Jupiter’s Atmospheric Variability from Long-term Ground-based Observations at 5 μm

In this study, we use images captured between 1984 June and 2018 July at wavelengths between 4.76 and 5.18 μm by eight different ground-based instruments mounted at the Infrared Telescope Facility (IRTF) on Maunakea, Hawai’i, and at the Very Large Telescope (VLT) in Chile. The 5-μm radiation, which originates in Jupiter’s tropospheric cloud decks, mainly reveals the thermal emission from the 1 to 4 bar level (below the ammonia ice clouds). The use of a single narrowband filter does not allow us to distinguish reflected sunlight from thermal emission, but the reflected sunlight makes only a small contribution to the 5-μm radiance (Giles et al. 2015). At this wavelength, cloudy regions appear dark at 5 μm, while cloud-free regions appear bright, due to the high aerosol opacities blocking the light. A summary of the data set used is shown in Table 1, and a brief description of the eight instruments used is given below:

• 1.  
  BOLO-1: The 2–30 μm Bolometer (“BOLO”) mounted at the IRTF on Maunakea, Hawaii, provided observations of Jupiter between 1980 and 1993, first by central meridian scans and later by raster scans, sampling at 1″ intervals in R.A. and decl. (Orton et al. 1994). In this study, M-band filter (4.80 μm) data covering Jupiter’s full disk are used to expand Jupiter’s brightness variability analysis to the 1980s (1984 June–1991 February). These data have not previously been studied, and we followed the same techniques as in Orton et al. (1991, for the stratosphere) and Orton et al. (1994, for the troposphere).
• 2.  
  ProtoCam: The infrared array camera ProtoCam was a 62 × 58 pixel InSb camera (Toomey et al. 1990) mounted at the 3-m IRTF in Hawaii and was sensitive to wavelengths between 1 μm and 5 μm. Its plate scale varied between 014 and 035, and due to the low number of pixels on the CCD, images of Jupiter’s whole disk had to be acquired by mosaicking (Harrington et al. 1996a, 1996b). Mosaicked ProtoCam data captured at 4.85 μm from 1992 January–April (Harrington et al. 1996a, 1996b) are used in this study to analyze, in particular, Jupiter’s EZ disturbance (Antuñano et al. 2018).
• 3.  
  NSFCam and NSFCam2: The NSFCam (Shure et al. 1994) was a 1–5.5 μm camera with a 256 × 256 InSb detector mounted at the IRTF between 1993 September and 2004 April. This camera captured Jupiter images with a very high temporal resolution, enabling us to analyze the brightness variability at the cloud-forming region in 4.78 and 4.85 μm in detail between 1994 April and 2004 February, with pixel scales of 015 and a field of view of 379 (smaller than Jupiter’s disk at opposition). In 2004, the NSFCam imager was upgraded with a 2048 × 2048 Hawaii 2RG detector array, giving way to NSFCam2. The pixel scale of the NSFCam2 is 004/pixel with a field of view of 81 × 81″, providing images with higher spatial resolution of Jupiter’s full disk. In this study, NSFCam2 images captured at 4.78 μm between 2006 April and 2008 October are used. These two cameras, like the SpeX, MIRSI, and VISIR instruments described below, captured data in a chopping-plus-nodding mode, where at each nod position a chop of the secondary mirror by a few arcseconds is performed in order to subtract the background and sky emission and detect Jupiter’s emission (Fletcher et al. 2009b).
• 4.  
  MIRSI: The Mid-Infrared Imager and Spectrometer (MIRSI; Deutsch et al. 2003) is a 2–20 μm camera and grism spectrograph with a 320 × 240 Si:As array detector. Its plate scale of 027/pixel provides a field of view of 85 × 64″, where Jupiter’s full disk can be captured, albeit with a lower spatial resolution than data obtained with NSFCam, NSFCam2, SpeX, TEXES, and VISIR. In this study, MIRSI images captured between 2005 January and 2006 April at 4.9 μm are essential to be able to fill the temporal gap between the higher resolution images captured by NSFCam and NSFCam2.
• 5.  
  SpeX: The SpeX 0.7–5.3 μm spectrograph (Rayner et al. 2003), mounted at the IRTF in Hawaii since 2000 May, is used in this study for images captured at 4.76 and 5.1 μm between 2009 and 2018 July. This guide camera, which was upgraded in 2014, used a Raytheon Aladdin 3 1024 × 1024 InSb array in the spectrograph with a pixel scale of 015 before the upgrade and a Teledyne 2048 ×2048 Hawaii 2RG array with spectrograph pixel scale of 012 after the upgrade.
• 6.  
  TEXES: The Texas Echelon Cross Echelle Spectrograph (TEXES; Lacy et al. 2002) is a cross-dispersed grating spectrograph mounted on the IRTF able to obtain spectra between 5 and 24 μm. In this study, we use data captured at the M band (5 μm) between 2014 December and 2018 July using the medium (R ∼ 15,000) and low (R ∼ 2000) spectral resolutions. Unlike the rest of the instruments’ data used in this study, these maps are produced by scanning the TEXES slit repeatedly across Jupiter’s disk. Detailed technical information on the instrument and the observations can be found in Fletcher et al. (2016).
• 7.  
  VISIR: The Very Large Telescope (VLT) Imager and Spectrometer for the mid-infrared (VISIR; Lagage et al. 2004) provides diffraction-limited imaging between 5 and 20 μm with a pixel size of 00453/pixel and a 38 × 38″ field of view, smaller than Jupiter’s disk at opposition. In this study, we use the data captured at 5 μm between 2016 December and 2017 July (Fletcher et al. 2018).

The data mentioned above were acquired under a vast array of weather conditions. In order to omit the poor-quality observations in our large data set, we analyze each of the data carefully, removing those of particularly poor quality. When data were acquired under average seeing conditions of ∼1″, the minimum spatial resolution of our data on Jupiter’s equator is ∼2850 km (23 longitude) at opposition and ∼3800 km (31 longitude) at quadrature (corresponding to the 3-m IRTF data because the 8-m VLT data display a higher spatial resolution), sufficient to perform a detailed analysis of the temporal variability of the cloud morphology and dynamics of Jupiter’s zone and belts.

