Here we report the discovery at the upper cloud level of an extremely narrow and strong prograde jet, centered in the middle of the broad equatorial jet. Measurements from images obtained by the Cassini Imaging Science Subsystem (ISS) show that the jet reaches 430 ms−1 ± 16 m s−1 with a peak speed difference of 180 ms−1 relative to nearby latitudes at 60 mbar and 390 ms−1 ± 23 m s−1 at depths > 500 mbar spanning ∼6° about the equator . Contrarily to what is observed in other latitudes, its velocity increases with altitude. This jet is the first of its kind to be discovered on any of the giant planets, and adds to our three-dimensional understanding of Saturn's stratospheric wind field.
 The intense equatorial eastward jets observed at cloud level in Jupiter and Saturn represent a major challenge for geophysical fluid dynamics. Most numerical models are based either in deep convection (“deep forcing”) or in a shallow external layer fed by solar radiation (“shallow forcing”), but the physical origin of this puzzling feature is still under debate [Ingersoll et al., 2007; Del Genio et al., 2009]. Saturn's equatorial jet is of particular interest in view of its three dimensional structure [Sánchez-Lavega et al., 2007], suspected large temporal variability [Sánchez-Lavega et al., 2003], and related stratospheric semiannual oscillation [Fouchet et al., 2008; Orton et al., 2008]. Observing the cloud structure and measuring the winds of the equatorial region of Saturn is not an easy task due to the obscuration of Saturn's rings and their shadow on the globe. Published works on Saturn's equatorial winds at cloud level provided only a partial latitudinal coverage. This was the case in 1980–81 during the Voyagers 1–2 flybys [Sánchez-Lavega et al., 2000], from 1990 to the present in studies performed with the Hubble Space Telescope [Sánchez-Lavega et al., 2003, 2007], and in previous analysis of Cassini-ISS images [Sánchez-Lavega et al., 2007; Porco et al., 2005; Vasavada et al., 2006; García-Melendo et al., 2009]. We selected Cassini-ISS images which comprise the long-term variations of Cassini spacecraft's orbital inclination relative to Saturn's equatorial plane, combined with the progressive changing effect of the rings projection plus their shadows in blocking the planet's equator view. This allowed us to cover, for the first time and in high resolution, the whole equatorial region of Saturn.
2. Observations and Wind Measurements
 Images were obtained by the Cassini Imaging Science Subsystem (ISS) Narrow and Wide Angle Cameras. We analyzed observations in three filters: MT3, centred at the 889 nm strong methane absorption band restricted to the upper tropospheric hazes, and CB2 and CB3 centred at the near infrared 752 nm and 939 Snm continuum [Porco et al., 2004], which penetrate deeper into the atmosphere showing discrete cloud features [Sánchez-Lavega et al., 2007; García-Melendo et al., 2009; Pérez-Hoyos and Sánchez-Lavega, 2006]. The CB2 and CB3 filters have a similar penetration in Saturn's atmosphere [Pérez-Hoyos and Sánchez-Lavega, 2006], so we will refer to the CB3 filter hereinafter. Examples of the morphology of the clouds and hazes observed in the three filters and their motions are shown in Figure 1.
