Ionospheric irregularities at Jupiter observed by JWST

Abstract

Jupiter’s upper atmosphere is composed of a neutral thermosphere and charged ionosphere. In the latter, the dominant molecular ion H₃⁺ emits in the near-infrared, allowing for the remote exploration of the physical properties of the upper atmosphere. However, the Jovian low-latitude ionosphere remains largely unexplored because H₃⁺ emissions from this region are faint and spectrally entangled with bright neutral species, such as CH₄. Here, we present James Webb Space Telescope H₃⁺ observations of Jupiter’s low-latitude ionosphere in the region of the Great Red Spot, showing unexpected small-scale intensity features such as arcs, bands and spots. Our observations may imply that the low-latitude ionosphere of Jupiter is strongly coupled to the lower atmosphere via gravity waves that superimpose to produce this complex and intricate morphology.

Main

The extent to which the non-auroral upper atmosphere of Jupiter is coupled to the convective-radiative atmosphere below and to the magnetosphere above is largely unknown because ground-based observations do not offer the sensitivity required to investigate this. The James Webb Space Telescope (JWST) performed observations of Jupiter’s Great Red Spot (GRS) with the NIRSpec Integral Field Unit (IFU) on 27 July 2022 as part of the Early Release Science programme (ERS #1373 (ref. ¹)). The F290LP/G395H filter/grating produced a spectrum between 2.87 and 5.27 μm with a spectral resolving power of R = 2,700. In this Article, we use the short-wavelength (NRS1) detector only, covering 2.85–4.01 μm. The observational setup is shown in Fig. 1a, with JWST acquiring six IFU spectral cubes over a period of 53 min, each with an effective integration time of 5.4 min. The observational details are summarized in Table 1. The NIRSpec IFU has 30 × 30 spaxels, each measuring 0.1^″ × 0. 1^″, producing a spatial resolution of ~300 km per spaxel on the planet. This wavelength region contains emissions from the molecular ion H₃⁺, a dominant ion in the charged-particle ionosphere, along with H⁺, situated at altitudes >300 km above the cloud deck². This ion is produced by both electron-impact ionization about the magnetic poles and solar photo-ionization on the dayside of the planet, which dominates as a production mechanism in the low-latitude ionosphere³. In the absence of dynamic interactions with the atmosphere and magnetosphere, the low-latitude ionosphere may be expected to emit in a relatively uniform manner, given that the incoming solar flux at these heliospheric distances will be constant across the disk at any given moment, with a peak production of H₃⁺ occurring at the subsolar point. At low latitudes, the H₃⁺ emissions are roughly ten times weaker than the auroral region⁴ and are blended with strong solar excited non-local thermodynamic equilibrium (LTE) methane emissions⁵. This makes any observations of this region from the ground challenging⁶ and therefore a suitable target for JWST. The integrated ionospheric H₃⁺ emissions are shown in Fig. 1b, and the analysis procedure is outlined below.

Fig. 1: JWST observations of Jupiter’s GRS.

a, The JWST NIRSpec IFU mosaic sequence used to capture Jupiter’s GRS area on 27 July 2022, shown together with the 3 μm aerosol scattering from the upper troposphere. The times are the midpoint of each observation, and the background image is a JWST Near Infrared Camera (NIRCAM) observation from 27 July 2022 20:27:10 using the F440W filter. Figure 2 shows spectra at points P₁ and P₂. b, The integrated H₃⁺ radiance extracted from the NIRSpec observations. The horizontal grey lines show the magnetic dip angle, and the vertical grey lines shows the orientation of the magnetic field lines as mapped above the surface, spaced by 10° longitude⁹. The oval shows the location of the GRS.