Images captured with NSFCam, NSFCam2, MIRSI, SpeX, and VISIR are reduced following Fletcher et al. (2009b), by subtracting the sky emission using the chop-nodded images, flat-fielding, navigating them by limb fitting, and projecting the data in cylindrical maps. TEXES data are reduced using the Pipeline Data Reduction software (Lacy et al. 2002), which performs the sky subtraction, flat-fielding, and radiometric calibration automatically. Information for the raster-scanning method used in BOLO-1 observations is described in Orton et al. (1994). Finally, we use the mosaicked ProtoCam data from Harrington et al. (1996a, 1996b). We made no attempt to calibrate the data because of the very different atmospheric conditions of the observations made during almost 40 yr, which often were carried out in the absence of a reliable standard star. However, a relative calibration, described in Section 2.3, is performed in order to be able to compare the data from different epochs.

When analyzing the temporal variability of the latitudinal brightness of Jupiter’s atmosphere at 5 μm, it is essential to take into account that the brightness is not homogeneously distributed in longitude, varying with emission angle and displaying limb-darkening. Therefore, understanding the dependence of the observed brightness on the emission angle is essential to choosing the right average value for each latitude and date. To do so, we follow Antuñano et al. (2018) and compute the average brightness by binning the brightness corresponding to emission angles smaller than 75° of all images captured in a single observation night, in latitudinal bins of 1° between ±70° latitude. Brightnesses corresponding to larger emission angles are excluded from the zonal mean in order to avoid most of the effect of limb-darkening.

This technique is then compared to the average brightness obtained by performing a second-order polynomial fitting of the brightness versus emission angle at each latitude and date, where the average (fitted) brightness corresponds to the smallest emission angle (i.e., the brightness value less affected by the dependency of the emission angle). Once the average brightness is scaled to a quiescent region (see below), both techniques show a similar temporal variability of the brightness, with only small differences of less than 5% of their peak values.

As mentioned above, due to the very different atmospheric conditions at which the data were captured and the large temporal coverage of the data set used in this study, we did not calibrate the data before computing the average brightness. However, in order to be able to compare Jupiter’s brightness at 5 μm from different epochs, a scaling of the average brightness to a quiescent latitudinal region must be performed (Fletcher et al. 2011, 2017b). As Jupiter’s zones are usually less variable than belts, they are better scaling regions, although the lower signal-to-noise ratio could be troublesome. Previous studies scaled Jupiter’s brightness to the typically dark EZ (Fletcher et al. 2011, 2017b). However, the EZ disturbance observed at 5 μm and reported by Antuñano et al. (2018) completely changes the appearance of the EZ during these events, making this region unsuitable for this purpose.

In order to identify the optimal scaling region (i.e., the most quiescent zone on Jupiter during the epochs under study), we first omitted the data corresponding to the dates where an EZ disturbance is present (i.e., 1992 January–April, 1999–2000 August, and 2006 April–2007 August). We then scale the average brightness to the temporal average brightness of the equator (±5° latitude of the equator, which is known to be a 5-μm-dark region away from the EZ disturbances) by applying the following equation:

$$\begin{eqnarray}&&{L}_{\phi ,t}={\mathrm{counts}}_{\phi ,t}\displaystyle \frac{\lt d\gt }{d}\end{eqnarray} \tag{ 1 }$$

where L_ϕ,t is the scaled brightness at each latitude ϕ and date t, counts_ϕ,t is the zonal mean brightness at each latitude and date, d is the averaged brightness over the range of latitudes ±5°, and the < > indicates the average over all dates.

The results are shown in Figure 1(A), where the normalized average brightness is shown against latitude (different colors corresponding to different dates and the shadowed regions indicating Jupiter’s zones) for the dates where no EZ disturbance at 5 μm was present. In Figure 1(D), we show the normalized mean-centered brightness as a function of time, showing Jupiter’s 5-μm appearance when scaling the zonally averaged brightness to the equator. The normalized mean-centered brightness allows us to identify the 5-μm changes more clearly, and it is computed by subtracting the mean of zonally averaged brightness to the data and then normalizing the maximum value to 1 (−1 in the case of negative values). Here, red represents 5-μm-bright regions (belts), while blue represents 5-μm-dark regions (zones). Figure 1(A) demonstrates the key challenge of this study: no zone or belt remained unchanged over the 34 yr under study, making it difficult to choose the right scaling region. However, the South South South Temperate Zone (S3TZ) at ∼46–48° S and the South Temperate Belt (STB) at ∼24–28° S underwent the smallest temporal variability, with a relative brightness variability smaller than 5% and 10%, respectively. A careful analysis of various potential scaling regions (e.g., the S3TZ, the STB, mid-latitudes above 50° N, and the NEB, among others) show that the belts, as well as the mid-latitudes above 50° N, are not good scaling regions because the obtained scaled and normalized brightness displays a strong dependence on these scaling regions, changing the scaled-normalized brightness completely from one case to the next. Therefore, here we analyze the S3TZ and the STB as potential scaling regions.

Zoom In Zoom Out Reset image size

Figures 1(B) and (C) show the normalized average brightness scaled to the STB and S3TZ, computed by using Equation (1), with d averaged between 24° and 28° S and 46° and 48° S, respectively. Figures 1(E) and (F) show the mean-centered brightness as a function of time. Note that the normalization of the mean-centered brightness is performed independently for each of the scaling regions. The STB scaling region seems to be a good choice for all of the dates except 2018, where the scaling drops the average brightness of 2018 significantly (see Figure 1(B)). This is due to an unusual 5-μm-bright STB observed in the 5-μm data from 2018, potentially related to the convective storms that erupted at the cyclonic region of the STBn known as the STB “Ghost,” leaving a very turbulent region (Inurrigarro et al. 2019). On the other hand, the S3TZ seems to be a valid scaling region at every epoch except 1992, where the Protocam data show a bright S3TZ not observed in other dates. Figures 1(B) and (C) show that both scaling regions give similar results over most of the studied epochs, showing Jupiter’s temporal overall variability. However, note that the intensity of the normalized average brightness varies with the chosen scaling region, due to the normalization performed, where we normalized the maximum average brightness to unity for each of the scaling regions. Therefore, we advise the reader to be cautious when comparing the intensity of the zones and belts.

