Composition of the lunar surface as will be seen from SMART-1:

A simulation using Clementine data

Yuriy G. Shkuratov,1 Dmitriy G. Stankevich,1 Vadim G. Kaydash,1

Vitaliy V. Omelchenko,1 Carle M. Pieters,2 Patrick C. Pinet,3 Serge D. Chevrel,3

Yves H. Daydou,3 Bernard H. Foing,4 Zoran Sodnik,4 Jean-Luc Josset,5

Lawrence A. Taylor,6 and Vladislav V. Shevchenko7

Received 23 August 2002; revised 21 January 2003; accepted 20 February 2003; published 3 April 2003.

[1] We present a new technique for remote sensing determination of lunar surfacecomposition in the context of the SMART-1 mission. The technique is based on spectraland composition data obtained by Lunar Soil Characterization Consortium for a fewparticle-size separates of lunar soils. We map the abundance of TiO2 and FeO, pyroxenecontent, maturity degree (Is/FeO), and a characteristic size of particles. Comparison of theTiO2 and FeO abundance maps with proper distributions obtained by Lucey et al. [2000a]shows high correlation. We found also an inverse correlation between the Is/FeOdistribution and Lucey et al.’s [2000b] parameter OM. Application of this approach showsthat fresh mare crater regolith is characterized with a higher abundance of pyroxenes,coarse particles, and low maturity degree. The pyroxene abundance map can be used toidentify pyroclastic regions. An excess of small particles is predicted for highland areas.We note appreciable variations of the characteristic size of particles in mare regions. Ourpreliminary results for the Reiner-g formation show that there is no composition anomalyfor the TiO2 and FeO abundance, in agreement with previous analyses. Our maps alsoindicate that the formation contains a surface material characterized with low maturity andhigh degree of crystallinity, consistent with the occurrence of immature regolith possiblycontaminated with dust. INDEX TERMS: 6250 Planetology: Solar System Objects: Moon (1221);

5420 Planetology: Solid Surface Planets: Impact phenomena (includes cratering); 5464 Planetology: Solid

Surface Planets: Remote sensing; 5470 Planetology: Solid Surface Planets: Surface materials and properties;

KEYWORDS: chemistry and mineralogy of the lunar surface

Citation: Shkuratov, Y. G., et al., Composition of the lunar surface as will be seen from SMART-1: A simulation using Clementine

data, J. Geophys. Res., 108(E4), 5020, doi:10.1029/2002JE001971, 2003.

1. Introduction

[2] Multispectral photometry is a powerful tool to exam-ine lunar surface chemistry, mineralogy, and physical prop-erties. Pioneers of quantitative lunar spectral studies [e. g.,Whitaker, 1972; McCord and Adams, 1973; Charette et al.,1974; McCord et al., 1976, 1981] clearly demonstrated thediagnostic potential of spectral reflectance measurementsfor the lunar surface. Multispectral photometric imaging

gives valuable information about the first few millimeters ofthe lunar regolith and provides the highest spatial resolutionamong remote sensing methods. Such data are very impor-tant for geologic studies of the lunar surface.[3] In 1994 the NASA-DoD Clementine spacecraft

acquired multispectral images for more than 90% of thelunar surface [Nozette et al., 1994]. In fact, Clementine hasopened a new era in geologic studies of the Moon withmultispectral images of high spatial resolution. It allowsconsiderable advance in understanding the history andevolution processes on the lunar surface [e.g., McEwen etal., 1994; Pieters et al., 1994; Shoemaker et al., 1994;Tompkins and Pieters, 1999; Blewett et al., 1997; Lucey etal., 1998, 2000a, 2000b; Le Mouelic et al., 2000; Pinet etal., 2000]. Multispectral Clementine data in combinationwith Earth-based observations and laboratory measurementsof lunar samples allowed progress in evaluating the chem-ical composition of the lunar surface. The most widely usedapproach to estimate composition (TiO2, FeO), and maturitydegree of the lunar surface was suggested by Lucey et al.[1995, 1998, 2000a, 2000b] and Blewett et al. [1997].However, attempts to improve this algorithm or develop

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 108, NO. E4, 5020, doi:10.1029/2002JE001971, 2003

1Kharkov Astronomical Observatory, Kharkov, Ukraine.2Geological Sciences, Brown University, Providence, Rhode Island,

USA.3University P. Sabatier Observatoire Midi-Pyrenees, Toulouse, France.4European Space Agency/European Space Research and Technology

Center, Noordwijk, Netherlands.5Cenre Suisse d’Electronique et de Microtechnique (CSEM), Neuchatel,

Switzerland.6Planetary Geoscience Institute, University of Tennessee, Knoxville,

Tennessee, USA.7Shternberg Astronomical Institute, Moscow, Russia.

Copyright 2003 by the American Geophysical Union.0148-0227/03/2002JE001971$09.00

1 - 1

an independent technique are continuing [e.g., Shkuratov etal., 1999a].[4] Detailed compositional and spectral analyses of lunar

soils carried out by the Lunar Soil Characterization Con-sortium (LSCC) [Taylor et al., 1999, 2000a, 2000b, 2000c,2001] provide a unique opportunity to develop an approachthat might be successfully used in remote compositionalanalysis of the lunar surface. Direct and indirect linksbetween the combined spectral parameters of soils and theircompositional characteristics have been evaluated through astatistical analysis of the variance across each type ofmeasurement using principal component analysis [Pieterset al., 2002]. Initial steps to establish links using theClementine UVVIS spectral bands and LSCC data havealso been carried out [Stankevich et al., 2002; Shkuratov etal., 2002a].[5] Exploration with the upcoming ESA SMART-1

spacecraft is expected to significantly extend knowledgeabout the Moon, particularly by providing new multispec-tral images with high resolution. The images will beacquired for a wider range of illumination/observationgeometry than in case of Clementine. This allows a quanti-tative analysis of the surface physical properties with phase-angle-ratio images. Usefulness of phase-angle-ratio imageshas been demonstrated with telescope images of the Moon[e. g., Shkuratov et al., 1994] and then with Clementine data[Shkuratov et al., 1999b]. This technique is planned to beapplied to SMART-1 data [Shkuratov et al., 2002b] to studylunar surface areas that have undergone a fresh disturbanceof the surface structure. Clementine phase-angle-ratioimages of the Apollo 15 landing site, for example, havealready allowed us to find recent disturbances of the regolithformed by the landing jets [Kreslavsky and Shkuratov, 2001,2003].[6] Another important and more conventional use of

SMART-1 multispectral data is remote composition analysisof the lunar surface. Spectral bands of the SMART-1 cameraare close to the spectral bands of the Clementine UVVIScamera. This permits us to plan joint compositional analysesof the lunar surface with Clementine and SMART-1 images.Moreover, this allows one to use current Clementine data inorder to study the applicability and efficacy of futureSMART-1 data for remote composition analyses.[7] This paper aims to present a technique for remote

sensing determination of lunar surface composition in thecontext of SMART-1 measurements. To develop and dem-onstrate this technique, Clementine UVVIS data and recentresults of the LSCC studies are used. To better understandthe capability of SMART-1 data, we give a short descriptionof the SMART-1 mission focusing on the imaging experi-ment. Then the analysis technique is described and examplemaps of chemical/mineral composition are presented.

