storm-induced H escape? Regardless of the exact mechanism by which this storm developed, the overall pattern of high-altitude water vapor increase in the event presented here, is similar to that observed during global and regional dust storms in the southern summer season. Once a significant amount of dust is introduced into the atmosphere, it will ascend to higher altitudes by the meridional circulation (see Fig. S6). The heating of the atmosphere by the dust also leads to an increase in water vapor saturation pressure, and hence to a reduction of water ice cloud formation. This in turn favors transport of water vapor to the middle atmosphere, as demonstrated in ref. 39 for the MY 34 GDS.

Finally, as photolysis rates of H2O peak in the middle atmosphere22,27,57, this leads to the increased production of atomic hydrogen. This mechanism, first identified as a driver of the hydrogen escape response to high-altitude water vapor by ref. 38, has since been further investigated in models.58 quantified the contributions of different altitudes to the exospheric hydrogen budget, highlighting the 60 km region as a key source of hydrogen transported to the upper atmosphere. Together, these processes explain the observations of atomic hydrogen in the upper atmosphere by EMUS (Fig. 2), following the observed increase of water vapor in the middle atmosphere by NOMAD illustrated in Fig. 1 and bottom panel of Fig. 2.

It is important to note that this event in MY 37 is exceptional in that it produced a clear increase in water vapor abundances in the middle atmosphere, reaching values close to those observed during the C-storm period in MY 36 at high northern latitudes (bottom row of Fig. 2). Hydrogen densities at the exobase were also similar between these two periods (first and second rows of Fig. 2). Such conditions of enhanced water vapor and exospheric H density are rarely observed outside the perihelion dusty season, and this represents the first example of an event where dust activity is seen to impact H loss outside of southern summer. The magnitude of the H loss rate, however, seems to be at least a factor 2 lower than the mean H escape observed during the C-storm period of MY 36, as shown in Fig. 2. Nevertheless, the MY 37 dust storm driven H escape presented here is comparable to the values reported by ref. 22 during previous C-storm periods close to 3 × 108 cm−2 s−1 escape rates. Notice however that some model-based studies report larger escape rates close to 109 cm−2 s−1 during C-storm periods58,59. Based on our results and these references, besides some discrepancies, a hierarchy could be established regarding the impact of lower atmospheric effects on the H escape: the perihelion dusty season being the most relevant22,58, followed by global dust storms, and finally type C and regional storms such as the one studied here.

What are the implications for the understating of water loss? From current water isotopic fractionation measurements, a 137 m deep GEL is expected to have been lost into space throughout recent Martian history4,5. To reach this estimate, a global planet-wide average escape rate to space of 4.9 x 109 molecules cm−2 s−1 would be needed60. However, GCM simulations and observations do not reach this value, and approach it only occasionally during global dust storms. Hence, additional present or/and past hydrogen escape mechanisms are needed to reconcile these results. A key aspect of the reported water escape event is that it was triggered by a strong local dust storm. This study demonstrates that localized dust storms, even during the aphelion season, can trigger the transport of water vapor to high altitudes and enhance hydrogen escape. However, unlike the perihelion dust season, which occurs annually and sustains larger and longer escape rates, northern summer regional storms appear to be much rarer, with only two such cases observed in the past ten MY (including this event and another in MY27;55). With current and future instrumentation systematically sounding Mars, it will be possible to confirm or refine these statistics. While their long-term contribution to Mars’ water loss budget is limited, these events remain important as they reveal an additional, if infrequent, pathway for atmospheric escape.

Could these strong northern summer storms affect the global budget of hydrogen lost to space? With Mars’ present-day obliquity, the hydrogen escape rates during the aphelion and perihelion seasons are orders of magnitude apart38,60,61, with the perihelion season, particularly during southern summer, driving the bulk of the escape. The anomalous aphelion escape enhancement induced by the MY 37 dust storm, while presenting a strong contrast to the expected escape during this season, is still far from the typical perihelion escape rates, both due to 1) the difference in maximum escape rates and 2) the short duration of the MY 37 storm compared to the whole dusty season, limiting the ability of this single event to make up for the seasonal discrepancy in escape. Moreover, unlike the perihelion dusty season, such aphelion dust events do not occur every MY and appear to be less frequent than global dust storms typically occurring every four MYs, further constraining their overall impact on long-term hydrogen escape. In addition, Mars’ current axial tilt, orbital eccentricity and argument of perihelion make the incident solar flux less intense during northern summer, so that hydrogen reaching the upper atmosphere is less prone to escape. However, the orbital configuration of Mars has not always been as it is today.

Mars’ obliquity varies significantly over time due to gravitational interactions with the Sun and other planets (in contrast to Earth, whose Moon stabilizes its axial tilt). As it is one of the main triggers of dust storms, surface winds and their associated dust lifting have a strong dependency on Martian obliquity. At higher obliquities, near-surface winds and dust lifting were possibly more intense62. This could have favored the formation of large dust storms more frequently, contributing to larger water loss rates. However, it remains uncertain whether the results for seasonal dust activity was actually more or less intense during periods of high obliquity in Mars’ past63.

