Icarus 221 (2012) 248–261

Contents lists available at SciVerse ScienceDirect

Icarus

journal homepage: www.elsevier .com/ locate/ icarus

Mapping zonal winds at Venus’s cloud tops from ground-based Doppler velocimetry

Pedro Machado a,b, David Luz a,⇑, Thomas Widemann b, Emmanuel Lellouch b, Olivier Witasse c

a CAAUL, Centro de Astronomia e Astrofísica da Universidade de Lisboa, Observatório Astronómico de Lisboa, Tapada da Ajuda, 1349-018 Lisboa, Portugalb LESIA, Laboratoire d’Etudes Spatiales et d’Instrumentation en Astrophysique, Observatoire de Paris, CNRS, UPMC, Université Paris-Diderot, 5 place Jules Janssen, 92195Meudon, Francec ESA/RSSD, ESTEC, 2200 AG Noordwijk, The Netherlands

a r t i c l e i n f o

Article history:Received 28 February 2012Revised 27 June 2012Accepted 6 July 2012Available online 25 July 2012

Keywords:Venus, AtmosphereAtmospheres, DynamicsSpectroscopy

0019-1035/$ - see front matter � 2012 Elsevier Inc. Ahttp://dx.doi.org/10.1016/j.icarus.2012.07.012

⇑ Corresponding author.E-mail address: dluz@oal.ul.pt (D. Luz).

a b s t r a c t

The most significant aspect of the general circulation of the atmosphere of Venus is its retrograde super-rotation. A complete characterization of this dynamical phenomenon is crucial for understanding its driv-ing mechanisms. Here we report on ground-based Doppler velocimetry measurements of the zonalwinds, based on high resolution spectra from the UV–Visual Echelle Spectrograph (UVES) instrumentat ESO’s Very Large Telescope. Under the assumption of predominantly zonal flow, this method allowsthe simultaneous direct measurement of the zonal velocity across a range of latitudes and local timesin the day side. The technique, based on long slit spectroscopy combined with the high spatial resolutionprovided by the VLT, has provided the first ground-based characterization of the latitudinal profile ofzonal wind in the atmosphere of Venus, the first zonal wind field map in the visible, as well as new con-straints on wind variations with local time. We measured mean zonal wind amplitudes between 106 ± 21and 127 ± 14 m/s at latitudes between 18�N and 34�S, with the zonal wind being approximately uniformin 2.6�-wide latitude bands (0.3 arcsec at disk center). The zonal wind profile retrieved is consistent withprevious spacecraft measurements based on cloud tracking, but with non-negligible variability in localtime (longitude) and in latitude. Near 50� the presence of moderate jets is apparent in both hemispheres,with the southern jet being stronger by �10 m/s. Small scale wind variations with local time are alsopresent at low and mid-latitudes.

� 2012 Elsevier Inc. All rights reserved.

1. Introduction

The atmosphere of Venus displays an unusual combination ofextremes of terrestrial-type atmospheres. The Venus obliquity of177.3�, with the rotation axis almost perpendicular to the ecliptic,is the highest of all the planets, and the atmospheric spin is in thedirection of the planetary rotation, so that seen from Earth theatmosphere rotates in the retrograde sense. Moreover, the atmo-sphere is in a state of superrotation, in which the long rotation per-iod of the solid body, of 243.02 days (Mueller et al., 2012),contrasts with the short period of atmospheric rotation at thecloud tops (4.4 days at the equator).

Due to the large atmospheric CO2 content (roughly 96.5%), adramatic runaway greenhouse effect was established in the Venusatmosphere during its evolution (Titov et al., 2007). Despite highback-scattering by the clouds (more than 75% of the incoming solarflux), the greenhouse effect raises the average surface temperatureto about 735 K. Hence, Venus presents the hottest surface in theSolar System, as well as a dramatic case study for global warmingand the ensuing climate change.

ll rights reserved.

The atmosphere is shrouded by a dense cloud layer of sulfuricacid haze and an as yet unidentified UV absorber within the upperregion of these clouds (Markiewicz et al., 2007). The contrastingclouds constitute nearly ideal tracers of the atmospheric motion,enabling the effective application of the cloud tracking techniqueto imaging data. The presence of radiatively active gases and aero-sols allows, in turn, to perform wind measurements based on thestudy of absorption and emission lines, such as for heterodyneDoppler techniques and Doppler velocimetry based on solar radia-tion scattered off the upper cloud layer, the method used here.

Three main mechanisms are present in the general circulationof Venus’ troposphere and lower mesosphere, particularly withrespect to cloud top dynamics. The low latitudes are characterizedby the presence of pronounced convection (Markiewicz et al.,2007). The most relevant altitude region where pronouncedconvection occurs, as determined by the Pioneer Venus and Vegamissions, is the range extending between the heights of 48 and55 km (Baker and Schubert, 1992). In the range of latitudes up toabout 60�N/S, a system of zonal retrograde (from East to West)winds prevails, flowing almost parallel to the equator. Polewardof 60� there is an abrupt transition to the major structures of thecircumpolar vortices in both the North and the South hemispheres(Suomi and Limaye, 1978; Taylor et al., 1980; Limaye et al., 2009).

P. Machado et al. / Icarus 221 (2012) 248–261 249

This transition zone is accompanied by a transient jet streamlocated around this latitude, whose temporal variability and verti-cal extension are still a matter of debate in relation with thermalfield measurements (Markiewicz et al., 2007; Tellmann et al.,2009; Piccialli et al., 2012).

Direct evidence of the retrograde, zonal super-rotation of theatmosphere of Venus was provided by Mariner 10 observationsin the ultraviolet (Limaye and Suomi, 1981) and by Pioneer Venusand Galileo SSI observations (Rossow et al., 1980, 1990; Beltonet al., 1991; Toigo et al., 1994; Limaye, 2007; Peralta et al., 2007).This retrograde zonal circulation is characterized by a monotoni-cally increasing zonal wind with height, which grows from�10 m/s in the lowest scale height to �100 m/s at the cloud tops(Gierasch et al., 1997). The meridional wind velocity is, in general,much lower than the zonal wind, with an upper limit of �10 m/s(Gierasch et al., 1997).

In the thermosphere, the flow is further characterized by a ther-mal circulation from the subsolar to the antisolar point (SS–AS),which can be inferred from the large contrasts in temperatures be-tween the day and night sides. The wind velocity on SS–AS cellsshow a strong variability on its zonal component over differenttime scales, as revealed by several complementary observing tech-niques such as ground-based wind measurements from Doppler-shifted infrared, sub-mm and millimeter lines (Goldstein et al.,1991; Lellouch et al., 1994, 2008; Clancy et al., 2008; Rengel et al.,2008; Sornig et al., 2008, 2012; Clancy et al., 2012), ground-basedand Venus Express/SPICAV observations of non-LTE airglow emis-sion variability on the night side (Ohtsuki et al., 2008; Gérardet al., 2009; Soret et al., 2010, 2012) as well as modeling constraintsof the upper mesosphere and lower thermosphere in terms of ver-tical and latitudinal extension of the cyclostrophic balance approx-imation (Bougher et al., 1997; Mueller-Wodarg et al., 2006; Brechtet al., 2012). In the upper cloud region the wind variability may becaused by intrusions of the SS–AS return branch, superimposed tothe mean zonal flow at low and midlatitudes (Widemann et al.,2007) or, in the polar region, by tidal effects (Peralta et al., 2012).

The state of atmospheric superrotation is characterized by a meanangular momentum which is much greater than that of the atmo-sphere at rest relative to the surface. Since equatorial superrotationcannot be maintained by an axially symmetric circulation alone(Hide, 1969), a possible explanation is the Gierasch–Rossow–Wil-liams mechanism, in which the Hadley circulation coupled with plan-etary waves produces a net transport of angular momentum fromhigh to low latitudes (Gierasch, 1975; Rossow and Williams, 1979).However, the mechanisms controlling this redistribution of momen-tum in the global circulation remain largely unconstrained.

Since 2006, Venus Express (VEx) has been the first mission in16 years dedicated to studying the Venus atmosphere. In thecontext of this mission, the zonal circulation has been the subjectof intense scrutiny both from the spacecraft instruments andground-based observations. The Venus Monitoring Camera (VMC)and the Visual and Infra-Red Thermal Imaging Spectrometer(VIRTIS) have been the main experiments on board for characteriz-ing the atmospheric dynamics (Svedhem et al., 2007). Importantadvances were made in characterizing the global circulation, withzonal wind measurements based on tracking the cloud motion inimages at visible, UV and infrared wavelengths (Markiewicz et al.,2007; Piccioni et al., 2007; Sánchez-Lavega et al., 2008; Luz et al.,2011). The various measurements by cloud tracking show asignificant dispersion of the instantaneous direct measurementsof the wind vectors, as well as variability with local time andtime-dependence of the mid-latitude jet (Sánchez-Lavega et al.,2008; Moissl et al., 2009; Hueso et al., 2012; Peralta et al., 2012).