2. Lunar Images to be Collected by SMART-1

2.1. SMART-1 Mission

[8] The SMART-1 mission is the first in the program ofESA’s Small Missions for Advanced Research and Technol-ogy [Foing and SMART-1 team, 1999]. Its objective is todemonstrate an electric primary propulsion for future mis-sions (such as Bepi-Colombo) and to test new technologiesfor spacecraft and instruments [Racca et al., 2001, 2002]. The

approved SMART-1 scenario includes 6 months operationsin a lunar polar orbit, with possible extension of the mission.[9] SMART-1 is a box-shaped spacecraft with a total

mass of about 350 kg. The spacecraft will be launched fromthe Ariane-5 booster into a transition orbit. The primarypropulsion is used to spiral out from the orbit (during 15–17 months depending on the precise launch date), thenachieve lunar swing-by, capture, spiraling in to a near-polarorbit with apolune of 10000 km and perilune of 300–1000km placed upon the south pole. The orbit plane of thespacecraft will be constant in space, and hence each threemonths solar rays will be approximately perpendicular orparallel to the plane (in sectors ±45�). To provide optimalillumination of the solar panels, the spacecraft will havetwo orientation modes, corresponding to the three monthsperiods.[10] The SMART-1 science payload, with a total mass of

about 15 kg, includes innovative instruments and advancedtechnologies [Almeda et al., 2002]. In particular, a cameraAMIE, a near-infrared point-spectrometer (SIR) for miner-alogy investigation, and a compact X-ray spectrometer (D-CIXS) providing fluorescence spectroscopy and imagery ofthe elemental composition will be used.

2.2. AMIE/SMART-1 Camera

[11] One of the main goals of the SMART-1 mission is toacquire lunar images with the AMIE camera. The camerawill be able to take images in NIR filters with high photo-metric precision and resolution [Josset et al., 2002]. TheAMIE/SMART-1 imaging system consists of two parts: theAMIE micro-imager camera unit and the AMIE electronicsunit. The micro-imager is placed on one of the lateral panelsof the spacecraft. The main functions of the system are totake images, to keep image data in a memory buffer, tocarry out image processing tasks (data compression, imagesubtraction), and to allow the downloading of the images bythe on-board computer.[12] The AMIE/SMART-1 will provide a high spatial

resolution on the order of 50–100 m/pixel. As the spacecraftmoves along the orbit, a sequence of images will covernarrow pieces of the lunar surface with the length notexceeding 300–500 km and width near 50 km. TheAMIE/SMART-1 camera has been equipped with fourfilters, in the NIR spectral range 965, 915, 830, and 750nm; these filters correspond to different surface portions ofCCD array, as shown in Figure 1. The CCD part denotedwith ‘‘no filter’’ operates in total light (with maximalsensitivity) trying to obtain images of permanently shad-owed regions that can nevertheless be slightly illuminated bysolar light reflected from neighbor places of the lunar sur-face. As seen in Figure 1, there are two sets of filterscorresponding to the two modes of spacecraft orientationsmentioned above. Images of the same regions in differentspectral ranges will be obtained sequentially as the space-craft moves along the orbit. Spectral curves of the filtersaccounting for the spectral response of the CCD array areshown in Figure 2. A small part of the CCD array has beencovered with a filter centered at 830 nm. This will providethe SMART-1 laser-link experiment that will be performedby pointing the spacecraft’s AMIE camera towards the Earth.The filter 830 nm can also be used for spectral imaging thelunar surface, though it will cover only a half area of each

1 - 2 SHKURATOV ET AL.: COMPOSITION OF LUNAR SURFACE FROM SMART-1

strip of images (see Figure 1). The micro-imager will work inclose conjunction with the SIR and D-CIXS instruments.Two AMIE filters overlap and complement the SIR spec-trometer (900–2400 nm).

2.3. Photometric Calibration

[13] Images from the AMIE camera will require a carefulphotometric calibration. The images can be transposed inthe Clementine UVVIS photometric system. Actually, thisis a necessity if we will implement the technique presentedbelow for chemical/mineral determination of the lunar sur-face using AMIE/SMART-1 images.[14] Images obtained by the UVVIS camera were pro-

cessed with radiometric calibration, normalization on photo-metric function, and transformation to obtain values ofalbedo [McEwen et al., 1998]. The transition to albedowas performed by comparison of Apollo-16 sampled siteimages obtained by Clementine with laboratory measure-ments of bidirectional spectral reflectance of Apollo-16 soilsamples using the RELAB instrument [Pieters et al., 1991].The laboratory sample used for calibration (62231) wascarefully selected as a representative of mature Apollo-16soil. It was measured multiple times at zero angle ofobservation, when incident (phase) angle equals 30�[Pieters, 1999].[15] Comparing Earth-based measurements by Saari and

Shorthill [1967] and Clementine data shows that in recentstudies the Clementine calibrated albedo scale is 1.5 timestoo large compared to that observed for ground-based data[Hillier et al., 1999; Shkuratov et al., 2000]. An additionalfactor of 1.6 is suggested to accommodate differences inphase angle (Clementine at 30� versus Earth-based at 5�). Areason for the difference (1.5 � 1.6 times) between theClementine and Earth-based observations can be that thecontribution of the shadow-hiding effect to brightness phasefunctions measured with a telescope and in the laboratoryshould be noticeably different. Any fine structure of thenatural regolith is destroyed when the samples are collected

and thus is not present in laboratory measurements. Fur-thermore, in the RELAB measurements the size of thesurface element under study is only several millimeters[Pieters et al., 1991]. Hence the contribution to total bright-ness of shadow-hiding on the lunar surface at the milli-meter-centimeter scales is absent in the laboratory case.[16] Here we deal with spectra of lunar samples obtained

by RELAB in the same condition (30� incident angle at zeroangle of observation) as the Apollo-16 sample used for theabsolute calibration of Clementine data. Therefore thephotometric calibration of our initial data is on the samescale, and the laboratory and Clementine data can be usedtogether.

3. Compositional Predictions

3.1. Data and Their Analysis

[17] Our analysis is based on laboratory studies of cur-rently available lunar soils. The mare soils include LSCC-characterized regolith samples of the Apollo-11, -12, -15,and -17 missions [Taylor et al., 2001]. The highland regolithis only represented by preliminary data for Apollo-16 soils[Fischer, 1995], which includes bulk chemical composition,but no appropriate mineral characterization. Therefore ourchemical/mineral predictions for the highland surface aswell as unusual mare regoliths (potentially swirl materials)are very preliminary. The highland predictions will becomemore reliable, when data for a range of Apollo-16 highlandsoils and perhaps Luna-20 regolith are completed by LSCCand incorporated into our algorithm.[18] The initial data for mare soils produced by LSCC

include mare regolith samples selected to be representativeof various lunar basalt compositions having differentdegrees of soil maturity. Coordinated compositional andspectral measurements of each soil were obtained for foursubsamples with controlled particle size: <45 mm, <10 mm,10–20 mm, and 20–45 mm. To minimize physical effects ofsample preparation, all size separates were prepared withultra-pure deionized water. The <45 mm sample is consid-

Figure 1. Scheme of CCD field of AMIE/SMART-1camera.

Figure 2. RELAB spectra of the soil sample 15071 andthe spectral bands of UVVIS Clementine and AMIE/SMART-1 cameras. The offset of the bands is just forpurpose of display.