Table 1 Summary of the JWST NIRSpec IFU observations

In Fig. 1a, two points are indicated: P₁ at the GRS and P₂ away from it. The NIRSpec spectra for these points are shown in Fig. 2b, together with a model H₃⁺ spectrum in Fig. 2a. The observed radiances are clearly much higher at the GRS than away from it, showing an enhancement in the reflective aerosols in the upper troposphere and lower stratosphere at the GRS, extending up to 40–50 mbar (ref. ⁷). This figure also illustrates the challenge in extracting the H₃⁺ spectrum from this region, which is rich in spectral features: the H₃⁺ signal is very weak compared to bright structures generated in the lower part of the atmosphere, with H₃⁺ peaking at ~1 mW m⁻² sr⁻¹ μm⁻¹ and the peak radiance at 3.9 μm reaching up to ~20 mW m⁻² sr⁻¹ μm⁻¹ where the H₃⁺ Q(1, 0⁻) line is situated. However, between 3.2 and 3.6 μm, deep CH₄ absorbs most of the incoming sunlight, rendering this region spectrally dark, and the discrete H₃⁺ emission can be readily extracted after the subtraction of the high-altitude solar-pumped non-LTE CH₄ spectrum. This process is repeated for each spaxel and is illustrated in Fig. 3 and expanded upon in the Methods. Note that the strong H₃⁺ line at ~3.46 μm is omitted in this analysis because it is blended with a bright feature produced by the absence of methane opacity in the lower atmosphere⁸, which is not readily modelled. The spectral regions used in this study are shaded grey in Fig. 2, containing 21 H₃⁺ emission lines (Supplementary Table 1). Overall, the solar-pumped CH₄ emissions do not show any substantial small-scale variation across the field of view (Supplementary Fig. 1).

Fig. 2: JWST NIRSpec Jupiter spectra.

a, A model H₃⁺ spectrum generated with values consistent with low-latitude emissions (T = 700 K and N = 5 × 10¹⁵ m⁻²). b, The NIRSpec IFU F290LP/G395H NRS1 spectra from individual spaxel points P₁ and P₂ shown in Fig. 1, from inside and outside the GRS. The vertical green lines indicate the locations of the brightest H₃⁺ lines. The grey shaded area shows the wavelength region considered for this analysis, from which the H₃⁺ spectrum is extracted.

Fig. 3: H₃⁺ extraction procedure.

The analysis sequence applied to retrieve H₃⁺ model fits to the NIRSpec IFU data. a, The spectrum from point P₂ in Fig. 1, from which regions with bright H₃⁺ emission lines are extracted. b, A CH₄ model is fitted to the extracted spectrum. c, Once the CH₄ model is subtracted from the data, isolating the H₃⁺ emission lines, the spectrum can be fitted using the Python package h3ppy to retrieve the temperature of the upper atmosphere and ionospheric H₃⁺ density.

The H₃⁺ emissions extracted from the NIRSpec data can be seen in Fig. 1b, together with the magnetic field topology across this region⁹. A striking result of JWST observations that was never seen in previous ground-based images⁴ is the presence of intricate small-scale structures. These are visible in this small patch on the planet, covering only 37° longitude and 27° latitude. In fact, ground-based observations need to average over several nights to bring out low-latitude features, preventing us from seeing the instantaneous ionospheric structure, which now reveals its complex morphology. There are discrete features that form bands and arcs ~1,000 km wide and up to ~6,000 km long. On average, H₃⁺ is relatively dark inside the GRS, with two bright regions to the northwest and southeast of it. Across the field of view, the integrated radiance varies by up to a factor 4. The intensity of H₃⁺ is governed both by ion density and by the temperature of the upper atmosphere, with which the ion thermalizes. Via H₃⁺ spectral retrievals, these two parameters can be disentangled and the driver of the ionospheric emission features can be determined. The extracted spectra for each spaxel (for example, Fig. 3c) are fitted using the open source H₃⁺ Python package h3ppy, and the results of these retrievals are shown in Fig. 4. The median temperature of all spaxels is 713 ± 20 K, with a median density of 4.5 ± 0.4 × 10¹⁵ m⁻², broadly consistent with previous studies^10,11.

Fig. 4: Temperature and H₃⁺ ion density retrievals.

The physical properties of the ionosphere >300 km (ref. ²) above the Great Red Spot area. a, the temperature, T, of the upper atmosphere, b, H₃⁺ column integrated density, N, c, the uncertainty on the retrieved temperature, ΔT, d, the uncertainty on the retrieved H₃⁺ column integrated densities, ΔN. a and b also show contours of the smoothed parameters to reveal broader-scale structures.