The morphology of the upper cloud layer of Venus is character-ized by three main types of cloud structures. Equatorial cloudsdisplay a mottled pattern of small scale features apparently caused

by local convection, with the horizontal length scales of the convec-tion cells �20 km (Markiewicz et al., 2007). In the middle latitudesa stable system of bands almost parallel to the equator prevails (Pic-cialli et al., 2008), while at circumpolar latitudes a large hurricane-like vortex is present, with highly variable cloud structures, chang-ing over time scales of a few days (Luz et al., 2011). This structure,which is similar to its northern counterpart, is warmer than its sur-roundings and is accompanied by a surrounding cold collar (Picci-oni et al., 2007; Taylor et al., 1980). The persistence of thesecomponents of the circulation highlights the question of the rela-tion between super-rotation and long-lived atmospheric features.

Independent ground-based observations have been of majorimportance to complement space-based observations and have con-tributed to a more complete picture of the Venus atmosphericdynamics. The early ground-based wind measurements using thecloud tracking method (Boyer and Camichel, 1961; Dollfus et al.,1975) were based on the monitoring of large-scale ultravioletmarkings, and were very sensitive to wave motions and convectionfeatures, rather than the real atmospheric particle velocity. This is apotential problem which affects all wind measurements based oncloud tracking (ground and space-based). Recently, cloud trackingin the near infrared using the observational window in the nightside of the continuum K near 2.26 lm (Young et al., 2010), aswell as quantification of middle and lower cloud variability andmesoscale dynamics from Venus Express/VIRTIS observations at1.74 lm (McGouldrick et al., 2012) allowed the monitoring of windconditions in the lowest layer of clouds (45 km). On the other hand,visible observations, or more specifically Doppler shift observationsof solar Fraunhofer lines and of CO2 lines in the lower mesosphere(65–85 km) have provided the only wind measurements near thecloud tops based on direct Doppler velocimetry (direct velocity ofthe cloud top atmospheric aerosols in the observer’s direction) inrecent years (Widemann et al., 2007, 2008; Gaulme et al., 2008). Thisaltitude region is essential, as it constrains the global mesosphericcirculation in which zonal winds generally decrease with heightwhilst thermospheric SS–AS winds increase.

The technique of Doppler measurements in the visible rangewas pioneered by Young (1975). Andrew Young first realized thatthis type of observations are affected by a systematic bias, due tothe finite angular size of the Sun as seen from Venus and to the fastequatorial solar rotation. The early visible Doppler measurementsshowed a large dispersion of results (Young, 1975, 1979; Trauband Carleton, 1979; Lellouch et al., 1997), and it is now clear thatthey were affected by what has been called the ‘‘Young effect’’.

The methods applied in recent planetary wind measurementsusing high-resolution spectroscopy in the visible range (Luzet al., 2005a, 2006; Widemann et al., 2007, 2008; Gabsi et al.,2008) all address the fundamental problem of maintaining a stablevelocity reference during acquisition. In order to measure the glo-bal circulation at cloud top altitude, we need to address windamplitude variations (or wind latitudinal gradients) on the orderof 5–10 m/s projected on the line-of-sight (Widemann et al.,2007). Such an accuracy cannot be achieved by single line fitting,even at high spectral resolution (with the exception of very highresolution heterodyne techniques, which require dedicated instru-mentation; Kostiuk et al., 2001, 2005; Sonnabend et al., 2006, 2008,2010). Therefore, it becomes necessary to optimally measure rela-tive Doppler shifts between two sets of absorption lines (Connes,1985), while simultaneously monitoring the change in spectral cal-ibration with time (Widemann et al., 2008).

There are two alternative ways to deal with the previous ques-tions. First, sequential Doppler shift measurements (Widemannet al., 2007; Gabsi et al., 2008), where a continuous monitoringof the pixel versus wavelength relation, prior to velocity measure-ments, is performed. Differential measurements of line shifts areperformed while the change of spectral calibration with time is

250 P. Machado et al. / Icarus 221 (2012) 248–261

monitored. This method has been applied in retrievals based on theDoppler shifts of CO2 absorption lines (Widemann et al., 2007,2008). The second option is to perform simultaneous measurementsof relative Doppler shifts between two points in the disk usinglong-slit spectroscopy. This automatically suppresses the Dopplershift due to the relative motion between the Earth and Venus, aswell as any systematic drifts due to time variation of the calibra-tion (Civeit et al., 2005; Luz et al., 2005a, 2006). This method hasthe advantage of providing an instantaneous, one-dimensional im-age of the wind Doppler shift, enabling a direct determination ofspatial variations in the flow, and to search for wave and tidalstructures if sufficient spatial resolution is possible. This methodhas been previously applied for determining the radial velocitiesof stars (Baranne et al., 1996), in asteroseismology (Martic et al.,1999), and was recently used in planetary wind measurements inthe visible range for Titan and Saturn, from the Doppler shifts ofthe backscattered solar spectrum (Luz et al., 2005a,b, 2006). Itsapplicability to the case of Venus atmospheric winds has beentested in Observatoire de Haute Provence by Martic et al. (2001).

Here we report on observations of Venus made with the Ultra-violet and Visual Echelle Spectrograph (UVES) at ESO’s Very LargeTelescope, with the objective of measuring the latitudinal profileof the zonal wind and searching for local disturbances at the uppercloud level. For Doppler velocimetry at visible wavelengths theoptical depth reaches unity at 70 km (Ignatiev et al., 2009), whichis also the altitude studied with cloud tracking measurementsbased on imaging in the UV and thermal infrared wavelengths,with both VEx/VMC and VEx/VIRTIS instruments (Moissl et al.,2009; Sánchez-Lavega et al., 2008; Peralta et al., 2012). This allowsa direct comparison of magnitudes and spatial variations obtainedwith VLT/UVES and with Pioneer Venus, Galileo (SSI), and VenusExpress. These observations were part of the first campaign ofcoordinated in-orbit and ground-based observations of Venus, tak-ing advantage of the simultaneous presence of the Venus Expressorbiter and of the Messenger flyby in 2007. The objective of thiscampaign has been to provide unique data from measurementsthat were not possible with Venus Express, to improve the tempo-ral baseline for time-varying phenomena, to allow cross-validatingmeasurements from different techniques, and to obtain simulta-neous measurements sampling a wide range of altitudes (Lellouchand Witasse, 2008).

The observations, the data analysis method and the correctionsof the Young and geometric projection effects are described in Sec-tion 2. In Section 3 we present the results of wind measurementswith the two types of observational geometries used. The interpre-tation of the results, in particular their comparison with previousspacecraft observations and with Venus Express observations bythe VIRTIS and VMC instruments, and with other ground basedtechniques is presented in Section 4. Sensitivity studies of the dataanalysis method can be found in Appendix A.

Fig. 1. Example image from the UVES slit viewer camera, showing one of thepositions of the spectroscopic slit on the disk of Venus, for one of the offsets withthe slit perpendicular to the rotation axis.

2. Observations and data

The observations were carried out at the Very Large Telescope(VLT) at Eso’s Chilean facility in Cerro Paranal, in two nights inMay and June 2007. Since there was no critical time window, theywere performed in service mode. The UVES instrument, a cross-dispersed, high-resolution echelle spectrograph mounted on theNasmyth platform of the UT2 (Kueyen) unit of the VLT has beenused. This combination of telescope and instrument has the capa-bility of simultaneous high spectral resolving power (R � 100,000)and high spatial resolution.

The advantage of this instrument is threefold: (i) the availabilityof a large band pass (the entire 300–1100 nm wavelength rangecan be covered with the appropriate combinations of dichroics

and cross-dispersers); (ii) the large collecting area of the telescope;(iii) the small pixel size (15 lm, which represents approximately0.2 arcsec in our observations) and the mapping capability of theinstrument. The first two aspects allow a large gain in rms accuracyin retrieving the wind field, while the third allows both a simulta-neous comparison of wind velocities measured in different pointsof the disk and seeing-limited spatial resolution. Therefore, smallscale spatial variability with latitude and local solar time can bestudied.

In UVES, the beam can be split by a dichroic into two beams:one covering UV and blue wavelengths (blue arm), and anothercovering the red region of the spectrum (red arm). It is possibleto select either one or both by using the dichroic splitter. A set ofcross-dispersers provides choice for both the central wavelengthand the range covered by each arm in the observations. The obser-vations with the UVES red arm were performed at a central wave-length of 580 nm. Additional observations were made in dichroicmode at central wavelengths 437 and 860 nm (using both the blueand red arms of UVES), but were severely underexposed and werediscarded from further analysis.

The UVES red arm detector is composed of a mosaic of two CCDchips, with 2048 � 4096 pixels each (one EEV CCD-44 and oneMIT/LL CCID-20 CCD). With the CD3 cross disperser used in theseobservations, the EEV detector covers the wavelength range be-tween 480 and 570 nm in 23 spectral orders, while the MIT detec-tor covers the region between 590 and 670 nm in 16 spectralorders.