SHKURATOV ET AL.: COMPOSITION OF LUNAR SURFACE FROM SMART-1 1 - 3

ered to represent the bulk soil since it has been demon-strated that the optical properties of bulk soils are dominatedby the smaller size fractions [Pieters et al., 1993]. This isbelieved to be due to larger particles being covered bysmaller particles in a natural soil. However, it may presentsome limitations in the case when the regolith is locallydisturbed, such as this can be in the situation occurring withswirls.[19] In this work we restrict our analysis to five character-

istics: the abundance of TiO2 and FeO, total pyroxenescontent, maturity degree (Is/FeO), and characteristic size ofparticles (d). Although other characteristics, such as contentof ilmenite, olivine, etc., are also available, the most reliableprediction can be made just for the mentioned five param-eters. An example of the composition determination of15071 soil separates from the Apollo-15 mission is givenin Table 1 (more details given by Taylor et al. [2001]). Datafor the few Apollo-16 soil samples included in our analysishave been characterized in literature; see details given byFischer [1995].[20] Bidirectional spectra were obtained for all samples

with the RELAB spectrometer in the spectral range 300–2600 nm [Pieters et al., 2000] at a phase (incident) angle30�. We have transposed the set of soil spectra into thespectral framework of the AMIE/SMART-1 (750, 915, and965 nm) and Clementine (415, 750, 900, 950, and 1000 nm)spectral bands (see Figure 2). All spectra of the sampleshave been convolved with the spectral curves of the filtersand the CCD response.[21] The main purpose of our statistical approach is to

find relationships that provide the highest correlation coef-ficients between different linear combinations of opticalparameters, on one hand, and chemical/mineral composi-tion, degree of maturity, and size of regolith particles, on theother hand. The choice of the optical parameters is notunique, and therefore the suggested technique is also non-unique. However we anticipate that different choices, if theyare not too artificial and exotic, will give similar results. Weconfirmed this, studying several suites of optical parameters[Shkuratov et al., 2002a, 2002b; Omelchenko et al., 2002].[22] The most natural way to demonstrate our technique

is to use Clementine 5-band images. Initial exciting resultshave already been obtained on this way [Stankevich et al.,2002; Shkuratov et al., 2002a]. Nevertheless, realizing theimportance of the SMART-1 mission we would prefer to usesimulated AMIE/SMART-1 data. To illustrate our techniquewe must transform Clementine images into the 3-bands ofAMIE/SMART-1. For this purpose we linearly interpolatedClementine data to the corresponding wavelengths. Weignore the spectral channel 830 nm as it covers only halfthe area of other SMART-1 images, though there is noproblem involving this in the analysis.[23] To simulate multispectral AMIE/SMART-1 data we

use the following optical parameters: albedo AR = A(750

nm) [%], two color-indexes: CIR1 = A(915 nm)/A(750 nm)and CIR2 = A(965 nm)/A(750 nm), and spectral bend D =A(750 nm) A(965 nm)/[A(915 nm)]2. While the choice ofthe color indexes CIR1 and CIR2 is quite rational, the choiceof the parameter of spectral bend D is not so evident. In afirst approximation it characterizes the depth of the pyrox-ene band. Mapping this parameter with Clementine datashows that it is useful for geologic studies of the lunarsurface, in particular, to find regions of nonmare volcanismon the Moon [Raitala et al., 1999].[24] Our studies show that a suitable expression for

deriving high correlation coefficients and small residualerrors is the following form:

logP ¼ aAR þ bCIR1 þ cCIR2 þ dDþ e; ð1Þ

where P is a studied parameter. The weight coefficients a, b,c, d, and e minimize the RMS deviations of the calculatedvalues of P from the measured ones. These coefficients aswell as the resulting correlation coefficients k for thepredicted and measured values of the five studiedparameters are given in Table 2. All cases reveal highcorrelation coefficients that confirms the usefulness of theprognoses presented below. It should be emphasized thatformula (1) is an empirical equation. The right side of theequation is a linear combination of well-known opticalcharacteristics. In the left side we used log to avoid apotential problem related to negative values of predictedparameters.[25] In spite of our intention to use simulated AMIE/

SMART-1 data, we have also made the same analysis forClementine (5-bands) data in order to test the stability of ourprognoses. In case of the Clementine data we deal with thefollowing optical parameters: AR = A(750 nm), color-indexes: CBR = A(415 nm)/A(750 nm), CIR1 = A(900 nm)/A(750 nm), CIR2 = A(950 nm)/A(750 nm), CIR3 = A(1000nm)/A(750 nm), and bend D = A(750 nm) A(1000 nm)/[A(900 nm)]2. The same type of the calibrating equation hasbeen used.[26] Results of our determinations of lunar surface com-

position are illustrated as maps. We use the spectral cube ofClementine UVVIS mosaics with 1-km resolution as initialreflectance data. We map abundance of TiO2 and FeO,pyroxenes content, Is/FeO, and d for a large portion ofwestern maria of the nearside (see Figure 3). The predictionmaps are presented in Figures 4–8. In all cases dark colorcorresponds to low values of the mapped parameter. Allmaps include small inserted frames (right side) that areratios of the distributions determined with Clementine (5-bands) and simulated AMIE/SMART-1 (3-bands) data.These ratios show detailed dissimilarity of images that

Table 1. Composition of 15071 Soil Separates

Size Range <45 mm 20–45 mm 10–20 mm <10 mm

TiO2, % 1.81 2.33 1.88 1.57FeO, % 14.9 15.6 15.4 9.59Is/FeO 71 49 80 159Pyroxene, % 22.07 16.71 10.92

Table 2. Data for the Coefficients in Equation (1) and Correlation

Coefficients k

a b c d e k

TiO2 �0.0446 70.505 �45.858 19.443 �42.88 0.91FeO �0.0256 21.129 �13.005 7.595 �14.277 0.88Pyroxene 0.0064 2.889 �4.503 �0.439 3.23 0.86Is/FeO �0.0166 �41.477 26.465 �14.856 31.67 0.81Particle size �0.0307 4.144 �4.737 0.771 1.25 0.91

1 - 4 SHKURATOV ET AL.: COMPOSITION OF LUNAR SURFACE FROM SMART-1

correspond to the different suites of optical parameters.Mostly, the differences are small (see below).[27] We also compare our results with similar results from

Lucey et al.’s [2000a, 2000b] approach. To find the abun-dance of TiO2 and FeO and optical maturity parameter OMwe use Lucey’s equations:

FeO% ¼ �17:43 arctanA 950ð Þ=A 750ð Þ � y

A 750ð Þ � x

� �� 7:56; ð2Þ

TiO2% ¼ 3:71 arctanA 415ð Þ=A 750ð Þ � z

A 750ð Þ

� �� �5:98

; ð3Þ

OM ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiA 750ð Þ � xð Þ2þ A 950ð Þ=A 750ð Þ � y½ �2

q; ð4Þ

where x = 0.08, y = 1.19, and z = 0.42.

3.2. Chemical/Mineral Composition

[28] We start with the estimation of TiO2. The earliestrelationship between the TiO2 abundance in mare soils andthe slope of spectral continuum in the visible range wasestablished by Charette et al. [1974]. In terms of UV/Visratios, say A(400 nm)/A(560 nm), relatively ‘‘blue’’ basalts

are Ti-rich and relatively ‘‘red’’ basalts are Ti-poor [see,e.g., Pieters, 1978]. More advanced methods to optimizethis correlation were used by Johnson et al. [1991], Blewettet al. [1997], Lucey et al. [1998, 2000a], and Shkuratov etal. [1999a]. One of the problems with this relationship is astrong nonlinearity of the correlation; for Ti-poor materialsthe correlation is practically absent. Various explanations ofthis empirical relation have been proposed [Pieters, 1978,1993; Shkuratov, 1982; Lucey et al., 2000a], but completeconsensus on the origin of this relationship has not beenforthcoming.[29] Using equation (1) with proper coefficients from

Table 2 and simulated AMIE/SMART-1 data we haveestimated the distribution of TiO2 on the lunar surface.Results are presented in Figure 4. Qualitatively, this distri-bution is very similar to what one can find in many otherworks [e.g., Johnson et al., 1991; Blewett et al., 1997; Luceyet al., 1998, 2000a; Shkuratov et al., 1999a]. The distributionis well correlated with UV/Vis color index [c.f. Whitaker,1972]. This is amazing, as the simulated AMIE/SMART-1data do not use UV range at all. Thus the information aboutthe UV-Vis behavior of lunar spectra is latently contained inthe NIR part of the spectra. The range of TiO2 variations near0.1–12% seems to be quite reasonable.[30] The small inserted frame in Figure 4, which presents

ratios of the TiO2 distributions that correspond to Clem-

Figure 3. Clementine mosaic with 1 km resolution for NW portion of the lunar nearside.