 We used image pairs separated by one planetary rotation (about 10 hours) to get into view the same cloud features in the Equator, measure their motions in longitude, and retrieve the zonal winds by employing two complementary techniques. On the one hand, we applied an automatic correlation technique that compares zonal scans of the same latitude in different images [García-Melendo et al., 2009]. On the other, we manually tracked those individual cloud features that we were able to unambiguously identify in the image pairs, measuring their displacements and deriving their zonal velocity. Tables 1 and 2 list the respective selected image datasets for the automatic correlation and image tracking data reductions. The images were navigated and contrast enhanced using a combination of unsharp mask and Butterworth filters in the Planetary Laboratory Image Analysis (PLIA) software [Hueso et al., 2010]. Automatic zonal wind retrieval was computed by cross-correlating brightness scans along a latitude circle, covering longitude intervals from ∼60° to 80° long obtained from image pairs of the same longitude sector. Each scan was corrected for limb-darkening effects and illumination conditions. In this way the equatorial region was meridionally scanned in 0.1° latitude steps. Images covered a time span of five years between September 2004 and January 2009 with a non homogeneous time sampling imposed by the observing conditions from Cassini's spacecraft. A complete retrieval of the zonal winds around Saturn's equator was possible from images taken on February 2005. Due to the low latitude of the subsolar point (∼22.5°S), the projected shadow by the rings did not obscure the equator, which was visible up to the planetographic latitude of ∼ 6.5°N. In addition, the spacecraft orbital plane was on Saturn's equatorial plane and the rings were edge-on allowing a complete view of the equator. Image resolution was 160 km pixel−1, so a one-pixel navigation error accounted for a velocity error of 4.2 m s−1. In 2004, 2008, and 2009, we examined either the north or south sides of the equator at image resolutions between 60 km pixel−1 and 80 km pixel−1, with one-pixel navigation errors between 1.6 m s−1, and 2.0 m s−1. In total, about 150 image pairs were used in automatic measurements. Manual tracking of individual cloud tracers allowed a more homogeneous time sampling between 2004 and 2008. We obtained manual measurements several times every year except in 2006. Only those images where the 0° latitude line was visible and image pairs separated by a Saturn rotation were selected. Frames used had a resolution between 80 km pixel−1 and 60 km pixel−1. A total of 250 measurements were collected. The overall time coverage was not completely uniform, but gave us a long term view from 2004 to 2009.
Table 1. Observing Dates and Wavelengths of the Cassini Images Used in the Correlation Method for Automatic Zonal Winds Determination
N stands for Narrow Angle Camera, W stands for Wide Angle Camera.
Image resolution is given in km pixel−1.
2004 Sept. 6–7
CB1, MT1, CB3, MT3
2004 Sept. 9–10
2004 Sept 30
2005 July 4–5
2005 Oct. 2–3
2007 Feb. 4
2007 May 24
2007 June 10
2007 Sept. 3–4
2007 Nov. 8–11
CB2, CB3, MT2, MT3
2008 Jan. 13–14
2008 Feb. 18
2008 Feb. 25–27
2008 Mar. 29–30
2008 April 17
2008 Dec. 28–30
3. Wind Measurement Results
 From latitudes 40°N to 40°S, but outside a band of ∼ 6° width centered at the equator, the eastward jet shows a double symmetric peak in CB3 and a more asymmetric peak in MT3 (Figure 2a). Both are highly reminiscent of what we see in Jupiter [Ingersoll et al., 2007]. Unlike Jupiter however, a few degrees around the equator the zonal winds form a narrow eastward jet spanning from 3°N to 3°S. This jet is seen in both filters but with different intensity. It is more prominent in the MT3 filter, that senses higher altitudes, where the peak speed reaches 430 ms−1, about 180 ms−1 above the velocity in nearby background latitudes (e.g., at 5° North and South). In the CB3 filter (lower altitudes) the peak speed reaches 390 ms−1 or about 50 ms−1 above the background flow.
 Manual individual measurements shown as single points in Figure 2a confirm the image correlation results. The correlation method has higher spatial resolution and yielded better results on the average, but the gross structure of the profile was also captured by cloud tracking with few discrepancies between both methods. In Figure 2b we show the distribution of the r.m.s. error as a function of latitude. The automatic wind retrieval gave an average r.m.s. scatter of 6.3 m s−1 for the whole equatorial band retrieved from CB3 images and 8.5 m s−1 for the MT3 measurements (from 40°S to 40°N). Around the equator, the average velocity errors in the latitude band from 5°N to 5°S where the central jet sits and resulting from the combination of the two methods, are 23 ms−1 in CB3 and 16 ms−1 in MT3. Thus the central jet is unambiguously detected in both filters above 10-σ r.m.s. error in MT3.
 Up to 2008 we detected no variations in the equatorial narrow jet at the dispersion level of the manual measurements. Results indicate that it has maintained its intensity, narrowness, and latitudinal position at least from 2005 to 2008. In those equatorial latitudes previously measured by other authors, we retrieved the same velocities [Sánchez-Lavega et al., 2007; Porco et al., 2005; Vasavada et al., 2006; García-Melendo et al., 2009], but now with a higher detail in the meridional structure and more precision in the wind determination in the central equator that allowed us identification of the detailed jet structure. Some indication of the central jet in CB2-CB3 filters is given by Sánchez-Lavega et al. [2007, Figure 2] but the dispersion of the data masked the fine structure, whereas in the MT2-MT3 filters there were just 3 to 5 points with high dispersion in velocity and too sparse for jet detection.
