Long-Term Behavior of Tropospheric Temperatures in Jupiter: Evidence for Quasi-Seasonal and Non- Seasonal Variability

Glenn Orton (  glenn.orton@jpl.nasa.gov ) Jet Propulsion Laboratory, California Institute of Technology https://orcid.org/0000-0001-7871-2823 Arrate Antuñano Universidad del PaÃ Â s Vasco https://orcid.org/0000-0001-9206-6960 Leigh Fletcher University of Leicester James Sinclair Jet Propulsion Laboratory/California Institute of Technology https://orcid.org/0000-0001-5374-4028 Thomas Momary Jet Propulsion Laboratory, California Institute of Technology Takuya Fujiyoshi National Astronomical Observatory of Japan Padma Yanamandra-Fisher Space Science Institute Padraig Donnelly Laboratoire de Meteorologie Dynamique Jennifer Greco 7Uppsala University; 8The Swedish Collegium for Advanced Study Anna Payne University of Hawaii https://orcid.org/0000-0003-3490-3243 Kimberly Boydstun Nationwide Insurance Laura Wake eld Northrup-Grummon Corporation

appear to be correlated with those 60-70 km higher in the stratosphere near ~10 mbar, with a ~2-year delay from the stratospheric to the tropospheric level, suggesting a top-down control of tropospheric temperatures by radiative processes in Jupiter's stratosphere. Correlations between temperature variability and changes in visible appearance of clouds were also detected in Jupiter's South Equatorial Belt (6°S-17°S) and North Equatorial Belt (6-15°N). No variability was detected at periods longer than 14 years.

Full Text
Images of Jupiter's thermal emission at two wavelengths sensitive to its upper-tropospheric temperatures were acquired over a span of 40 years. The longitudinal-mean variability of the emission and the 330mbar temperatures derived from these observations ±30° of latitude from the equator are shown in Figure   1, together with their associated power spectra. Several features are prominent in the power spectra.
Variations at speci c latitudes, shown in Figure 2, reveal correlations with changes Jupiter's visible appearance. An apparent anticorrelation of temperature variability in northern and southern hemispheres at speci c latitudes in this gure is found to be true overall, as shown in Figure 3. Variability at 300 mbar appears to be correlated with temperature variations higher in the stratosphere, but with a 2-year delay, suggesting a top-down control of temperature variations. Further details on this correlation, as well as the data acquisition and processing are provided in the Supplementary Information. Temperature and its variability constitute an essential element of the climatology of a planetary atmosphere, inextricably linked to planetary dynamics and chemistry. Initial studies of long-term variability of outer-planet temperatures focused on the stratosphere of Saturn 1,2,3,4 . Studies focused on Jupiter detected a stratospheric temperature oscillation 5,6 which -like Saturn -results from the vertical propagation of waves 7 . The earliest study of Jupiter's tropospheric temperatures examined 1-and 2dimensional scans of its disk over 1980-1990 8 , later extended in time to include 2-dimensional midinfrared imaging through 2001 9 . Although snapshots of Jupiter's tropospheric temperatures have been investigated during epochs of visible change in Jupiter's banded features 10,11,12 there has been no systematic assessment of long-term tropospheric temperature variations.
Here, we extend the record of Jupiter's infrared variability to cover 1978-2019, allowing us to diagnose Jupiter's upper-tropospheric temperature eld for three and a half of its 11.9-year solar orbits, separating seasonal and non-seasonal variability unambiguously. The images were obtained using discrete lters at 17-25 µm, where the emission is sensitive to temperatures in the upper troposphere (100-400 mbar). In order to cover variability over such a long time frame, we chose a subset of these lters providing the longest continuous record, with central wavelengths of 18.72 and 20.50 µm. Some of the earliest observations that were made with only a broad lter near 18 µm between 1978 and 1983 8 were analyzed independently, as discussed in the Supplementary Information. Figure 1 shows the 1983-2019 longitudinal-mean brightness temperatures of both lters. Brightness temperatures were smoothed and interpolated from an irregular sampling to a 60-day time grid. In order to remove any systematic emission-angle dependences in the apparent brightness temperature, we derived physical temperatures over pressures of 100-400 mbar from an inversion of ltered radiances measured at 18.72 and 20.50 µm, using the NEMESIS atmospheric retrieval code 13 following procedures for discrete-ltered observations 14 (see Supplementary Information for details). Temperatures were allowed to vary at all altitudes in a constrained way; the retrieved temperatures near 330 mbar had the smallest uncertainties. To minimize any dependence on emission angle further, we limited our analysis to planetocentric latitudes within 30° of the equator, focusing on a region where planetary-scale changes in Jupiter's bands (its visibly dark 'belts' and bright 'zones') appear most frequently and where they are adequately resolved, even in our earliest data. Adding more lters to constrain the inversion better would have provided a more limited temporal sampling. Figure 1C shows the retrieved temperatures and Figure  1D the residual from a longitudinal mean. Figure 2 illustrates temperatures at speci c latitudes.
