North-south roughness anisotropy on Venus from the Magellan Radar

Altimeter: Correlation with geology

N. V. Bondarenko,1,2 M. A. Kreslavsky,1,3 and J. W. Head1

Received 22 September 2005; revised 8 November 2005; accepted 21 November 2005; published 28 April 2006.

[1] Analysis of the Doppler spectrum of Magellan Radar Altimeter echoes has revealedthe spatially coherent deflection of the echo center from the nadir point (Tyler et al., 1992).We present and analyze a global map of the Doppler centroid of the radar altimeter echo,which, for the plains of Venus, characterizes the degree of anisotropy of the surfaceroughness in the north–south direction at scales from decimeters to kilometers. We founda global hemispherical trend of roughness anisotropy: on average, equator-facing slopesare steeper than pole-facing slopes. We searched globally for correlations betweencontrasts in roughness anisotropy and surface geology. We found that sharp contrasts inanisotropy often coincide with boundaries of geological units, although not consistently.We found a number of examples where radar-bright (rough) lava flows have higherasymmetry than their surroundings. Diffuse radar-dark crater-related parabolic featuresand halos often have a symmetric roughness signature, while typical regional plains awayfrom these diffuse features often have an anisotropic roughness. There is not, however,any direct and uniform correlation between the roughness anisotropy signature andgeomorphologic terrain type or appearance in the radar images. We believe thatasymmetric topography of deposits of windblown material, e.g., microdunes or small-scale wind streaks in the wind shadow of obstacles, is the main contributor to the observedroughness anisotropy.

Citation: Bondarenko, N. V., M. A. Kreslavsky, and J. W. Head (2006), North-south roughness anisotropy on Venus from the

Magellan Radar Altimeter: Correlation with geology, J. Geophys. Res., 111, E06S12, doi:10.1029/2005JE002599.

1. Introduction

[2] Because of the thick cloudy atmosphere of Venus,radar probing is the main source of geomorphologicalinformation about the surface of this planet. Within thewidely used monostatic radar probing scheme, the radarcross section depends on two angles: the incidence angle qand the incidence azimuth. When interpreting radar obser-vations of rocks and soils, it is often assumed that thesurface within the resolution element is statistically isotro-pic, and hence the radar cross section does not depend onthe incidence azimuth. This assumption is not always valid,and dependence on the azimuth does exist. Such anisotropyof the radar cross section bears information about thesubresolution-scale anisotropy of the surface, and may beof great interest for geological interpretation. Anisotropy ofradar cross section has been identified for different terrainson Venus, as described below. In this paper we consider newaspects of this phenomenon using Magellan data.

[3] The Magellan spacecraft [Saunders et al., 1992]orbited Venus in a quasi-polar elliptical orbit with thepericenter above �10�N latitude. The surface was probedwith two monostatic radar systems: a side-looking syntheticaperture radar (SAR) and a radar altimeter (RA). The SARincidence was oriented approximately in cross-track(approximately east-west) directions. The SAR operatedduring three cycles of the Magellan survey of Venus [Fordand Plaut, 1993]. During the 1st and 3rd cycles the incidentsignal was directed approximately to the west (‘‘left-look-ing’’ geometry). About 90% of the surface was imaged inthis mode during the 1st cycle. During the 2nd cycle theincidence was to the east (‘‘right-looking’’ geometry), andabout 30% of the surface was covered. About 25% of thesurface of Venus was viewed from both directions, mainlyin the southern hemisphere.[4] Weitz et al. [1994] compared SAR images from the

1st and 2nd cycles. They found three areas on the surfacewith very different appearance under left- and right-lookinggeometry. One area is located near the crater Stowe (43.2�S,233.0�E) and is seen as a number of adjacent bright patches(up to 20 km in size) in the right-looking images. This areacannot be distinguished from the radar dark parabola in theleft-looking 1st cycle SAR images. The difference in theradar cross section of the site between the 1st and 2ndcycles is about 7.5 dB. The two other areas are locatedsoutheast of crater Guan Daosheng (61.1�S, 181.8�E) andnear crater Eudocia (59.1�S, 201.9�E) and are brighter in the

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 111, E06S12, doi:10.1029/2005JE002599, 2006

1Department of Geological Sciences, Brown University, Providence,Rhode Island, USA.

2Institute of Radiophysics and Electronics, National Academy ofScience, Kharkov, Ukraine.

3Astronomical Institute, Kharkov National University, Kharkov,Ukraine.

Copyright 2006 by the American Geophysical Union.0148-0227/06/2005JE002599$09.00

E06S12 1 of 10

left-looking images. These areas are not distinguishablefrom the background in the right-looking images. Thedifference between the cross sections is 9 dB and 8.1 dB,respectively. As in the previous case, the unusual areas arewithin dark parabolas [Campbell et al., 1992] associatedwith the craters.[5] Several factors can be responsible for differences in

apparent intensity between left-looking and right-lookingSAR images. One is the scale of the resolved topographicslopes that affect the local incidence angle and, hence,change the apparent radar cross section. These differencesactually permit one to see topography on SAR images.Resolved topography, however, is obviously not the causefor these areas of strong backscattering anisotropy. Thedifference between the incidence angles during the 1st and2nd cycles for those latitudes was small (the incidenceangles in the 1st and 2nd cycles, respectively, 26.0� and25.3� for Stowe, and 20.4� and 23.9� for Guan Daoshengand Eudocia); given the known variability of backscatteringfunctions on Venus, this difference cannot account for theobserved strong effect. Grating-like diffraction on small-scale regular surface features, mentioned by Weitz et al.[1994], cannot play any role, because it requires far-rangeorder (‘‘coherence’’) of wavelength-scale features withwavelength-scale accuracy (<10 cm) over the scale of theresolution element (�200 m). Such high order never occursfor geological features. The most likely reason for theobserved anisotropy of the radar cross section is the trueanisotropy of the surface properties, in particular, subreso-lution-scale surface topography. This explanation was pro-posed by Weitz et al. [1994], who suggested the presenceof strongly asymmetric microdunes with east and westfacing slopes. There are many wind streaks around everyanomalous area, which is consistent with active eolianprocesses that might be responsible for microdune formation.Kreslavsky and Vdovichenko [1999] simulated backscatter-ing by dune-like forms and found that the observed high levelof differences between left-looking and right-looking crosssections are only possible if the microdunes cover a signif-icant proportion of the surface and their steeper slopes are assteep as the angle of repose (�30�).[6] Kreslavsky and Vdovichenko [1999] performed a