4. Radiative Transfer Analysis
 To obtain the altitude locations of the wind tracers, the vertical cloud structure of Saturn's equatorial region was analyzed using a radiative transfer code. The method consists in fitting the observed zonal brightness scans from images in different observing wavelengths to model profiles. A subset of 68 images from April 2007 to January 2009 which fulfilled the restrictions of this analysis (continuum and methane image pairs and enough longitude coverage far from ring shadowing) were calibrated using CISSCAL software [Porco et al., 2004]. The radiative transfer numerical model we employed is described by Acarreta and Sánchez-Lavega  and Pérez-Hoyos et al. , and the reference cloud model is based in our studies of Saturn's Equatorial Zone [Sánchez-Lavega et al., 2007; Pérez-Hoyos et al., 2005]. The cloud model consists of two detached haze layers. The upper one is located in the stratosphere between ∼ 20 mbar and 50 mbar and is formed by small particles, showing only a small influence on the retrieved results. The second one is a vertically extended and thick tropospheric haze where the cloud features we tracked sit at least at two altitude levels [Sánchez-Lavega et al., 2003; Sánchez-Lavega et al., 2007; García-Melendo et al., 2009]. In order to fit the reflectivity scans we left two free parameters to model the tropospheric haze: its top pressure and total optical depth.
 In order to determine the heights at which the tracked features were located we followed the same method discussed by Sánchez-Lavega et al.  and Pérez-Hoyos et al. . The features tracked in the MT3 filter are usually at the top of the tropospheric haze, above or close to the tropopause at 90 mbar [Fletcher et al., 2007]. For the features in the CB3 filter, we calculated the expected contrast for an idealized tracer [Pérez-Hoyos et al., 2005] as it is embedded in various pressure levels in the model atmosphere retrieved above. Minimum observable contrast gives an upper value for the pressure value, normally in the top of or within the ammonia cloud deck, which in the end results in an upper limit for the vertical wind shear. To estimate the lower limit, a similar calculation was performed this time to find the pressure level at which the feature becomes visible in adjacent methane filters [Pérez-Hoyos et al., 2005]. Table 3 gives the values obtained for the cloud top altitudes, i.e., MT3 features, with their corresponding error bars and the maximum height for the deeper CB3 features, together with the upper and lower values for vertical wind shear for equatorial latitudes with substantial wind speed differences. Accordingly, outside the band encompassed between the latitudes 3°N and 3°S, the winds decrease with altitude at a maximum rate of ∼−40 ms−1 per scale height H (one scale height is 43 km at Saturn's equatorial troposphere), in agreement with our previous findings [Sánchez-Lavega et al., 2007; Pérez-Hoyos et al., 2005]. This also follows the general trend found in Saturn's non-equatorial eastward jets [García-Melendo et al., 2009]. However, the reverse occurs within the central equatorial jet peak (from latitude 3°N to 3°S). Here, the winds increase with altitude within the troposphere at a maximum rate ∼ +12 m s−1 per scale height H.