In Fig. 4, the oval indicates the location of the GRS (from Fig. 1), and the contours show a Gaussian smoothing with a width of 2° longitude and latitude of each fitted parameter, showing the broad-scale structure present in the data. The temperature distribution in Fig. 4a shows higher temperature toward higher latitudes, and three regions appear hotter, with a median temperature of 738 K: (1) at the centre of the GRS and (2) northwest and (3) southeast of it. Northwest of the GRS, the troposphere is extremely turbulent¹² and could be a source of gravity waves. The elevated temperatures at the GRS have larger uncertainties associated with them due to the elevated background emissions (Fig. 4c), but this could be related to an increased homopause altitude above the GRS, which destroys colder H₃⁺ at low altitudes via charge exchange, producing column observations of high temperatures and low densities, matching what is seen in Fig. 4. The higher noise levels observed in the GRS might be associated with the presence of elevated hazes that increase the brightness of the GRS at 3.3–3.6 μm. The H₃⁺ column-density retrievals in Fig. 4b show patchy structures that are largely mirrored in the integrated radiance (Fig. 1b). Next, we quantitatively investigate the drivers of the discrete H₃⁺ radiance features.

From the set of NIRSpec observations presented here, mosaic tile (vi) (Fig. 1) contains the clearest (dark) banded structures and is located away from the region with large errors at the centre of the GRS and regions of localized heating. To investigate whether these are produced by reduced temperatures or lower H₃⁺ densities, the relationship between these parameters and the integrated H₃⁺ radiance are plotted in Fig. 5. A third-order polynomial is fitted to each correlation (black line), with the shaded region showing the extent of the median uncertainty. The relationship between temperature and radiance in Fig. 5b shows that the temperatures are not strongly correlated with the observed H₃⁺ radiance, having a Pearson correlation coefficient of 0.15. However, Fig. 5c shows that as the radiance increases, the column density also increases, well outside the density uncertainties, with a correlation coefficient of 0.66. This shows that the observed changes in the H₃⁺ radiance is driven primarily by the H₃⁺ column density. The correlation relations for the other six mosaic tiles are shown in Supplementary Fig. 2. The Pearson correlation coefficients indicate that the radiance is better correlated with density than the temperature for all but one tile (tile (iii)). This may suggest that although the radiance is generally driven by density, there are small areas where the temperature is a significant driver, perhaps driven by gravity-wave heating¹³.

Fig. 5: Drivers of H₃⁺ radiances.

a, The integrated H₃⁺ radiance for NIRSpec IFU mosaic tile (vi) (Fig. 1). b,c, Correlation between temperature and radiance (b) and column-integrated H₃⁺ density and radiance (c). The black line shows the third-order polynomial fit to the data, and p indicates the Pearson correlation coefficient.

Figure 5 also shows that higher temperatures are generally associated with lower column densities. This anticorrelation is well established and expected (for example, refs. ^14,15) and is produced by the fact that the intensity of H₃⁺ emissions is driven exponentially by temperature and linearly by density¹⁶, so they are intrinsically anticorrelated. In general, the spatial differences in temperature and density can be trusted if the differences are larger than the derived uncertainties¹⁴, which is not the case for temperatures in Fig. 5b but is the case for densities in Fig. 5c.

In Fig. 4b, the H₃⁺ column densities vary by a factor of ~4. In the polar region, the column densities are higher by a factor of ~10, but the column-density variability across the H₃⁺ aurorae varies by similar factors¹⁷. The H₃⁺ density variations at high latitudes are related to the discrete auroral field-aligned currents generating vast amounts of H₃⁺ via impact ionization. Such large density gradients are a surprise at low latitudes, however, where ion production is dominated by the smoothly varying incident solar flux. High-altitude in situ Juno plasma observations have also suggested that there are significant fluctuations in the H₃⁺ density at low latitudes¹⁸, which may be related to the density variations seen here.

The results of this study form the starting point for understanding the short-term and small-scale drivers of Jupiter’s low-latitude ionosphere. At Earth, the low-latitude ionosphere is dynamic and rich in features, dominated by electrodynamics¹⁹. Neutral winds play an important role in the behaviour of this region, carrying the ionospheric plasma with them away from the subsolar point. This generates a strong duskward electric field (E), known as the equatorial electrojet. Close to the equator, the magnetic field (B) runs parallel to the surface and the magnetic field-line dip angle is zero, also called the magnetic dip equator. At this location, the E × B force moves ions up in altitude, generating the ‘plasma fountain’, producing a depletion of plasma at the magnetic dip equator and enhancements in plasma density at either side of it as the plasma diffuses back down due to gravity, known as the equatorial ionization anomaly (EIA). The magnetic field of Jupiter has a magnetic field polarity opposite that of the Earth, reversing the direction of B at the equator. This also likely reverses the equatorial electrojet to run dawnward, still resulting in an Earth-like upward E × B force, producing a plasma fountain, removing H₃⁺ ions from the magnetic dip equator. However, our observations show no clear decrease in the H₃⁺ density close to the dip equator, nor any evidence for EIAs (Fig. 4b and Supplementary Fig. 3).