Finally, in order to analyze the robustness of the results shown in this study, we perform the same analysis using each of these three scaling regions independently (i.e., EZ for the dates without a EZ disturbance, STB, and S3TZ), and we then compare the results. We also analyzed different potential scaling regions. All figures in Sections 4 and 5 are computed by scaling the average brightness to the S3TZ (i.e., 46–48° S; see the supplemental material for the results obtained for scaling at the STB and at the EZ).

The Lomb–Scargle periodogram (Scargle 1982) is a widely used tool in the study of planetary science that characterizes periodic signals in unevenly sampled data, by computing least-squares fits with sinusoidal functions. Previous studies of Jupiter have mainly used this technique to study possible wavenumbers of diverse tropospheric and stratospheric waves (e.g., Harrington et al. 1996b; Ortiz et al. 1998; Arregi et al. 2006; Barrado-Izagirre et al. 2009; García-Melendo et al. 2011; Simon-Miller et al. 2012), although it has also been used to analyze possible periodicities in the changes of the tropospheric zonal wind structure (e.g., Simon-Miller & Gierasch 2010; Tollefson et al. 2017), among others. In this study, we use the Lomb–Scargle analysis to study possible periodicities in the brightness variability of Jupiter’s atmosphere below the ammonia cloud level, at the 1–4 bar region.

In particular, we use the “Scargle” function written in the IDL. This algorithm computes possible periodicities automatically and allows the setting of a desired false-alarm probability to control the significance of the different periodicities returned by the algorithm. In this study, we compute the Lomb–Scargle periodogram analysis for each latitude by setting a false-alarm probability of 0.001, or 99.99% significance, and a minimum period of 1 yr periodicity. The uncertainty of the obtained periods is assumed to be equal to the FWHM of the power spectrum peak. Results are shown in Section 5. However, we urge caution in the interpretation of the results as the Lomb–Scargle analysis expects sinusoidal changes in the data, something not observed in our data.

In order to validate the results obtained from the Lomb–Scargle analysis and to be able to discard possible spurious results, we also analyze possible periodicities of the 5-μm brightness variability by performing a Wavelet Transform analysis. This is a powerful tool for analyzing temporal variations within a time series that enables determination of the dominant modes of variability and their temporal change (Torrence & Compo 1998). This analysis has been widely used in geophysics because, unlike a Fourier analysis or Lomb–Scargle periodogram that expands functions in terms of sines and cosines, the wavelet analysis expands functions in terms of a set of functions (called wavelets) that are built by translating and dilating a fixed mother wavelet chosen by the user (Lee & Yamamoto 1994).

The continuous Wavelet Transform of a real signal s(t) with respect to a mother wavelet ϕ(t) is given by the convolution of the real signal with a set of wavelets:

$$\begin{eqnarray}&&S(b,a)=\displaystyle \frac{1}{\sqrt{a}}{\int }_{-\infty }^{\infty }\phi ^{\prime} \left(\displaystyle \frac{t-b}{a}\right)s\left(t\right){dt}\end{eqnarray} \tag{ 2 }$$

(Lee & Yamamoto 1994), where ϕ′ is the complex conjugate of ϕ, a is the scale factor (dilating value) that controls the width of the window and the oscillation period of the real signal, and b is the time shift (translation) of the mother wavelet along the time axis. The correct choice of the mother wavelet is essential, as this choice influences the time and frequency resolution of the results. In this study, we use the Morlet wavelet as the mother wavelet, which consists of a plane wave modulated by a Gaussian (Chapa et al. 1998; Huang et al. 1998), and it is given by

$$\begin{eqnarray}&&\phi \left(t\right)={\pi }^{-1/4}{e}^{i{\omega }_{0}t}{e}^{-{t}^{2}/2}\end{eqnarray} \tag{ 3 }$$

where ω₀ is a wavenumber. This is the most commonly used mother wavelet as it is very well localized in frequency, where the scale factors nearly represent the Fourier period of the real signal (the period obtained by computing a Fourier Transform to the Morlet wavelet of a specific scale “a”; Torrence & Compo 1998).

In this study, we compute the Wavelet Transform analysis using the function “wv_cwt.pro” written in IDL. This code dilates and translates the mother wavelet given in Equation (3) to compute a set of Morlet wavelets with different scale values and then performs the convolution of the normalized average brightness and each wavelet. Because the wavelet analysis requires the data to be evenly spaced in time, we interpolate the normalized average brightness by a linear interpolation, and then apply a boxcar average of 10 to smooth the interpolated, normalized average brightness. Additionally, in order to minimize the errors introduced at the boundaries of our study, we set the keyword PAD, which pads the normalized average brightness with enough zeros to make the total length of the normalized average brightness array equal to the next-higher power of 2. Finally, we compute the wavelet power spectrum as $| S\left(b,a\right){| }^{2}$. As in the Lomb–Scargle analysis, the uncertainty of the obtained periods is assumed to be equal to the FWHM of the wavelet power spectrum peak. The results of this analysis are shown in Section 6.

With the aim of analyzing possible correlations or anticorrelations in the 5-μm changes at different zones and belts, we also compute a PCA. This technique decomposes a data set into a set of basis vectors or principal axes that recognize the major sources of variance within the data set (e.g., Jolliffe 2002). With this technique, each point in a data set can then be represented in terms of its components (i.e., principal components) along each of the principal axes. The advantage of this technique is that, in practice, the first few axes represent the majority of the variability within the data set, and less important axes can be ignored, reducing the dimensionality of the data.

