MAJIS is one of the four remote sensing instruments included in the JUICE payload. Together, this suite conducts observations from the ultraviolet to millimeter wavelengths to characterize the geology, surface composition, and exospheres of the Jovian satellites, as well as Jupiter's atmosphere, with high spatial and spectral resolution. The scientific objectives and core specifications of MAJIS are detailed in Poulet et al. (2024b). Here, we briefly describe the instrument, highlighting specific instrumental characteristics that were relevant to or validated thanks to the LEGA dataset, and refer the reader to Poulet et al. (2024b) for a comprehensive overview of its technical design and science goals.

The spectral range is covered by two channels: the VISNIR (Visible and Near-Infrared) channel (0.5–2.35 µm) and the IR (Infrared) channel (2.28–5.56 µm). The spectral sampling is about 3.6 nm per band and 6.5 nm per band for the VISNIR and IR channels, respectively. MAJIS is a cryogenic instrument that requires the optics and VISNIR detector to operate at temperatures below ∼135 K, and the IR detector below ∼95 K. This is achieved via a passive cooling system that uses two dedicated radiators: one cools the entire optical head (OH), while the other is connected directly to the IR detector via a thermal strap.

Teledyne H1RG detectors were selected for the two focal plane arrays (FPAs), each optimized for a different wavelength cutoff. These detectors feature a resolution of 1024×1024 pixels with an 18 µm pixel pitch (hereafter referred as to pixel_18 µm). The effective photon collection area covers 800 lines of 1016 pixels. A 2× binning is applied in the spatial direction, yielding an instantaneous field of view (IFOV) of 150 µrad (corresponding to an effective pixel size of 36 µm) and a total field of view (FOV) of 60 mrad corresponding to 400 IFOVs along the Xsc axis. A windowing mode allows tunable readout from 400 to 32 binned pixels by step of 16. Windowing will be particularly useful when the target (such as an icy satellite, Io, or Jupiter) occupies only part of the FOV, or in specific observation geometries, including exosphere studies near the limb and stellar occultations. In addition to optimizing spatial coverage, windowing enables significantly shorter integration times within a single frame of a given data cube by reducing the conversion and transfer of data from the detector electronics to the proximity electronics. During the LEGA observations, this capability was critical for mitigating saturation effects caused by the high radiance of targets such as the Moon and Earth (Sect. 2.2) compared to that of Jupiter or its moons. In the spectral dimension, the nominal operating mode also includes a 2× binning in the spectral direction, providing 508 spectral samples. This nominal mode was selected during the LEGA observation (Sect. 2.2).

In addition to the windowing functionality described above, the electronics can initiate detector readout from a configurable starting row, independently for the VISNIR and IR channels. This feature was originally designed to compensate for a potential spatial offset between the two channels along the slit direction (Filacchione et al., 2024). It also enables the acquisition of out-of-FOV pixels for calibration and straylight assessment (Langevin et al., 2024, 2026). This capability was also employed during the LEGA observations (Table 1).

MAJIS can operate in push-broom mode, scanning mode, and motion compensation mode, depending on the scientific objectives and target characteristics. Optimum spatial sampling is achieved when the MAJIS line-of-sight (LOS) drifts by exactly one IFOV during one frame acquisition: in this configuration, the along-track spacing between lines matches the across-track spacing between data elements (binned pixels or samples). When JUICE points to the target, the across-slit FOV is built by acquiring successive frames either by actuating the scan mirror (within ± 2° limits corresponding to ±4° on the LOS) or by taking advantage of the slew of the spacecraft.

During a nominal observation, so-called “browse” data products can be requested by telecommand (TC) to the instrument's on-board main electronics (Poulet et al., 2024b). Each browse product contains a maximum of 32 wavelengths (out of the nominal 508 spectral channels) selected for their relevance to the science goals of the target being observed (Ganymede, Europa, Jupiter…). The general purpose of this MAJIS capability is to mitigate the mission's data volume constraints during the nominal science phase. A lower spatial resolution and a more effective compression scheme compared to the nominal dataset can also be selected so as to meet a 3 % data volume allocation (Poulet et al., 2024b). The low data volume browse cubes are planned to be downloaded within a few days using the X-band link, so that each of these quick-look products will help assess the relevance of the observation and determine whether the full spectral resolution data cube (“science data”) should be downloaded using the Ka-band link. As an example, this makes it possible for MAJIS to perform more observations during the low orbital phase around Ganymede than can be accommodated in the MAJIS data volume allocation, selecting for downlink those with the highest science ranking on the basis of their browse dataset. The LEGA observations offered the first opportunity to test this on-board data processing functionality in-flight using an external source, with one browse product acquired for each observation and each channel (Sect. 3.3).