systematic quantitative comparison of 1st and 2nd cycleMagellan SAR images in the 40�S–50�S zone, where theincidence angle difference is less than 5�. Coincidence of asharp contrast in the cross section difference with geologicalboundaries indicated that weak east-west asymmetry (0.5–2.0 dB) on the plains is ubiquitous. This has been attributedmostly to small-scale dunes and ripples. The coincidence ofasymmetry contrasts with a few lava flow boundaries wasinterpreted as a possible anisotropy of surface texturecaused by flow during solidification.[7] The Magellan RA antenna was looking close to nadir

and had a wide bandwidth. Analysis of the frequencyspectrum of the echo permitted one to distinguish betweenthe northern and southern parts of the footprint and, hence,to undertake the study of backscatter anisotropy in thenorth-south direction. Such anisotropy was noted by Tyleret al. [1992].[8] We apply these data to studies related to surface

geology. We first describe the data set we used and thetechnical details necessary for the correct interpretation;

then we analyze the RA-derived N-S anisotropy and corre-late it with geological features.

2. Mapping of the Doppler Centroid

[9] Magellan RA measurements were processed indepen-dently by teams at MIT and at Stanford University. The firstteam produced the well-known topography maps of Venus[Ford and Pettengill, 1992], the GTDR data set from thePlanetary Data System (PDS). They also generated maps ofFresnel reflectivity (GReDR) and Hagfors’ roughnessparameters (GSDR data set from PDS). The team atStanford University focused on inverting the measurementsto recover the surface scattering function. The techniqueused for altimeter data processing and initial results fromthat work were discussed in detail by Tyler et al. [1992].The whole set of results obtained were archived in the PDSas the SCVDR data set (Surface Characteristics Vector DataRecord), where data are stored as points along orbits, andthe GVDR data set (Global Vector Data Record), where dataare presented in a map-projected gridded format [Maurerand Simpson, 1994; Simpson et al., 1995]. As far as weknow, these data were only occasionally used [Campbelland Shepard, 2001] and have not been used in any geologicassessment, and we have begun to undertake such ananalysis. Here we report on our results on the study of theDoppler centroid fD, one among a number of parametersfrom the SCVDR and GVDR.[10] The Doppler centroid fD, was derived by the Stanford

University team in the following way [Tyler et al., 1992].The Doppler frequency corresponding to the radial velocityof the spacecraft was removed from the radar altimeterecho. Then the echo was sampled according to the Dopplershift into 16 bins, Df = 935 Hz per bin, from �7Df to +8Df.The echo strength of +8Df bin was assigned to the aliasedfrequency of �8Df. These 17 frequencies were averagedwith the weights proportional to echo strength in each bin;this weighted average gave the central echo frequency, theDoppler centroid fD.[11] This procedure is equivalent to subdivision of the

altimeter footprint into 17 stripes approximately normal tothe orbit track. The single-burst footprint size is �25 kmalong track near the Magellan orbit periapsis (�10�N) andup to 220 km at high latitudes. The algorithms used forSCVDR data set production included the averaging ofechoes obtained for 5 consecutive bursts, so the finalresolution along orbit was about 80 km at the periapsisand about 260 km near the poles. Descriptions of resolutioncells depending on latitude are presented in Table 1. Theorbit evolved during the Magellan survey; actual numbersfor a particular orbit differ by small amounts from thetypical numbers listed.[12] The GVDR data set includes a gridded map of the

Doppler centroid; the gridding algorithm accounted for theeffective footprint size, which provides a theoreticallyoptimal balance between noise suppression and preservationof spatial information. We, however, generated anotherversion of a gridded map of fD (Figure 1), using a simpleralgorithm which ignores the footprint size: for each map cell(in the simple cylindrical four-cells-per-degree projection)we took a weighted average of all SCVDR data pointswhose footprint centers are within the cell; we used the

E06S12 BONDARENKO ET AL.: ANISOTROPY ON VENUS FROM MAGELLAN ALTIMETER

2 of 10

E06S12

inverse formal fD error from SCVDR as the weight. Typi-cally, there are from 1 to 4 data points per map cell. Ourmap does not have more information than the analogousmap from GVDR; however, it provides better apparentsharpness, which is useful for morphological comparisonwith radar images.

3. Physical Meaning of the Doppler Centroid

[13] For a globally horizontal surface with an isotropicbackscattering function, the echo spectrum is symmetricwith respect to the Doppler frequency corresponding to thenadir, and fD = 0. For many areas on Venus, however, theobserved Doppler spectrum is systematically biased towardeither positive or negative frequencies. This effect is corre-lated among different orbits over hundreds of kilometers on

Venus [Tyler et al., 1992], and among different passes overthe same place in sequential mapping cycles. Uncertaintiesin the Doppler shift due to inaccuracy of the orbit andknowledge of the gravity field is at least two orders ofmagnitude smaller than the observed effect [Tyler et al.,1992]. Thus the effect is considered to be real: the observeddeflection of the spectrum center means that there is a north-south asymmetry of the backscattering cross section of thesurface. This means that the strongest echo in the along-track direction is often coming from either ahead (fD > 0) orbehind (fD < 0) the nadir, respectively. Since Magellan wasprobing the surface moving approximately from the north tothe south, a positive fD means a stronger echo from thesouthern part of the footprint, and a negative fD means astronger echo from the northern part.[14] The frequency distribution of fD over the surface of