Recent studies based on GCM simulations reveal that when increasing the obliquity by 5° (from 25° to 30°), the hydrogen escape rate during the perihelion increases by one order of magnitude63. This same study shows that even a higher obliquity of 35° completely changes the seasonal cycle of hydrogen escape, increasing the escape rates by more than three orders of magnitude during aphelion and perihelion. At these high obliquities, with such a large background hydrogen escape, dust storms during the northern summer could have been extremely significant for the total hydrogen escape budget.

Regardless of uncertainties about past climates, this MY 37 northern dust summer dust event reveals that even during low-obliquity periods like that on present-day Mars, when dust is expected to be more settled, a strong localized regional storm can still induce significant water escape.

Methods

Water vapor retrieval scheme

The water vapor profiles presented in this work are obtained from level 1.0 NOMAD SO infrared transmittances, following a two stages retrieval scheme. This methodology has been tested in previous studies and is fully described in refs. 18,25. Here we present a brief summary highlighting the most relevant processing steps.

First, a pre-processing of the NOMAD data is required in order to remove systematic bendings and shifts in the spectral features and to characterize the instrumental noise. Second, a state-of-the-art retrieval control program (RCP) which incorporates the line-by-line and layer-by-layer Karlsruhe Optimized and Precise Radiative Transfer Algorithm (KOPRA,64) is used to invert the water vapor vertical profiles. The RCP processor relies on multiparameter nonlinear least squares fitting of measured and modeled spectra65. Further information about RCP can be found in ref. 66 and in ref. 67. We perform the inversion of the NOMAD data available from five different diffraction orders: 134 (3011–3035 cm−1), 135 (3034–3058 cm−1), 136 (3056–3081 cm−1), 168 (3775–3805 cm−1) and 169 (3798–3828 cm−1). The first three contain relatively weak absorption lines (S ~ 10−21 cm−1/(molecule cm−2)), whereas the others show much stronger lines (S ~ 10−19 cm−1/(molecule cm−2)) from the spectral region close to the ν3 fundamental band which saturate at low altitudes. The effect of saturation in absorption lines introduces uncertainties and instability in the retrievals. In order to minimize this issue, we only use measurements obtained in optically thin conditions. This way we differentiate between diffraction orders at high tangent altitudes (with strong lines) and at low tangent altitudes (with weak lines) combining up to five different pairs of orders, 134, 135 or 136 with 168 or 169. Typically, we use low altitude orders (134/135/136) below 60 km and high altitude orders (168/169) above 60 km. This transition altitude is not fixed and it is fine tuned for each occultation analyzing the equivalent width Wλ of the strongest lines in the spectral region between 3770-3830 cm−1 (ν3 band), as described in ref. 26. With this technique, RCP performs an elegant combination (global inversion) of the information content in both diffraction orders.

As in previous studies18,25,26, we use pressure and temperature profiles from specific simulations of the Mars Planetary Climate Model (MPCM;68) to define the a priori atmospheric state during the inversion. These simulations are matched to the exact time and location at 50 km altitude of each solar occultation provided by the NOMAD telemetry. They incorporate the latest implementations of the dust and water cycles described in ref. 28 with a comprehensive photochemical scheme from69 covering all atmospheric layers from the surface to the exobase. The MPCM simulations use the most accurate estimates of dust optical depth available for each Martian Year (MY), based on11. To prevent non-physical oscillations in the retrieved profiles, we apply first-order Tikhonov regularization, optimized for each diffraction order looking for a tradeoff between error propagation, vertical resolution, and convergence rate. More details on retrieval sensitivity tests and performance as well as comparison and validation exercises between NOMAD water vapor retrievals from14 and17 can be found in refs. 18,25. These exercises revealed an excellent agreement between the different teams regardless the retrieval methodology used to obtain NOMAD water vapor vertical profiles.