The instrument was calibrated during a dedicated campaign at instrument level using various calibration sources and subsequently verified during the initial payload checkouts with the internal calibration unit (Langevin et al., 2024, 2026). The results were found to be highly consistent with radiometric, spectral, and geometrical performance models (Poulet et al., 2024a; Haffoud et al., 2024; Langevin et al., 2024; Royer et al., 2025). As briefly noted in the introduction, the LEGA offered a unique opportunity for post-launch calibration verification using external sources, aimed at validating and monitoring in-flight the results obtained during ground calibration and early checkouts. During the radiometric calibration of the VISNIR channel, straylight was identified across wavelengths from 0.5 to 1.4 µm. Observations using sources with the Sun black body temperature (∼5800 K) were not available during ground calibration. These are essential to confirm that the legitimate signal contribution significantly exceeds straylight contamination, in contrast to ground tests conducted with sources at lower black body temperatures (350–2800 K). One of the main objectives of the LEGA was thus to investigate straylight behavior using resolved observations of the Moon and Earth, illuminated by sunlight.

Observational planning accounted for various operational constraints, including encounter geometry, fixed attitude and rapid motion of the spacecraft, data volume limitations, inter-instrument interference, target brightness, and the instrument's capabilities. The main constraints and the corresponding mitigation strategies are reviewed below.

In hot conditions (corresponding to spacecraft-to-Sun distance <1.34 AU for the JUICE cruise), the spacecraft is continuously protected from high solar flux and elevated temperatures, by orienting the high gain antenna towards the Sun. This configuration was maintained throughout the LEGA window, thus preventing true nadir pointing, which would have been more optimal for remote sensing instruments like MAJIS. Consequently, the actual observation window was only open for a short period, only when the Zsc (boresight of MAJIS) crossed the target surface (Fig. 1). The closest approach (C/A) distances were 750 km for the Moon and 6840 km for the Earth. Both C/As occurred near the terminator, with a transition from night to day for both targets, and a phase angle of approximately 90°, decreasing along the observation path (Fig. 2).

Integration times of a single MAJIS frame during LEGA were primarily set to 11 ms, with a few cases at 22 and 70 ms (Table 1). As a comparison, nominal MAJIS integration times at Jupiter will reach 800 ms or more for long dwell observations (e.g., during the 5000 km Ganymede orbit or for exospheres) and 100 ms for short dwell cases (e.g., high resolution observations of icy moons). High flux levels can lead to detector saturation in less than 200 ms, particularly when observing icy moons in the VISNIR or hot spots on Jupiter in the IR. The much shorter exposure times during LEGA were also necessary to prevent saturation, as MAJIS was designed for radiance levels of targets with low to bright albedo located at 5 AU. This was successful, in particular for EGA cubes obtained at relatively high incidence (65° for C10) with signal levels ranging up to 60 % of the full well capacity. The last two observations of the lunar surface reach saturation even in 11 ms at long wavelengths as the surface temperatures (up to 380 K) are much higher than that of Earth's clouds or oceans.

As introduced in Sect. 2, these short integration times can be achieved by reducing the instrument FOV to 64 or 128 samples (compared to the nominal 400 samples) for most acquisitions (Table 1), and by using the 1 MHz readout mode. This readout mode was specifically developed for MAJIS to read out the full projected MAJIS FOV (400 IFOVs) in less than 100 ms. This is required for short dwell times or high collection rates (Jupiter and icy satellites in the VISNIR range and Jupiter hot spots in the IR range) where rapid readout of the detector is essential to avoid saturation (Langevin et al., 2022) as stated above.

For all MAJIS observations, a dark signal acquisition must be available with the same operating parameters (FOV, integration time, spatial binning, spectral mask, frame binning…) before and after the observation. The dark signal is determined by closing a shutter and then averaging 9 frames. For the LEGA operations, the dark acquisition strategy was robust, with dark measurements requested both before and after each observation (except for C2 of the LGA, where only a pre-observation dark was performed in order to minimize the overhead during the very short lunar flyby). This enables testing the critical dark subtraction procedure (Sect. 2.3.1).