Venus is shown in Figure 2. This distribution has a unim-odal shape with the maximum at fD = 0. The distribution,however, is not symmetric about fD = 0; the median value of

Table 1. Effective Along-Track Size of the Single-Burst RA

Footprint, Lburst, Effective Along-Track Resolution of a Single

Data Point Obtained From Five Consecutive Bursts, Lpoint, and

Surface Tilt a1 Equivalent to One Doppler Shift Bin Size, 935 Hz,

for Different Latitudes

Latitude, deg Lburst, km Lpoint, km a1, deg

80 64 101 0.6570 50 93 0.5760 42 84 0.5250 38 80 0.4840 33 67 0.4530 28 62 0.4320 25 61 0.4210 24 56 0.420 25 61 0.42�10 28 62 0.43�20 32 66 0.45�30 37 79 0.48�40 42 84 0.53�50 48 90 0.57�60 53 109 0.65�65 58 113 0.73

Figure 1. Map of the Doppler centroid fD.

Figure 2. Incremental fD-frequency distribution.

E06S12 BONDARENKO ET AL.: ANISOTROPY ON VENUS FROM MAGELLAN ALTIMETER

3 of 10

E06S12

fD is positive. Its width at half of the maximum frequency isfrom �560 Hz to 740 Hz (Figure 2). For the total surface ofVenus we calculated that 57.7% showed positive fD, whilenegative values correspond to 42.3%; 81.3% of the surfacehas fD between ±935 Hz (two bins near zero).[15] The Doppler-sampled echo profile inevitably has a

high noise level due to the coherent nature of signalregistration (so-called ‘‘speckle noise’’ or ‘‘fading noise’’),which causes some noise in fD estimates. This noisecomponent of fD, however, does not correlate betweenneighboring footprints and does not form spatially consis-tent patterns.[16] For each individual footprint, true variations of

scattering properties within the footprint may cause asym-metric backscattering, as noted by Tyler et al. [1992].Consider, for example, a sharp boundary between smoothand rough terrains. At a first approach, a rougher surface haslower backscatter. If the northern part of a footprint isoccupied by rougher terrain, the contribution of this partis smaller than the contribution of the southern part. Thesouthern part of a footprint gives positive Doppler shifts,and the resulting fD is positive. Thus we expect a narrowband of positive fD along southern sharp boundaries ofrough terrains. Analogously, there should be a band ofnegative fD along the northern boundaries. We do seenumerous examples of this kind in the map of fD (one ofthem is shown in Figure 10 with the wide arrow).[17] The obvious possible reason for higher scattering

from off-nadir parts of the footprint is general surface tiltwithin the footprint [Tyler et al., 1992], as illustrated inFigure 3a. For north facing slopes the normal incidence ofthe probing radar signal, and hence the strongest echo,would occur southward from the nadir point, which wouldlead to a positive centroid, fD > 0. Analogously, south facingslopes cause a negative centroid, fD < 0.[18] The value fD equal to one Doppler bin (935 Hz)

corresponds to a surface tilt of a1 = 0.4� close to the orbitperiapsis (low latitudes). This number characterizes thesensitivity of fD to the surface tilts. Far from the periapsisthis tilt is larger due to slower orbital motion and higheraltitude over a spherical surface. It exceeds 0.7� in the high-latitude regions (Table 1). Thus the sensitivity of fD tosurface tilts at high latitudes is almost two times lower.Surface tilts on the order of 0.4� at the baselines of tens ofkilometers can be found on Venus, but for plains (that is, for

the majority of the surface), the slopes at these baselines aremuch gentler. Figure 4 compares the frequency distributionsof slopes and fD for the whole planet. We calculated globalslopes in the south-north direction at a baseline of about26 km (about 0.25� of latitude) along meridians using theMagellan topography map [Ford and Pettengill, 1992].Then we converted slopes into Doppler offsets using thelatitude-dependent sensitivity. The frequency distribution ofthese topography-caused Doppler shifts is shown in Figure 4(thin line). The observed frequency distribution of fD fromFigure 2 is shown in Figure 4 (bold line) for comparison. Itis clearly seen that the slope distribution is much narrower,and large-scale slopes alone cannot account for the observedfD.[19] The other physical explanation for the nonzero fD is

anisotropic subresolution topography, as schematically il-lustrated in Figure 3b [Tyler et al., 1992]. If the area withinthe footprint covered by north facing a few degree steepslopes exceeds the area covered by the south facing slopesof the same steepness, the radar echo shows positive fD.Since the whole footprint should be horizontal in average,the excess of gentle north facing slopes should be compen-sated by an excess of steep south facing slopes: longer northfacing slopes automatically mean steeper south facingslopes. Thus the positive fD means that south facing slopesare generally steeper than the north facing ones. Analo-gously, the negative fD means that south facing slopes arelonger, and north facing slopes are steeper. Thus in flatareas, the Doppler centroid is a measure of the north-southasymmetry of surface topography (roughness).[20] A wide range of horizontal scales of topography can

potentially contribute to the observed scattering anisotropy:these could range from several wavelengths to just belowthe footprint size, that is from decimeters to kilometers. Thesensitivity of fD to slope asymmetry is somehow character-ized by the same characteristic surface tilt a1 correspondingto typical fD values (Table 1); thus, fD is more sensitive tothe slope asymmetry at low latitudes and less sensitive athigh latitudes. The footprint size also plays some role.Larger footprints at higher latitudes cause averaging over a

Figure 3. Schemes showing emergence of fD due to (a)regional (footprint-scale) slope and (b) small-scale slopes(roughness) anisotropy. Adapted from Tyler et al. [1992].

