The familiar axisymmetric zones and belts that characterize Jupiter’s weather system at lower latitudes give way to pervasive cyclonic activity at higher latitudes1. Two-dimensional turbulence in combination with the Coriolis β-effect (that is, the large meridionally varying Coriolis force on the giant planets of the Solar System) produces alternating zonal flows2. The zonal flows weaken with rising latitude so that a transition between equatorial jets and polar turbulence on Jupiter can occur3,4. Simulations with shallow-water models of giant planets support this transition by producing both alternating flows near the equator and circumpolar cyclones near the poles5,6,7,8,9. Jovian polar regions are not visible from Earth owing to Jupiter’s low axial tilt, and were poorly characterized by previous missions because the trajectories of these missions did not venture far from Jupiter’s equatorial plane. Here we report that visible and infrared images obtained from above each pole by the Juno spacecraft during its first five orbits reveal persistent polygonal patterns of large cyclones. In the north, eight circumpolar cyclones are observed about a single polar cyclone; in the south, one polar cyclone is encircled by five circumpolar cyclones. Cyclonic circulation is established via time-lapse imagery obtained over intervals ranging from 20 minutes to 4 hours. Although migration of cyclones towards the pole might be expected as a consequence of the Coriolis β-effect, by which cyclonic vortices naturally drift towards the rotational pole, the configuration of the cyclones is without precedent on other planets (including Saturn’s polar hexagonal features). The manner in which the cyclones persist without merging and the process by which they evolve to their current configuration are unknown.
NASA’s Juno spacecraft10,11 has been operating in a 53-day highly elliptical polar orbit of Jupiter since 5 July 2016. The spacecraft has passed close to Jupiter six times now, on five of which occasions instruments on board were able to sound the planet and observe many interesting atmospheric structures12,13,14,15. The Juno spacecraft is in a high-inclination orbit with perijove (the point in its orbit nearest Jupiter’s centre) approximately 4,000 km above the cloud tops, passing from pole to equator to pole in about two hours. From their unique vantage point above the poles, JIRAM16,17 (Jupiter InfraRed Auroral Mapper) and JunoCam18, onboard Juno, obtained unprecedented views of Jupiter’s polar regions. JIRAM is an infrared imager suitable for atmospheric mapping and JunoCam is a pushframe visible camera. Jupiter fly-bys took place during perijove passes PJ1 on 28 August 2016, PJ3 on 11 December 2016, PJ4 on 2 February 2017 and PJ5 on 27 March 2017 (no remote-sensing observations were collected during PJ2).
The atmospheric structure in Jupiter’s polar regions is very different from the well known axisymmetric banding of alternating belts and zones at lower latitudes. The polar turbulence predicted by models is consistent with initial close-up observations in the visible part of the spectrum15. Cyclones, as opposed to anticyclones, were expected in the polar regions as a result of the Coriolis β-effect9,19,20. What was unexpected is their stable appearance, close clustering and symmetry around each of the poles.
The Northern Polar Cyclone (NPC, Fig. 1) has a diameter of approximately 4,000 km (on the JIRAM infrared images). It is offset relative to the geographic north pole of Jupiter by about 0.5° and is surrounded by eight circumpolar cyclones (henceforth referred to as just ‘cyclones’) in a double-squared geometric pattern (Figs 1 and 2). Counting alternating cyclones, four are centred at about 83.3° N and the other four are centred at about 82.5° N. The square formed by the latter four cyclones is shifted with respect to the square formed by the former four cyclones by 45° longitude, forming a ‘ditetragonal’ shape, in which the angular distances between the centre of one cyclone and the next vary from 43° to 47°. All cyclones have similar dimensions with diameters ranging from 4,000 km to 4,600 km. Spiral arms are prominent in their outer regions, but tend to disappear in their inner regions except in the NPC itself. These arms define an additional sphere of influence beyond the cores of the cyclones in which co-rotating material can be found. The four cyclones furthest from the NPC have broad cloud-covered inner regions with sharp oblate boundaries. The four cyclones interspersed between them have more diverse and irregular inner regions, with very small-scale cloud textures; some of them appear chaotic and turbulent.