In order to cover a representative range of latitudes and longi-tudes on Venus’ dayside hemisphere (Fig. 1) a long slit configura-tion was used (0.3 arcsec-wide and 20 arcsec-long for the entireslit length). The narrow slit width compared to the large angularsize of the planet allows a direct determination of latitudinal(Fig. 2a) or longitudinal (Fig. 2b) variation of the zonal winds inboth the northern and southern hemispheres.

2.1. Description of observations

The observations were made when Venus was near greatesteastern elongation (phase angle 82.9� and 87.7�, see Table 1),

Fig. 2. Geometry of the observations, showing the various slit position offsets on the planetary disk. (a) Case PL: slit parallel to the planetary rotation axis. (b) Case PP: slitperpendicular to the rotation axis.

Table 1North pole position angle, apparent radius of Venus, latitude of sub-solar point andphase angle at the dates of the observations.

Date Positionangle (�)

Apparent radius(arcsec)

Sub-solarlatitude (�)

Phaseangle (�)

27 May2007

+7.84 10.23 �1.83 82.9

4 June2007

+10.92 11.23 �1.36 87.7

P. Machado et al. / Icarus 221 (2012) 248–261 251

allowing observation of both the planetary limb and terminatorsimultaneously. As the sub-solar point was near the limb, the cen-tral meridian (and Venus’ rotation axis) was close to the eveningterminator. The seeing conditions were good on the night of 05/27 (between 0.54 and 0.58 arcsec), but considerably poorer onthe night of 06/04 (between 1.23 and 1.54 arcsec) and sky trans-parency in the latter was also affected by passing clouds.

Two major observational configurations were set up by usingthe derotator (a prism enabling field rotation by any given positionangle relative to the celestial North–South direction). In the firstconfiguration (hereafter called PL) the slit was oriented parallelto the Venus rotation axis, whereas in the second configuration(hereafter called PP) it was aligned perpendicularly to the axis. Inaddition, the slit was offset in the East–West or the North–Southdirection in order to cover different regions of the disk (see Fig. 2).

In the PL configuration, three offsets were applied: two with theslit placed over the central meridian and centered at 7 arcsec be-low and above the equator, referred to as positions 1 and 2, anda third offset by 8 arcsec to the West with the slit centered at theequator (position 3). Three exposures were obtained for each offsetin this configuration (see Table 2).

In the PP configuration, the first offset placed the slit center10 arcsec South of the equator and 7 arcsec West of the rotationaxis, and in the following offsets the slit was moved North by 2 arc-sec (�10� latitude at low latitudes) relative to its previous position.Two exposures were obtained for each offset in this configuration.Only a subset of these offsets were used for analysis (see Table 2).

The target acquisition and guiding were done automatically,based on ephemeris data provided in advance to the telescope oper-ator. After acquisition, the position of the slit was also visually

confirmed in an image of the slit viewer camera, and maintainedvia pointing corrections performed by the secondary guiding sys-tem. Since the Venus brightness implies very short exposure times(1 s in this work, using the neutral ND3 pre-slit filter and theSHP700 filter), target drifts are negligible. VLT pointing and UVESoffset uncertainties are both equal to 0.1 arcsec (nominal value),and we take as the global positioning error of the slit

rtot ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffir2

pointing þ r2offset

q, or 0.14 arcsec. The 0.3-arcsec slit width

thus corresponds to an angle of 2.6 ± 1.2� in latitude at disk center.As Venus was an extended target and the derotator was used in

order to orient the slit in each specific configuration, the atmo-spheric dispersion corrector (ADC) was employed. The ADC partlycompensates for differential refraction in the terrestrial atmo-sphere, which is highest in the elevation direction and at the blueend of the visible range.

In order to achieve accurate wavelength calibration, to check forinstrument stability and to correct for optical slit curvature, expo-sures of the built-in Thorium–Argon (Th–Ar) lamp were taken afterthe science exposures for each offset in case PP, and before andafter each observing block in case PL. These calibration exposureswere important in correcting for the slit image curvature effect.

2.2. Data analysis and method

The data analysis method is based on the algorithm of Connes(1985) for retrieving the line of sight Doppler shift between twosets of absorption lines. For the data analysis procedure weadapted a software package previously validated in Civeit et al.(2005) and subsequently used for retrieving Titan’s winds (Luzet al., 2005a, 2006). The method has been adapted to long-slitmeasurements of a large target body and to perform severalcorrections, in particular: (a) by introducing a georeferencingcorrespondence between pixel location and geographic coordi-nates; (b) to change the size of the active sounding window ofthe slit; (c) to correct for slit image curvature; (d) to correct forthe systematic shift known as the Young effect (Y); (e) to correctfor the specific geometry of the observations; and (f) to removetelluric lines from the spectra (though this correction proved

Table 2Summary of the geometry and conditions of the observations. The cases PL and PP correspond to slit orientations parallel and perpendicular to the Venus rotation axis (slitposition angle of 7.84� and 100.92�), respectively. The offset number represents a position on the disk, N is the number of exposures taken for each offset and the times are givenat the start of the first exposure. The integration time for all exposures was of 1 s. Lat and Long are the coordinates of the slit’s central point for each offset. The latitude andlongitude values given are affected by the VLT/UVES nominal pointing and offset uncertainty, with a total uncertainty ’0.14 arcsec.

Slit Offset Date Time UT N Lat Long Airmass Seeing (arcsec)(dd-mm-yy) (hh:mm)

PL 1 27-05-2007 22:50 3 46.2S 0 2.36 0.54PL 2 27-05-2007 22:54 3 40.2N 0 2.41 0.58PL 3 27-05-2007 23:03 3 1.9S 51.5W 2.56 0.58PP 1 04-06-2007 23:02 2 34.3S 48.9W 2.38 1.25PP 2 04-06-2007 23:03 2 23.1S 42.7W 2.40 1.23PP 3 04-06-2007 23:05 2 12.6S 39.7W 2.43 1.38PP 4 04-06-2007 23:07 2 2.3S 38.6W 2.46 1.54PP 5 04-06-2007 23:09 2 7.9N 39.0W 2.49 1.35PP 6 04-06-2007 23:11 2 18.6N 41.1W 2.52 1.38

252 P. Machado et al. / Icarus 221 (2012) 248–261

negligible in comparison to other effects). A batch of new testswere also performed.

The data were first de-biased and flat-fielded based on sets offive bias and five flat-field images obtained on each night of obser-vation. Master bias and master flats were constructed by comput-ing the median of each set.

Due to a gap of 10 nm between the two CCD chips in the detec-tor, the spectra recorded in the EEV and MIT chips need to be ex-tracted and treated separately. Spectra of different spectralorders were also extracted and analyzed separately (Fig. 3). The

Fig. 3. Steps for obtaining spectra from a UVES echellogramme (example). (a) Raw echelpart of one order, where absorption lines (dark vertical bands) are visible. From each orspectra, with each one corresponding to one pixel in the slit’s active window. (d) Eachdetector. The plot shows an example of the 16 components of an MIT spectrum, each comin the Venus disk.

extraction has been performed by a spline interpolation alongthe order center. In order to avoid overlaps of spectral orders, theactive sounding window of the slit was restricted to a size of 61pixels on the detector (which corresponds to a section ofapproximately 11 arcsec of the total 20 arcsec slit length), with aspatial resolution of �108 km/pixel at disk center. Since the aver-age pixel scale for the UVES red arm is 0.183 arcsec/pixel, the spa-tial extension on the planetary disk corresponds, in fact, to arectangle with dimensions 0.183 arcsec � 0.3 arcsec. Indeed eachretrieved spectrum comes from an integrated field which is a

logramme showing the spectral orders for one of the detectors. (b) Magnification ofder, a stack of 61 spectra are extracted (the active window of the slit). (c) Set of 61

spectrum is divided into 16 orders in the MIT detector and 23 orders in the EEVing from one spectral order. (e) Example spectrum from one order and one location

(a)

−80−60−40−2000

10

20

30

40

50

60

70

Longitude (deg.)

Youn

g ef

fect

(m/s

)(b)

Fig. 4. (a) Young effect as a function of longitude from the central meridian, along the equator. The dashed line marks the position of the central pixel position of the slit. (b)Schematics of the Doppler effect in the single scattering approximation. The dotted arrow at the sub terrestrial point indicates a redshift in the absorption of solar radiation bythe atmospheric aerosols in this region, due to the atmosphere’s retrograde rotation. The dashed arrow at the sub solar point indicates a blueshift in the solar radiationscattered towards the observer. Thin arrows indicate the direction of the zonal wind. Adapted from Gabsi et al. (2008).

Fig. 5. Isolines of geometric projection factor. The Doppler shift has been computedassuming a solid body rotation of the atmosphere with an equatorial velocity of100 m/s. Note the existence of a local time meridian where the projection factor isnull due to the compensation between the redshift induced by absorption and theblueshift induced by emission towards the observer. For the June 4 observations,the bisector is at �43.85� longitude.