Figure 4. Comparison of frequency distributions of theDoppler centroid and large-scale slopes. The bold line is thefD-frequency distribution, the same as in Figure 2. The thinline is the frequency distribution of the north-south slopesrecalculated into Doppler shift using known latitude-dependent sensitivity.

E06S12 BONDARENKO ET AL.: ANISOTROPY ON VENUS FROM MAGELLAN ALTIMETER

4 of 10

E06S12

wider area, which additionally reduce the sensitivity to slopeasymmetry. The angle a1, however, does not have themeaning of a typical surface slope, as it did for large-scaletopography. Typically we expect that the slope asymmetryinvolves longer gentler slopes a few times steeper than a1.

4. Observations and Interpretations

4.1. Correlation With Topography

[21] As we saw above, large-scale topographic slopesshould influence the fD values. We studied the relationshipbetween the Doppler centroid map and global topography ofVenus. There are many examples where large-scale slopes ofmeridional direction or pronounced isolated topographicfeatures amid plains are clearly seen in the fD maps(Figure 1). Generally these features on the map look liketopography illuminated from the north, so the north facingslopes are brighter, and south facing slopes are darker.Topographic slopes have the greatest effect on the Dopplercentroid, when they are long (their length is comparable to thefootprint size) and gentle (their tilt is comparable to a1).[22] The map of fD for Artemis Chasma, together with the

SAR image of the area, are presented in Figure 5. ArtemisChasma is a �30-km-wide trough that is 2–4 km deeperthan the adjacent plains. The trough is paralleled by ridgesas high as 4 km above the plains both inward and outwardof the trough. The outward ridge is wide and has gentleslopes; its typical north facing (trough-facing) slope is 1.1�.The observed Doppler centroid value in the same location isfD � 2.8 kHz, which corresponds to a surface slope a �1.2�, remarkably similar to the actual slope. In this case thesurface topography is undoubtedly the factor responsible forthe observed scattering anisotropy.[23] In large tessera areas, relatively steep large-scale

slopes are ubiquitous, and fD for each radar altimeter burst

is defined by the actual balance of 10-km-scale slopes ofdifferent orientation within the footprint. This leads to anextremely noisy, spotty appearance of tesserae in the Dopp-ler offset map. Tellus Tessera (Figure 6) provides a goodexample.[24] In a wide lower-latitude zone (Figure 1), the map

reveals numerous spatially consistent flat areas of ratherhigh positive and negative values of fD that do not resembleilluminated topography and are caused by roughness an-isotropy or slope asymmetry.

4.2. Roughness Anisotropy in the Plains

[25] The plains areas characterized by pronounced non-zero fD are concentrated in a wide equatorial belt (�40�S–60�N). This is probably a result of the latitude trend of fDsensitivity to roughness anisotropy and of the footprint size.[26] We searched for correlations of contrasts in the fD

map (Figure 1) with geologic unit boundaries, using Magel-lan SAR mosaics. We found that there is no direct globalcorrespondence between fD and surface morphology. Inmany places, however, we do see a correspondence betweensharp contrasts in the fD map and boundaries of geologicalunits. This gives the strongest observational evidence that fDin the plains indeed reflects some intrinsic anisotropy of thesurface.[27] As we noted above, the topography or roughness

responsible for the observed anisotropy can be of verydifferent scales ranging from centimeters to kilometers.The largest scales in this range, kilometers and hundredsof meters, are resolved by SAR images. Numerous tectonicfeatures characterized by these scales are seen amid theplains in SAR images. There is no clear indication ofpronounced asymmetry of these features. The most abun-dant resolved tectonic features in the plains are the wrinkleridges. Their three-dimensional shapes are rarely resolved;however, the largest examples, as well as analogy withsimilar features on the other planets [e.g., Watters, 1988]indicate that the ridges often have asymmetric profiles

Figure 5. (a) Map of the Doppler centroid fD and(b) synthetic aperture radar (SAR) mosaics of ArtemisChasma (large arcuate feature) and surroundings. The grayscale of fD is the same as in Figure 1.

Figure 6. (a) Map of the Doppler centroid fD and (b) SARmosaics of Tellus Tessera and surroundings. The gray scaleof fD is the same as in Figure 1.

E06S12 BONDARENKO ET AL.: ANISOTROPY ON VENUS FROM MAGELLAN ALTIMETER

5 of 10

E06S12

[Kreslavsky and Basilevsky, 1998]. The sense of this asym-metry, however, often changes along individual ridges, andthere is apparently no consistent asymmetry throughoutlarge areas. Measurements by Kreslavsky and Basilevsky[1998] showed that the spacing of the ridges is typically�10 times wider than the ridges themselves, which meansthat the asymmetric topography of the ridges occupies nomore than 10% of the footprint area. Thus we do not expectwrinkle ridges to play a significant role in the formation ofscattering anisotropy. We performed visual comparisons ofthe Doppler centroid map (Figure 1) with the distributionof wrinkle ridges over the planet [Bilotti and Suppe, 1999].Wrinkle ridges oriented approximately latitudinally (that iswith a noticeable north-south slope component) are seenalmost everywhere in the plains. We did not observe anyapparent correlation between areas of high scatteringanisotropy and the presence or absence of wrinkle ridges.[28] Therefore it is more probable that the observed

scattering anisotropy is caused by slope asymmetry atsmaller scales, from tens of meters down. As we discussedin the introduction, wind action is the most probable agentcausing asymmetry at these scales. Loose material thatcan potentially make asymmetric topographic forms ispresent on Venus, as seen at the Venera landing sites [e.g.,Florenskii et al., 1977] and from the presence of eolianfeatures in SAR images [Greeley et al., 1995, 1997] and aspredicted theoretically [Garvin, 1990]. Surficial deposits ofloose material can produce different types of asymmetrictopography, as summarized in Figure 7. Formation ofmicrodunes is possible under Venus atmospheric conditions[Greeley et al., 1995, 1997]; these produce the same slopeasymmetry as do terrestrial dunes: gentle upwind-facing andsteep downwind-facing slopes (Figure 7a). Deposition ofwindblown loose material in wind shadows of localobstacles on a rough surface produces slope asymmetry ofthe opposite sense: it creates long gentle downwind-facing