In this study, we perform PCA following the next steps. First, we compute the mean of all of the 5-μm data used in this study to obtain a vector that represents the center of the distribution of points in n-dimensional space. This mean is then subtracted from the data, such that the data set is now “mean-centered.” This allows us to analyze the variability of the 5-μm brightness with respect to the averaged state of Jupiter. Once the mean-centered brightness is obtained, we create an n × m matrix (where n = 141 is the number of latitudes, and m corresponds to the number of dates in our data set), which is then multiplied by its transpose to form the n × n covariance matrix of the data. From this, the n principal axes are computed as the eigenvectors of the covariance matrix, and the corresponding eigenvalues represent the contribution of each axis to the variance within the data set. The principal components of each profile are the inner products of the corresponding vector with each of the principal axes.

The real data set can then be reconstructed exactly as a linear combination of all of the principal axes, or approximated by a linear combination of just the most important principal axes. A choice must be made as to the number of principal axes to retain, which depends on the amount of variance that should be represented in the approximation. In this study, we analyze the first three principal axes, as they display most of the major changes observed at 5 μm and described in Sections 4 and 5. The results of the PCA are shown in Section 6.

The NTR, located between 21° N and 31.5° N, is formed by the NTB (i.e., 21–28° N) and by the North Temperate Zone (NTZ) at 28–31.5° N. Here we also include the North North Temperate Belt (NNBT, i.e., 31.5–35.5° N). These regions have been observed to undergo large temporal variations since the early 20th century, where the NTB fades, revives, and drifts in latitude (Peek 1958; Sánchez-Lavega et al. 1991, 2008, 2017; Rogers 1995).

In Figures 3(A) and (B) we represent the 5-μm brightness temporal variability of the NTR between 1984 April and 2018 July as a function of latitude and time, showing the large temporal variability of the 5-μm emission at the cloud-forming region during those epochs. In Figure 3(B), we show the relative variability of brightness at each latitude with respect to their darkest state, computed by comparing the normalized brightness at each latitude and date to the temporal mean of the smallest 10% of brightness values of each latitude. This technique allows us to show the brightness variability at each latitude independently in the same figure, with the changes of the darker latitudes not being suppressed by the changes of the brighter latitudes. In this figure, the zero corresponds to the temporal mean of the smallest 10% of brightness values of each latitude. From these figures, we see that this region underwent at least four kinds of major changes (labeled by numbers in Figures 3(A) and (B)) in its brightness contrast in the last 34 yr. The first major change (labeled as 1 in Figures 3(A) and (B)) is observed between 1992 January and April, where the NTR displays a wide increase of its brightness between 23° and 30° N, reaching its maximum brightness at 26° N. During this epoch, the NTB seems to be moved poleward, brightening part of the usually dark NTZ. The formation of this belt is quite unique, as it is the widest belt developed in the NTR between 1984 and 2018, as shown in Figure 3(a). The lack of infrared data between 1991 February and 1992 January prevents us from analyzing the formation of this belt, but it might be related to the eruption of two outbreaks at the zonal jet at ∼21° N observed in early 1990, which started a revival of the previously faded NTB, creating low- and high-albedo regions at the southern edge of the NTB (NTB(S); Sánchez-Lavega et al. 1991; Rogers 1992; García-Melendo et al. 2005).

Zoom In Zoom Out Reset image size

Additional major changes are observed in 1996, 1998–1999, and 2001, when a notable increase of the 5-μm emission is observed at the NTB(S), shown in Figures 3(A) and (B) by green and yellow colors and labeled as number 2. Unlike the formation of the NTB observed in 1992, these brightness increases were not preceded by outbreaks at the NTBs. During these epochs, the normalized brightness of the NTB(S) and of the NNTB at 31–35° N seems to increase and decrease contemporaneously, in time intervals of two years, as shown in Figure 3(B) (blue colors represent the NTB(S) and red/black colors represent the NNTB), suggesting that the NTB and the NNTB are somehow connected to each other. The 5-μm images of these epochs show that at the times of normalized average brightness maxima in this region, the NTB appears as a narrow, bright region at 21–25° N, followed by a dark region between ∼25° N and ∼30° N (NTB(N) and NTZ) and accompanied by a non-homogenous, bright NNTB between ∼30° N and ∼34° N. An example of this is shown in Figure 4(a). At the epochs of the brightness minima, a slightly bright narrow band is observed at 21–23° N. These 5-μm changes were not apparent at visible wavelengths, where the NTB appeared normal (i.e., broad and dark gray) during those years.

Zoom In Zoom Out Reset image size

During 1997, 2007, and 2016, the NTR displays a completely different appearance at 5 μm, with a dark (cloud-covered) region between 21° N and 25° N (blue crosses in Figure 3(B)) and a 5-μm-bright region spanning from the usually dark North Temperate Zone up to the NNTB (i.e., 28–34° N, orange and red crosses in Figure 3(B)), forming a new North Temperate Zone Belt (NTZB), as shown by the number 3. Figure 4(b) shows that during those epochs the NTZB presents a chain of 5-μm-bright ovals, not present at the times where the NTZ is observed dark at 5 μm. The formation of this NTZB seems to occur in time intervals of 8–10 yr, between 1984 and 2018, in agreement with previous studies (Rogers 1995; Fletcher 2017), with a missing NTZB in the BOLO-1 data from 1987. The exact periodicity of these changes is analyzed in Section 6.

Finally, during 1984–1989, 2003–2005, 2009, and 2011–2014, the entire region was mainly dark at 5 μm, displaying a remarkably low brightness at all latitudes (number 4 in Figure 3(B)). Jupiter images at 5 μm from those dates show that the NTB was partially faded (5-μm-dark), with some 5-μm-bright spots of various sizes found at different latitudes intermittently, especially during 2003–2005 and 2009–2012 (see Figure 4(c)). The lower resolution of the BOLO-1 data between 1984 and 1989 prevents us from observing any bright spots at the NTR. These fading events do not seem to be related to the outbreaks that occur at ∼21° N creating a planetary-scale disturbance known as the North Temperate Belt Disturbance (NTBD; Peek 1958; Sánchez-Lavega et al. 1991, 2008, 2017; see Table 2 where a list of observed NTB outbreaks is given). Finally, the contemporaneous fading of the NTB and the NNTB suggests that these two belts are connected to each other. Both the fading of the NTB and NNTB and their lifetimes appear to be random with no clear periodicities. This is discussed further in Section 6.