Straylight effects were identified in the VISNIR channel during on-ground calibration, with implications for radiometric calibration of this channel (Langevin et al., 2024). The straylight signal extends beyond the edge of the FOV, defined by the starting pixel_18 µm 100 and 88 for the VISNIR and IR channels respectively. It exhibits spectral shapes similar to those of adjacent in-FOV pixels, though with slightly lower intensity. To assess this VISNIR straylight component, one LGA observation and 2 EGA observations were configured such that the FOV window, defined by the first or last pixels, included out-of-FOV pixels (Table 1). This approach enables straylight correction and evaluation of its impact on the legitimate signal (Langevin et al., 2026; Sects. 2.3.1 and 3.4).

During the first payload checkout, a dedicated test campaign was conducted to evaluate potential interferences from the RIME (Radar for Icy Moons Exploration) instrument on MAJIS. While RIME passive mode operation showed no impact on MAJIS data quality, interference tests revealed that the MAJIS infrared channel experiences perturbations when RIME emitter modes are activated, for both tested bandwidths (1 and 2.8 MHz; see Sect. 3.5). A complementary test using an external interference source was recommended to further characterize the impact under different configurations and to validate the joint performance of MAJIS and RIME during coordinated science operations onboard JUICE. In this context, the Moon flyby, with a C/A distance below 1000 km (the maximum range at which RIME can operate) offered a unique opportunity to perform this test. RIME was operated in both passive and emitter modes during the MAJIS observation window, and the acquired signals were representative of those expected during nominal FB C/A operations (e.g. the pushbroom phase of close flybys of satellite such as the first Europa flyby “7E1” or Ganymede circular 500 km orbit phase).

Lastly, it is important to report that a dry run of the LEGA operations was successfully performed in July 2024 to validate the full sequence of instrument operations, including those involving MAJIS.

During the LEGA, MAJIS ultimately acquired 5 and 19 hyperspectral cubes of the Moon and Earth, respectively, for a total data volume of ∼7.5 Gbit. Only the first cube of LGA was acquired during a spacecraft slew; all subsequent cubes were obtained using the scan mirror. Lossless compression without despiking, corresponding to the (N, M, K) parameter set (1, 1, 0) as defined in Langevin et al. (2020), was applied to all cubes. Spatial resolution ranges roughly from 130 to 300 m for the Moon, and from 1 to 3 km for Earth observations.

The four MAJIS image cubes targeting the lunar surface are concentrated along the equatorial band, with longitudes ranging approximately from 90 to 20° E (Fig. 3). Although spatial coverage is limited, the cubes encompass a variety of lunar terrains, including cratered highlands, low-albedo volcanic Mare Fecunditatis and southern portion of Mare Tranquillitatis, as well as impact structures such as the 137 km-diameter Langrenus crater, located on the eastern margin of Mare Fecunditatis. Owing to the spatial resolution of MAJIS, spectral analysis of small-scale features such as km-sized craters and ray systems is achievable (Zambon et al., 2026). The first cube crosses the terminator, providing ideal conditions for geometric calibration, as shadows cast by craters and other geological features are elongated, enhancing the visibility of these features in the image (Seignovert et al., 2026). In addition, the wide range of solar incidence angles enables investigation of the temperature profile from the terminator in C1 to the dayside covered by the three other cubes (Tosi et al., 2026).

MAJIS imaged the Earth over a 25 min interval near the closest approach of 6840 km, acquiring spectra across 1016 spectels (spectral elements). The observation sequence began on the nightside, inbound over the southern Bay of Bengal, then concluded on the dayside in the middle of the Pacific Ocean very close to the Hawaiian Islands (Fig. 4). MAJIS obtained four nightside observations, with the first two capturing land surfaces. One observation crossed the terminator between Taiwan and the Philippines, followed by a series of cubes over the Pacific Ocean under tangent illumination. Later morning observations of tropical convective clouds reveal elongated shadows, enabling cloud-top altitude retrieval based on shadow geometry projected onto the ocean surface (Sect. 5.3). The pointing of Pacific ocean observations is limited to ocean and clouds, even if some small islands are in principle covered by the swath of some cubes. However, due to the illumination conditions and to the presence of clouds, it is very difficult to identify their signal in the data. Three observations over the ocean were acquired with long integration times to estimate in-band trace signatures. The last two observations were acquired in limb geometry (MAJIS slit in a tangent orientation wrt limb) for straylight assessment.

One of the primary objectives of the LEGA campaign was to compare MAJIS data with contemporaneous datasets from other instruments. The relevant datasets are presented below.