Another important driver of ionospheric density perturbations at Earth are gravity waves, which produce travelling ionospheric disturbances (TIDs), where a single travelling wave produces ordered density perturbations in the ionosphere¹⁹. Above Jupiter’s GRS, the ionosphere is likely subject to a complex barrage of upward travelling gravity waves originating from different tropospheric storms^20,21,22. This process might explain the complex structures see in Fig. 1b—the superposition of multiple gravity waves changing the H₃⁺ density of the ionosphere—and would also explain the arc-like structures seen in the data. For example, the right-angled feature in mosaic tile (vi) could be the result of two perpendicular waves.

A further process that shapes the ionosphere of the Earth at lower latitudes is the propagation of Rayleigh–Taylor instabilities, especially in the post-dusk sector²³. This process is capable of generating extended features and requires a trigger, such as the removal of the low-altitude ionosphere after dusk or a density fluctuation caused by gravity waves. However, the main H₃⁺ density peak at Jupiter is in photochemical equilibrium²⁴, which will act to limit the development of Rayleigh–Taylor instabilities on the dayside; the relatively short lifetime of H₃⁺ means that these cannot be propagated effectively. Therefore, we suggest that TIDs generated by gravity waves are the most likely mechanism responsible for the small-scale features seen in the JWST data.

At Earth, another source of ionospheric irregularities is meteoritic seeding of the upper atmosphere that produces vertical stratification of the ionosphere, known as sporadic E layers, which generates localized increases in the ionospheric density¹⁹. Jupiter’s large mass makes these impacts more common than at the Earth and could therefore be an important contributor to shaping the ionospheric structure, manifesting as H₃⁺ density perturbations close to the homopause²⁵. The meteoritic ions (Mg⁺ and Fe⁺) are long lived, with lifetimes of ~300 Jovian rotations; the sporadic layers can also be shaped by gravity waves, and the spatial variability of electron density can significantly alter the lifetimes of H₃⁺, shaping the observed densities.

Gravity waves in Jupiter’s atmosphere, analogous to TIDs at the Earth, have been modelled²¹, showing that they can produce H₃⁺ density variations of over an order of magnitude, as well as much smaller perturbations in temperature, up to ~30 K. Because the temperature uncertainties in Fig. 4c are of a similar magnitude, we expect to resolve gravity-wave perturbations only in the H₃⁺ column density and not in temperature. The dominant ions in the ionosphere of Jupiter are H⁺ and H₃⁺, with lifetimes of ~1 Jovian day and ~1,000 s, respectively, at ~300 km (ref. ²¹). For waves with periods close to the H₃⁺ lifetime, only very small H₃⁺ density perturbations are expected; but for longer waves with periods closer to the H⁺ lifetime, significant H₃⁺ variations are predicted. This is because the chemical reactions that govern H₃⁺ (that is, production and destruction) occur on much shorter time scales, which means that H₃⁺ remains in chemical equilibrium along the wave, so the H₃⁺ radiance features are a result of the gravity-wave-induced density perturbations in the H₂ source population. The gravity waves seen in the Galileo entry probe thermal profile²⁶ are predicted to generate H₃⁺ density variations of 5% to 23%, whereas here we see variations of up to a factor of ~4 (for example, see Fig. 5c). Note, however, that the modelled H₃⁺ density variations are highly dependent on the wave parameters used, and they do not consider the superposition of multiple gravity waves.