P. Machado et al. / Icarus 221 (2012) 248–261 253

rectangle with these dimensions. At the extraction stage, bad pix-els in the detector (pixels with an abnormal response or pixels hitby cosmic rays) were discarded by using a pre-prepared mask.

In a grating echelle spectrometer, the monochromatic image ofa long slit will appear curved. This slit image curvature, on the or-der of a fraction of a pixel, induces non-negligible errors in thewavelength calibration. This effect has been corrected by measur-ing the artificial spectral shift due to the slit image curvature, usingthe Th–Ar spectrum, and subtracting the result from the values ofthe spectral shifts (due to the motion of cloud particles) measuredon Venus. Both shifts were measured using the velocimetry algo-rithm described in Connes (1985).

The Doppler measurement is based on an optimal weighting ofthe Doppler shifts of all the lines present in the spectrum, withrelation to a reference spectrum. In this work the measured andreference spectra are taken simultaneously, with the latter being ta-ken at the center of the slit (for a detailed description of the meth-od, its implementation and applications see Connes (1985), Luzet al. (2005a, 2006) and Civeit et al. (2005)). The algorithm gener-ates line of sight Doppler shifts as a function of pixel number alongthe slit length, which is converted as a function of latitude (in thePL case) or longitude (PP case), for each exposure. We then com-puted the average of the set of two or three velocity profiles asso-ciated with each slit position offset.

The velocity profile has been measured as a weighted average ofthe shifts for the various spectral orders. The dispersion of themeasurements obtained from the different orders is the mainsource of uncertainty. The combined pointing and offset error ofthe VLT/UVES is around 0.14 arcsec. The PFS error due to the see-ing, around 0.6 arcsec in the PL case and 1.3 arcsec in the PP case,is dominant.

These profiles of Doppler shift are impacted by two observa-tional biases affecting the measurement of the zonal winds: thegeometric projection factor (F) and the Young effect (Y), which re-sult from the observational geometry and the finite solar diameteras seen from Venus. Furthermore, the line-of-sight Doppler shift ateach pixel on the planetary disk results from the combination ofthe relative Earth–Venus orbital motion (��14 km/s), the retro-grade planetary rotation (1.81 m/s at the equator, which is negligi-ble in comparison to wind velocities at low latitudes) and the cloudparticles’ motion relative to the ground. Orbital velocities were re-trieved from the NASA Horizons ephemerides website. Overall, the

absolute Doppler shift measured along the line of sight at any givenpoint in the slit is

DV ¼ F � V þ Y þ OS: ð1Þ

The orbital shift (OS) induced by the relative Earth–Venus orbitalmotion is the same for every point of the slit and cancels when rel-ative Doppler shifts are computed between two spectra acquiredsimultaneously at different points—at the sounding point and at areference point at the center of the slit. The remaining effects aredescribed below.

254 P. Machado et al. / Icarus 221 (2012) 248–261

2.2.1. Doppler shift geometric projection factorThe goal of this work is to measure the Doppler shift due to the

motion of cloud particles in the direction of the line of sight, in or-der to subsequently determine the wind velocity. The problem canbe reduced to an understanding of the interaction between solarradiation and cloud particles embedded (in the rotating atmo-sphere) during the scattering process.

We use the single-scattering approximation and consider thescattering process as a sequence of absorption and re-emission ofthe solar radiation by cloud particles in the upper cloud deck. Thissimplification allows computing a correction factor as a function oflongitude. Assuming pure zonal motion in the retrograde sense anda 90� phase angle (since Venus was near quadrature), two extremecases can be considered (see Fig. 4b). One, at the sub-solar point,where there is no Doppler shift in the absorption (null incidenceangle) but there is a blueshift in re-emission towards Earth (radia-tion is emitted in a direction which is tangent to the surface). Theother, at the sub-terrestrial point, is the converse situation wherethere is a redshift at the moment of radiation absorption (90� inci-dence angle), but no shift in the emission towards the Earth (direc-tion normal to the surface).

The exact geometric projection factor (F) affecting the Dopplershift can be calculated using the bisector theorem (Eq. (2), see Gab-si et al. (2008) for a detailed explanation). It is a function of thespecific observing geometry and of the longitude. The line of sightDoppler shift is proportional to the projection of wind velocity onthe bisector phase angle.

DV ¼ F � V ¼ V � 2cosðU=2Þsinðu�U=2Þcosb; ð2Þ

where U is the phase angle at which the observation was made, b isthe latitude of the sub-terrestrial point and u is the longitude of thepoint being measured (note that the convention applied to plane-tary longitude is that it increases in the direction of the rotation;since the Venus rotation is retrograde, the eastern and westernhemispheres are symmetric to the terrestrial case). Given the geom-etry of the observations, the projection factor vanishes at a longi-tude close to �40�, where the redshift produced in the absorptionof incoming solar radiation and the blueshift1 in the emissiontowards Earth cancel each other, Fig. 5 illustrates the effect of thegeometric projection factor.

2.2.2. Young effectYoung (1975) discussed a systematic Doppler shift affecting scat-

tered solar radiation, which is caused by the finite angular size of theSun as seen from Venus, leading to points near the terminator of Ve-nus being unequally illuminated by the approaching and recedinglimbs of the rapidly rotating Sun. The solar rotation induces a Dopp-ler shift in the solar radiation: a blueshift in the West limb of the Sun,and a redshift in the East limb. Since the solar zenith angle varieswith longitude, points of the disk at different longitudes are illumi-nated asymmetrically by the rim of the Sun that is blueshifted and bythe opposite rim of the Sun that is redshifted. The sunlight reachingVenus is already affected by this Doppler shift, which will be addedto the scattering-induced shift due to the motion of the cloud parti-cles. This effect is of the order of the Sun’s equatorial velocity(2 km s�1), multiplied by the ratio of its apparent radius as seen fromVenus, to the angular distance from the target point to the termina-tor, and can be empirically approximated by:

Y ¼ 3:2 tanðSZAÞ; ð3Þ

where SZA is the solar zenith angle (Young, 1975). The Young effectbecomes significant near the terminator (which was near the cen-tral meridian in our observations), where it increases substantially

1 For interpretation of color in Figs. 2–11, A12 and A13, the reader is referred to theweb version of this article.

(see Fig. 4). In principle, Eq. (3) would lead to Y =1 at the termina-tor. In reality, divergence does not occur because at the same timethe illumination tends to zero. Assuming to first approximation thatthe solar intensity varies as cos(SZA), we calculated the Young effectby weight-averaging Eq. (3) over the effective pixel size (i.e. the ac-tual pixel size convolved by the seeing), and finally corrected ourDoppler shift results for this effect.

3. Results

In this section we present the velocity measurements retrievedfor each of the two observing geometries, with the position angleof the slit parallel (PL) and perpendicular (PP) to the axis ofrotation.

3.1. Slit parallel to rotation axis

In the case of the PL configuration, with the spectroscopic slitparallel to the planetary rotation axis, three offsets were used, withthree exposures for each one. In this case, due to the latitudinallyvarying zonal wind along the slit, the method yields the relativeDoppler shift and zonal velocity between each pixel along the slitand the central pixel, but not the absolute zonal velocity.

For the third position offset (measurements near the limb), wecan take advantage from the fact that, due to the low obliquity, pix-els which are symmetric relative to the slit center fall on the samemeridian. Therefore they are subject to the same projection factor,allowing to measure the latitudinal asymmetry of the zonal wind.On the other hand, for positions 1 and 2, the observational scans donot cover North and South latitudes simultaneously due to the slitlength limitation.

The observations obtained on May 27 had a signal-to-noise ra-tio between 40 and 80 (PL geometry), much higher than the spectrafrom June 4 (SNR between 2 and 6, PP geometry). The reason wassky transparency variations and passing clouds affecting the signal.The better conditions allowed retrieving Doppler shifts with errorbars of just a few m/s in the former date, since the noise levelcan have a considerable impact in the accuracy of this technique(see Appendix A for a study of the impact of noise level on theresults).

In positions 1 (latitude range: �79–12�S) and 2 (�8–78�N), allthe points sounded on the disk are along the central meridian,hence affected by the same projection factor. Therefore, after sub-tracting the Young effect, and assuming a purely zonal wind, wecan obtain the differential velocity between each point along theslit and the central point.

Fig. 6 presents Doppler shift curves for one of the exposures foroffset 1. Fig. 6a shows the differential Doppler shifts between thespectrum coming from each pixel along the slit and the referencespectrum. The vertical dotted line represents the apparent plane-tary limb near the South pole. The curves from the two parts ofthe detector are consistent, even in some of the small scale fea-tures, except for pixels falling outside the disk, where they diverge.Pixels to the right of the plot are in fact sounding the sky and pro-duce spurious Doppler shifts without any physical meaning. InFig. 6b we performed the Doppler shifts’ weighted average fromthe three exposures taken at this offset, the pixel positions havebeen converted to latitude and spectra from the portion of the slitfalling outside the planetary disk were discarded.