We assess the MAJIS performances in terms of SNR from the EGA cubes. As shown in Fig. 7, the radiances measured by MAJIS on clouds and the Pacific Ocean at high incidence present many similarities with that expected in the Jupiter system. Significant absorption bands from gases and ices reduce the spectral radiance (or albedo/reflectivity) compared to the solar reflectance spectrum: H₂O (gas/ice/water) and CO₂ gas for the Earth, H₂O ice for icy satellites, CH₄ gas for Jupiter. For the Earth and Jupiter's infrared hotspots (gaps in the overlying clouds), the radiance is dominated by the thermal contribution for wavelengths larger than 4.5 µm (this is not the case for icy satellites, with much colder surfaces). It is interesting to note that the albedo of Jupiter (∼0.65) and icy satellites (∼0.65 to 0.75) in the visible range is similar to that of Earth clouds (up to 0.9), so that the radiances expected at low phase angles in the VISNIR range for Jupiter and icy satellites are typically a factor of 10 lower than the radiance observed for Earth clouds for an incidence of 65° (multiplied by 2.5 due to the cosine of the incidence and divided by 25 for the reduction of the solar radiance at 5 AU). The cloud-free ocean is much darker than clouds at short wavelengths, but it is hotter so that the signal at long wavelengths is larger. It is also interesting to note that for the Earth as well as for Jupiter and icy satellites there is a very large dynamic range along the spectrum, with signals reduced by factors of up to 100 in strong absorption bands.

As mentioned in Sect. 2.2, high resolution observations in the Jupiter system will be performed with an integration time of 100 ms, so that the SNR performance for the Earth observations at high incidence in 11 ms (with nominal spatial and spectral sampling) is representative of what can be expected for high resolution observations with nominal sampling during the science phase of MAJIS.

The readout noise (RON) of the 1 MHz readout mode at the H1RG pixel_18 µm level is ∼30 e⁻ (Langevin et al., 2024). The noise (then the SNR) can be assessed for H1RG pixel_18 µm by combining the shot noise and the RON, then dividing it by 2 for MAJIS data element with nominal sampling (2×2 H1RG pixel_18 µm). The results for the most relevant EGA observation are shown in Fig. 8. The signal collected by MAJIS data elements ranged from a few 10 e⁻ in deep absorption bands up to 250 000 e⁻ (clouds near 1000 nm). Figure 8 shows that, except in very deep absorption bands, the SNR for high resolution observations in the Jupiter system will range from good (∼50) to very good (∼100) and excellent (∼500) even without stacking. In the crossover range (2280 to 2370 nm), the best SNR will be obtained with the IR channel. It is important to note that on-board stacking will be applied on most science data as a result of de-spiking strategies (Langevin et al., 2020) or when long repetition times are available (7.8 s for the orbit phase at an altitude of 5100 km over Ganymede). As an example, the most frequently used Jupiter observation mode (“disk scan”) averages 8 out of 12 independent samplings into a MAJIS data element, improving by a factor of 3 the SNR as displayed in Fig. 8 for representative Earth spectra. This estimate is consistent with previous radiometric modeling (Poulet et al., 2024b; Royer et al., 2025). For regional mapping of Ganymede, the SNR will exceed 1000 over a large part of the wavelength range due to longer integration time. Additional stacking can be considered on ground for further improving the SNR to detect very weak signals (rings, exospheres) or absorption bands.

The very large dynamic range of signal levels for a single acquisition raises a major challenge for the MAJIS data compression strategy. It implements reversible compression after shifting the DN levels to the right by 0 to 7 (Poulet et al., 2024b). This reduces the number of bits required for coding each value. Specifically, this procedure saves 1 bit/data per step as long as only noise bits are shifted out of the signal in DN. However, a shift by 1 to 7 multiplies the quantization noise (∼1.2 e⁻ for the data as acquired) by 2 to 128, so that implementing a large shift to a relatively low signal increases the total noise. Data compression with a single value of the shift parameter for the full data cube would have been either very ineffective or very penalizing for spectral ranges with lower signal. Therefore, a specific shift can be selected by TC for up to 16 spectral bands per channel so as to adjust the level of shift to the signal level, maintaining the quantization noise at a lower level than the physical noise (Poulet et al., 2024b). For the EGA, a shift of 0 (reversible compression) was applied to all spectral bands (8 per channel for the EGA), leading to an average data volume of 6.43 bits per datum after compression (already a significant reduction from the 16 bits before compression). An optimum set of shifts, as will be applied during the JUICE science phase, would have reduced the data volume to 3.5 bits per datum. As shown in the right panel of Fig. 8, the SNR (which is good to excellent) is reduced by at most a factor 1.1 except in the deepest absorption bands. This is definitely a small price to pay for nearly doubling the number of observations which can be implemented within the MAJIS data volume allocation.