slopes (Figure 7b). Accumulation of material on the upwindsides of obstacles is also possible, which may make upwind-facing slopes gentler (Figure 7c). However, if material isaccumulated on both sides of obstacles one should expectgentler slopes on downwind side, as in Figure 7b.[29] Deposition of windblown material is probably the

most effective way to produce the observable surfaceanisotropy on Venus. Wind erosion is also capable ofproducing slope asymmetry; terrestrial wind erosion land-forms, however, are generally characterized by much lessslope asymmetry than dunes. Flow of lavas itself canintroduce directional asymmetry in the roughness patternof the flow surface.

4.3. Global Hemispherical Trend

[30] A global hemispherical trend of scattering anisotropyis clearly seen in the map (Figure 1): in the northernhemisphere larger areas typically have fD > 0, while in thesouthern hemisphere the sign is typically the opposite(although numerous exceptions to this trend exist). Figure 8illustrates this trend with the latitudinal dependence of thepercentage of area occupied by significantly positive (fD >450 Hz) and significantly negative (fD < �450 Hz) Dopplercentroids. The observed hemispheric signature correspondsto relatively steeper equator-facing slopes of small-scaletopography. There is no perfect hemispherical symmetry;the most prominent global feature is a global band of positivefD at low northern latitudes (see Figures 1 and 8).[31] If we assume that the roughness anisotropy is due to

microdunes (Figure 7a), the observed hemispheric signaturecorresponds to an equatorward global wind pattern. If weassume deposition in the wind shadows (Figure 7b) as themechanism of anisotropy, the observed signature wouldcorrespond to poleward winds. Greeley et al. [1995] haveobserved a hemispherical trend of the wind directionsinferred from wind streaks not related to parabolic features.In that study, the wind streaks showed the dominance ofequatorward winds. Thus the global trend in wind streakdirections is in agreement with the global trend in slopesymmetry under the assumption of microdunes rather thanwind shadows.[32] Although there are global hemispherical trends both

in the wind streak orientation and in the north-south slope

Figure 7. Sketches showing different types of asymmetriclandforms formed by deposition of wind-blown material:(a) microdunes; (b) deposits in wind shadow of obstacles;(c) deposits on upwind side of sharp-edged obstacles.

Figure 8. Percentage of areas with a noticeable positive(fD > 450 Hz) (bold line) and negative (fD < �450 Hz) (thinline) Doppler centroid plotted as a function of latitude.

E06S12 BONDARENKO ET AL.: ANISOTROPY ON VENUS FROM MAGELLAN ALTIMETER

6 of 10

E06S12

asymmetry, the actual correlation between them is not high.We studied this in detail using the database of eolianfeatures on Venus from Greeley et al. [1997]. Wind streaksneither tend to concentrate in the areas with high degree ofslope asymmetry, nor do they tend to avoid such areas.There are numerous examples where local relationshipsbetween wind streak orientation and slope asymmetryfollow the global trend, and there are examples where localrelationships are the opposite.

4.4. Correlation With Volcanic Units

[33] We searched for correlations of contrasts in the fDmap (Figure 1) with geologic unit boundaries, using Magel-lan SAR mosaics. We found that there is no direct globalcorrespondence between fD and surface morphology, withthe only exception being the spotty pattern of large topo-graphically dissected tesserae mentioned above. In manyplaces, however, we do see a correspondence between sharpcontrasts in the fD map and boundaries of geological units.This provides the strongest observational evidence that fD inthe plains indeed reflects some intrinsic properties of thesurface.[34] There are a number of examples, where the distinc-

tive contrasts in fD correspond to boundaries of volcanicunits in SAR images. We found 24 examples of lava flowsshowing rather uniform fD distinctly different from thesurroundings. These sites are listed in Table 2 together withthe mean value of fD for each flow and for its surroundingsurface. All flows found are rather small; the largest ofthem, the bright lava flow in Sedna Planitia (Figure 9;number 1 in Table 2), is about 105 km2. This flow, located athigh northern latitudes, has a strong slope asymmetry (fD ��2 kHz) with steeper north facing slopes, contrary to thehemispherical trend. Its surroundings show variable asym-metry, on average small but also negative (fD � �0.2 kHz).A few wind streaks found by Greeley at al. [1997] in the

vicinity of this flow have the inferred wind direction to thenorth-west, also contrary to the hemispheric trend.[35] Five sites (numbers 12, 13, 15, 17 and 18 in Table 2)

form a tight cluster of similar flows in Rusalka Planitia.These flows are in the southern hemisphere and are char-acterized by positive fD contrary to the global hemisphericaltrend. Other sites are scattered (they do not occur inclusters). The remaining nine flows in the southern hemi-sphere follow the global trend. The northern hemisphereflows exhibit both positive and negative Doppler centroidvalues. Thus as a general rule, the distinctive lava flows donot follow the fD global trend.[36] All flows listed in Table 2 are rather bright and some

of them are very bright in SAR images (that is, they have ahigh radar cross section) and hence, have rather rough(sometimes very rough) surfaces at the scales of centi-meters. All flows show a degree of slope asymmetry higherthan their surroundings. There is no general correlationbetween the sign of fD and wind direction from the closestwind streaks. Also, no apparent correlation is seen betweenthe sign of fD and the flow direction; this direction,however, cannot be reliably identified for some sites.[37] The coincidence of flow fronts and contrasts in fD

suggests that the flow itself may cause the observed surfaceroughness anisotropy. However, it is not necessarily thecase. The rough flow surface may favor accumulation ofwindblown material and may produce the anisotropy bymaking deposits in wind shadows (Figure 7b) or beforesharp-edged obstacles (Figure 7c) more pronounced. Sincethe tilt of surfaces of such deposits is small, the depositsrelated to centimeters-high obstacles will have meter-longfacets; such facets would effectively influence the RAbackscattering pattern. Uniformly strong anisotropy overthe whole flow is more consistent with this explanation,because for flowing lavas we would expect variations offlow direction and flow properties across the flow area.