SHKURATOV ET AL.: COMPOSITION OF LUNAR SURFACE FROM SMART-1 1 - 5

entine (5-bands) and simulated AMIE/SMART-1 (3-bands)data, reveals variations correlated with the predicted TiO2

distribution. It appears that the AMIE/SMART-1 (3-bands)distribution of TiO2 has muted contrast. On average thisratio equals 0.81; it means that the simulated AMIE/SMART-1 (3-bands) data predict 19% higher content ofTiO2 than in case of Clementine (5-bands) data. The rangeof variations of the ratio is rather small; it averages near 16relative percents. It should be also emphasized that thecorrelation coefficient for TiO2 distributions generated withthe Clementine and simulated AMIE/SMART-1 data israther high; it equals 0.94.[31] For further comparison we used the prediction of

TiO2 distribution by Lucey et al. [2000a] (see equation (3))for the same region. We have studied the linear regression(TiO2)L = A(TiO2)S + B, where (TiO2)L and (TiO2)S corre-spond to Lucey et al.’s [2000a] and simulated AMIE/SMART-1 (3-bands) data, respectively. Results are givenin Table 3, which shows that the correlation coefficient ofthe regression is very high, though the scales of the TiO2

distributions are rather different. In our case the scale isnoticeably compressed. For the region under study one doesnot meet values of TiO2 more than 12%, whereas Lucey et

al.’s [2000a] prediction shows values up to 20%. The latterseems to be too large.[32] Iron is a very important rock-forming element. It is

incorporated into several minerals and influences albedo ofrocks and soils and produces absorption bands correspond-ing to specific minerals, such as pyroxene and olivine. Ahighly diagnostic absorption band due to Fe2+ is seen inlunar spectra near 1 mm [e.g., Burns, 1993]. Therefore onewould expect to find some correlation of albedo (and/orratios) near 1 mm with FeO abundance.[33] The results of our prediction for FeO are shown in

Figure 5. As would be anticipated highlands have signifi-cantly lower FeO abundance than in the case of maria.Distinct boundaries can be found between different mareareas. In many cases the boundaries on the map of FeO andTiO2 are the same that provides correlation between thesetwo parameters. An example of such a common boundarycan be found southward of the crater Aristarchus.[34] By analogy with the titanium case we compared our

results with those of Lucey et al. [2000a] using the follow-ing linear regression (FeO)L = A(FeO)S + B. To estimate(FeO)L we use equation (2). The coefficients A and B aregiven in Table 3. The range of FeO variations is approx-

Figure 4. Distribution of TiO2 for NW portion of the lunar nearside. Dark color means small content ofTiO2. The inserted frame (right side) presents the ratio of the TiO2 distributions originated withClementine (5-bands) and simulated AMIE/SMART-1 (3-bands) data.

1 - 6 SHKURATOV ET AL.: COMPOSITION OF LUNAR SURFACE FROM SMART-1

imately 3–18%; that is reasonable. The scale of our FeOdistribution is somewhat larger than in case of Lucey et al.’s[2000a] prediction, but on average (FeO)L = 16%, whereas(FeO)S = 13%. The reason for this difference can resultfrom that the suites of samples used in the LSCC analysisand for obtaining equation (2) are too different. Thecorrelation coefficient for the two compared distributionsof FeO is as high as 0.92. Thus our FeO distribution turnsout to be rather close to that of Lucey et al. [2000a].[35] The inserted fragment in Figure 5 presents variations

of the ratio of FeO distributions corresponding to Clem-entine (5-bands) and simulated AMIE/SMART-1 (3-bands)data. On average this ratio equals 0.99. The range of ratiovariations is small; it does not exceed 10%. Like in the caseof TiO2 the correlation coefficient for FeO distributionsgenerated with the Clementine (5-bands) and simulatedAMIE/SMART-1 (3-bands) data is very high; it equals0.97. Thus the FeO distributions are very similar to eachother.[36] The new maps of FeO and TiO2 abundance presented

here are very informative and their detailed analysis will bepresented elsewhere. We note here two results. Most of theyoung craters having bright albedo halos almost disappearon these maps. That is quite natural since the craters show

up in albedo mostly due to their immature soils. Thedisappearing means that their ejecta materials have approx-imately the same composition as surrounding areas. How-ever not all fresh craters disappear; some of them are seenon the FeO and TiO2 abundance maps indicating verticaldifferences in composition. Thus high-resolution imagesthat will be obtained with AMIE/SMART-1, using thesuggested method, provide an opportunity to study chemicalinhomogeneities of the lunar surface in vertical cross sectionat small scales. Another important observation for the FeOand TiO2 abundance maps is that the Reiner-g formation isalmost invisible. Hence this swirl is definitely not a com-positional anomaly on the lunar surface, at least for the ironand titanium parameter, in agreement with earlier inves-tigations [e.g., Bell and Hawke, 1981; Pinet et al., 2000].[37] The LSCC data, which includes detailed mineral

analyses, allows prediction of pyroxene abundance for mareregions. Figure 6 shows a distribution of total pyroxeneabundance obtained with equation (1). This is a first attemptto map pyroxene distribution, and therefore we have noopportunity to compare our results with data of otherworkers. We can only make a comparison for pyroxenedistributions generated with the Clementine (5-bands) andsimulated AMIE/SMART-1 (3-bands) data. We present their

Figure 5. Distribution of FeO for NW portion of the lunar nearside. Dark color means small content ofFeO. The inserted frame presents the ratio of the FeO distributions originated with Clementine (5-bands)and simulated AMIE/SMART-1 (3-bands) data.

SHKURATOV ET AL.: COMPOSITION OF LUNAR SURFACE FROM SMART-1 1 - 7

ratio with the inserted frame in Figure 6. As can be seen,spatial variations of this ratio are small; on average it doesnot exceed 10 relative percents. The correlation coefficientof these two suites of data is 0.92; the scales of thedistributions are very close to each other. The average ofthe pyroxene distribution for this region is 16%; variationsare approximately 3–20% that seems to be satisfactory.[38] The map of total pyroxene abundance illustrates

several aspects of soil formation as well as distinctions ofrock type across the Moon. Relatively fresh mare craters,e.g., the crater Aristarchus, have a higher abundance ofpyroxenes in their ejecta than in their surroundings. This isdue to the fact that crystalline material in surrounding areasis consumed during the agglutinate formation process. Wesee that Reiner-g formation contains crystalline material.This consists with facts suggesting the presence of a youngage for the formation [e.g., Shevchenko et al., 1993; Pinet etal., 2000]. Although the initial algorithm for estimating thepyroxene abundance is derived only from the mare soilsuite, the estimation works also fairly well for highlandsoils, which are known to exhibit lower pyroxene abun-dance. Lunar pyroclastic regions also exhibit low pyroxenesabundance (as they should). For instance, in Figure 6 onecan easily find the mantled Aristarchus Plateau and dark

mantle areas to the east of the crater Copernicus. Thus thesuggested technique provides a good tool to identify addi-tional regions with pyroclastics using Clementine and/orfuture AMIE/SMART-1 data.