For observations with selective downlink, two files are produced by MAJIS: a nominal data file with the full spatial and spectral resolution, and a “browse” file. LEGA data provided the opportunity to validate the browse production during the LEGA by checking that the browse dataset is correctly extracted on-board from the nominal dataset, and that a more effective compression scheme does not negatively impact the relevance of the browse dataset for selection on the basis of the science interest of the data. Selective downlink is not implemented during cruise, so that downloading of the nominal dataset for all 24 MAJIS observations (5 Moon, 19 Earth) was planned. However, the selective downlink procedure has been tested by preparing browse files for each LEGA observation. The browse spectral table (which can be selected by TC) included 52 wavelengths for this test (32 for the VISNIR channel, 20 for the IR channel). The spatial resolution was retained for the browse dataset, but a more effective compression scheme was selected, with the browse data shifted by 4 to the right before reversible compression while reversible compression was applied to the nominal data as such.

Figure 9 (top panel) displays the DN levels of the raw data for the decompressed nominal and browse datasets. It shows that the browse spectels were correctly extracted from the nominal dataset. The shift by 4 to the right before reversible compression reduced the data volume of the browse file from ∼7.7 to ∼4 bits per datum, 2.8 % of the nominal data file (reversible compression, ∼7 bits per datum), in line with the 3 % target defined for the data volume allocated to browse data. The error introduced by this shift ranges from -7 to +8 DN. As can be seen on the top panel, this error is small compared to a signal from 900 to 12 000 DN for all but one of the selected browse spectels so that the ratio between browse and nominal DN levels is close to 1 (blue line). In the VISNIR spectral range, the last selected browse spectel (2.309 µm) in the VISNIR spectral range had a very low signal hence the error introduced by compression is ∼10 % (due to the low SNR, this spectel will not be selected in updated versions of the browse spectral tables). As can be seen in Fig. 9 (bottom panel), the browse data for the 51 remaining wavelengths would have made it possible to assess the science interest of radiance spectra from this observation of the Moon in the selected spectral ranges.

The 9 cubes extending beyond the FOV show that the relative contribution of straylight observed at short VISNIR wavelengths is larger for LGA data than for EGA data (see Langevin et al., 2026). This confirms that the VISNIR straylight photons have wavelengths in the 1.5–2.5 µm range, as shown during calibration (Langevin et al., 2024). Due to the strong absorption bands of H₂O and CO₂ at wavelengths larger than 1 µm, Earth spectra exhibit a blue slope in the near-IR, with higher reflectance at short wavelengths (impacted by VISNIR straylight, Sect. 5.2) than in the source region of straylight photons, while Moon spectra exhibit a red slope in the near IR (Pieters et al., 2013).

Calibration results have shown that the spectral shape of the VISNIR straylight exhibits only minor changes in spectral shape between neighbor pixels, with a signal level slowly increasing toward the center of the detector. This made it possible to consider a straylight mitigation approach for regions close to the edge of the FOV by applying a coefficient to the spectral shape observed out of the FOV. Three conditions must be met:

The read-out window must extend beyond the edge of the FOV. This was the case for 9 cubes (see Table 1): C4 of the LGA (extending left of the FOV) and C8 for the EGA (4 extending left of the FOV and 4 extending right of the FOV)

Several thermal sensors are available to monitor the thermal behavior of the instrument, which critically affects its performance in terms of operability and dark signal level and subtraction procedure (Langevin et al., 2024). The LGA flyby serves as a representative case of a typical satellite flyby during the nominal science phase, characterized by arrival from the nightside, a spacecraft velocity of ∼3 km s⁻¹ (compared to ∼3.7 km s⁻¹ for the first Europa flyby, “7E1” or 5.2 km s⁻¹ for most of the Callisto flybys), and signal levels reaching saturation. Figure 12 shows the temperature evolution of the OH and the IR focal plane array over a few days surrounding the LEGA event. During LGA operations that lasted about 8 min, the IR FPA temperature increased from 88 to 88.7 K. Subsequently, a ∼30 K increase was observed about 1 h later (with a spacecraft being at a distance of ∼ 1000 km from the Moon at this time) due to radiator illumination by the dayside lunar disk. Using Stefan-Boltzmann's law, this heating can be extrapolated to a ∼1 K increase during 7E1, assuming subsolar temperatures of 380 K for the Moon and 160 K for Europa. While a more refined estimate will require dedicated thermal modeling, this preliminary analysis is reassuring regarding the thermal stability of the instrument during critical events such as close satellite flybys. A detailed assessment of the impact of the thermal evolution on the dark signal and instrument operability is presented in Langevin et al. (2026).