The Southern Polar Cyclone (SPC, Figs 1 and 2) is surrounded by five large circumpolar cyclones in a quasi-pentagonal pattern. They are of similar size, but are generally bigger than the northern cyclones, with diameters ranging between 5,600 km and 7,000 km. The southern cyclones present a range of morphologies, although the differences are much less distinct than in the north. In particular, some of them display a quasi-laminar circulation: the SPC and two adjacent cyclones have cloud spirals converging to the centre, while the other three cyclones appear to be very turbulent along their spiral cloud branches. The SPC has an offset of about 1°–2° relative to the geographic south pole and the angular distance between two adjacent cyclones is not as regular as in the north: it can vary from 65° to 80° relative to the centre of rotation of the SPC.
Figures 1 and 2 show the correspondence between the features in JIRAM maps and in JunoCam images. Regions that are relatively bright in the JunoCam images are cool in the JIRAM thermal infrared images and regions that are relatively dark in the visible are warm. Because the JIRAM thermal radiance in the approximately 5-μm M-band is primarily governed by cloud opacity, regions that appear warm can be interpreted as relatively clear of clouds, allowing radiance from deeper, warmer regions to be detected, and regions that appear cold must be cloudier. Thus, the visibly bright discrete features in the JunoCam images in Figs 1 and 2 correspond to high-altitude clouds, while the darker background corresponds to a deeper cloud deck. This corresponds to a general qualitative result from JunoCam observations made during PJ1, that visually bright regions correspond to regions that are also relatively bright in the 890-nm band, which is sensitive to absorption by methane gas, implying high-altitude clouds in those regions14. Figure 3, with the highest-resolution maps of the polar regions, gives a detailed view of the polar morphologies, showing JIRAM images corresponding to brightness temperatures in the range 190–260 K.
In most cases, the cyclones are essentially in contact if the spiral arms that extend beyond the core are included. In some cases, a single cloud streak connects the outer spiral arms of adjacent cyclones and can be seen to be continuously stretched. The SPC is in contact with the five cyclones around it. By contrast, the PJ4 JIRAM animation (see Supplementary Videos) reveals a chaotic zone between the NPC and the eight surrounding cyclones, within which there is a largely continuous westward (clockwise) flow at about 86° N; poleward of this, the eastward (anticlockwise) flow of the NPC begins. This chaotic zone appears to contain turbulent small-scale cloud textures and a few small anticyclonic vortices. The largest of these, located between the NPC and the cyclones, may be identical to a similar anticyclonic vortex at PJ5, having moved westward by 31° longitude in Juno’s 53-day orbital period. JIRAM data acquired during PJ4 cover a time span of about 2 h at each pole, enabling us to monitor the movements of the clouds and other structures that are evident within each cyclone, which in turn permits the identification of cyclonic and anticyclonic zones. The velocity field inside each cyclone is not straightforward to evaluate from these data, both because the pointing inaccuracy of JIRAM is not negligible when dealing with fine-scale structures inside cyclones, and because detailed structures whose movement is visible are not scattered uniformly. Table 1 provides a summary of preliminary JIRAM measurements of the rotational speeds of individual cyclones at 1,500 km from their respective centres.
The changes occurring in the polar polygons can be seen by the JIRAM observations with a time lapse of about 53 terrestrial days between PJ4 and PJ5 (Extended Data Fig. 1). By this analysis, the northern eight cyclones appear to drift very slowly around the north pole (or around the NPC), by approximately 2.6° eastward in System III longitude. The changes between the observations taken by JunoCam are based on images of the sunlit side of the poles during PJ1, PJ3, PJ4 and PJ5. On the other hand, JunoCam polar images have an overlap of about 90°–180° in sunlit longitudes between successive perijove observations in similar 53-day intervals, and can also be compared with the complete map from JIRAM at PJ4. They show that the eight cyclones are preserved throughout the entire seven-month period, retaining their individual morphological characteristics, and showing only minor movements (Fig. 2). The visible sector of the octagon rotated around the north pole as follows (positive is westward): PJ1 to PJ3, about +2°; PJ3 to PJ4, −4° to −7.5°; PJ4 to PJ5, 0° to −3.5°. Thus the octagon has not shown any progressive rotation about the pole in System III longitudes. Both instruments observe small meridional displacements of individual cyclones of the same order of magnitude. For the cyclones at the south pole, JunoCam comparisons between perijove passes do suggest a progressive anticlockwise zonal rotation relative to the SPC of +1° every 53 days, as well as some wandering of individual cyclones. There are large variations in the spacing of the cyclones around the pentagon, associated with the opening and closing of a gap that is always present between two of the cyclones. Just as in the north, the cyclones have preserved their individual morphologies over the seven months of observations.