The NTrR, formed by the usually brownish (5-μm-bright) NEB at 7–15° N and by the typically 5-μm-dark NTrZ at 15–21° N, displays a large variety of cloud morphology and wave phenomena that change the appearance of this region, both at the cloud tops (∼700 mbar) and at the cloud-forming region at 1–4 bar (Rogers 1995; Fletcher et al. 2017c). Visible observations have shown that the boundary between the NEB and the NTrZ varies in latitude from year to year, due to cloud-clearing events at the NTrZ, called NEB expansions (NEEs) that take place every 3–5 yr (Rogers 1995, 2017b; García-Melendo & Sánchez-Lavega 2001; Simon-Miller et al. 2001; Fletcher et al. 2017c; Rogers & Adamoli 2019).

In Figure 5, we represent the normalized average brightness of the NTrR, between 7° N and 20° N over the same 34 yr as discussed for the NTR (Section 4.1). As in Figure 3(A), different colors indicate different latitudes in Figure 5(A). In Figure 5(B), we represent the normalized brightness residuals by showing the deviation of the normalized brightness from the temporal mean of the smallest 10% brightness values of each latitude. We observe that the southern edge of the NEB, between 7° N and ∼9° N (represented by dark blue crosses in Figure 5(A)), displays the largest brightness intensity of the entire region, potentially due to the absence of the 5-μm-dark ovals and rifts (turbulent convective regions) that are usually present at higher latitudes (Rogers 1995) and the presence of the 5-μm hot spots located at ∼7° N (Keay et al. 1973; Terrile & Westphal 1977; Beebe et al. 1989; Rogers 1995; Ortiz et al. 1998; Hueso et al. 2017). Poleward of these latitudes, the brightness contrast decreases with latitude, reaching its minimum value at the northern edge of the NTrZ.

Zoom In Zoom Out Reset image size

The temporal variability of the 5-μm emission due to the latitudinal expansions or contractions of the NEB (i.e., cloud clearing or covering at the southern edge of NTrZ) is clearly observed in our results, where the brightness between 15° N and 18° N is observed to increase and decrease significantly every 3–5 yr (see arrows in Figures 5(A) and (B)), in agreement with previous studies (Rogers 1995, 2017b; Rogers et al. 2004; Fletcher et al. 2017c). The latitudinal coverage of the NEB expansion (i.e., brightness increases of the NEB(N) and NTrZ(S)) is observed to vary from event to event, with the events from 1989, 1992, and 2001 expanding from 7° N to ∼20° N (see Figure 6(b)), while the 1996, 2006–2007, 2009–2010, 2012, and 2015–2017 events expanded from 7° N to ∼18° N. This is in agreement with the latitudinal coverage of the NEB expansion from 2015 to 2016 given by Fletcher et al. (2017b, 2017c). Interestingly, during the events from 1992, 2007–2008, and 2015–2017, not only was the NEB observed to be expanded, but it was also observed to be brighter than usual and also brighter than the rest of the NEB expansions. This suggests that the NEB expansions differ from one event to the next, with some producing “stronger” cloud clearings of the NTrZ than others.

Zoom In Zoom Out Reset image size

Finally, during 1984–1988, 1995, 1998, 2005, 2009, and 2012–2013, the boundary between the NEB and the NTrZ at 5 μm seems to be moved equatorward, varying between 10° N and 15° N from event to event. During these epochs, the NEB was not only narrower but also darker at 5 μm than during the expansions. This is particularly remarkable in 2005 and 2011–2012, where Figure 5(B) shows that during these events the NEB displayed the lowest brightness in the 34 yr (i.e., the brightness did not deviate from the temporal average brightness of the smallest 10% brightness values). Additionally, in late 2011 and early 2012, the NEB seemed to be bright at 5 μm only between 7° N and 10° N (see Figure 6(a)), in agreement with the analysis of the NEB fade in 2012 at visible wavelengths reported by Rogers (2019), where the NEB seemed to be the narrowest and faintest it had ever been since 1920.

Zoom In Zoom Out Reset image size

Jupiter’s EZ, covering latitudes within 7° of the equator, displays unique atmospheric phenomena at its northern and southern boundaries, where regions of depleted volatiles and aerosols known as 5-μm hot spots (Keay et al. 1973; Terrile & Westphal 1977; Beebe et al. 1989; Ortiz et al. 1998; Wong et al. 2004; de Pater et al. 2016; Fletcher et al. 2016; Hueso et al. 2017; Rogers 2017b) and small, periodic chevron-like features (Simon-Miller et al. 2012) are observed at its northern and southern edge, respectively. Additionally, a recent study by Antuñano et al. (2018) has shown that the usually 5-μm-dark EZ undergoes strong variations every 6–8 yr, where its appearance changes completely from the darkest region on Jupiter to a region with a bright band south of the equator and large, bright filaments connecting the NEB and this new bright band (see Figure 8(a)).

Zoom In Zoom Out Reset image size

In Figure 7(A) we represent the 5-μm brightness of the EZ, expanding Figure 2(c) in Antuñano et al. (2018) to show the new data between 2017 August and 2018 July. In Figure 7(B) the normalized brightness residuals are shown. Both figures show the rapid 5-μm brightness increases during the EZ disturbances of 1992, 1999–2000, and 2006–2007. Figure 7(B) also shows strong brightness increases and decreases at 6–7° N, related to the presence or absence of the 5-μm hot spots. Most of these brightness increases seem to be related to periods when an EZ disturbance was occurring. However, a strong increase is also observed at 6–7° N in 1990, 2013, and 2015–2017 when no EZ disturbance was observed. Finally, Figures 8(b) and (c) show that, as predicted by Antuñano et al. (2018), a new EZ disturbance has been developing since 2018 August, with the latest images (2019 February) suggesting that the EZ disturbance is not yet completely developed.