Cruise-phase temperature measurements also provide insights into the thermal conditions anticipated for the science phase. The temperatures recorded since the launch closely follow a linear trend with the solar flux scaled as 1/r², where r is the heliocentric distance. Extrapolation to Jupiter's distance by assuming the same spacecraft attitude yields temperatures of about 82 K for the IR FPA and 129 K for the OH. These values fall within the range predicted during the design phase, at the lower end for the IR FPA and the higher end for the OH (Poulet et al., 2024b).

Previous lunar hyperspectral observations by Cassini (Bellucci et al., 2002) and Juno (Adriani et al., 2016) provided valuable VISNIR spectral data but were limited by coarse spatial resolutions that restricted detailed mapping. The Moon Mineralogy Mapper (M³) on Chandrayaan-1 (Pieters et al., 2009) provided the most impressive near-global mineral mapping at ∼140 m per pixel but covered only wavelengths <3 µm. MAJIS mitigates these limitations by providing sub-kilometer-scale hyperspectral reflectance and thermal imaging capabilities, thereby addressing an observational gap. Table 3 summarizes the key metrics that can be derived from MAJIS data, their relevance to calibration and lunar science objectives, and the corresponding publications within the ANGEO special issue.

In the VISNIR range, MAJIS measurements capture common diagnostic absorption features near 1, 1.25, and 2 µm, attributable to the presence of olivine, low- and high-calcium pyroxenes, anorthosite, arising from both electronic and vibrational transitions. These absorption bands form as solar radiation interacts with multiple randomly oriented mineral grains within the uppermost mm of the regolith, producing reflectance spectra that integrate signals from all surface particles. As lunar soils mature under space weathering, individual grains develop silicate glass coatings enriched in nanophase metallic iron (npFe°), which significantly darkens the soils and reduces the depth of absorption features, particularly in FeO-rich mare regions (e.g., Noble et al., 2007). Spectral indicators such as band depths and continuum slopes can thus be used to assess soil maturity. Another distinctive phase can be observed in the VISNIR range (Mustard et al., 2011). It is characterized by a generally featureless continuum, except for a prominent reflectance maximum near 0.75 µm. In addition, it displays a broad, poorly defined absorption band near 1 µm and a very weak feature near 2 µm. These signatures are typical of that expected for synthetic glasses as well as some lunar pyroclastic glasses (Tompkins and Pieters, 2010). Finally, the 3 µm spectral region enables the detection of hydration features due to hydroxyl (OH) and water (H₂O) molecules (Pieters et al., 2009). As indicated in Sect. 2.3, the lunar flyby was restricted to equatorial latitudes, where surface hydration is known to be the lowest or nearly absent (Li and Milliken, 2016). A detailed analysis of the mineralogical indicators corresponding to the spectral features discussed above is provided in the companion paper by Zambon et al. (2026), while Langevin et al. (2026) examines the detection limits for hydration features, leveraging the high signal-to-noise ratio of the MAJIS data.

Lunar spectra acquired by MAJIS are influenced by thermally emitted radiation beyond 2 µm. Sun visibility and scattering geometry due to topography determine the observed surface temperature. The challenge of quantifying and removing the thermal contribution complicates the identification of spectral features in the shorter wavelength range. The importance of thermal emission has been previously recognized in M³ data (e.g. Li and Milliken, 2016), where initial corrections relied on empirical approaches due to the absence of independent surface temperature measurements acquired at the same local time. Despite its limited spatial coverage, MAJIS data offer the opportunity to investigate thermal emission across a wide range of local times, allowing for an assessment of the accuracy of temperature retrievals and thermal correction methods. An additional outcome of this analysis is the derivation of surface emissivity. Tosi et al. (2026) present a comparison of three methods for extracting surface temperature, and explore the sensitivity of both emissivity and thermally corrected spectra to the adopted retrieval approach.

We refer the reader to the MAJIS papers in this special issue for detailed analyses of the LGA dataset. Below, we present a few representative results that illustrate the instrument's ability to retrieve valuable spectral properties of the lunar surface despite not being specifically designed for this purpose.