Figure 9. Magellan SAR mosaic of Sedna Planitia (50�N,345�E). Inset shows the Doppler centroid map of the samearea. Arrow shows a radar-bright (rough) lava flow with awell-pronounced anisotropy signature.

Table 2. Distinctive Lava Flows That Have a Prominent North-

South Slope Asymmetry Signature

No.Latitude,

degLongitude,

degfD,

kHz, FlowfD,

kHz, AroundArea,

103 km2

1 49.8 346.5 �1.98 �0.85 1052 38.8 209.5 1.80 0.87 203 36.5 337.0 �1.19 0.27 224 35.5 227.0 �0.61 �0.07 115 26.8 292.0 1.03 0.36 646 26.5 255.0 1.47 0.26 117 24.0 297.5 �1.00 0.05 188 19.8 288.5 �1.31 �0.24 349 16.8 286.0 �1.44 �0.50 4910 12.5 278.0 �0.88 �0.12 3511 �4.0 148.5 1.54 0.26 3012 �4.3 147.0 1.25 0.30 1313 �6.8 183.0 1.15 0.64 614 �7.0 150.5 1.31 0.04 2315 �8.0 288.0 �1.22 �0.08 2116 �8.3 151.0 1.18 0.04 917 �10.0 152.5 0.74 �0.23 818 �10.3 293.5 �1.47 �0.25 4719 �15.0 14.0 �1.80 �0.80 4920 �16.0 301.0 �2.03 �0.91 3121 �21.5 355.0 �1.43 �0.41 5822 �22.0 302.0 �1.46 �0.57 12923 �34.3 106.5 �1.70 �0.81 5924 �43.5 341.0 �1.11 �0.29 92

E06S12 BONDARENKO ET AL.: ANISOTROPY ON VENUS FROM MAGELLAN ALTIMETER

7 of 10

E06S12

[38] Table 3 lists eight typical examples of relatively large(>1.5 � 105 km2) uniformly anisotropic areas in the verytypical regional plains. These areas are far from apparentparabolas, diffuse halos and other features attributed to thepresence of surficial deposits [e.g., Basilevsky et al., 2004,and references therein]. All areas contain well-developednetworks of wrinkle ridges. The areas are composed ofdifferent flow units; their boundaries have low or moderatecontrast in the SAR images and do not appear in fD maps.There is no clear coincidence seen between the flowboundaries and the edges of uniform areas in the fD map.These typical uniform areas have rather high (although notextreme) values of fD. Seven of the eight follow thehemispherical trend.[39] As we discussed above, the wrinkle ridges are

unlikely to be responsible for the observed anisotropy. Wefavor an explanation for the anisotropy as being related todeposition of windblown material. We interpret the unifor-mity of the anisotropy to be related to the uniformity ofsubstrate characteristics and the uniformity of winds in thevery flat, very homogeneous areas of regional plains.

4.5. Correlation With Surficial Deposits

[40] We searched for correlations of contrasts in the fDmap (Figure 1) with diffuse features in SAR images, e.g.,crater-related radar-dark parabolas, halos, craterlesssplotches, etc., features presumably related to surficialdeposits of loose material ejected by impact processes[e.g., Basilevsky et al., 2004, and references therein]. Wefound that there is no universal correlation between fD andsuch features. However, we see a number of exampleswhere crater-related parabolas or other diffuse features havefD � 0, while the plains unit over which the parabola issuperposed has a pronounced N-S slope asymmetry. One ofthe best examples of this kind is shown in Figure 10.Generally, the absence of roughness asymmetry is typicalfor the radar-dark parabolas and halos, although it is notuniversal; sometimes it is valid only for some part of surfacecovered by radar-dark material.[41] Roughness isotropy of the parabolas is consistent with

the hypothesis that they are formed by a meter-scale-thickflat-surface mantle of loose material deposited after theimpact [Campbell et al., 1992; Bondarenko and Head,2004]. Recently multipolarization Arecibo radar observa-tions provided independent support for this model [Carteret al., 2004]. Thus after emplacement, the mantle remains flatfor a long time before dunes form. The characteristic time-scale of this is the time necessary to form a few large craters.[42] Splotches are circular diffuse bright or dark features

in SAR images [e.g., McKinnon et al., 1997] interpreted as

‘‘failed craters,’’ the results of projectile explosion in thelower atmosphere. There are a few hundred of these onVenus. Comparison between fD map and SAR mosaicsshowed that several areas of peculiar fD are associated withsome of the largest splotches. However, not all splotches,including the large ones, have any expression in the fDmaps; the majority of them are simply not resolved. Wefound 17 clear examples of correlation between a splotchand a peculiar feature in the fD map, and these are listed inTable 4. In the majority of these cases (14 of 17), splotches

Table 3. Areas of Typical Regional Plains With a High Uniform

North-South Slope Anisotropy

PlanitiaLatitude,

degLongitude,

degArea,

103 km2Mean fD,

kHzSD,kHz

Sedna 46 351 559 �0.78 0.54Sedna 41 330 814 1.41 0.44Guinevere 38 313 637 1.48 0.54Kawelu 30 253 167 1.15 0.65Bereghinya 21 12 665 1.50 0.45Llorona 10 125 467 1.59 0.46Hinemoa 5 281 279 1.45 0.52Tahmina �16 68 343 �1.46 0.39

Figure 10. Magellan SAR mosaic of the crater Bassi(35 km diameter, 19�S, 64.7�E, narrow arrow) and vicinity.Inset shows the Doppler centroid map of the same area. Thewide arrow shows a rough ridge with clear expression in theDoppler centroid map due to its roughness contrasts. T,Ovda Tessera.