The temporally averaged view of Jupiter’s ionosphere produced by adding almost 50 nights of observations⁴ showed that the intensity of H₃⁺ is strongly linked to the topology of the magnetic field, suggesting that electrodynamic processes shape the structure of Jupiter’s low-latitude ionosphere. In stark contrast, the JWST observations presented here suggest that the instantaneous structure of the ionosphere is largely driven by wave dynamics. The equator-ward edge of IFU mosaic tile (ii) reaches the magnetic dip equator (Fig. 1b), but there is no H₃⁺ density decrease present to suggest the existence of a plasma fountain or an EIA (Fig. 4 and Supplementary Fig. 3). This is not, however, inconsistent with the electrodynamics revealed by the long-term study. Wave dynamics must govern the short-term behaviour of the low-latitude ionosphere, and only when these are averaged out over longer time scales can we reveal features associated with electrodynamics. In other words, in the low-latitude ionosphere of Jupiter, wave dynamics appear to dominate over electrodynamics in the short term, at least in the GRS region. The opposite is true at Earth¹⁹. Additionally, Jupiter’s low-latitude ionosphere has long been considered quite bland²⁴, especially when contrasted to the dynamic auroral zones. The observations presented here show that this is far from true, and there is a richness in features that have never been seen before.

Although solar photo-ionization is expected to dominate the production of H₃⁺ at low latitudes³, there has been speculation that particle precipitation could also occur in this region²⁷. An increase in particle impact ionization would serve to increase the observed H₃⁺ density, due to increased ion production. However, the very complex small-scale pattern seen in Fig. 1b would imply an equally complex set of field-aligned currents across this small region, and there is hitherto no evidence of such currents away from the auroral region.

One of the largest outstanding questions concerning the upper atmospheres of the giant planets is the energy crisis, wherein the observed low-latitude temperatures of these regions are hundreds of kelvin hotter than what the solar energy flux can provide (for example, ref. ²⁸). The two main candidate resolutions to this crisis are equator-ward redistribution of auroral heating and the breaking of acoustic or gravity waves generated in the turbulent troposphere. Both of these potential solutions have challenges to overcome. According to some models²⁹, the movement of energy from the pole down to the equator is largely prohibited by the strong Coriolis forces generated by the large size and rapid rotation of these planets, and the effectiveness of gravity waves as a heat source remains largely unknown. Pole-to-pole observations of the upper atmosphere of Jupiter and Saturn^11,30 show that both have a temperature gradient that very gradually decreases from the pole to the equator, indicative of global transport of heat away from the auroral region. The somewhat higher temperatures seen at higher latitudes in this study are broadly consistent with this scenario.

At Saturn, momentum loading of the upper atmosphere, which could be driven by gravity waves, has been proposed as a mechanism that could reduce the efficacy of the Coriolis force in preventing the movement of heat from the pole down toward the equator³¹. The disturbances observed here are highly suggestive of gravity waves present in Jupiter’s ionosphere. Having an upper atmosphere replete with these waves could similarly reduce the efficacy of Coriolis forces at Jupiter, enabling the equator-ward transport of heat, explaining the gradual pole-to-equator temperature gradient reported previously¹¹, suggestive of a global redistribution of auroral heating. In addition, the large-scale wave of elevated temperatures previously seen at the polar region¹¹ may also contribute to the wave activity seen at lower latitudes when it propagates towards the equator, although this was largely affecting the observed temperatures, not the densities, as is seen here. Although there is observational evidence of redistribution of auroral energy, reproducing these results with numerical modelling is very challenging, and the exact mechanism that allows heat to be redistributed is still an open question. By varying the incoming solar-wind pressure, and therefore the auroral heating, modelling has been able to show that some energy is able to move equator-ward²⁹, albeit only at levels of tens of kelvins.

The temperatures derived from the JWST data are relatively high, as expected from the ‘energy crisis’, but they are also very uniform across the field of view. Although heating has previously been observed above the GRS¹³, no extreme temperatures were found in the present study. This potentially indicates that extreme GRS heating is time-variable or that upwelling from below can occasionally carve out the bottomside of the ionosphere³², resulting in only high-altitude, high-temperature H₃⁺ remaining. However, the presence of multiple interfering gravity waves in the JWST data may have important implications for ionospheric heating. Turbulence is present in the troposphere almost everywhere, so the breaking of gravity waves could produce localized heating at a large number of locations across the planet. As it stands, there is no modelling consensus to what degree gravity waves heat the upper atmosphere³³. Further detailed modelling is required to determine the physical properties of the gravity waves responsible for shaping the low-latitude ionosphere: only certain combinations of wave parameters (for example, amplitude and wavelength) can explain the complex structures seen here, so these observations impose important constraints on future modelling efforts. Ultimately, by deriving those wave parameters based on observations such as these, better estimates of wave-induced heating can be made.