Among the striking properties of the brightness-temperature time series ( Fig. 1E and 1F) are apparent 10-14 year periodicities of temperatures (the yellow, orange and red boxes in Fig. 1H), all intriguingly close to Jupiter's 11.9-year orbital period. To ensure that this quasi-annual behavior is not an artifact of the seasonal variations of the emission angle at a xed latitude, we estimated the size of the brightness temperature change due solely to the change of emission angle: even at 30° from the equator, the change (0.3 K) is much smaller than the observed variations. Furthermore, the apparent warmest and coldest temperatures are not coincident with the solstices, L s = 90° and 270° (see Figs. 2, S4 and S5), which would result from a purely geometric effect. In addition, the hemispheric temperature contrasts are consistent with those derived from independent studies of Voyager IRIS 15 and Cassini CIRS 16 observations (Fig. 2). The temperature variations are unlikely to result directly from radiative heating, as recent models 19 predict peak-to-peak tropospheric temperature seasonal variability of only 0.4 K or less, given Jupiter's small axial tilt. The models are also signi cantly offset in time from the measurements (Fig. S4). Time-dependent oscillations do exist in other planetary atmospheres that are related to seasonal cadences, such as Saturn's semi-annual equatorial oscillation 20 , or perhaps more loosely, the Earth's 28-29 month Quasi-Biennial Oscillation (QBO) 21 . Both are tied to narrow low-latitude regions, which might imply similar mechanisms for Jupiter, but they do not explain one other striking characteristic: the variability of these waves is anticorrelated between the northern and southern hemispheres. This can be seen at the discrete latitudes shown in Figure 2B and 2C. Examining the Pearson correlation coe cient associated with conjugate latitudes yielded negative correlations over a broad range of latitudes, the largest at 16° (Fig. 3A), as well as 22° and 30° from the equator (Fig. 3B).
This suggests teleconnected patterns, such as the Earth's El Niño Southern Oscillation (ENSO) 22  A prominent period near 8-9 years con ned to ±10° of the equator, and a fainter period of ~4 years, are both detectable at 8°S-22°N (the green and white boxes, respectively, in Fig. 1G). These tropospheric periodicities are similar to and may be related to the ~4-year equatorial stratospheric oscillation 26 . Temperature oscillations at 330 mbar ( Fig. 2A) appear to lag equatorial zonal-mean stratospheric temperature oscillations in 1980-2011 26 by ~2 years, consistent with the shorter study of 1978-2001 data 9 . This implies that the 330-mbar oscillation is compatible with "top-down" control of tropospheric temperatures by the dynamics of the stratosphere, similar to "sudden warming" events in the Earth's atmosphere 27 . If both the ~4-and 8-9-year periodicities are related to the equatorial stratospheric oscillation, the cause of the major difference in their latitudinal extent remains unresolved. Jupiter's zonal-mean winds have 13.8-year and 7.6-year variabilities 28 , both con ned to within 5° of the equator; the latter is near our low-latitude 8-9 year periodicity. A period of ~7 years is also detectable over a wide range of latitudes (dark orange box in Fig 1H). Intriguingly, although apparently unrelated to variability in the major axisymmetric bands, a 7-year period is also detected in 5-µm studies near the equator 18 , suggesting a possible correlation between temperature variability and condensate levels there.