Two questions arise from these data. The first is why the pentagon and octagon drift so slowly or not at all. By Stokes’ theorem, net cyclonic vorticity at the centre would imply cyclonic circulation around the periphery. The other question is why the vortices do not merge. Saturn has a single cyclonic vortex at each pole. By analysing the conditions for formation of each Saturn vortex and comparing them with the conditions on Jupiter, it was predicted that the polar circulation could be different on Jupiter9. Some studies21, applying the theory to the merger of Jupiter’s white ovals in 1998–2000, have also shown that like-signed vortices merge on a fast, advective timescale of four months when they are no longer separated by opposite-signed vortices in a single path, a ‘vortex street’. Mergers of the polar cyclones are possible, but they have not occurred over seven months of observation, nor is there any evidence of new structures appearing inside the cyclone polygons. Finally, on the other hand, other studies22,23 show that polygonal vortex patterns (vortex crystals) can develop owing to interaction with a background of weaker vorticity and last indefinitely in a two-dimensional Euler flow.
The Jovian InfraRed Auroral Mapper
The Jovian InfraRed Auroral Mapper (JIRAM) is composed of an imager and a spectrometer that share the same telescope16,17. The imager focal plane is equipped to observe the planet through a bandpass filter centred at 4.78 μm with a 480-nm bandwidth (M-band) and a bandpass filter centred at 3.45 μm with a 290-nm bandwidth (L-band). The spectrometer’s slit is optically co-located with the imager’s field of view and its spectral range covers the 2–5-μm interval in 336 spectral bins (bands) resulting in a spectral sampling of 8.9 nm per band across the full spectral range. The instrument design allows for acquisition of simultaneous imager and spectrometer observations: in this study, we used the data from the M-band filter, which covers a field of view of about 1.75° by 6° with 128 × 432 pixels. The instantaneous field of view is 240 μrad (see ref. 16 for instrumental details).
At the time of the observations, the Juno spacecraft was spinning almost perpendicular to the orbital plane. For each spin, JIRAM takes two images: one to the target (nadir direction), and one to the anti-nadir direction, to evaluate the background, which is removed onboard. JIRAM is also equipped with a de-spinning mirror that compensates for the spacecraft rotation and enables it to keep the target image in the field of view during the data acquisition. The de-spinning mirror may also be activated at different times with respect to the nadir direction, allowing a scan of the planet in the spacecraft’s spinning plane. No pointing outside the spinning plane is permitted.
The data shown in this paper (integrated radiance from 4.5 μm to 5 μm) have been taken with 12 ms of integration time, resulting in a noise-equivalent radiance lower than 5 × 10−5 W sr−1 m−2. Table 1 shows exact times of observations (start time, stop time and number of observations in that sequence or scan). JIRAM observed both poles with high image quality and spatial resolution during PJ4 and PJ5. Polar coverage during PJ4 was complete for regions within 30° latitude of both poles with a spatial resolution varying between 12 km per pixel and 96 km per pixel for the north pole and between 21 km per pixel and 62 km per pixel for the south pole. JIRAM coverage of the poles during PJ5 was incomplete, limited by JIRAM’s field of view and Juno’s spin axis orientation at perijove.
For PJ4 at the north pole we have complete data coverage with good emission angle (that is, close to 90°), but to cover other parts of the planet we also use radiance emitted at lower angles. Such radiance is partially depleted because the absorption due to cold clouds occurs over a longer atmospheric path. A simple correction was applied, mainly with the purpose of better identifying the same features at both PJ4 and PJ5. Since during PJ4 JIRAM observed the same regions of the north pole at different emission angles, those data were used to compile a look-up table that, given radiances measured at emission angles lower than 90°, permits scaling of the measured values to the radiances expected at 90° to make the measurements more comparable to each other.