In the auroral region, the large-scale structure of H₃⁺ emissions is governed by the interaction between the magnetosphere, ionosphere and atmosphere. However, it is possible that gravity waves would still provide small-scale ionospheric disturbances at high latitudes that are superimposed on the auroral emissions, so that the polar ionosphere is shaped by forcing both from above (magnetosphere) and from below (gravity waves generated in the troposphere). Overall, however, the polar ionosphere is predominantly shaped by magnetospheric processes, whereas the low-latitude ionosphere is driven largely by the coupling with the lower atmosphere. This highlights the importance of the ionosphere as an interface layer in a highly coupled neutral atmosphere–charged-particle ionosphere–magnetosphere system.

The fact that the lower and upper atmospheres are so strongly coupled at Jupiter may also imply strong coupling at other giant planets, with momentum transfer via gravity waves being an important mechanisms to shape the global ionosphere. Because these waves alter the ionospheric density, they also change the conductivity, which in turn will regulate the efficacy of any interaction with magnetospheric currents, particularly at high latitudes. These JWST observations form the proof of concept for future investigations of this region: for example, if these ionospheric irregularities were monitored over time, the progression of these waves could be directly measured, providing a view of temporal variability and an opportunity to characterize the wave amplitudes and wavelengths.

Methods

The GRS of Jupiter was observed by JWST NIRSpec³⁴ on 27 July 2022 using the IFU and the F290LP/G329H grating/filter as part of ERS programme #1373: ‘ERS Observations of the Jovian System as a Demonstration of JWST’s Capabilities for Solar System Science’ (Co-Principal Investigators: I. de Pater, T. Fouchet). The IFU produces a three-dimensional cube with two spatial dimensions covering 3^″ × 3^″ field of view, with each spaxel covering 0.1^″ × 0. 1^″ and a third spectral dimension covering 2.87–4.01 μm (using the NRS1 detector only) at a spectral resolution, at wavelength λ, of $R=\frac{\lambda }{\Delta \lambda } \approx$ 2,700. The NIRSpec IFU data was calibrated using the Python jwst calibration pipeline (v.1.9.5.dev3+g22acc46) up to Level 2 with Calibration References Data System context defined by jwst_1045.pmap. The MAST observation ID is jw01373-o005_t004_nirspec_g395h-f290lp. Geometry, for example latitude and longitude, was assigned to the data using the Python class JWSTSolarSystemPointing (https://github.com/JWSTGiantPlanets/NIRSPEC-Toolkit).

Six IFU mosaic tiles were acquired to capture the full extent of the GRS, illustrated in Fig. 1 with details listed in Table 1. Each tile is a single observation made up of 30 groups giving an effective exposure time of 322 s (5.4 min). The use of a large number of groups was specifically to capture the relatively weak H₃⁺ emissions with a high signal-to-noise (S/N). Between 3.3 and 3.6 μm, the mean S/N is 55, whereas at 3.0 μm it is 274 (Fig. 2). The full 2 × 3 mosaic took a total of 53.0 min to complete, with a mean time separation of 3.9 min between each tile. Table 1 also lists the local time at the centre of each tile, showing that the prenoon to mid-afternoon sector was sampled. The maps in Fig. 1 and Fig. 4 were projected with a latitude and longitude resolution of 0.05°.

The Level 2 calibrated NIRSpec data contains a considerable amount of arc-like fringing, which appears to be especially apparent where the radiance is low, which is the case for the regions that contain emissions from H₃⁺. To address this, a flat field was constructed from the six mosaic tiles, with the flat field of each wavelength bin being equal to the median of the six mosaic tiles. This works well when the spatial variability across the bin is small, for example for the H₃⁺ bins, but does not work at all when there are bright discrete features, for example discrete cloud features at 3 μm. The flat-field procedure for a wavelength bin that contain strong H₃⁺ emissions is illustrated in Supplementary Fig. 4. Ultimately, the process used to remove the fringing does clearly improve the fidelity of the data and does not introduce any significant artificial features in the resulting maps.