3.3. Maturity and Particle Size

[39] There are different definitions of maturity degree oflunar soils [e.g., see Lucey et al., 2000b]. We use here thematurity characteristics Is/FeO that was introduced andstudied by Housley et al. [1973] and Morris [1977, 1980].This characteristic is the ratio of nanophase metallic ironcontent Is determined by measurements of ferromagneticresonance to bulk abundance of iron oxide in a sample.Nanophase iron grains are formed by a few processes suchas in-melt reduction with hydrogen, selective sputtering,and selective condensation of vaporization products result-ing from microimpacts [Morris, 1977, 1980; Hapke, 2001].Ferromagnetic resonance measurements are able to detectmetallic iron inclusions with sizes more than 40A but lessthan 300 A [Morris, 1977, 1980]. If the value Is is related tosurface condensation processes, it would correlate with thetotal area of surface particles. FeO bulk abundance is afunction of the total volume of particles. Thus the ratioIs/FeO should be a function of particle size. If Is is strictly

Figure 6. Distribution of pyroxene for NW portion of the lunar nearside. Dark color means smallcontent of pyroxene. The inserted frame presents the ratio of the pyroxene distributions originated withClementine (5-bands) and simulated AMIE/SMART-1 (3-bands) data.

1 - 8 SHKURATOV ET AL.: COMPOSITION OF LUNAR SURFACE FROM SMART-1

proportional to the surface area, then the ratio Is/FeO shouldbe inversely proportional to average particle size of regolithsamples. Morris [1977] found that for mature soils the ratioIs/FeO is actually proportional to d0.18, whereas for imma-ture regoliths this is d0.75. The first value for mature soilsmeans that the nanophase iron is basically distributed in theparticle volume; the second for immature soils correspondspredominantly to near-surface iron concentration.[40] Several attempts have been made to map the degree

of soil maturity [e.g., Lucey et al., 2000b; Shkuratov et al.,1999a], but they all are still preliminary. The techniquesuggested by Lucey et al. [2000b] uses the so-called opticalmaturity (parameter OM) instead of the ratio Is/FeO. Thistechnique suggests the optical maturity is obtained onlyfrom the diagram color ratio (950 nm/750 nm) versusalbedo (750 nm). If one uses the color ratio (415 nm/750nm) to map the optical maturity, the results turn out to bevery different [Kaydash and Shkuratov, 1998]. An improve-ment of Lucey’s technique has been made with the combi-nation of Clementine UVVIS and NIR data sets to betterevaluate the spectral alteration effect [LeMouelic et al.,2000]. Another technique offered by Shkuratov et al.[1999a] allows simultaneous determination of FeO, TiO2

abundance, and maturity degree, but it uses too poorly

constrained data for chemistry and Is/FeO values for thelanding sites of Apollo, Luna, and Surveyor missions.[41] New LSCC measurements may make it possible to

improve the Is/FeO prediction. Figure 7 shows a distributionof Is/FeO values mapped with equation (1) and simulatedAMIE/SMART-1 (3-band) data. Young mare craters (theirejecta and rims) have very low maturity degree. The Reiner-g formation is also an example of a region with veryimmature soil, with the possibility that the average maturityindex Is/FeO be as low as 25–40 [Pinet et al., 2000]. Formare regions the map of Is/FeO exhibits a similarity to thepyroxene map, when comparing craters to background. Thisis natural, as the amount of crystalline material shouldanticorrelate with regolith maturity. All pyroclastic deposits,such as the Aristarchus Plateau, exhibit distinctly highvalues of Is/FeO. This is hardly related to a real excess ofmaturity. The physical and compositional properties ofpyroclastic soils are perhaps too unusual to obey the samerules as other lunar materials. Note that Lucey et al. [1998,2000b] also suggested that pyroclastic materials may violatethe assumptions on which the FeO, TiO2, and opticalmaturity mapping algorithms are based. Thus the Is/FeO,as well as the pyroxene content parameter, is a goodempirical indicator for finding new sites of pyroclastic

Figure 7. Distribution of Is/FeO for NW portion of the lunar nearside. Dark color means small values ofIs/FeO. The inserted frame presents the ratio of the Is/FeO distributions originated with Clementine(5-bands) and simulated AMIE/SMART-1 (3-bands) data.

SHKURATOV ET AL.: COMPOSITION OF LUNAR SURFACE FROM SMART-1 1 - 9

materials. Even though these results are not yet optimizedfor highland soils, we note that the difference between mareand highland regions in the parameter Is/FeO is small.[42] We compared the Is/FeO distributions generated with

the Clementine (5-bands) and simulated AMIE/SMART-1(3-bands) data, and their ratio is given in Figure 6 (insertedframe). On average the ratio equals 1.2. It means that thesimulated AMIE/SMART-1 (3-bands) data predict 20%lower values for the ratio Is/FeO as compared with Clem-entine (5-bands) data. Spatial variations of the ratio arerather large and reach up to 25%. The correlation coefficientof these two suites of data is comparatively low, 0.80. Thusthe prediction of Is/FeO appears not as reliable as the otherparameters, though it predicts a reasonable average of Is/FeO (near 60) and variations of the parameter Is/FeO, fromseveral units to approximately 100.[43] We compared the distributions of Is/FeO predicted by

the LSCC data and the optical maturity parameter (OM)found with the technique by Lucey et al. [2000b] (seeequation (4)). The correlation coefficient is not high (near�0.78). This again suggests that the objective mapping ofthe parameter Is/FeO is in fact a difficult problem. Theregression coefficients A and B for this correlation are given

in Table 3 (all OM values are multiplied by the factor 1000at the calculations).[44] Although we have in our disposal only three particle

size-separates of the samples, it allows us to consider thecharacteristic particle size d as an independent parameter,and values linked to relative particle size of soils can at leastroughly be estimated by the suggested technique. Results ofsuch a prediction is shown in Figure 8. The produced mapof the parameter d is very exciting. Relatively coarseparticles appear to be characteristic of rims and ejecta offresh craters material that is quite natural. Small particles arerepresentative of highlands. Although our technique mightbe exploited with the restrictions, yet, the latter result islikely rational. The reason for some of the differencebetween highland and mare regoliths is known: plagioclase,

Figure 8. Distribution of d (the characteristic particle size) for NW portion of the lunar nearside. Thedark color means small values of d. The inserted frame presents the ratio of the d distributions originatedwith Clementine (5-bands) and simulated AMIE/SMART-1 (3-bands) data.