Due to the high spatial resolution of the observations (pixelscale is on the order of 108 km at disk center) small scale featurescan be seen in the velocity curves that indicate higher relativevelocities on the range �10–30 m/s over spatial scales of approxi-mately 216 km (two pixels, such as at 60�S). These have been

(b)

−80−70−60−50−40−30−20−10−50

0

50

100

Latitude (deg.)

Shift

(m/s

)

(a)

Fig. 6. Example of velocity curves produced in two stages of the data analysis, from one of the exposures in slit position 1 (case PL, night of May 27). (a) Doppler shift as afunction of the pixel. The dashed line marks the position of the slit’s center point used as the reference spectrum. The dotted line marks the position of the southern limb.Doppler shifts retrieved from the MIT CCD are shown in red, and from the EEV CCD are shown in blue. (b) Velocity curves mapped to planetocentric latitude. The latitudinalcoverage of the effective pixel increases toward higher latitudes due to the planetary curvature.

P. Machado et al. / Icarus 221 (2012) 248–261 255

measured in spectra recorded simultaneously on both detectorsand are unlikely to be instrumentally generated.

The fact that the pixels near the limb are affected differently forthe two CCDs, as we can see in Fig. 6 close to the limb between75�S and 85�S, may be caused by the difficulty to correct for theYoung effect near the terminator (which for high latitudes is alsonear the limb) in the presence of multiple scattering. We haveadopted a purely geometric correction of the Young effect (Eq.(3)), weighted by the illumination. If multiple scattering occurs,the scattered intensity is not proportional to the illumination, soit is likely that our description of the correcting factor fails nearthe northern and southern limbs. Furthermore, differences in thealtitudes sounded for the two wavelength ranges covered by eachCCD detector, leading to different wind retrievals due to a possiblevertical wind shear, may be another possible reason for these dif-ferences. It is clear that the technique may be improved to extendmeasurements closer to the limb if a more realistic approach to thenature of the scattering process, allowing to accurately computethe contribution function for the wavelength range of each detec-tor, is introduced.

Fig. 7 shows the final relative velocity curves from all measure-ments in case PL. In addition to Fig. 6b, deprojection and subtrac-tion of the Young effect Doppler shift was made, and a weightedaverage was computed of the results from the three exposures ineach offset and for both detectors. These curves represent relativezonal velocity (measured relative to the reference point defined asthe center of the active window), and therefore require an absolutebaseline value to be interpreted, which is provided by measure-ments made with the PP geometry. This will be done in the discus-sion of Fig. 11.

In slit position 3 (latitude range: 33�S–31�N), any two pointswith symmetric latitudes are also on the same meridian and solarzenith angle. Therefore, their projected velocities can be correctedfor the same projection factor and Young effect, which allowsderiving the differential velocity between symmetric points inthe northern and southern portions of the slit. In this case it is pos-sible to compare the wind in low northern and southern latitudes(Fig. 8). The maximum asymmetry in the equatorial region is6 ± 5 m/s. On average, southern zonal winds are faster than north-ern ones by less than 5 m/s, and within error bars the asymmetry isnegligible.

3.2. Slit perpendicular to rotation axis

Observations made with the slit orientation perpendicular tothe planetary rotation axis allow measuring the absolute magni-tude of the wind velocity. In this case pixels along the slit fall onthe same latitude circle, and assuming that the zonal wind dependsonly on latitude one can obtain the zonal wind directly from thelongitudinal profile of the relative velocity (see Eqs. (1) and (2)).This requires, however, a correction for geometric projection andYoung effects, since the measured spectrum and the referencespectrum (spectrum from the slit’s center) are not at the sameSZA. Since the geometric projection factor falls to zero and changessign close to the center of the slit, the uncertainty of the measure-ment increases significantly for points in this region, because thecorrection is made by dividing the relative shift measured by thegeometric projection factor.

Fig. 9 and Table 3 present the zonal wind magnitudes obtainedin this case for the six slit offsets, between 18�N and 34�S. Each pa-nel in Fig. 9 shows the weighted average of the velocity curves ob-tained from the MIT and EEV CCDs and from the two exposurestaken for each slit offset. These results are also displayed in amap projection in Fig. 10. The mean velocities range from 106 to127 m/s, with an uncertainty that is directly correlated with theSNR of the observation, but which also has a contribution fromlocal variations. In general, the zonal wind is approximatelyuniform at each latitude offset (validating our initial assumptionto first order), with stronger winds close to the equator than athigher latitudes. However, at longitudes between �25� and �35�,non-negligible medium to large scale oscillations, spanningapproximately two pixels, or �216 km, are apparent for slit posi-tions between 8N and 23S (Fig. 9 panels b–e). The magnitude ofthese (second order) oscillations cannot be quantified directly,however, since our retrievals are based on the working hypothesisthat the zonal wind does not depend on longitude.

Although we also made observations at latitudes higher thandisplayed in Table 3, they have been discarded because of thelow signal to noise ratio, which made it impossible to obtain theDoppler shifts. The same problem of high noise affected pixels atlongitudes near the limb, as can be seen in the size of the errorbars. Thus the pixels East of the slit center were not consideredfurther.

(a)

−80 −70 −60 −50 −40 −30 −20 −10−100

−80

−60

−40

−20

0

20

Latitude (deg.)

Shift

(m/s

)

(b)

0 10 20 30 40 50 60 70 80 90−100

−80

−60

−40

−20

0

20

Latitude (deg.)

Shift

(m/s

)

(c)

−40 −30 −20 −10 0 10 20 30 40−20

−15

−10

−5

0

5

10

15

20

Latitude (deg.)

Shift

(m/s

)

Fig. 7. Differential velocities between each pixel and the central pixel of the slit,deprojected and corrected for the Young effect, for the observations in the PLorientation (night of May 27). The plots show the weighed average of the curvesobtained with the MIT and EEV CCDs. (a) Slit position 1, covering latitudes from79�S to 12�S. (b) Position 2, 8�N to 78�N. (c) Position 3, 33�S to 31�N. The dashed lineindicates the reference point at the center of the slit. Note different scales.

0 10 20 30−20

−10

0

10

20

Degrees of latitude from equator

Rel

ativ

e he

mis

pher

ic s

hift

(m/s

)

Fig. 8. Relative velocity between points symmetric relative to equator (case PL,night of May 27). A weighted average has been computed of the results from theMIT and EEV detectors and from the three exposures in this offset.

256 P. Machado et al. / Icarus 221 (2012) 248–261

3.3. Analysis of spatial variations

As explained above, the observing geometry with the slit ori-ented parallel to the axis of rotation (case PL) yields relative windmeasurements, whereas a perpendicular slit (case PP) providesabsolute wind measurements. Therefore, it is possible to use themean values of the winds measured for each offset of the PP obser-vations to calibrate the baseline of the PL observations, i.e., thewind at the reference point. Fig. 11 summarizes the results after

doing this. The first step of the procedure has been to calibratethe relative velocities from offset PL3, which covers the equatorialregion in both hemispheres, using the average along the slit for thesix offsets from the PP geometry (marked in the figure as blacksquares). The second step has been to fit the relative velocitycurves from offsets PL1 and PL2 to the curve from offset PL3 intheir common (overlapping) ranges of latitude. This allows obtain-ing a simultaneous latitudinal profile extending from high north-ern to high southern latitudes.

The zonal wind is stronger in mid-latitudes than in the equato-rial region by 10–20 m/s. These are moderate jets, with the south-ern one being stronger than its northern counterpart by �10 m/s.The latitudinal wind shear is not very pronounced (�0.35 m/s perdegree at low latitudes), and stronger shear (�3 m/s per degree)is limited to narrow regions at 60�S, 10�S, 30�N and 40�N.

4. Discussion

Doppler velocimetry based on long slit spectroscopy has en-abled the first ground-based, direct measurement of the latitudinalprofile of instantaneous zonal wind in the atmosphere of Venus.The technique, based on the method of absolute accelerometry de-vised by Connes (1985), allowed (1) the simultaneous determina-tion of the zonal wind across a wide range of latitudes and localtimes and (2) to determine the presence of local, small scale vari-ations relative to the mean. The combination of the 8-m class VeryLarge Telescope UT2 with the UVES spectrograph allowed the re-trieval of the solar Fraunhofer lines in Venus’ scattered spectrumwith a spectral resolution of �105 in the visible range, leading toa precision of a few m/s in zonal wind retrievals (but dependingstrongly on the signal-to-noise ratio) and with a spatial resolutionof �108 km/pixel at disk center. The observations made with theslit perpendicular to the rotation axis all have a much lower SNR(4–10, caused by intermittent high clouds at Paranal) than the onesperformed with the parallel slit (40–80); this is the main reason forthe larger error bars in the former case.

Our Doppler retrievals (Fig. 11 and Table 3) are in general goodagreement with previous measurements based on cloud tracking(Del Genio and Rossow, 1990; Limaye, 2007; Peralta et al., 2007;Sánchez-Lavega et al., 2008; Moissl et al., 2009). We retrieve thesame order of magnitude and a similar latitudinal variation to Pio-neer Venus, Galileo and VEx/VIRTIS measurements, which cross-validates both techniques and provides reasonable confirmationthat cloud tracking and Doppler methods both retrieve the veloci-ties of air masses to first order. Wind magnitudes at low latitudes

(a)

−45−40−35−30−25−20−15−10−50

0

50

100

150

200

250

300

Longitude (deg.)