Data are jovian-located and then re-projected in System III planetocentric geographical coordinates, using a polar orthographic projection. Geometric information was obtained by using ad hoc algorithms based on the NAIF-SPICE tool24 for each image. JIRAM raw data are calibrated in units of radiance (W m−2 sr−1) as described16,17. The responsivity used in this study has been revised to a flat value of 2 × 106 digital numbers (DN) per (W m−2 sr−1) by using the cruise-calibration campaign data, performed by using the orange giant star Aldebaran (α Tauri) as a reference target.
Finally, the diameters of the cyclones are calculated on the JIRAM infrared images, defining the outer border of the cyclone where the smaller, anticyclonic structures form and planetocentric coordinates are used throughout this report. Images from JIRAM were processed using Matlab (Fig. 1 and Extended Data Fig. 1) and ENVI-IDL (Fig. 3).
Processing of consecutive images allows for animations revealing motion, as well as for quantitative analysis of cloud velocities. JIRAM data in Extended Data Table 1 have been arranged in animations that show the movement of single vortices during PJ4 observations. Each sequence or scan has been composed into a mosaic, and then each mosaic became a frame of the video. We provide nine Supplementary Videos for the north pole (eight for circumpolar cyclones plus the NPC); each video is made of 11 images. We also provide six Supplementary Videos for the south pole (five for circumpolar cyclones plus the SPC); each video is made of 6 images.
JunoCam is a visible-spectrum camera designed to acquire images through broadband red, green and blue filters mounted directly on a CCD detector, with an 889-nm methane absorption band filter acquiring an image on a separate rotation typically 30 s later. JunoCam is rigidly mounted on the spinning spacecraft. That way, it uses the spacecraft rotation to take a full panorama within about 30 s consisting of up to 82 narrow exposures, referred to as the ‘pushframe’ mode. Usually, it takes partial panoramas of the target of interest. The camera has a horizontal field of view of about 58°, and a Kodak KAI-2020 charge-coupled device (CCD) sensor with four filter stripes, a red, a green, a blue (RGB) and a narrow-band 890-nm infrared filter attached on the 1,600 × 1,200 light-sensitive pixels. For each of the four filters, there is an corresponding readout region of 1,600 × 128 pixels which can be transferred into the resulting raw image. This transfer is not immediate, but the 12-bit data number of each pixel is encoded as an 8-bit value, and tiles of 16 × 16 pixels are compressed in either a lossy or lossless manner. Usually, the encoding of the 12-bit data as an 8-bit value is nonlinear, according to a ‘companding’ function. Motion blur is mostly avoided by a technique called time delay integration. In colour (RGB) mode, for each exposure, three of the four readout regions are added as stripes to the raw image. Full details of the instrument and its operation are available in ref. 18.
JunoCam observed the same polar regions as JIRAM on PJ4 and PJ5—as well as complementary longitudinal regions on PJ1 and PJ3—but as a visible imager, it acquires images in reflected light. A complete polar view must be pieced together from the unshadowed portions of images collected during multiple perijove passes.
Observations were made in both north and south polar regions during PJ1, PJ3, PJ4 and PJ5. Polar imaging in PJ5 was scheduled over extended periods of time to cover more longitudes as the planet rotates through daylight, which enabled time-lapse measurments that include measurements of rotation of the cyclones.
With an approximate geometrical camera model, including its pointing for each exposure, the appropriate three-dimensional vector was calculated for each pixel in a given reference frame, for example, J2000. Position and pointing information are inferred from SPICE data24, with some manual adjustment. Jupiter is modelled as a MacLaurin spheroid on Jupiter’s 1-bar level. A planetocentric coordinate system assigns a three-dimensional position to each longitude/latitude pair. The three-dimensional vector, pointing from Juno to the three-dimensional position, completes the connection of each longitude/latitude pair to colour information. With this method, each raw JunoCam image of Jupiter is reduced to an approximately geometrically calibrated polar-map projection.