Table 3. Parameters of Linear Regression Comparing Results

Described Here With Those of Lucey et al. [2000a, 2000b]

Parameters A B k

TiO2 1.135 �0.22 0.94FeO 0.92 4.2 0.92Is/FeO and OM �0.0058 2.25 �0.78

1 - 10 SHKURATOV ET AL.: COMPOSITION OF LUNAR SURFACE FROM SMART-1

which dominates the highland soil and rocks, can becrushed by micrometeorite impacts more efficiently thanmafic minerals occurring in mare materials [e.g., Engelhardtet al., 1976].[45] We notice appreciable variations of the parameter d

in mare regions. Areas with possibly coarsely grainedregolith are southwest of Plateau Aristarchus and southwestof the crater Kepler. These regions can also be characterizedas coarsely grained in the mapping of the average particlesize with a polarimetric method [Shkuratov and Opana-senko, 1992]. These regions can be found as photometricanomalies in Earth-based telescope phase-ratio images[Shkuratov et al., 1994].[46] In Figure 8 we also find almost no traces of the

Reiner-g formation, suggesting that the characteristic par-ticle size of the formation would be the same as forsurrounding areas. This is a surprising result, as thisformation contains relatively crystalline material and showslow values of Is/FeO (low maturity is normally character-istic of coarsely grained regolith [Morris, 1977, 1980]). Thisresult is somewhat in contrast also with the previousdetailed studies proposing a local modification of the upperlayer of the regolith [e.g., Bell and Hawke, 1981; Pinet etal., 2000], with the removal of the finest fraction (<45 mm)in the soil and its redistribution in the vicinity with alaterally variable proportion and local accumulations [Pinetet al., 2000].[47] One possible explanation of the low contrast of the

Reiner-g formation is that the formation could have imma-ture coarsely grained regolith with interstices filled with finedust. On average this bimodal mixture would have theparticle size close to that for materials of surrounding areas.This would be consistent with earlier photometric studies ofthe region carried out at higher spatial resolution, whichshowed that the Reiner-g regolith surface has smoothedroughness [Kreslavsky and Shkuratov, 2003]. Anotherexplanation is that in the case of Reiner-g, the implicitassumption made that in a natural soil the larger particles arecovered by smaller particles with a typical size <45 mm maynot hold and consequently there may be some restrictionsfor interpretation relying on the sample suite.[48] As the map of characteristic particle size is in fact a

first reliable map of such a kind, there are no possibilitiesfor comparing our results with other data. However, asbefore, we have compared distributions generated with theClementine (5-bands) and simulated AMIE/SMART-1 (3-bands) data in Figure 8. On average this ratio is of about 1,though its variations are rather high, up to 30%. Thecorrelation coefficient of the two suites of data is 0.84.[49] Taken at face value, the characteristic size of par-

ticles over the region under study would be 21 mm; thescale of variations is approximately from 10 to 45 mm.Such estimates are not reasonable, if one considers thecharacteristic size as the average size of particles. Thecharacteristic size are at least 3 times smaller than well-known laboratory determinations of the average size ofparticles [e.g., McKay et al., 1991]. This is most likelylinked to the fact that the LSCC size separates are notnatural soils. However we anticipate that our characteristicparticle size map gives, at least qualitatively, adequateinformation about distribution of the parameter over themost portion of the lunar surface.

[50] We note also that the map of Is/FeO does notcorrelate almost with the distribution of characteristic par-ticle size. It indicates that the poor correlation betweenmean grain size and Is/FeO found by Morris [1977] formature regolith is mostly valid for the whole lunar surfaceexcluded perhaps only fresh craters.

4. Conclusion and Future Work

[51] 1. In this paper we represent a new technique forremote sensing determination of lunar surface compositionin the context of the SMART-1 mission. The technique isbased on spectral and compositional data for soil samplesobtained by the Lunar Soil Characterization Consortium.Since we used principally mare soils, our results are stillpreliminary. This concerns especially our prediction forhighlands. Although in future we plan to make improve-ments (extending the sample suite to include soils fromhighland regions) our technique is now able to give rationalresults. Being very simple, this technique can be applied inpractice. Thus when AMIE/SMART-1 images of the Moonare acquired and brought to the same photometric systemused by Clementine (RELAB lunar soils), one can directlyuse equation (1) and proper coefficients from Table 2 tomap the abundance of TiO2 and FeO, pyroxenes content,maturity degree (Is/FeO), and characteristic particle size forstudied lunar regions. We have presented examples of suchpreliminary maps for the five surface characteristics.[52] 2. Comparison of the TiO2 and FeO abundance maps

with distributions obtained with the approach by Lucey etal. [2000a] shows high correlations, though the actualalgorithms are very different. On average our estimationgives smaller values of TiO2 and FeO abundance for mareregions. We compare also the obtained Is/FeO distributionwith the parameter OM by Lucey et al. [2000b]. There is aninverse correlation between these two characteristics, asexpected. Unfortunately the correlation coefficient in thiscase is not high, only �0.78. After refinement with addi-tional highland samples, our Is/FeO characteristics may bemore suitable for applications.[53] 3. Our mapping of pyroxene abundance shows that

fresh mare craters have a higher abundance of pyroxenesin their ejecta and rims than in surroundings, in accord-ance with low maturity degree of the young crater rego-lith. Indeed, during short exposure time on the surface,crystalline material (e.g., pyroxenes) is much less con-sumed by agglutinate formation and the accompaniedgrowth of Is/FeO does not occur. The pyroxene abundancemap also clearly identifies pyroclastic regions as very lowpyroxene abundance. We plan to use this approach to findsmall unknown pyroclastic areas.[54] 4. We map the characteristic particle size using the

LSCC data for particle size separates. This attempt to studythe surface texture is very interesting and useful. Relativelycoarse particles appear to be characteristic of rims and ejectaof young craters, as would be expected. An excess of smallparticles is predicted for highland areas. This looks alsoreasonable as the plagioclase component, which dominatesthe highland soil and rocks, is crushed by micrometeoriteimpacts more effectively than mare material.[55] 5. We recommend special attention be devoted to

description of specific regions of the Moon using such

SHKURATOV ET AL.: COMPOSITION OF LUNAR SURFACE FROM SMART-1 1 - 11

maps. One example could be the Reiner-g formation. Ourpreliminary results confirm that this formation is not achemical anomaly, at least in TiO2 and FeO abundance, inagreement with previous analyses. Our maps also indicatethat the formation contains a surface material with unusualproperties characterized with low maturity and high degreeof crystallinity, consistent with the occurrence of immatureregolith possibly contaminated with dust.[56] 6. Our analysis and modeling with LSCC data and

Clementine multispectral images show that SMART-1 mul-tispectral data can be successfully used for estimating andmapping important parameters of the mare surface. How-ever, data rate limitations for AMIE/SMART-1 do not allowobtaining a global image set for the Moon. This requirescareful planning of sites to be imaged with the AMIEcamera. Indeed, unusual areas, the regolith of which isperhaps not represented in the lunar sample collection,should be the most interesting. There are many examplesof such areas, like swirls or red spots. On the other hand,our technique is developed using ‘‘ordinary’’ mare samples,and therefore our prediction is more reliable for typical mareregions. Thus prospective regions, which could be sug-gested for SMART-1 investigations, should include ‘‘stand-ard’’ areas that are the Apollo and Luna landing sites.

[57] Acknowledgments. This work is partially supported by INTASgrant 2000-0792, by the French Space Agency CNES, and by the PaulSabatier University of Toulouse, with the attribution of visiting positions toVVS, YGS, and DGS. Research support from NASA grant NAG5-10469(CMP) is gratefully acknowledged. This paper has also benefited from thepreparatory work achieved by the AMIE/Smart-1 Science team. Authorsthank D. Blewett for the thoughtful analysis of the paper.

ReferencesAlmeda, M., B. Foing, D. Heather, A. Marini, R. Lumb, and G. Racca,SMART-1 Science experiments co-ordination and expected outputs, pa-per presented at European Geophysical Society XXVII General Assem-bly, Nice, France, April 2002.

Bell, J. F., and B. R. Hawke, The Reiner Gamma Formation: Compositionand origin as derived from remote sensing observations, Proc. LunarPlanet. Sci. Conf. 12th, 679–694, 1981.

Blewett, D. T., P. G. Lucey, B. R. Hawke, and B. L. Jolliff, Clementineimages of the lunar sample-return stations: Refinement of FeO and TiO2mapping techniques, J. Geophys. Res., 102, 16,319–16,325, 1997.

Burns, R. G., Mineralogical Application of Crystal Field Theory, pp. 551,Cambridge Univ. Press, New York, 1993.