Zona

l vel

ocity

(m/s

)(b)

−45−40−35−30−25−20−15−10−50

0

50

100

150

200

250

300

Longitude (deg.)

Zona

l vel

ocity

(m/s

)

(c)

−45−40−35−30−25−20−15−10−50

0

50

100

150

200

250

300

Longitude (deg.)

Zona

l vel

ocity

(m/s

)

(d)

−45−40−35−30−25−20−15−10−50

0

50

100

150

200

250

300

Longitude (deg.)

Zona

l vel

ocity

(m/s

)

(e)

−45−40−35−30−25−20−15−10−50

0

50

100

150

200

250

300

Longitude (deg.)

Zona

l vel

ocity

(m/s

)

(f)

−45−40−35−30−25−20−15−10−50

0

50

100

150

200

250

300

Longitude (deg.)

Zona

l vel

ocity

(m/s

)

Fig. 9. Zonal wind versus longitude for the observations made at June 4 with the spectroscopic slit parallel to the equator (case PP). Each point is the weighted mean of theretrievals from two exposures and from the EEV and MIT detectors. The latitudes are: (a) 33.5S; (b) 23S; (c) 13S; (d) 2S; (e) 8N; and (f) 18N.

P. Machado et al. / Icarus 221 (2012) 248–261 257

(perpendicular slit configuration case), however, are about 20–30 m/s higher than these previous cloud tracking measurements.The cause for this (second order) discrepancy is twofold, and stemsfrom the differences between the two methods. On the one hand,latitudinal wind profiles based on cloud tracking are constructedby zonally averaging sets of many measurements of the velocitiesof individual cloud features, made over a wide range of longitudes.On the other hand, with our Doppler technique we directly measurethe instantaneous velocity of cloud particles within a particularrange of local solar time (between 15 h and 17 h) close to the even-ing terminator. Since our measurements only cover this limitedrange of local times, it is impossible to produce a zonal mean to di-rectly compare with cloud tracking results. It is likely that this local

character of Doppler measurements accounts for the discrepancy,due to the presence of thermal tides. Del Genio and Rossow(1990) and Rossow et al. (1990) detected a solar-locked componentwith a 10 m/s amplitude at low latitudes which peaked close to theevening terminator, and Sánchez-Lavega et al. (2008) measuredwinds stronger at 15 h than at noon by �10 m/s, close to 40S. Thisthermal tide contribution would be averaged out if data were avail-able over a complete longitude circle, but local measurements suchas ours are sensitive to solar-locked extrema.

The latitudinal profiles measured with the parallel slit geometryare also local and instantaneous, and therefore the sameconsiderations apply. Nonetheless, there is a good agreement withthe VEx/VMC (Moissl et al. (2009), Fig. 4), Galileo (Peralta et al.

Table 3Summary of latitudes covered, and mean zonal wind velocities measured with the PPgeometry on June 4 (Fig. 9). The zonal velocities are the weighted means of themeasurements made at each offset. The weighting coefficients were the inverse of thevariance obtained for each pixel. Std. dev. is the standard deviation of the weightedmean. Latitudes are given for the slit center with an uncertainty of 1.2�; the0.3 arcsec-slit covers a latitude band 2.4� at disk center.

Latitude Zonal velocity (m/s) Std. dev. (m/s)

18� North 116 538� North 106 212� South 127 1413� South 126 1123� South 122 1433.5�South 116 29

0° 10° 20° 30° 40° 50° 60° 80°E

90° S

60° S

45° S

30° S

15° S

0°

15° N

30° N

45° N

60° N

75° N

0

20

40

60

80

100

120

140

160

180

200

Fig. 10. Map projection of the zonal wind for the observations of June 4. Thelatitudinal extent of the slit pixels has been exaggerated for clarity. Estimated pixeldimensions are approximately 108 km in longitude at disk center. See Table 3 andFig. 9 for the latitudes and longitudes covered.

−80 −60 −40 −20 0 20 40 60 800

20

40

60

80

100

120

140

160

180

200

Latitude (deg.)

Velo

city

(m/s

)

Offset PL 1Offset PL 2Offset PL 3PP Geometry

Fig. 11. Summary of the measurements made with the slit perpendicular to therotation axis (zonal wind velocity, squares) and parallel (relative zonal windmeasurements, lines). Curve PL3 (green) has been offset vertically in order to best-fit the absolute zonal wind measurements closest to the equator. Curves PL1 (blue)and PL2 (red) were offset in order to align them with curve PL3 in their regions ofoverlap.

258 P. Machado et al. / Icarus 221 (2012) 248–261

(2007), Fig. 3) and Pioneer Venus (Del Genio and Rossow (1990),Fig. 12 and Rossow et al. (1990), Fig. 11) cloud tracking measure-ments, whereas the VEx/VIRTIS profile (Sánchez-Lavega et al.,2008, Fig. 2) has somewhat lower winds at high latitudes. Thismay be due to an incomplete correction of the Young effect, whichis quite strong near the terminator where these velocities were re-trieved, and in particular at high latitudes.

Previous measurements by Pioneer Venus, reanalyzed inLimaye (2007), indicated the presence of high latitude zonal jetsclose to 50� latitude in both hemispheres, with a slight asymmetrybetween the northern and southern ones. However, longer

temporal averages of cloud-tracked winds by the Galileo SSI instru-ment (Peralta et al., 2007), and by the Venus Express VMC and VIR-TIS instruments (Sánchez-Lavega et al., 2008; Moissl et al., 2009)do not display any clear evidence for high latitude jets at cloudtops, although shorter time scale averages of VMC measurementsin Moissl et al. (2009) indicate that jets may occur but are shortlived. We believe that, rather than being discrepant, different windmeasurements provide important insight into the variability inher-ent to the atmospheric circulation of Venus. Our direct measure-ments of instantaneous zonal winds provide additional evidencefor the occasional presence of jets and for variability in general.The realization that a latitudinal wind profile with jets such asmeasured by Pioneer Venus is likely to be barotropically unstable(Limaye et al., 2009) is an argument in favor of variability, sincesuch a profile generates wave motions that carry zonal momentumfrom the jets and into slow-moving latitudes, thus reducing theinstability. Our observations of small scale perturbations (Figs. 9and 10) seem to point to small scale wave motions as the mostlikely processes by which the instability unfolds. Further investiga-tions regarding wave motions and their latitudinal fluxes of angu-lar momentum should privilege the midlatitudes as the focus ofthis activity. The more general question remains concerning therole of the mean meridional circulation on the buildup of these jets,and their link to long-lived structures such as the polar vortex, itssurrounding cold collar and the extended clouds at high latitudes.

Acknowledgments

This article is based on observations collected at the EuropeanOrganisation for Astronomical Research in the Southern Hemi-sphere, Chile (run 079.C-0344). The authors are grateful to Dr. ToddClancy and an anonymous reviewer for their careful reading of themanuscript. They are also grateful to Drs. D. Berry and J. Peralta forhelpful discussions and to Dr. F. Patat (ESO) and ESO staff for theirhelp with the data acquisition. Fig. 2a and b courtesy of the Institutde mécanique céleste et de calcul des éphémérides (http://www.imc-ce.fr). P. Machado acknowledges the support of the Observatoire deParis-LESIA, and of the Portuguese Foundation for Science andTechnology (FCT, Ph.D. Grant Reference: SFRH/BD/66473/2009).P.M., D.L., and T.W. also acknowledge FCT funding through ProjectGrants POCI/CTE-AST/110702/2009, PEst-OE/FIS/UI2751/2011 andPessoa-PHC programme.

(a)

−30 −20 −10 0 10 20 30−300

−200

−100

0

100

200

300

Slit region (pixel)

Shift

(m/s

)(b)

−30 −20 −10 0 10 20 30−300

−200

−100

0

100

200

300

Slit region (pixel)

Shift

(m/s

)

Fig. A.12. Sensitivity to the signal-to-noise ratio. The plots show the velocity curves relative to the central spectrum, with prescribed shifts between �300 m/s and +300 m/s.The red and blue curves are from the MIT and EEV detector, respectively. (a) Order reconstructed from a spectrum from the order center, SNR = 54. (b) Spectrum withSNR = 12. Both spectra are from one of the exposures from offset 1, case PL.

−30 −20 −10 0 10 20 30−300

−200

−100

0

100

200

300

Slit region (pixel)

Shift

(m/s

)

Fig. A.13. Doppler shift curve for spectra shifted with a 10 m/s step relative to thecentral spectrum, between �300 m/s and +300 m/s. This curve has been retrievedfrom a synthetic spectrum.