The recorded signal ranges from noise levels (as observed in the first lines of C1, where no illumination is present) to saturated values in some spectels, particularly in the IR, due to thermal emission. This variability is driven by the Sun's illumination path being nearly orthogonal to the instrument's pointing direction. The area imaged by MAJIS includes a variety of distinct spectral units, enabling the identification of diverse surface compositions. Figure 13 presents four examples of VISNIR spectral diversity identified in LGA C4. The spectra on the top panel compare fresh materials from both mare and highland regions with those of more mature soils. All reflectance spectra exhibit a characteristic 1 µm absorption band associated with Fe²⁺ in pyroxene and/or olivine, along with a generally red spectral slope, which is more pronounced in mature terrains as expected (Taylor et al., 2001). The middle panel shows that the normalized highland mature spectrum is characterized by weaker or absent 1 µm absorption features and steeper red continuum slopes compared to the mature spectrum from Mare Tranquillitatis. This trend is observed throughout the entire scene (Zambon et al., 2026). The normalized spectrum extracted from a fresh mare crater (black curve) shows the deepest 1 µm mafic absorption. Although these features are visible in the raw reflectance spectra, residual calibration artifacts can introduce systematic errors, which are reduced through spectral ratioing. In this approach, the spectrum from a specific region of interest is divided by the scene-averaged spectrum, enhancing absorption features diagnostic of mineralogy and maturity (Fig. 13, bottom). The near-zero slope in spectra of mature areas reflects the dominance of these terrains within the scene. The slight blue slope observed in the spectrum of the fresh mare crater is interpreted as an indication of immaturity, consistent with the reduced reddening typically seen in less weathered materials. Notably, the red slope observed in the short-wavelength region (<1 µm) of the fresh mare material (green curve) is interpreted as a signature of TiO2 depletion; lower titanium content is known to increase the spectral reddening in this short wavelength range when mixed with pyroxene (Robertson et al., 2022). The 2 µm pyroxene band is barely discernible in the spectrum of selected fresh zones, especially in the black spectrum of the fresh Mare crater. A more in-depth analysis of this band is carried out in Zambon et al. (2026). Overall, the spectral impact of soil maturity, associated with the accumulation of submicroscopic iron in agglutinates and glassy coatings formed through micrometeorite bombardment and solar wind sputtering is clearly expressed through general darkening (especially in the highlands), attenuation of absorption features, and steeper positive spectral slopes.

Daytime surface temperatures were estimated from three temperature retrieval methods (Tosi et al., 2026). The first approach (referred to as Method 1) is based on Kirchhoff's law, in which the thermal emission component in the MAJIS spectra of the Moon is modeled using a Planck function to fit the measured radiance spectrum. Two free parameters (surface temperature and emissivity) are allowed to vary within predefined ranges and within the bounds of in-flight instrumental noise, until convergence is achieved using a Bayesian algorithm (Adriani et al., 2016; Tosi et al., 2014, 2026). A second empirical method of thermal correction (Method 2) is also examined, in which the fraction of incident solar flux F reflected by the surface (hereafter I/F) values at 1.55 µm were fitted using a correlation of reflectance at 1.55 and 2.54 µm observed in laboratory reflectance spectra of Apollo and Luna soil and glass-rich samples to predict I/F values at a wavelength of 2.54 µm (Li and Milliken, 2016). Observed I/F values exceeding this predicted value are attributed to thermal emission. This empirical method shows limitations at high incidence angles, where the laboratory-based correlation does not hold. A third approach builds upon the roughness-informed model introduced by Wohlfarth et al. (2023), originally developed to improve our understanding of the lunar diurnal water cycle based on M³ observations. Finally, the three retrieval methods are compared against simulations from a thermal model developed by Vasavada et al. (1999).

A few examples of best-fit results from two retrieval methods (1 and 2) are here shown in Fig. 14. Temperatures derived using Method 1 are consistently lower than those obtained with Method 2. A sensitivity analysis was conducted to assess the impact of uncertainties in the absolute radiometric calibration of the MAJIS radiance data. The propagation of this uncertainty into the temperature retrieval can lead to deviations of up to several K. A comprehensive analysis of the thermal analysis is presented in Tosi et al. (2026). A comparison with the Diviner Lunar Radiometer Experiment aboard NASA's Lunar Reconnaissance Orbiter (LRO) (Paige et al., 2010) was also successfully performed. Overall, the lunar surface temperatures observed by MAJIS, as derived from three independent model derivations, are found to be first order self-consistent and in agreement within ±10 K with temperatures measured by Diviner and the Vasavada thermal model as a function of solar incidence.