Charette, M., T. McCord, C. Pieters, and J. Adams, Application of remotespectral reflectance measurements to lunar geology classification anddetermination of titanium content of lunar soils, J. Geophys. Res., 79,1605–1613, 1974.

Engelhardt, W., H. Hurrle, and E. Luft, Microimpact-induced changes oftextural parameters and modal composition of the lunar regolith, Proc.Lunar Planet. Sci. Conf. 7th, 373–392, 1976.

Fischer, E. M., Quantitative compositional analysis of the lunar surfacefrom reflectance spectroscopy: Iron, aluminium, and a model for remov-ing the optical effects of space weathering, Ph.D. thesis, Brown Univ.,Providence, R.I., 1995.

Foing, B. H., and SMART-1 team, Laboratory Astrophysics and SpaceResearch, edited by P. Ehrenfreund et al., pp. 529–536, Eur. SpaceAgency, Paris, 1999.

Hapke, B., Space weathering from Mercury to the asteroid belt, J. Geophys.Res., 106, 10,039–10,073, 2001.

Housley, R., R. Grant, and N. Paton, Origin and characteristics of excess Femetal in lunar glass welded aggregates, Proc. Lunar Planet. Sci. Conf.4th, 2737–2749, 1973.

Hillier, J. K., B. J. Buratti, and K. Hill, Multispectral photometry of theMoon and absolute calibration of the Clementine UV/Vis Camera, Icarus,141, 205–225, 1999.

Johnson, J. R., S. M. Larson, and R. B. Singer, Remote sensing of potentiallunar resources: 1, Near-side compositional properties, J. Geophys. Res.,96, 18,861–18,882, 1991.

Josset, J.-L., et al., Asteroid Moon micro-Imager Experiment (AMIE) forSMART-1 mission, science objectives and development status, paperpresented at European Geophysical Society XXVII General Assembly,Nice, France, April 2002.

Kaydash, V. G., and Y. G. Shkuratov, Fe, Ti, and Is/FeO maps for the lunarnearside: new estimations by optical data (abstract 1089) [CD-Rom],Proc. Lunar Planet. Sci. Conf. 29th, 1998.

Kreslavsky, M., and Y. Shkuratov, Anomalies of the lunar regolith structurein the vicinity of Apollo-15 landing site: Results from photometric ana-lysis of Clementine UVVIS images (abstract 1075) [CD-Rom], Proc.Lunar Planet. Sci. Conf. 32nd, 2001.

Kreslavsky, M., and Y. Shkuratov, Photometric anomalies of the lunar sur-face: Results from Clementine data, J. Geophys. Res., 108, doi:10.1029/2002JE001937, in press, 2003.

Le Mouelic, S., Y. Langevin, S. Erard, P. Pinet, Y. Daydou, and S. Chevrel,Discrimination between maturity and composition of lunar soils fromintegrated Clementine UVVIS-NIR data: Application to Aristarchus Pla-teau, J. Geophys. Res., 105, 9445–9455, 2000.

Lucey, P., G. Taylor, and E. Malaret, Abundance and distribution of iron onthe Moon, Science, 268, 1150–1153, 1995.

Lucey, P., D. Blewett, and B. Hawke, Mapping the FeO and TiO2 content ofthe lunar surface with multispectral imagery, J. Geophys. Res., 103,3679–3699, 1998.

Lucey, P. G., D. T. Blewett, and B. L. Jolliff, Lunar iron and titaniumabundance algorithms based on final processing of Clementine ultravio-let-visible images, J. Geophys. Res., 105, 20,297–20,306, 2000a.

Lucey, P. G., D. T. Blewett, G. J. Taylor, and B. R. Hawke, Imaging of thelunar surface maturity, J. Geophys. Res., 105, 20,377–20,386, 2000b.

McCord, T. B., and J. B. Adams, Progress in remote optical analysis oflunar surface composition, Moon, 7, 453–474, 1973.

McCord, T. B., C. M. Pieters, and M. A. Feierberg, Multispectral mappingof the lunar surface using ground-based telescope, Icarus, 29, 1 –34,1976.

McCord, T. B., R. N. Clark, B. R. Hawke, L. A. McFadden, P. D. Owensby,C. M. Pieters, and J. B. Adams, Moon: Near-infrared spectral reflectance,a first good look, J. Geophys. Res., 86, 10,883–10,892, 1981.

McKay, D., G. Heiken, A. Basu, G. Blanford, S. Simon, R. Reedy, B. French,and J. Papike, The Lunar Regolith, Lunar Source-Book, pp. 285–356,Cambridge Univ. Press, New York, 1991.

McEwen, A. S., M. S. Robinson, E. M. Eliason, P. G. Lucey, T. C. Dux-burry, and P. D. Spudis, Clementine observations of the Aristarchusregion of the Moon, Science, 266, 1858–1862, 1994.

McEwen, A., E. Eliason, P. Lucey, E. Malaret, C. Pieters, M. Robinson, andT. Sucharski, Summary of radiometric calibration and photometric nor-malization steps for the Clementine UVVIS images (abstract 1466) [CD-Rom], Proc. Lunar Planet. Sci. Conf. 29th, 1998.

Morris, R., Origin and evolution of the grain-size dependence of the con-centration of fine-grained metal in lunar soils: The maturation of lunarsoils to a steady-state stage, Proc. Lunar Planet. Sci. Conf. 8th, 3719–3747, 1977.

Morris, R., Origin and size distribution of metallic iron particles in the lunarregolith, Proc. Lunar Planet. Sci. Conf. 11th, 1697–1712, 1980.

Nozette, S., et al., Clementine mission to the Moon: Scientific overview,Science., 266, 1835–1839, 1994.

Omelchenko, V., Yu. Shkuratov, D. Stankevich, C. Pieters, and P. Pinet, Acomparison of two approaches using three NIR-VIS wavelengths forpredicting the lunar surface composition, paper presented at 36th Inter-national Microsymposium on Planetology, Moscow, 14–16 October2002.

Pieters, C. M., Mare basalt types on the front side of the Moon: A summaryof spectral reflectance data, Proc. Lunar Planet. Sci. Conf. 9th, 2825–2849, 1978.

Pieters, C. M., Compositional diversity and stratigraphy of the lunar crustderived from reflectance spectroscopy, in Remote Geochemical Analy-sis: Elemental and Mineralogical Composition, edited by C. M. Pietersand P. A. J. Englert, pp. 309–339, Cambridge Univ. Press, New York,1993.

Pieters, C. M., The Moon as spectral calibrations standard enabled by lunarsamples: The Clementine example, in Workshop on New Views of theMoon II: Understanding the Moon Through the Integration of DiverseDatasets, pp. 47–49, Lunar and Planet. Inst., Houston, Tex., 1999.

Pieters, C., S. Pratt, H. Hoffmann, P. Helfenstein, and J. Mustard, Bidirec-tional spectroscopy of returning lunar soils: Detailed ‘‘Ground Truth’’ forplanetary remote sensors, Proc. Lunar Planet. Sci. Conf. 22nd, 1069–1070, 1991.

Pieters, C., E. Fischer, O. Rode, and A. Basu, Optical effects of spaceweathering: The role of the finest fraction, J. Geophys. Res., 98,20,817–20,824, 1993.

Pieters, C., M. Stade, E. Fischer, S. Tompkins, and G. He, A sharper viewof the craters from Clementine data, Science, 266, 1844–1848, 1994.

1 - 12 SHKURATOV ET AL.: COMPOSITION OF LUNAR SURFACE FROM SMART-1

Pieters, C., L. Taylor, D. McKay, S. Wentworth, R. Morris, and L. Keller,Spectral characterization of lunar mare soils (abstract 1865) [CD-Rom],Proc. Lunar Planet. Sci. Conf. 31st, 2000.