The STrR, formed by the SEB at 6°–17° S and by the STrZ at 17°–25° S (Rogers 1995), undergoes the most spectacular variations of all the regions, with the SEB changing in a very turbulent way from being one of the broadest dark brown belts of Jupiter to a whitish zone-like appearance (e.g., Peek 1958; Sánchez-Lavega 1989; Rogers 1995; Sánchez-Lavega & Gómez 1996; Fletcher et al. 2011, 2017b; Pérez-Hoyos et al. 2012).

In the last 34 yr, four fade-and-revival cycles have been observed at the SEB in (1) 1989–1990 (Yanamandra-Fisher et al. 1992; Kuehn & Beebe 1993; Satoh & Kawabata 1994), (2) 1992–1993 (Sánchez-Lavega & Gómez 1996; Moreno et al.1997), (3) 2007 (Baines et al. 2007; Reuter et al. 2007; Rogers 2007a, 2007b), and (4) 2009–2011 (Fletcher et al. 2011, 2017b; Pérez-Hoyos et al. 2012; Rogers 2017a, 2017c). These fading episodes of the SEB are also observed in our analysis of the 5-μm brightness variability, indicated by strong decreases in the brightness during 1989–1990, the second half of 2007, and 2010, due to the formation of ammonia ice clouds in this region. The gap in the 5-μm data between mid-1992 and early 1994 prevents us from observing the 1992–1993 fading event. However, our data show a low 5-μm brightness in 1994, right after the SEB revival, corresponding to a quiescent period of the SEB that seems to occur rapidly after the SEB revivals (Rogers 2017a). All these brightness decreases are shown in Figure 9 indicated by black arrows. An example of what Jupiter looked like during the 2010 SEB fading event is shown in Figure 10(a). In the case of the 2007 fading event, the 5-μm brightness at 7°–9° S did not decrease as much as during the other fading events (where all the latitudes present a strong decrease in their brightness), indicating that in 2007 the SEB did not fade completely, displaying a bright SEB(N) (see Figure 10(b)). This is in agreement with observations reported in Rogers (2007a), where a violent revival was observed before the SEB was completely faded. Note that the brightness decrease observed in 2005 in Figure 9 depends on the scaling region, and it is not related to a faded event of the SEB (see the Appendix for the brightness scaled at the STB).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The normalized brightness residuals shown in Figure 9(b) show that the brightness of the SEB varies from date to date when it is not faded. In particular, it is observed that at some particular epochs (1996–1998, 2000–2001, 2008, 2015–2016) the SEB displayed bright regions at ∼8°–12° S and 16°–18° S, separated by a low-brightness and cool region named the South Equatorial Belt Zone (SEBZ; Fletcher et al. 2011), while in 1998–2000, 2004, and 2014, the SEB was observed to be bright continuously between 8° S and ∼18° S. Figures 10(c) and (d) show two examples of how the SEB looked during these epochs. Periodicities of these changes are analyzed in Section 6.

The STR comprises the STB at 25–29° S and the South Temperate Zone (STZ) at 29–32° S, and we also include the South South Temperate Belt (SSTB) at 32–35.5° S. As shown in Figure 2, overall this region displays a completely different appearance at visible and 5-μm wavelengths between 1984 and 2018, when the belt and zone locations at visible wavelengths do not follow the 5-μm-bright and -dark regions. An example of this is the STB, which, as described in Section 2.3 and shown in Figure 1, is one of the darkest regions of Jupiter at 5 μm (after the EZ), with a very small temporal variability.

In Figure 11 we represent the normalized brightness of the STR as a function of time and the normalized brightness residuals. Overall, the entire region displays a low 5-μm brightness over the 34 yr under study, and compared to the other regions described in this section, the STR is one of the least variable. The largest brightness variability is found at the SSTB (blue crosses in Figure 11(A)), whose brightness increases and decreases gradually. Figure 12 shows two 5-μm images of Jupiter from 2013 to 2016 March, showing that the SSTB sometimes displays a number of small 5-μm-bright ovals (not related to the regular white anticyclones at 40° S; Rogers 1995; de Pater et al. 2010). These ovals are not present at all of the longitudes, making the belt artificially bright at the dates where the observations capture them. Because the normalized brightness shown in Figure 11 corresponds to the zonally averaged brightness of each date, the increases and decreases of the SSTB observed in our results need to be interpreted with caution, because dates where a complete global map of Jupiter is not available would result in a totally different zonal mean brightness than for dates where we do have global maps available.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 14 shows the Lomb–Scargle periodogram ((A)–(C) panels; Scargle 1982) and the Wavelet Transform analysis periodogram ((D)–(F) panels) of the 5 μm brightness variability between 1984 April and 2018 July, for the three scaling regions analyzed in this study. Below we analyze the most significant peaks with a false-alarm periodicity of 0.01, or a confidence greater than 99.99%, that remain invariant irrespective of the scaling region and the analysis type (i.e., appear both in the Lomb–Scargle periodogram and the Wavelet Transform analysis for the three scaling regions with similar periods). These are indicated by numbers and represent the most trustworthy periodicities. Periodicities larger than 9 yr are not discussed in this study as they strongly depend on the scaling region and technique. The Lomb–Scargle periodogram and the Wavelet Transform analysis are described in Sections 2.4 and 2.5, respectively. The uncertainties on the periodicities shown below are assumed to be equal to the FWHM of the power spectrum peak.

