derivation of elemental abundance maps at intermediate resolution from optical interpolation of...

ARTICLE IN PRESS

0032-0633/$ - se

doi:10.1016/j.ps

�CorrespondE-mail addr

Planetary and Space Science 53 (2005) 1287–1301

www.elsevier.com/locate/pss

Derivation of elemental abundance maps at intermediate resolutionfrom optical interpolation of lunar prospector gamma-ray

spectrometer data

Yuriy G. Shkuratova,�, Vadim G. Kaydasha, Dmitriy G. Stankevicha, LarissaV. Starukhinaa, Patrick C. Pinetb, Serge D. Chevrelb, Yves H. Daydoub

aAstronomical Institute of Kharkov National University, 35 Sumskaya St. Kharkov, 61022, UkrainebUniversity P. Sabatier Observatory Midi-Pyrenees, 14 Av. E. Belin, 31400 Toulouse, France

Received 17 February 2004; received in revised form 3 January 2005; accepted 7 July 2005

Available online 25 August 2005

Abstract

We propose a technique that interpolates available lunar prospector gamma-ray spectrometer (GRS) data using Clementine

UVVIS spectral reflectance images. The main idea is to use low resolution GRS data as a ‘‘ground truth’’ to establish relationships

linking optical data and geochemical information maximizing the respective correlation coefficients. Then the relationships and

Clementine UVVIS data are used to derive elemental abundance maps with significantly improved spatial resolution. The main

limitation of the technique is its dependence on how well the abundance of the elements correlates with the Clementine UVVIS data.

The technique can also be applied to analysis of coming D-CIXS/Smart-1 and AMIE/Smart-1 data to increase resolution of lunar

compositional maps. As an illustration of the suggested technique, maps for the elements Fe, Ti, O, Al, Ca, and Mg with pixel size

15 km� 15 km are presented. The Fe and Ti distributions resemble qualitatively to the maps obtained with the well-known

technique by lucey et al. (2000a. Lunar iron and titanium abundance algorithms based on final processing of Clementine ultraviolet-

visible images. J. Geophys. Res. 105, 20,297–20,306), though in our case the ranges of Fe and Ti variations are, respectively, wider

and narrower than for lucey’s maps. New maps for the elements Fe, Ti, O, Al, Ca, and Mg appear to be informative. For instance,

the map of oxygen abundance demonstrates an anomaly in the crater Tycho. The maps of Fe and Al contents show for highland

regions slight variations related to maturity degree. Reliability of this relation is confirmed with lunar sample data. The reason of the

correlation between chemical composition and exposition age of the lunar surface can be the global transport of the lunar surface

material due to meteorite impacts.

r 2005 Elsevier Ltd. All rights reserved.

Keywords: The moon; Gamma-ray spectrometry; Spectral reflectance; Chemical composition; Lunar surface

1. Introduction

The chemical elements Si, O, Fe, Ti, Al, Ca, and Mgare major elements in the lunar rocks and soils.Information about the abundance and distribution ofthese and some other elements over the lunar surfacewas obtained with different remote sensing techniques:gamma-ray, neutron, X-ray, and optical spectroscopy.

e front matter r 2005 Elsevier Ltd. All rights reserved.

s.2005.07.001

ing author. Tel.: +38057 719 2883.

ess: [email protected] (Y.G. Shkuratov).

First gamma-ray data allowing geochemical mappingof the Moon were obtained while measuring with theApollo orbital modules (e.g., Arnold et al., 1977). Asmall portion of the lunar surface near the equator wasstudied with effective spatial resolution approximately100 km� 100 km, providing distribution of Fe and Ti.Global gamma-ray spectrometer (GRS) and neutronspectrometer (e.g., Feldman et al., 1999; Lawrence et al.,2002, 2003; Elphic et al., 2002; Prettyman et al., 2002)data were acquired during the low- and high-altitudeportions of the Lunar Prospector mission. The data

www.elsevier.com/locate/pss

ARTICLE IN PRESSY.G. Shkuratov et al. / Planetary and Space Science 53 (2005) 1287–13011288

were presented as preliminary maps with different sizesof pixels, respectively, 0.51� 0.51 (15 km� 15 km at theequator) and 51� 51 (150 km� 150 km at the equator).The initial GRS data went through complicated proces-sing (Feldman et al., 1999; Lawrence et al., 2002, 2003;Prettyman et al., 2002). The data will certainly beimproved in the future. However, at the present time,the current maps are unique determinations of the lunarsurface composition, i.e., distributions of Si, O, Fe, Ti,Al, Ca, and Mg, with direct remote sensing technique.