Pieters, C. M., D. G. Stankevich, Y. G. Shkuratov, and L. A. Taylor,Statistical analysis of the links between lunar mare soil mineralogy,chemistry and reflectance spectra, Icarus, 155, 285–298, 2002.

Pinet, P., V. Shevchenko, S. Chevrel, Y. Daydou, and C. Rosemberg, Localand regional lunar regolith characteristics at Reiner Gamma formation:Optical and spectroscopic properties from Clementine and Earth-baseddata, J. Geophys. Res., 105, 9457–9475, 2000.

Racca, G., B. Foing, and M. Coradini, Smart-1: The first time Europe to theMoon; wandering in the Earth-Moon space, Earth Moon Planets, 85,379–390, 2001.

Racca, G. D., B. H. Foing, and Smart-1 Project team, Smart-1 projectdevelopment status, paper presented at European Geophysical SocietyXXVII General Assembly, Nice, France, April 2002.

Raitala, J., M. Kreslavsky, Y. Shkuratov, L. Starukhina, and V. Kaydash,Non-mare volcanism on the Moon: Characteristics from remote sensingdata (abstract 1457) [CD-Rom], Proc. Lunar Planet. Sci. Conf. 30th,1999.

Saari, I. M., and R. W. Shorthill, Isotermal and isophotic atlas of the Mooncontours through lunation, NASA Conf. Publ., 855, 186 pp., 1967.

Shevchenko, V. V., P. C. Pinet, and S. D. Chevrel, Remote sensing studiesof immature lunar soils (Reiner-Gamma Formation), Sol. Syst. Res., 27,310–321, 1993.

Shkuratov, Y. G., Color differences and chemical abundance in lunar soils,Sol. Syst. Res., 16, 51–56, 1982.

Shkuratov, Y. G., and N. V. Opanasenko, Polarimetric and photometricproperties of the Moon: Telescope observation and laboratory simulation.2, The positive polarization, Icarus, 99, 468–484, 1992.

Shkuratov, Y. G., L. V. Starukhina, M. A. Kreslavsky, N. V. Opanasenko,D. G. Stankevich, and V. G. Shevchenko, Principle of perturbation in-variance in photometry of atmosphereless celestial bodies, Icarus, 109,168–190, 1994.

Shkuratov, Y. G., V. G. Kaydash, and N. V. Opanasenko, Iron and titaniumabundance and maturity degree distribution on lunar nearside, Icarus,137, 222–234, 1999a.

Shkuratov, Y. G., M. A. Kreslavsky, A. A. Ovcharenko, D. G. Stankevich,E. S. Zubko, C. M. Pieters, and G. Arnold, Opposition effect from Clem-entine data and mechanisms of backscatter, Icarus, 141, 132 –155,1999b.

Shkuratov, Y. G., V. G. Kaydash, C. M. Pieters, and N. V. Opanasenko, Acomparison of absolute calibrations of Clementine UVVIS and Earth-based data for the Moon [CD-Rom], Proc. Lunar Planet. Sci. Conf.31st, 2000.

Shkuratov, Y., D. Stankevich, C. Pieters, L. Taylor, V. Kaydash, andV. Omelchenko, New estimates of the lunar surface composition withClementine images and lunar sample data, paper presented at 36th Inter-national Microsymposium on Planetology, 14–16 October 2002a.

Shkuratov, Y. G., M. A. Kreslavsky, D. G. Stankevich, V. G. Kaydash, P. C.Pinet, V. V. Shevchenko, B. H. Foing, and J.-L. Josset, Photometricstudies of the Moon from SMART-1, Sol. Syst. Res, in press, 2002b.

Shoemaker, E. M., M. S. Robinson, and E. M. Eliason, The south polarregion of the Moon as seen by Clementine, Science, 266, 1851–1854,1994.

Stankevich, D. G., Y. G. Shkuratov, C. M. Pieters, and L. A. Taylor,Statistical Analyses of mare soil chemistry using Clementine band con-straints (abstract 1859) [CD-Rom], Proc. Lunar Planet Sci. Conf. 33rd,2002.

Taylor, L. A., C. M. Pieters, R. V. Morris, L. P. Keller, D. S. McKay,A. Parchen, and S. J. Wentworth, Integration of the chemical andmineralogical characteristics of lunar soils with reflectance spectroscopy(abstract 1859) [CD-Rom], Proc. Lunar Planet Sci. Conf. 30th, 1999.

Taylor, L. A., R. V. Morris, C. Pieters, A. Patchen, D. H. Taylor, L. P.Keller, S. Wentworth, and D. S. McKay, Chemical characterization oflunar mare soils (abstract 1697) [CD-Rom], Proc. Lunar Planet. Sci.Conf. 31st, 2000a.

Taylor, L. A., R. V. Morris, C. M. Pieters, A. Patchen, D. H. Taylor, L. P.Keller, S. Wentworth, and D. S. McKay, Major contributions to spectralreflectance opacity by non-agglutinitic, surface-correlated nanophase iron(abstract 1842) [CD-Rom], Proc. Lunar Planet. Sci. Conf. 31st, 2000b.

Taylor, L. A., A. Patchen, D. H. Taylor, R. V. Morris, C. Pieters, L. P.Keller, S. Wentworth, and D. S. McKay, Mineralogical characterizationof lunar mare soils (abstract 1706) [CD-Rom], Proc. Lunar Planet. Sci.Conf. 31st, 2000c.

Taylor, L. A., C. M. Pieters, R. V. Morris, L. P. Keller, and D. S. McKay,Lunar mare soils: Space weathering and the major effects of surface-correlated nanophase Fe, J. Geophys. Res., 106, 27,985–28,000, 2001.

Tompkins, S., and C. Pieters, Mineralogy of the lunar crust: Results fromClementine, Meteoritics Planet. Sci., 34, 25–41, 1999.

Whitaker, E. A., Lunar color boundaries and their relationship to topo-graphic features: A preliminary survey, Moon, 4, 348–355, 1972.

�����������������������B. Foing and Z. Sodnik, European Space Agency/European Space

Research and Technology Center, Noordwijk, Netherlands. (bfoing@estec.esa.nl; Zoran.Sodnik@esa.int)J.-L. Josset, Jaquet-Droz 1, CSEM, CH-2007, Neuchatel, Switzerland.

( jean-luc.josset@csem.ch)V. G. Kaydash, V. V. Omelchenko, Y. G. Shkuratov, and D. G. Stankevich,

Astronomical Observatory of Kharkov National University, 35 Sumskaya St.Kharkov, 61022, Ukraine. (Yuriy_Shkuratov@hotmail.com; stankevich@astron.kharkov.ua; vkaydash@astron.kharkov.ua; omelchenko@astron.kharkov.ua.)C. Pieters, Geological Sciences, Brown University, Providence, RI

02912, USA. ( pieters@mare.geo.brown.edu)P. Pinet, S. Chevrel, and Y. Daydou, University P. Sabatier Observatory

Midi-Pyrenees, 14 Av. E. Belin, 31400 Toulouse, France. (Patrick.Pinet@cnes.fr; Serge.Shevrel@cnes.fr; Yves.Daydou@cnes.fr)V. Shevchenko, Shternberg Astronomical Institute, Moscow, 119899,

Russia. (Shev@sai.msu.ru)L. Taylor, Planetary Geoscience Institute, University of Tennessee,

Knoxville, TN 37996, USA. (ltaylor@utk.edu)

SHKURATOV ET AL.: COMPOSITION OF LUNAR SURFACE FROM SMART-1 1 - 13