Table 4. Splotches That Have a Distinctive Signature in

Roughness Anisotropy

No.Latitude,

degLongitude,

degfD, kHz,Splotch

fD, kHz,Around

1 36.0 347.0 1.22 0.372 6.0 266.0 0.75 0.193 4.0 273.0 2.15 0.754 3.8 173.0 1.40 0.285 2.0 158.0 1.31 0.106 1.8 152.0 1.03 0.007 1.5 266.0 1.78 0.478 1.5 275.5 1.12 0.289 1.0 271.0 2.06 0.4710 �0.8 270.5 1.78 0.4711 �1.8 148.5 0.84 0.2812 �2.8 161.5 1.31 0.2813 �8.0 288.0 �1.22 0.1914 �18.0 15.0 �0.37 �1.5015 �19.5 13.5 0.00 �1.0316 �19.5 15.0 �0.37 �1.6817 �21.8 278.5 �1.59 �0.10

E06S12 BONDARENKO ET AL.: ANISOTROPY ON VENUS FROM MAGELLAN ALTIMETER

8 of 10

E06S12

have a higher anisotropy than the surroundings; in all thesecases the sign of the anisotropy follows the hemisphericaltrend. The remaining 3 splotches appear in the fD mapbecause they have a weak anisotropy on a highly aniso-tropic background.

5. Discussion

[43] We believe that the deposits of windblown materials,either microdunes or deposits in wind shadows of topo-graphic obstacles or both, are the best candidates for theexplanation of the observed ubiquitous roughness anisotropyin plains of Venus. So far, we have not found any observa-tions that would contradict this suggestion. The absenceof correlation of anisotropy with wind streaks is easilyexplainable. First, they are probably related to differentscales: the most probable scales responsible for the forma-tion of Doppler centroid are on the order of decimeters andlarger, while scales on the order of tens of centimeters areresponsible for the wind streaks. Second, they can form notsimultaneously, and conditions for their formation may bedifferent. Third, the anisotropy can be produced by thindeposits that are not seen or are not recognizable as windrelated in the SAR images.[44] In the framework of the eolian hypothesis, it is

natural that we observe some correlation of anisotropy withgeology, but this correlation is not universal and notuniform. Formation of anisotropic deposits is a complexprocess that depends on surface roughness characteristics(which is often seen but may be unseen in SAR images),availability of necessary loose material and its properties(which is only partly seen in the radar images as hard tointerpret diffuse features), and wind patterns during theepoch of emplacement (which is not seen at all). Further-more, formation of an anisotropic surface is interleaved withepisodes of removal of anisotropy by emplacement of freshlava flows, which are usually seen in the images, debris fallsfrom large impacts, which are seen in the images asparabolas, and very likely, eolian erosion, whose results(except in a few minor cases) we cannot yet identify inimages. This general picture explains why some, but farfrom all, contrasts in roughness anisotropy correlate withsurface features seen in SAR images.[45] The presence of a well-pronounced uniform anisot-

ropy in vast areas of regional plains indicates the presenceof surficial deposits formed by loose material. Thus even farfrom diffuse features, the surface is likely to be coveredwith deposits of loose, probably chemically altered andmaybe indurated material rather than with bedrock [e.g.,Basilevsky et al., 2004]. This conclusion is important inconsidering prospective future landing missions to Venusdesigned to undertake geochemical analyses of igneousmaterials. Models of radar scattering by microdunes andother hypothetical anisotropic forms should help to con-strain the possible amounts (characteristic thicknesses) ofthese surficial deposits.[46] The absence of anisotropy for a number of radar-dark

parabolas indicates that the processes of eolian transport onVenus are geologically slow on average, and that the loosematerial of the parabolas has a flat surface for a long time. Itis possible that eolian transport is active only episodicallyduring geologically short periods, for example only after

large impact events, as has been hypothesized by Schultz[1992]. Alternatively, the parabola material may need somesurface exposure age to produce a large number of grains ofproper size for eolian transport; for example, disintegrationof pebbles by chemical weathering or induration of fine dustmay be necessary to initiate eolian transport.

6. Conclusions

[47] We considered the global Doppler centroid map andcorrelated it with surface geology. We conclude the following.[48] 1. In the plains regions the Doppler centroid is a

measure of surface roughness anisotropy in the north-southdirection at scales from centimeters to kilometers. The mostimportant scale range for this anisotropy is likely to rangefrom decimeters to decameters.[49] 2. The most probable cause of observed surface

roughness anisotropy is surficial eolian deposits of loosematerial, for example, asymmetric microdunes or depositsin wind shadows of obstacles.[50] 3. There is a global hemispherical trend of roughness

anisotropy; in general, equator-facing small-scale topo-graphic slopes are steeper than pole-facing slopes.[51] 4. Roughness anisotropy in the plains of Venus is

ubiquitous. Typical regional volcanic plains often have amoderately high degree of roughness anisotropy, whichsuggests the presence of surficial deposits.[52] 5. There is no universal correspondence between the

signature of roughness anisotropy and surface morphology.However, some geological boundaries are associated withsharp contrasts in anisotropy.[53] 6. Typical radar-dark diffuse crater-associated parab-

olas and halos often do not show anisotropy.[54] 7. The overall scenario is consistent with very slow

or episodic eolian alteration of the surface interleaved withsurface modification by volcanism and large impact events.[55] Our analysis showed that Magellan radar data still

have unused potential for the study of geological processeson Venus, especially the analysis of surficial deposits andprocesses. Analysis of the Doppler centroid of the MagellanRA echo adds a new dimension to the reconstruction of thesmall-scale surface structure and evolution of Venus. Futurework in this direction should include (1) models of radarscattering, which should relate the geometric parametersof roughness and the observed Doppler centroid and(2) correlative studies of Doppler centroid data withother radar data sets, including Arecibo multipolarizationobservations.