Temperatures derived from MAJIS wavelengths are expected to be more sensitive to the hotter surface components compared to those retrieved from longer wavelengths, due to surface roughness effects, particularly at high incidence angles (Bandfield et al., 2015). At sub-millimeter, millimeter, and longer wavelengths, absorption losses are reduced, allowing subsurface layers to contribute significantly to the observed thermal emission. Complementary temperature measurements were obtained by the SWI instrument, which operated at 250 and 500 µm during the lunar flyby. The SWI ground track intersected MAJIS C4 at three locations. As expected, SWI-derived temperatures are lower by a few tens of K, consistent with a stronger contribution from cooler subsurface layers (P. Hartogh et al., private communication, 2025).

Eventually, one major by-product of the MAJIS temperature retrieval is the emissivity. The analysis of this metrics, also presented in the companion paper of Tosi et al. (2026), reveals a clear dichotomy between basaltic maria and feldspathic highlands, with higher emissivity in the former and lower values in the latter.

Given the wealth of the Earth dataset and near-coincident measurements by Earth-observing satellites, additional results along with more detailed retrieval of the key parameters and their contextualization for Jupiter and satellite science are described in a series of papers (Table 4). In the following sections, we present a few examples illustrating how different components of the Earth's atmosphere are manifested in the MAJIS spectra, highlighting the capabilities of the instrument: gaseous absorption and thermal profiles in Sect. 5.2, cloud phase and cloud-top properties in Sect. 5.3, and the first space-based spectroscopic detection of terrestrial lightning in Sect. 5.4. Note that some science topics are discussed in greater detail than others because they are not covered in the associated companion papers, whereas results such as lightning observations are described more concisely here and expanded upon elsewhere

MAJIS mapped cloud systems during these opportunistic Earth observations to reveal the complexity of their structures, whilst demonstrating MAJIS' spectral and imaging capabilities. In this section, we focus on two overlapping cubes obtained just west of the Hawaiian Islands, C16 and C17 taken at a local time around 10:00 a.m., with a resolution of ∼1.4 km per pixel. Average reflectance maps for the VISNIR and IR channels are displayed in Fig. 19. The two cubes showcase the ability of MAJIS to map a variety of cloud structures above the low-reflectivity ocean surface. This preliminary view indicates localized convective structures reminiscent of cumulus towers; diffuse high-altitude clouds; and thin cirroform cloud structures. Some of the clouds can be seen to cast shadows on the ocean surface.

The acquired spectroscopic measurements can be used to diagnose the variety of cloud properties and their vertical distribution. Below, we apply a series of spectral indices to perform an initial inspection of these two MAJIS spectral cubes, using the depths of gaseous absorption bands to show the capability of the instrument to characterize clouds. The key parameters to be considered include the thermodynamic phase of the clouds (mixtures of liquid droplets and ice crystals), the cloud-top altitude, temperatures, and cloud optical thickness (COT). These spectral indices are based on knowledge of sunlight scattering within gaseous and aerosol media, and provide preliminary assessments of the types of environments sampled by the MAJIS cubes before applying more advanced spectroscopic inversions as addressed in Oliva et al. (2026).

In the EGA C1 acquired over the storm-filled western Pacific, clear visible-wavelength emissions were detected at night off the coast of Sumatra (Fig. 25). This serendipitous signal constitutes the first unambiguous space-based spectroscopic detection of a terrestrial lightning flash (D'Aversa et al., 2026). Actually, up to four individual lightning flashes were identified in the VISNIR channel. Their interpretation relies on the simultaneous detection of atomic O and N emission features, indicative of plasma temperatures of several thousand kelvins. Complementary brightness-temperature maps from the IR channel locate the events at the edge of a deep convective system (Fig. 25). Although MAJIS is not optimized for lightning studies, D'Aversa et al. quantified biases related to spectral undersampling and integration times longer than the flash duration. After applying correction factors, the retrieved energies are consistent with average terrestrial lightning (0.7–1.3 MJ in the O I 777 nm line). Temperature estimates based on line ratios span 5000–20 000 K, reflecting uncertainties due to unresolved events.

These observations provide a valuable benchmark for future lightning searches on Jupiter, where H I emissions, faintly but possibly detected here near Hα, are expected to be key diagnostics. Extrapolating from the terrestrial case, we identify H I lines near 650 and 1870 nm as the most promising for Jovian lightning investigations.