Retrieval of hydrogen density and loss rate

Temporal variation of the Lyman beta (Lyβ) line at 102.6 nm, which is a proxy for Mars hydrogen escape, is measured by the Emirates Mars Ultraviolet Spectrometer (EMUS)70 onboard Emirates Mars Mission (EMM)71,72. Compared to the stronger Lyman alpha (Lyα) line, analysis of the weaker Lyβ line is more straightforward because the line is closer to being optically thin, and its brightness is closer to being directly proportional to the column density of the hydrogen atom than Lyα. We present the average Lyβ brightness observed by EMUS over the course of a complete spacecraft orbit for pointing altitudes greater than 500 km, representing the average brightness of the hydrogen corona as a function of time. The observed brightnesses are converted into retrieved parameters for the hydrogen corona using a data analysis process described in ref. 33 which was extended to retrieve values from Lyα and Lyβ simultaneously using the well-known cross sections of these emission lines. We retrieve three parameters: H density and temperature at 200 km (the nominal exobase altitude), and a Lyα model scale factor used to account for a well-characterized degradation of the EMUS instrument sensitivity at Lyα as a function of time. Together, the retrieved H density and temperature are used to derive the thermal H escape rate to space, assuming that the velocity at the exobase follows a Maxwellain distribution. Retrieved densities and escape rates are presented in Fig. 2. As explained in ref. 33, the uncertainty in the retrieved escape rate is smaller than the uncertainty in both the H exobase density and temperature, because the coronal brightness observed and the escape rate derived have a similar functional dependence on the exobase H parameters. All retrieved values are available on the EMM Science Data Center website.

EMM/EMIRS retrieval of column integrated dust optical depth and temperature

The atmospheric temperatures and dust optical depths derived from EMIRS observations use a constrained linear inversion algorithm originally developed for TES45,73, with additional updates for EMIRS74,75. The retrieval utilizes a forward radiative transfer model that includes multiple scattering by aerosol particles. Surface and atmospheric temperatures are first retrieved from the CO2 band centered at 667 cm−1. Dust optical depths are then obtained as part of the aerosol retrieval using spectral regions between 100 and 1315 cm−1 (excluding the CO2 band). Surface pressure is obtained from the Mars Climate Database. The model employs aerosol optical properties with an assumed dust effective radius of 1.5 μm, following established approaches for Martian atmospheric studies76. Dust is assumed to follow a Conrath vertical profile that varies seasonally and latitudinally77. A smoothness constraint is applied to the temperature profile retrieval using a two-point correlation matrix with a correlation length of 0.75 scale heights73. Uncertainties in retrieved atmospheric temperature are typically around 2 K in the middle atmosphere, with slightly higher values at the higher altitudes shown here. The estimated uncertainty in dust optical depth for all retrievals shown here is less than 0.05, with most individual values in the 0.01–0.03 range. Further details on the processing and retrieval algorithm for EMM/EMIRS data can be found in ref. 74.

NOMAD UVIS dust retrieval

The retrieval procedure to obtain dust opacities from NOMAD/UVIS nadir measurements is based on the same scheme used for the retrieval of clouds, dust and ozone abundances from SPICAM/UV spectra78. It combines the use LIDORT (Linearized Discrete Ordinate Radiative Transfer)79,80 to perform the radiative transfer calculation, the Mars Climate Database68 that provides the a priori atmosphere and the Optimal Estimation Method81 for the inversion. The procedure has been updated to include recent dust82, ice clouds and surface reflectance models83. The method takes advantage of the fact that ice clouds appear bright in the UV compared to the “dark” dust or surface to perform the retrieval. However, bright surface features like ice caps can mimic these features, so regions expected to contain ice are excluded. Further details on this retrieval scheme can be found in ref. 78.

NOMAD SO CO2 and temperature retrieval

The NOMAD CO2 densities and temperatures presented in this study have been obtained by performing retrievals of SO measurements from diffraction orders 132 (2966–2990 cm-1) below around 50 km and 148 (3326–3352 cm−1) above around 50 km. The CO2 lines in the former order appear below 50 km while those of the latter appear below 100 km and saturate below 50 km. The retrieval process is divided into three primary stages. First, the Atmospheric Spectra Inversion Modular Utility Tools (ASIMUT) program is used to determine the total CO2 column densities along the line of sight for each spectrum. Next, these slant column density profiles are converted into vertical profiles of local number densities using an approach similar to that of84. Finally, temperature profiles are calculated based on the hydrostatic equilibrium equation. Once this sequence is complete, the algorithm returns to the initial step, using the updated temperature profile. Several iterations are performed to ensure convergence and to improve the reliability of the retrieved values. For both orders, only a few iterations are required as the CO2 lines are very weakly dependent on the temperature. Further details on the retrieval of atmospheric temperatures from NOMAD SO observations can be found in85.

MCS temperature, dust and water ice

The Mars Climate Sounder86 is a passive infrared radiometer onboard Mars Reconnaissance Orbiter (MRO). It has 5 mid-IR, 3 far IR, and one broadband visible/near-IR channels, which view the atmosphere and surface in limb, nadir, and off-nadir geometries. Each spectral channel uses a linear detector array, which provides an instantaneous radiance profile when pointed at the Mars limb. From the measured radiance profiles, vertical profiles of atmospheric temperature, dust extinction and water ice extinction are retrieved in an altitude range between the surface and 80–90 km with  ~ 5 km vertical resolution87. The retrievals used for this work utilize a combination of mid-IR and far IR channels that enables aerosol profile retrievals from limb measurements that reach deeper into the atmosphere than would be possible using mid-IR channels only and allow the derivation of aerosol column optical depths through vertical integration88.