We found several correlations between temperatures and the visual appearance of Jupiter's bands. We searched rst for correlations with quasi-periodic changes in Jupiter's belts and zones. The red horizontal bars shown in Fig. 2B indicate the duration of South Equatorial Belt (SEB, 6°S-17°S) brightening and re-darkening ("fade" and "revival") episodes. All except a very brief sequence in the rst half of 2007 coincide with the coldest periods at 16°S, which would be consistent with more aerosol condensation and visible whitening. Blue horizontal bars in Fig. 2B denote epochs when the typically dark ~6-15°N North Equatorial Belt (NEB) expanded northward to cover 18.5°N 10,16 . The periods of expansion in 2002, 2010 and 2018 appear to be coincident with prominent maxima of the NEB upper tropospheric temperatures at 16°-30° N, consistent with the removal of aerosols by heating and sublimation. Equatorial Zone (EZ) disturbances 17,18 in 1992, 1999-2000 and 2006-2007 (denoted by the black horizontal bars in Fig. 2A) appear to be contemporaneous with decreases in 330-mbar temperatures, in general agreement with a detailed study of atmospheric properties for the 2006-2007 event 29 . This cadence is also consistent with the 6-7 year period of these events 12 , even though not all of the equatorial temperature changes are tied to full-scale cloud-disturbance events. On the other hand, periods when the prominent dark North Temperate Belt (NTB, ~22°-24°N) underwent lengthy brightening and re-darkening episodes related to spectacular plume activity (not shown in Fig. 2) appear to have no correlation with temperatures or their variability, at least at the resolution of our observations.
Our study of long-term zonal-mean tropospheric temperature variability in Jupiter has yielded evidence for both non-seasonal and quasi-seasonal periodicities at temperate to tropical latitudes along with associated puzzles. Although intriguingly close to Jupiter's 11.9-year revolution, the distinct 10-14 year periodicities are unlikely to be the direct result of radiative forcing in view of Jupiter's weak seasons, particularly given the pronounced hemispherical asymmetry of temperatures peaking at 16°, 22° and 30°f rom the equator. Such an asymmetry is mirrored in cloud opacity detected by 5-µm imaging, most prominently between the NEB and SEB 18 . The presence of 4-and 8-9-year periodicities suggests a relationship with stratospheric temperature variability. More detailed correlation between their periodicity and phase is needed to validate that connection, particularly to test the suggestion of "top-down" mechanisms, such as descending waves. The 7-year periodicity over a broad latitude range may also be related to the quasi-periodic equatorial disruption with the same cadence 18 . Although we found no straightforward correlations between periods and the latitudes of Jupiter's belts and zones, correlations with known global-scale changes of cloud morphology suggest at least some thermal modulation of aerosol condensation and sublimation cycles, which deserves more detailed quantitative scrutiny.
Realistic global climate models for Jupiter must address the origins of these unexpected seasonal and non-seasonal periodicities on a virtually aseasonal Gas Giant in preparation for their eventual extension to a wider array of gas-giant planets outside our solar system. Author contributions: Orton wrote all of the main text and most of the Supplementary Information, was responsible for the general organization, led many of the observing runs and the initial reduction of the observations. Antuñano organized the observations from the original measurements, performed the calibrations, executed the temperature retrievals and wrote a part of the Supplemantary Information. Fletcher guided the spectral retrieval methodology, led many of the observing runs and the initial reduction of the observations, and helped to draft the manuscript. Sinclair, Momary, Fujiyoshi, Yanamandra-Fisher and Donnelly constituted part of the teams making the observations since 2002. Greco, Payne, Boydstum and Wake eld were interns at the Jet Propulsion Laboratory, and were responsible for examining the consistency of the calibrations, testing stellar calibrations, and testing retrieval approaches on subsets of the data addressed here. All authors reviewed and commented on the manuscript. We acknowledge the work of those who made observations prior to 2002 and the authors of previous work 5,9 that inspired this study. showing all results exceeding a 1% false-alarm probability. Detection of important but fainter features in the power-spectra panels is enabled by raising them to the 0.7 power. The power spectra in panels G and H are identical, but in Panel H we identify period that are discussed in the text. The white box bounds the latitude and period range ~4 years, the dark orange box ~7 years, the green box ~8-9 years, the light orange boxes 9-12 years, the yellow boxes ~10-14 years, and the red box 11-14 years. There are no signi cant periods longer than 14 years. Figure 2