Because Jupiter is rotating and Juno is moving rapidly, the illumination for each JunoCam image changes rapidly. Comparison of images requires approximate normalization of the images. For now, this is achieved in a heuristic way, essentially stretching contrast over regions of approximately similar solar incidence angles, subtracting the mean brightness for these bins, and accounting for changing light scattering of a presumed haze layer as a function of emission angle, which can be obtained for sufficiently small crops by high-pass filtering. For the JunoCam maps shown in Figs 1 and 2, this correction was made down to a maximum solar illumination angle of 66°, above which the signal-to-noise ratio drops below a value of 3 per pixel. Further nonlinear brightness stretching and saturation enhancement brings out detail.
Time sequences of 2 to 3 frames of polar images were made on all perijove passes to track cloud motions. After PJ1, it was clear that the limited time sequence on that orbit verified the visible impression that the features surrounding the poles whose ‘arms’ implied cyclonic motion really were cyclones. A special effort was made in PJ5 to create longer sequences of time-lapse images that would illustrate subtler motions of the polar features by imaging over a longer time interval. The sequences given below represent three of the best of those animations. A sequence of images from the north pole is available at http://junocam.pictures/gerald/uploads/20170424/anim/jnc_pj05_N_089_to_105_blend4_enh.html. A sequence of images from the south pole is available at http://junocam.pictures/gerald/uploads/20170331/anim/jnc_pj05_polarS_60px_lin_interpolated_21frames_1200x1200.html. A close-up version of that south polar sequence is available at http://junocam.pictures/gerald/uploads/20170406/anim/jnc_pj05_south_polar_animation_111_to_121_8frames_20fps_1200px.html.
History and terminology of cyclone clusters
The term ‘ditetragonal’ has been introduced in the context of crystallography, since it is one of the ten two-dimensional crystallographic point groups (see table 10.1.2.1 in ref. 25). In the non-Euclidean geometry of the curved polar region, a two-dimensional pentagonal rather than a hexagonal pattern would be conceivable, similar to the surface of a pentagon-dodecahedron (see table 10.1.2.2 in ref. 25). But since the size of the vortices does not fit exactly to the geometry of a pentagon-dodecahedron, an unstable structure switching between a hexagon and pentagon could occur, or an oscillating pentagon for vortices of similar sizes. Besides the ‘vortex crystals’22,23 mentioned above, similar vortex patterns also occur in rotating superfluid helium II for quantum-mechanical reasons26. Theoretical predictions of such quantized vortices reach back to Onsager27 and Feynman28, although Landau29 introduced rotons in 1941. The first experimental observations30 were in 1979.
The data used for this study will be available once the proprietary period ends, namely about six months after the data were collected by Juno, from the NASA’s Planetary Data System at https://pds.jpl.nasa.gov/tools/data-search/. The JunoCam data are all available for direct download from the Mission Juno web site in both raw and processed form: https://www.missionjuno.swri.edu/junocam/processing.
The JIRAM project is founded by the Italian Space Agency (ASI). In particular this work has been developed under the ASI-INAF agreement number 2016-23-H.0. The JunoCam instrument and its operations are funded by the National Aeronautics and Space Administration. A portion of this work was supported by NASA funds to the Jet Propulsion Laboratory, to the California Institute of Technology, and to the Southwest Research Institute. A.P.I. was supported by NASA funds to the Juno project and by NSF grant number 1411952.
Extended data figures
This zipped file contains 9 videos for the North Pole (8 CPC, plus the NPC); each video is made of 11 images. It also contains 6 videos for the South Pole (5 CPC, plus the SPC); each video is made of 6 images. References are to Figure 1. 0°W in System III is positioned on the right centre of the images. Both Poles are seen from above, namely, progressing counter clockwise you move towards east in the North and towards west in the South. For the North Pole videos, north0.avi.mp4 is the North Polar Cyclone, north1.avi.mp4 is the cyclone at 0°W and for the other videos north[X].avi.mp4, the numbering [X=2 to 8] proceeds counter clockwise. For the South Pole videos, south0.avi.mp4 is the South Polar Cyclone, south1.avi.mp4 is the cyclone at 150°W and for the other videos south [X].avi.mp4, the numbering [X=2 to 5] proceeds counterclockwise.