Figure 2a shows a model H₃⁺ spectrum generated using h3ppy with parameters broadly appropriate for Jupiter’s low-latitude ionosphere (temperature T = 700 K, column density N = 5 × 10¹⁵ m⁻² (ref. ¹¹). Figure 2b shows the full NRS1 spectra from both outside and inside the GRS (points P₁ and P₂ in Fig. 1). A significant number of H₃⁺ lines are available in this wavelength range, but disentangling weak lines from a very bright background is extremely difficult. However, between 3.2 and 3.6 μm there is significant absorption of sunlight by CH₄ in the troposphere and stratosphere, rendering this region very dark (indicated as grey in Fig. 2b and shown in Fig. 3a). Within this band, there are two dominant spectral features: (1) non-LTE solar excited CH₄ emission contained within 3.2 and 3.4 μm and (2) discrete emission lines from H₃⁺. To isolate the H₃⁺ spectrum, the solar excited CH₄ spectrum needs to be subtracted. The full radiative transfer modelling required to model the full CH₄ spectrum⁵ is outside the scope of this study, and a simplified, yet effective, approach is adopted here. The spectral shape of the non-LTE CH₄ emissions can be approximated by modelling two separate LTE components: the fundamental transitions (ν₃ → ground) and the hot-band transitions (ν₃ + ν₄ → ν₄). Using the HITRAN³⁵ CH₄ line list, the radiance of the two components can be modelled separately at a fixed temperature (300 K), as well as a third-order polynomial background, which works well over limited wavelength ranges. Here, we use four separate narrow wavelength windows for the CH₄ fitting: (1) 3.23–3.31 μm, (2) 3.37–3.40 μm, (3) 3.40–3.43 μm and (4) 3.528–3.56 μm. An example CH₄ model fit for point P₂ is shown in Fig. 3b. The morphology of the integrated CH₄ radiance can be seen in Supplementary Fig. 1. The residual between the NIRSpec data and the CH₄ model then contains the H₃⁺ spectrum (for example, Fig. 3c), which is then fitted with h3ppy. The fitting program implements the H₃⁺ line list of ref. ³⁶ and the partition function of ref. ³⁷.

The analysis sequence illustrated in Fig. 3 was repeated for all IFU spaxels in this dataset, generating 6,927 individual H₃⁺ spectral fits. Note that this is larger than the nominal 5,400 number of spaxels contained within six NIRSpec IFU observations and is due to the fact that the JWST pipeline drizzles the cubes onto a grid that is slightly larger than the 30 × 30 spaxel IFU when using the ifualign coordinate system.

The magnetic field modelling and field-line tracing for Fig. 1 were performed using the JupiterMag model³⁸, using the latest magnetic field model derived from the Juno magnetometer observations from 33 perijoves of the primary mission⁹.

For a broader context, Supplementary Fig. 5 shows the magnetic field strength and the smoothed temperature distribution derived from the data, and Supplementary Fig. 6 shows the wavelength-integrated H₃⁺ total emissions³⁷ on top of the H₃⁺ imaging observations of ref. ⁴. Supplementary Fig. 2 shows the unprojected integrated H₃⁺ radiance and the correlation between the temperature and H₃⁺ density for tiles (i)–(v).

Supplementary Fig. 7 cuts along −18° latitude and 291°W longitude of the H₃⁺ radiance, temperature and column density, emphasizing that the radiance features are largely driven by column density, not by temperature. The small-scale features are clearly seen, with large variability present within these profiles.

The gaussian filter applied to Fig. 4 and Supplementary Figs. 5 and 6 is the ndimage.gaussian_filter function available in scipy.

The JWST ERS Jupiter programme also included both NIRSpec FL170P/G235H filter/grating and MIRI Channel 1 Short observations of the GRS, both capturing a number of H₃⁺ emission lines at ~2 μm and ~5 μm, respectively. These could have provided a means through which to examine the longer-term spatial variability of the ionospheric features, because they were obtained at different epochs. However, the large magnitude and intricate structure of the underlying signal (deep thermal emissions and aerosol scattering) and the low S/N of the H₃⁺ emission lines renders this mapping impossible. Only the NIRSpec FL290LP/G395H filter/grating was optimized for the relatively weak H₃⁺ emissions and used an effective exposure time roughly ten times longer than the other settings.