Table 3. Cloud Altitude Levels and Vertical Wind Shears in Saturn's Equator
60 ± 10
+0 – −6
45 ± 5
−25 – −35
60 ± 10
+5 – +12
70 ± 10
−30 – −40
70 ± 10
−20 – −35
5. Equatorial Jet Characterization
 To characterize the central part of the broad equatorial jet we considered the superposition of a parabolic function (outside the central jet) and a Gaussian function (the central jet). We also included for the vertical velocity a linear dependence with upward decreasing winds in the parabolic part and increasing winds in the Gaussian according to the measured shears,
Here u and P are the zonal velocity and pressure (altitude) and ϕ is the latitude. We fit the jet using u0 = 345 ms−1, u1 = 0.35 ms−1, u2 = 50 ms−1, which are the speeds in CB3 that define the parabolic fit (u0, u1) and the Gaussian amplitude (u2, see equation (2)). For the reference pressure level we take the upper value from our radiative transfer model in the CB3 filter, P0 = 500 mbar. A small latitude offset ϕ0 = ±0.5° is required with different signs for the parabolic and exponential terms, but it is so small that we cannot say if it is real or not within the measurement errors. For the central Gaussian jet we take a width Δϕ = 2°. The vertical shears are introduced by the parameters Δu = (∂u/∂z)H = 42 ms−1 in the parabolic term and by δ = = 1.2 in the Gaussian term where we have used (∂u/∂z)H =12 m s−1 and ln(P0/P1) = ln(500/60) from the estimated levels in the CB3 and MT3 filters. This representation fits reasonably well the equatorial jet peak at both atmospheric levels (Figure 2c), which allows us to represent he meridional structure of the Gaussian jet in terms of dynamical parameters as [Allison et al., 1990]
where y = aϕ is the meridional distance from the equator (a = 60330 km is Saturn's equatorial radius), and b = 2u2/Le2 = 2.25 × 10−11 m−1s−1 is the peak latitudinal curvature where Le = a(πΔϕ/180) = 2105 km is the associated e-folding scale for the latitudinal variation of the central jet. Note that u2 is the jet peak at 500 mbar relative to the background parabolic jet. The relative vorticity gradient of the jet, averaged over its meridional e-folding interval is <d2u/dy2>e ≈ b/e = 8.3 × 10−12 m−1s−1 ∼ 1.5β, where β ≅ 2Ω/a = 5.4 × 10−12 m−1s−1 is the planetary vorticity gradient at the equator (Ω = 1.64 × 10−4 s−1 for System III rotation reference frame). From these estimates we conclude that the relative vorticity gradient will play a similar role to the planetary vorticity gradient β in the dynamics associated to the jet as for example in wave formation in the upper troposphere [Allison et al., 1990]. We have run periodogram analysis of the East to West brightness patterns on different dates that do not reveal any frequency indicative of permanent waves. However, the wave detected in the thermal infrared by Li et al.  and the vertically propagating QBO or QQO oscillation reported by Fouchet et al. , both in the stratosphere, suggest that equatorially trapped waves could play an important role in generating the jet.
 The picture that emerges is that, while Saturn and Jupiter have similar broad eastward equatorial jets with a double symmetric peak structure in their troposphere, Saturn's equatorial jet is more complex with the presence of an unexpected equatorially centered, narrow and strongly peaked jet that increases in intensity with altitude. This could be a real dynamical difference between both planets or just that Saturn equatorial hazes are located at higher altitude than in Jupiter, allowing for the jet detection. In Saturn the central jet is located at the base of a region where a biannual oscillation in the temperature field has been detected [Fouchet et al., 2008; Orton et al., 2008] and where a high altitude (1 mbar) stratospheric jet and a thermal wave have been reported [Li et al., 2008]. Our results fit well the wind vertical structure reported in these works and serve as reference for their results. The central jet peak increases its eastward velocity with altitude within the upper troposphere from 390 m s −1 at ∼500 mbar to 430 m s−1 at ∼50 mbar with an estimated vertical shear of +12 m s−1 H−1, then it oscillates vertically in the stratosphere with speed 330 ms−1 at 10 mbar and 500 ms−1 at 1 mbar. It remains to be known whether this equatorial jet structure, now described in detail in three dimensions, is permanent or variable with the seasonal solar insolation cycle, including the variable shadow cast by the rings. It is also an unresolved problem how the sporadic giant convective storms [Sánchez-Lavega et al., 1991, 1996] and other moist convective phenomena that take place in the equator, as the series of bright plumes observed during the Voyager 2 encounter [Sánchez-Lavega et al., 2000], influence wave generation and the structure of the central equatorial jet [Sayanagi and Showman, 2007]. According to our analysis, the jet has been present from 2004 to 2009, a period during which no large storm activity occurred in Saturn equator, but during which the region suffered different ring shadowing periods. Only long-term observations by Cassini and theoretical modeling can clarify the nature of this unique feature.
 This work has been funded by the Spanish MICIIN AYA2009-10701 with FEDER support and Grupos Gobierno Vasco IT-464-07.