Zoom In Zoom Out Reset image size

Overall, all of the indicated peaks appear at periods between 4 and 8 yr. A peak at 8.5 ± 0.5 yr is observed at 30–33° N, in the southern edge of the NNTB and the northern edge of the NTZ. As described in Section 4.1, these variations correspond to the northward motion of the NTB, brightening the usually dark NTZ to form a continuous belt that spans the latitude range between the NTB and the NNTB. Rogers (1995) described these kind of events observed during the 20th century and reported a ∼10 yr periodicity, similar to the one observed in this study. This periodicity does not seem to be related to NTB outbreaks at 21° N, where a 4.8 ± 0.6 yr period is found in this study (number 2). The convolution between the brightness data at 32° N and the selected wavelet with an 8.5 yr periodicity, represented in Figure 15, shows a high correlation between 1995 and 2018, confirming the 8.5 yr periodicity obtained in this study. However, this is not true between 1984 and 1995, where our periodogram analysis suggests a brightness increase at the NTZ in 1990, which is not observed in the BOLO-1 data. Nevertheless, this might be due to the lower spatial resolution of the BOLO-1 data (raster maps with a 1″ step and a circular field of view of 2″, rather than 2D images) compared to the newer data. Our analysis suggests a new 5 μm brightening of the NTZ in 2024–2025. At 21–22° N, there is no correlation between the convolution and the real data (see Figure 15), implying that the derived 4.8 ± 0.6 yr periodicity is not real and should not be taken into account. This result is particularly intriguing as NTB outbreaks tend to occur every ∼5 yr and implies that NTB outbreaks do not directly change the 1–4 bar level. This is also observed in Figure 3, where no particular change is observed at 5 μm at the epochs of NTB outbreaks. A continuation of the 5-μm brightness time series will allow us to determine the cyclicity of this perturbation more accurately.

Zoom In Zoom Out Reset image size

Peaks of 4.4 ± 0.4 and 4.5 ± 0.8 yr are observed at the NEB (number 3 in Figure 14) and at the SEB (number 4), respectively. The former is in agreement with Fletcher et al. (2017c), who suggested that the NEB expansions occur every 3–5 yr. Tollefson et al. (2017) and Simon-Miller et al. (2007) also observed a 4–5 yr periodicity in the zonal wind profile at 18° N and ∼10° S, respectively, in agreement with the 4.4 ± 0.4 and 4.5 ± 0.8 yr found at the NEB and SEB in this study, as changes in the 5-μm brightness (e.g., cloud clearing or formation of higher clouds, thus brightening or darkening at 5 μm) affect the altitude of the wind measurements. The convolutions between the brightness data at 16° N (NEB) and 10° S (SEB) and the selected wavelet with their respective periodicities are represented in Figure 15. The comparison between the brightness data and its convolution with the selected wavelet gives us the trustworthiness of the derived period (i.e., a similar trend of these two confirms the veracity of the derived period). A high correlation at 16° N is observed at all the studied epochs, confirming the 4.4 ± 0.4 yr periodicity obtained. At 10° S, this correlation is not as clearly seen, but the convolution follows the overall trend of the brightness variability at this latitude.

The observed periodicities of the NEB and SEB coincide with the ∼4.5 yr variability found in the equatorial stratospheric temperatures, known as the Quasi Quadrennial Oscillation (QQO; Leovy et al. 1991; Cosentino et al. 2017), where a vertical alternating pattern in the zonally averaged temperature and winds is observed between 3 and 20 mbar at the equatorial and off-equatorial latitudes. This phenomenon resembles the quasi-biennial oscillation (QBO) in Earth’s equatorial stratospheric winds and temperatures (e.g., Baldwin et al. 2001) and Saturn’s quasi-periodic equatorial oscillation (QPO; Fouchet et al. 2008; Orton et al. 2008), which has been observed to span almost down to Saturn’s tropopause (Fletcher et al. 2017a). The relationship between the QQO and the 5 μm brightness variability observed at the NEB and SEB is not yet understood, as no direct relation is observed so far. Simon-Miller et al. (2006) reported noticeable variations in the thermal wind profile in the upper troposphere at ±15°–20° latitude and the equator, retrieved from temperature measurements using Voyager IRIS and Cassini CIRS data. These authors suggested that these changes may reflect a QQO response, suggesting that, like the QPO on Saturn, the QQO could also extend down as far as the upper troposphere.

Establishing whether the variations observed at the NEB, SEB, and EZ and at 7–9° N are related to the QQO is a challenge. Leovy et al. (1991), Orton et al. (1991), and Cosentino et al. (2017) showed a clear anticorrelation between the equatorial and off-equatorial latitude temperatures at the upper stratosphere, where the main equatorial features of the QQO have off-equatorial secondary circulations that descend at about the same rate. If the periodicities observed in this study at the NEB and SEB were related to the QQO, one would also expect to see such an anticorrelation at 5 μm too. However, this relationship is not observed in our results, where both periodograms show a periodicity of 6.6 ± 0.5 yr south of the equator (number 5 in Figure 14 corresponding to EZ disturbance events, in agreement with Antuñano et al. 2018), and a peak at 7.4 ± 1 yr at the southern edge of the NEB, at 7–9° N, in agreement with the ∼7 yr periodicity observed in the changes of the equatorial wind profile at cloud level (Simon-Miller et al. 2007; Simon-Miller & Gierasch 2010; Tollefson et al. 2017). Nevertheless, Cosentino et al. (2017) and Simon-Miller et al. (2006) suggested the existence of two different periodicities of the QQO, with the period at ∼4 mbar being shorter than the period at ∼14 mbar, although more observational data would be needed to verify this. If the QQO indeed extends to the troposphere, as it appears to do on Saturn, longer periods than 4.5 yr might be then expected in the troposphere. Future observations and numerical simulations are essential to understanding the relationship between the observed tropospheric variations and the QQO and, in the case of a connection between these two, to identifying whether the QQO is responsible for the tropospheric variations (i.e., the warm and cool regions descending from the stratosphere into the upper troposphere could be influencing the condensation of volatiles in this region), or on the contrary, the tropospheric variations are the cause of the QQO (i.e., deeper processes such as changes in the gravity-wave forcing affecting the stratosphere).