P. Machado et al. / Icarus 221 (2012) 248–261 259

Appendix A. Sensitivity tests

In order to test the robustness of the algorithm relative to anumber of variables, a batch of sensitivity tests have been per-formed. The variables investigated were the noise level, the densityof spectral lines, the symmetry of lines, the wavelength range andthe presence of telluric lines. The tests were made, on one hand,from reconstructed spectra based on the observed Venus spectraand, on the other, from fully synthetic spectra, to which a prede-fined Doppler shift was imposed in both cases. Removal of the tel-luric lines from the observed spectra showed this factor to have anegligible impact on the velocity curves. The remaining tests aredescribed below.

A.1. Tests based on reconstructed spectra

In these tests, all spectra in each order of the echellogram werereplaced with the reference (central pixel) spectrum, therebyreconstructing the spatial structure along the slit in order to re-move the relative shift that was originally present in the data.We then used these reconstructed data as input to the retrievalalgorithm.

In the first test the first step consisted of replacing all the 61spectra by the central spectrum, to ensure a relative null velocitybetween each of the spectra and the reference spectrum. The algo-rithm retrieved null Doppler shifts, as expected. In the second testall the spectra except the reference one were shifted by a pre-scribed shift. The spectra were shifted with 10 m/s increments, be-tween �300 m/s and 300 m/s from top to bottom of the slit. Thistest confirmed that the velocity curves for the reconstructed spec-tra were consistent with the imposed deviations (Fig. A.13). On theother hand, reconstructing the orders from a spectrum with a low-er signal to noise ratio led to retrieved velocities lower than theprescribed ones (Fig. A.12). The error bars were also higher for testswith spectra from the MIT detector, due to the higher noise of theobservations at longer wavelengths in this observational offset(Fig. A.12b). Additional tests with lower SNR spectra, from observa-tions of June 4, confirmed the sensitivity of the velocity curves tonoise.

To test the algorithm’s response to wavelength range, recon-structed spectra were also produced in which only one of the spec-tral orders was reconstructed as previously described, while allother orders were replaced by an average of the continuum level.Since there was no significant discrepancy between velocity curvesobtained from different spectral orders, we conclude that ourmethod is robust with respect to the wavelength range of the data.This is not surprising since the density of spectral lines has littlevariation between spectral orders.

A.2. Tests based on synthetic spectra

These tests were based on fully synthetic spectra constructedfrom gaussian line shapes. The reference spectrum was con-structed by defining the line width, depth and density (i.e. theaverage number of spectral lines per nanometer) so as to matchthose of the data, and adding random noise at appropriate level.The remaining spectra within each order were defined by replicat-ing this reference spectrum. The tests were performed by numeri-cally shifting each spectrum by a prescribed shift relative to thereference one.

As in the case of the spectra reconstructed from real data, threetypes of fully synthetic spectra were generated: one with a null rel-ative shift, one with a 100 m/s shift between all the spectra in eachorder and the central spectrum, and a third one with a relative shiftof 10 m/s between consecutive spectra, ranging from �300 m/s to300 m/s. In the baseline case, in which the synthetic spectra weresimilar to the data, the velocity curves were consistent with the

260 P. Machado et al. / Icarus 221 (2012) 248–261

imposed shift (see Fig. A.13). The impact of test variables such asthe noise level, the line density and the symmetry of the spectrallines was studied by varying each of these factors independently.Decreasing the signal to noise ratio or decreasing the line densityboth led to velocity curves lower than the prescribed shift, witha higher discrepancy, in velocity, for spectra farther from the refer-ence one. Increasing the noise level by a factor of four leads to a20% reduction in the retrieved Doppler shift, pixel by pixel, in thevelocity profiles obtained from the synthetic spectra relative tothe baseline case.

Changing the line symmetry, by widening each line in its redside (by increasing the full width at half maximum of the gaussianin its high wavelength side from the line center) led to highervelocities, whereas wider lines in the blue side led to velocitieslower than the imposed Doppler shift. Line symmetry is an impor-tant factor in the visible wavelength range, since it can be affectedby differential atmospheric refraction if not totally compensatedby the ADC prism.

References

Baker, D., Schubert, G., 1992. Cellular convection in the atmosphere of Venus.Nature 355, 710–712.

Baranne, A. et al., 1996. A spectrograph for accurate radial velocity measurements.Astron. Astrophys. Suppl. Ser. 119, 373–390.

Belton, M.J.S. et al., 1991. Images from Galileo of the Venus cloud deck. Science 253,1531–1536.

Bougher, S., Hunten, D., Philips, R., 1997. Upper atmosphere dynamics. In: Tucson,A.Z., Bougher, S.W., Hunten, D.M., Phillips, R.J. (Eds.), Global Circulation andGravity Waves, Venus II: Geology, Geophysics, Atmosphere, and Solar WindEnvironment. University of Arizona Press, pp. 259–291.

Boyer, Ch., Camichel, H., 1961. Observations photographiques de la planète Vénus.Ann. Astrophys. 24, 531–535.

Brecht, A.S. et al., 2012. Atomic oxygen distributions in the Venus thermosphere:Comparisons between Venus Express observations and global simulations.Icarus 217, 759–766.

Civeit, T. et al., 2005. On measuring planetary winds using high-resolutionspectroscopy in visible wavelengths. Astron. Astrophys. 431, 1157–1166.

Clancy, R.T., Sandor, B.J., Moriarty-Schieven, G.H., 2008. Venus upper atmosphericCO, temperature, and winds across the afternoon/evening terminator from June2007 JCMT sub-millimeter line observations. Planet. Space Sci. 56, 1344–1354.

Clancy, R.T., Sandor, B.J., Moriarty-Schieven, G.H., 2012. Circulation of the Venusupper mesosphere/lower thermosphere: Doppler wind measurements from2001–2009 inferior conjunction, sub-millimeter CO absorption lineobservations. Icarus. 217, 794–812. http://dx.doi.org/10.1016/j.icarus.2011.05.021.

Connes, P., 1985. Absolute astronomical accelerometry. Astrophys. Space Sci. 110(2), 211–255, ISSN:0004-640-X.

Del Genio, A.D., Rossow, W.B., 1990. Planetary-scale waves and the cyclic nature ofcloud top dynamics on Venus. J. Atmos. Sci. 47, 293–318.

Dollfus, A. et al., 1975. Photometry of Venus. I – Observations of the brightnessdistribution over the disk. Icarus 26, 53–72.

Gabsi, Y. et al., 2008. Measuring Venus winds using the absolute astronomicalaccelerometer: Solid super-rotation model of Venus clouds. Planet. Space Sci.56, 1454–1466.

Gaulme, P., Schmider, F.-X., Grec, C., Ariste, A.L., Widemann, T., Gelly, B., 2008. Venuswind map at cloud top level with the MTR/Themis visible spectrometer,instrumental performance and first results. Planet. Space Sci. 56, 1335–1343,Special issue on ground-based measurements in support of Venus Express.

Gérard, J.-C. et al., 2009. Concurrent observations of the ultraviolet nitric oxyde andinfrared O2 nightglow emissions with Venus Express. J. Geophys. Res. 114,E00B44.

Gierasch, P., 1975. Meridional circulation and Venus atmospheric rotation. J. Atmos.Sci. 32, 1038–1044.

Gierasch, P. et al., 1997. The general circulation of the Venus atmosphere: Anassessment. In: Tucson, A.Z., Bougher, S.W., Hunten, D.M., Phillips, R.J. (Eds.),Venus II: Geology, Geophysics, Atmosphere, and Solar Wind Environment.University of Arizona Press, pp. 459–500.

Goldstein, J. et al., 1991. Absolute wind velocities in the lower thermosphere ofVenus using infrared heterodyne spectroscopy. Icarus 94, 45–63.

Hide, R., 1969. Dynamics of the atmospheres of the major planets with an appendixon the viscous boundary layer at the rigid bounding surface of an electrically-conducting rotating fluid in the presence of a magnetic field. J. Atmos. Sci. 26,841–853.

Hueso, R., Peralta, J., Sánchez-Lavega, A., 2012. Assessing the long-term variability ofVenus winds at cloud level from VIRTIS-Venus Express. Icarus 217, 585–598.

Ignatiev, N.I. et al., 2009. Altimetry of the Venus cloud tops from the Venus Expressobservations. J. Geophys. Res. 114, E00B43. http://dx.doi.org/10.1029/2008JE003320.

Kostiuk, T. et al., 2001. Direct measurement of winds on Titan. Geophys. Res. Lett.28, 2361–2364.

Kostiuk, T. et al., 2005. Titan’s stratospheric zonal wind, temperature, and ethaneabundance a year prior to Huyghens insertion. Geophys. Res. Lett. 32, 4.

Lellouch, E. et al., 1994. Global circulation, thermal structure, and carbon monoxidedistribution in Venus’ mesosphere in 1991. Icarus 110 (2), 315–339.

Lellouch, E. et al., 1997. Monitoring of mesospheric structure and dynamics. In:Bougher, S.W., Hunten, D.M., Phillips, R.J., Tucson, A.Z. (Eds.), Venus II. TheUniversity of Arizona Press, pp. 295–324.