[56] Acknowledgments. We greatly appreciate constructive reviewsby Peter Ford and Richard Simpson. A grant from the National Aeronauticsand Space Administration Planetary Geology and Geophysics Program (toJ.W.H.) is gratefully acknowledged.

ReferencesBasilevsky, A. T., J. W. Head, and A. M. Abdrakhimov (2004), Impactcrater air fall deposits on the surface of Venus: Areal distribution, esti-mated thickness, recognition in surface panoramas, and implications forprovenance of sampled surface materials, J. Geophys. Res., 109, E12003,doi:10.1029/2004JE002307.

Bilotti, F., and J. Suppe (1999), The global distribution of wrinkle ridges onVenus, Icarus, 139, 137–157.

Bondarenko, N. B., and J. W. Head (2004), Radar-dark impact crater–related parabolas on Venus: Characterization of deposits with Magellanemissivity data, J. Geophys. Res., 109, E09004, doi:10.1029/2004JE002256.

E06S12 BONDARENKO ET AL.: ANISOTROPY ON VENUS FROM MAGELLAN ALTIMETER

9 of 10

E06S12

Campbell, B. A., and M. K. Shepard (2001), Near-nadir radar scatteringfrom Venus, Lunar Planet. Sci. [CD-ROM], XXXII, Abstract 1298.

Campbell, D. B., N. J. S. Stacy, W. I. Newman, R. E. Arvidson, E. M.Jones, G. S. Musser, A. Y. Roper, and C. Schaller (1992), Magellanobservations of extended impact crater related features on the surfaceof Venus, J. Geophys. Res., 97, 16,249–16,277.

Carter, L. M., D. B. Campbell, and B. A. Campbell (2004), Impact craterrelated surficial deposits on Venus: Multipolarization radar observationswith Arecibo, J. Geophys. Res., 109, E06009, doi:10.1029/2003JE002227.

Florenskii, K. P., A. T. Basilevskii, and L. B. Ronca (1977), Geomorphicdegradations on the surface of Venus—An analysis of Venera 9 andVenera 10 data, Science, 196, 869–871.

Ford, P. G., and G. H. Pettengill (1992), Venus topography and kilometer-scale slopes, J. Geophys. Res., 97, 13,103–13,114.

Ford, J. P., and J. J. Plaut (1993), Magellan image data, inGuide to MagellanImage Interpretation, report, pp. 7–18, Jet Propul. Lab., Pasadena, Calif.

Garvin, J. B. (1990), The global budget of impact-derived sediments onVenus, Earth Moon Planets, 50–51, 175–190.

Greeley, R., K. Bender, P. E. Thomas, G. Schubert, D. Limonadi, and C. M.Weitz (1995), Wind-related features and processes on Venus: Summary ofMagellan results, Icarus, 115, 399–420.

Greeley, R., K. C. Bender, R. S. Saunders, G. Schubert, and C. M. Weitz(1997), Aeolian processes and features on Venus, in Venus II: Geology,Geophysics, Atmosphere, and Solar Wind Environment, edited by S. W.Bougher, D. M. Hunten, and R. J. Philips, pp. 547–590, Univ. of Ariz.Press, Tucson.

Kreslavsky, M. A., and A. T. Basilevsky (1998), Morphometry of wrinkleridges on Venus: Comparison with other planets, J. Geophys. Res., 103,11,103–11,112.

Kreslavsky, M. A., and R. V. Vdovichenko (1999), Anisotropy of Venussurface properties inferred from Magellan side-looking-radar data, SolarSystem Res., 33, 110–119.

Maurer, M. J., and R. A. Simpson (1994), GVDR data product: A summaryof all Magellan observations of Venus surface characteristics, LunarPlanet. Sci., XXV, 855.

McKinnon, W. B., K. J. Zhanle, B. A. Ivanov, and H. J. Melosh (1997),Cratering on Venus: Models and observations, in Venus II: Geology,Geophysics, Atmosphere, and Solar Wind Environment, edited by S. W.Bougher, D. M. Hunten, and R. J. Philips, pp. 969–1014, Univ. of Ariz.Press, Tucson.

Saunders, R. S., et al. (1992), Magellan mission summary, J. Geophys. Res.,97, 13,067–13,090.

Schultz, P. H. (1992), Atmospheric effects on ejecta emplacement and craterformation on Venus from Magellan, J. Geophys. Res., 97, 16,183–16,248.

Simpson, R. A., M. J. Maurer, and G. L. Tyler (1995), Application of theMagellan GVDR to Venus surface studies, Lunar Planet. Sci., XXVI,1307–1308.

Tyler, G. L., R. A. Simpson, M. J. Maurer, and E. Holmann (1992),Scattering properties of the Venusian surface: Preliminary results fromMagellan, J. Geophys. Res., 97, 13,115–13,139.

Watters, T. (1988), Wrinkle ridge assemblages on the terrestrial planets,J. Geophys. Res., 93, 10,236–10,254.

Weitz, C. M., J. J. Plaut, R. Greeley, and R. S. Saunders (1994), Dunes andmicrodunes on Venus: Why were so few found in the Magellan data?,Icarus, 112, 282–295.

�����������������������N. V. Bondarenko, J. W. Head, and M. A. Kreslavsky, Department of

Geological Sciences, Brown University, Providence, RI 02912, USA.(james_head@brown.edu)

E06S12 BONDARENKO ET AL.: ANISOTROPY ON VENUS FROM MAGELLAN ALTIMETER

10 of 10

E06S12