Finally, the ∼7 yr peak observed at 36–41° S might not be real. An example of the 5-μm brightness variability is shown in Figure 13. The data show a high variability of the brightness in very short periods of time (days to weeks) with relatively large error bars. This is a clear sign of large longitudinal variations of the 5-μm brightness at this latitude, related to the small white ovals usually observed at 38–40° S. Therefore, one must be cautious because the brightness increases that the convolution shows (see Figure 25 in the Appendix) are related to dates with the largest amount of data available, making the temporal variation noisier. A summary of the measured periodicities described here is shown in Table 4.

The overall 4–8 yr periods found in this study could hint at some moist convective process, where convective accumulated potential energy could be periodically released. Sugiyama et al. (2014) performed numerical simulations of moist convection in Jupiter, showing periods of intermittency between convective eruptions of the order of tens of days. This is in agreement with some observed short-term scale events in Jupiter (e.g., the eruption of plumes in the SEB during its revival process), but cannot reproduce the long-term intervals found in this study. Li & Ingersoll (2015) analyzed moist convection in Saturn as an explanation of the observed frequency in Saturn’s giant storms. In their study, they show that the cooling phase of the atmosphere due to thermal radiation at the top affects the stable stratified layers, causing convective inhibition. The 4–8 yr periodicities found here are comparable to the radiative time constant (τ_r) from Conrath et al. (1990), who suggested a τ_r between 4 and 6 yr at 1 bar, and to the τ_r ∼ 4 yr at 5 bar given by Li et al. (2018). The similarities between the observed periodicities and the radiative time constant are therefore compelling, but not conclusive, evidence that the changes described here could also be related to moist convection. Future numerical simulations will be needed to probe this hypothesis.

The analyses of the Lomb–Scargle and Wavelet Transform periodograms show a very similar periodicity in the 5-μm brightness variability of the NEB and SEB (see Section 6.1). However, these analyses do not give us any information on whether these changes occur simultaneously or with a temporal shift and whether a correlation or anticorrelation exists in their brightness variations, information that is needed to understand the relation and the origin of these changes. Here, we assess the latter question by analyzing the results of a PCA we performed (described in Section in 2.6), shown in Figure 16.

Zoom In Zoom Out Reset image size

Figure 16 shows the first three eigenvectors of the PCA (A)–(C) and the reconstructed mean-centered brightness for each eigenvector (D)–(F). The red colors correspond to brightnesses larger than the meridional-mean brightness at each date, while blue colors represent the contrary (see Section 2.6 for a detailed description of the methodology of the PCA). In the PCA, one would need as many vectors as the number of data points to be able to reconstruct the signal entirely. However, in this study, we only show and analyze the first three eigenvectors, which display most of the major changes described in Section 4, like the EZ disturbances, NEB expansion, SEB fades (highlighted by stars, arrows, and diamonds, respectively, in Figure 16(F)), enough to study the correlation or anticorrelation between the NEB and SEB. Higher eigenvectors give us information on the finer-scale variations, like the NTZ and NTB variations described in Section 4.

Figures 16(B) and (C) show a clear anticorrelation in the brightness of the NEB and the SEB at all of the studied epochs, suggesting that the brightness increases and decreases of these belts are related. This is also observed in Figure 17, where we present the 5-μm brightness at 16° N (NEB) and 10° S (SEB) between 1984 and 2018. As indicated by the dark gray regions, at least five of the six brightness decreases observed at 16° N (that is, when the NEB is not expanded) are related to strong brightness increases at 10° S. In contrast, four of the seven brightness increases at 16° N are observed contemporaneously with brightness decreases at 10° S (indicated by the light gray bands). In Figure 17 we also show what Jupiter looked like at 5 μm at six particular dates, where the NEB–SEB anticorrelation is clearly visible. These results, combined with the periodogram analysis, suggest a cyclic anticorrelation of the 5-μm brightness at the tropical belts of Jupiter. This anticorrelation differs from “global upheaval” events, which correspond to episodes where the SEB fade/revival cycle, the strong EZ coloration, and the NTB outbreaks occur contemporaneously, spreading the coloration from one hemisphere to the other (Rogers 1995, 2007a, 2007b). These occurred in 1990 and 2007 and, unlike the anticorrelation we observed, are not cyclic. So far, the physical processes underlying the tropical belt anticorrelation and global upheavals are not understood, and no circulation model can explain this phenomenon. Fletcher et al. (2017a) observed that a storm that erupted at Saturn’s northern mid-latitudes strongly affected the stratospheric temperatures at the equatorial latitudes, significantly perturbing the wind structure in the middle atmosphere. These authors suggested that this could be originated by westward momentum being injected into the equatorial winds from waves created by the storm. Something similar could also happen in Jupiter, where outbreaks at tropical and mid-latitudes could influence the equatorial latitudes via meridional wave propagation. The development of new general circulation models and numerical simulations will be essential to understanding this phenomenon.

Zoom In Zoom Out Reset image size

Finally, Figure 16(A) also shows a clear asymmetry between the northern and southern hemispheres at latitudes greater than ∼45°, where the northern region appears to be brighter at 5 μm than the southern region during the entirety of the 34 yr analyzed in this study. This is also observed at all of the reconstructed mean-centered brightness maps shown in Figure 16 (the first reconstructed mean-centered brightness represented in Figure 16(D) is the same as Figure 16(A)) and in Figure 2. Previous studies by Irwin et al. (2004), Fletcher et al. (2009a), and Giles et al. (2015) show a north/south asymmetry in phosphine distribution at mid-latitudes, with higher amounts of phosphine detected at the northern mid-latitudes above 40° than in the south. Similarly, Zhang et al. (2013) and Achterberg et al. (2006) reported north/south asymmetries in the aerosol and ammonia cloud distribution, respectively, with higher amounts at the northern mid-latitudes. This north/south asymmetry observed in phosphine, aerosols, and ammonia distributions cannot explain the observed differences at 5 μm in the mid-latitudes, where a brighter southern mid-latitude at 5 μm would be expected because of the phosphine, aerosols, and ammonia blocking more of the 5-μm radiance from escaping in the northern mid-latitudes. Future spectroscopy analysis using the Jovian Infrared Auroral Mapper (JIRAM) data will be essential to understanding the nature of the 5-μm north/south asymmetry.