Lellouch, E. et al., 2008. Monitoring Venus’ mesospheric winds in support of VenusExpress: IRAM 30-m and APEX observations. Planet. Space Sci. 56, 1355–1367.

Lellouch, E., Witasse, O., 2008. A coordinated campaign of Venus ground-basedobservations and Venus Express measurements. Planet. Space Sci. 56, 1317–1319.

Limaye, S.S., 2007. Venus atmospheric circulation: Known and unknown. J.Geophys. Res. 112, E04S09. http://dx.doi.org/10.1029/2006JE002814.

Limaye, S.S. et al., 2009. Vortex circulation on Venus: Dynamical similarities withterrestrial hurricanes. Geophys. Res. Lett. 36, L04204. http://dx.doi.org/10.1029/2008GL036093.

Limaye, S.S., Suomi, V.E., 1981. Cloud motions on Venus – Global structure andorganization. J. Atmos. Sci. 38, 1220–1235.

Luz, D. et al., 2005a. Characterization of zonal winds in the stratosphere of Titanwith UVES. Icarus 179, 497–510.

Luz, D. et al., 2005b. UVES measurements of Saturn’s winds with the absoluteaccelerometry technique. Geophys. Res. Abstracts 7 (SRef-ID: 1607-7962/5).

Luz, D. et al., 2006. Characterization of zonal winds in the stratosphere of Titan withUVES: 2. Observations coordinated with the huygens probe entry. J. Geophys.Res. 111 (CiteID E08S90).

Luz, D. et al., 2011. Southern polar vortex reveals precessing circulation. Science332, 577–580.

Markiewicz, W. et al., 2007. Morphology and dynamics of the upper cloud layer ofVenus. Nature 450, 633–636.

Markiewicz, W. et al., 2007. Monitoring camera for Venus Express. Planet. Space Sci.55, 1701–1711.

Martic, M. et al., 1999. Evidence for global pressure oscillations on Procyon. Astron.Astrophys. 351, 993–1002.

Martic, M., Lebrun, J.-C., Schmitt, J., Lebreton, J.-P., Appourchaux, T., 2001. FeasibilityDemonstration of Titan Wind Measurement Technique at OHP using absoluteAccelerometry Method. Technical Report. Observatoire de Haute Provence.

McGouldrick, K., Momary, T.W., Baines, K.H., Grinspoon, D., 2012. Quantification ofmiddle and lower cloud variability and mesoscale dynamics from VenusExpress/VIRTIS observations at 1.74 lm. Icarus 217, 615–628.

Moissl, R. et al., 2009. Venus cloud top winds from tracking UV features in Venusmonitoring camera. J. Geophys. Res. 114, E00B31.

Mueller, N.T., Helbert, J., Erard, S., Piccioni, G., Drossart, P., 2012. Rotation period ofVenus estimated from Venus Express VIRTIS images and Magellan altimetry.Icarus 217, 474–483.

Mueller-Wodarg, I.C.F., Forbes, J.M., Keating, G.M., 2006. The thermosphere of Venusand its exploration by a Venus Express accelerometer experiment. Planet. SpaceSci. 54, 1415–1424.

Ohtsuki, S. et al., 2008. Distributions of the Venus 1.27-lm O2 airglow androtational temperature. Planet. Space Sci. 56, 1391–1398.

Peralta, J., Hueso, R., Sánchez-Lavega, A., 2007. A reanalysis of Venus winds at twocloud levels from Galileo SSI images. Icarus 190, 469–477.

Peralta, J. et al., 2012. Solar migrating atmospheric tides in the winds of the polarregion of Venus. Icarus 220, 958–970.

Piccialli, A. et al., 2008. Cyclostrophic winds from the visible and infrared thermalimaging spectrometer temperature sounding: a preliminary analysis. J.Geophys. Res. 113, E00B11.

Piccialli et al., 2012. Dynamical properties of the Venus mesosphere from theradio-occultation experiment VeRa onboard Venus Express. Icarus 217, 669–681.

Piccioni, G. et al., 2007. South-polar features on Venus similar to those near theNorth pole. Nature 450, 637–640.

Rengel, M., Hartogh, P., Jarchow, C., 2008. Mesospheric vertical thermal structureand winds on Venus from HHSMT CO spectral-line observations. Planet. SpaceSci. 56, 1368–1384.

Rossow, W., Williams, G., 1979. Large-scale motion in the Venus stratosphere. J.Atmos. Sci. 36, 377–389.

Rossow, W., Del Genio, A., Limaye, S., Travis, L., 1980. Cloud morphology andmotions from Pioneer Venus images. J. Geophys. Res. 85, 8107–8128.

Rossow, W., Del Genio, A., Eichler, T., 1990. Cloud tracked winds from Pioneer VenusOCPP images. J. Atmos. Sci. 47, 2053–2084.

Sánchez-Lavega, A. et al., 2008. Winds on Venus mapped in three dimensions.Geophys. Res. Lett. 35, L13204.

Sonnabend, G., Sornig, M., Krtz, P.J., Schieder, R.T., Fast, K.E., 2006. High spatialresolution mapping of Mars mesospheric zonal winds by infrared heterodynespectroscopy of CO2. Geophys. Res. Lett. 33 (CiteID L18201).

Sonnabend, G., Sornig, M., Scheider, R., Kostiuk, T., Delgado, J., 2008. Temperaturesin Venus upper atmosphere from mid-infrared heterodyne spectroscopy of CO2

around 10 lm wavelength. Planet. Space Sci. 56, 1407–1413 (Special issue onground-based measurements in support of Venus Express).

Sonnabend, G., Kroetz, P., Sornig, M., Stupar, D., 2010. Direct observations of Venusupper mesospheric temperatures from ground based spectroscopy of CO2.Geophys. Res. Lett. 37 (CiteID L11102).

P. Machado et al. / Icarus 221 (2012) 248–261 261

Soret, L. et al., 2010. Venus OH nightglow distribution based on VIRTIS limbobservations from Venus Express. Geophys. Res. Lett. 37, L06805.

Soret, L. et al., 2012. Atomic oxygen on the Venus nightside: Global distributiondeduced from airglow mapping. Icarus 217, 849–855. http://dx.doi.org/10.1016/j.icarus.2011.03.034.

Sornig, M. et al., 2008. Venus upper atmosphere winds from ground-basedheterodyne spectroscopy of CO2 at 10 lm wavelength. Planet. Space Sci. 56,1399–1406.

Sornig, M. et al., 2012. Direct wind measurements from November 2007 in Venus’upper atmosphere using ground-based heterodyne spectroscopy of CO2 at10 lm wavelength. Icarus 217, 863–874. http://dx.doi.org/10.1016/j.icarus.2011.03.019.

Suomi, V., Limaye, S., 1978. Venus – Further evidence of vortex circulation. Science201, 1009–1011.

Svedhem, H., Titov, D., Taylor, F., Witasse, O., 2007. Venus as a more Earth-likeplanet. Nature 450, 629–632.

Taylor, F. et al., 1980. Infrared remote sensing of the atmosphere of Venus from thePioneer 12 orbiter, space research. In: Proceedings of the Open Meetings of theWorking Groups on Physical Sciences, Bangalore, India, vol. 20. Pergamon Press,Oxford and Elmsford, NY, pp. 227–230.

Tellmann, S., Paetzold, M., Hausler, B., Bird, M.K., Tyler, G.L., 2009. Structure ofVenus neutral atmosphere as observed by the radio science experiment VeRa onVenus Express. J. Geophys. Res. (Planets) 114, E00B36.

Titov, D., Bullock, M., Crisp, D., Renno, N., Taylor, F., Zasova, L., 2007. Radiation in theatmosphere of Venus. Geophys. Monogr. Ser. 176. http://dx.doi.org/10.1029/176GM08.

Toigo, A., Gierasch, P., Smith, M., 1994. High-resolution cloud feature tracking onVenus by Galileo. Icarus 109 (2), 318–336.

Traub, W., Carleton, N., 1979. Retrograde winds on Venus – Possible periodicvariation. Astrophys. J. Pt. 1 227, 329–333.

Widemann, T. et al., 2007. New wind measurements in Venus lower mesospherefrom visible spectroscopy. Planet. Space Sci. 55, 1741–1756.

Widemann, T. et al., 2008. Doppler winds at cloud tops observed with ESPaDOnS atCFHT. Planet. Space Sci. 56, 1320–1334.

Young, A., 1975. Is the four-day ‘‘Rotation’’ of Venus illusory? Icarus 24, 1–10.Young, A., Schorn, R., Young, L., Crisp, D., 1979. Spectroscopic observations of winds

on Venus, I-technique and data reduction. Icarus 38, 435–450.Young, E., Bullock, M., Limaye, S., Bailey, J., Tsang, C., 2010. Evidence for and against

8-day planetary waves. Ground-based Cloud-Tracking Observations of Venus’Nightside, vol. 42. Bulletin of the American Astronomical Society, p. 975.