Imaging X-ray spectrometry was also carried outfrom the Apollo orbital modules (e.g., Bielefeld et al.,1977) to estimate regional variations of the ratios Al/Siand Mg/Si. The effective spatial resolution of the X-raydata is approximately 30 km� 30 km, which is some-what higher than in the case of Apollo gamma-raymeasurements. Comparison of the X-ray and gamma-ray data showed generally good mutual agreement(Bielefeld, 1977). The Compact Imaging X-ray Spectro-meter (D-CIXS) proceeds the X-ray measurements ofthe Moon. The instrument is on board of ESA space-craft Smart-1 (e.g., Grande et al., 2002; Dunkin et al.,2003). The D-CIXS instrument will provide maps of thelunar surface geochemistry with resolution up to10 km� 10 km. The instrument has a large effectivearea, which provides high signal-to-noise ratio for thelunar surface that is a very weak source of fluorescenceX-rays.

There are also optical techniques to estimate chemicalcomposition of the lunar regolith. Iron and titanium arethe main chromophore (transition) elements in the lunarmaterial (e.g., Burns, 1993). They influence albedo anddifferent colour indexes, which give a potential oppor-tunity to determine concentrations of the elements. Atechnique to estimate the titanium content using colourindex (0.42/0.56 mm) maps was first suggested byCharette et al. (1974). Iron and titanium are veryimportant discriminators to classify lunar rocks and,therefore, their remote measurement draws attention ofmany workers. A prospective and widely used approachto determine TiO2 and FeO abundance in the lunarsurface was suggested by Lucey et al. (1995, 1998,2000a, b) and Blewett et al. (1997) using ClementineUVVIS spectral reflectance (CSR) data. Attempts toimprove this algorithm (Gillis et al., 2003) or developindependent techniques (e.g., Shkuratov et al., 1999 a,2003a, b; Le Mouelic et al., 2000) are continuing. TheLucey et al. (1995, 1998, 2000a, b) and Blewett et al.(1997) algorithms are considered to be empirical,however, as has been shown (Starukhina and Shkur-atov, 2001) it can be derived from a theoretical model oflight scattering in the lunar regolith (Shkuratov et al.,1999b). Owing to correlations between concentrations ofthe chromophores and other elements, e.g., Al, Ca, andMg (Pieters et al., 2002), there is an opportunity to mapthe elements using the Clementine optical data (e.g.,

Fischer and Pieters, 1995; Shkuratov et al., 2003a, b),taking advantage of their high spatial resolution, up to100m for CSR.

More information about character and distribution ofchemical elements on the lunar surface can be obtainedin joint or integrating analyses of CSR and LunarProspector GRS data (e.g., Clark and McFadden, 2000;Chevrel et al., 2002a; Lawrence et al., 2002). This ispossible as CSR measurements and GRS provideinformation, respectively, from the upper 1mm of thelunar surface and the upper 20–30 cm layer. Unfortu-nately, spatial resolutions of available CSR and GRSdata are very different and, therefore, their comparison,which is important, is not a simple task.

It would be very interesting to develop a reliabletechnique to amplify the spatial resolution of the LunarProspector GRS data. Of course, such a techniqueshould be used with a great caution, as it must inevitablyuse additional data that may produce uncertainties.What we suggest here is an empirical technique tointerpolate available GRS data using CSR imagesacquired with UVVIS camera (Shkuratov et al., 2004).It should be specially emphasised that we do notsimply compare CSR and GRS determinations ofchemical elements as has been done in other works(e.g., Chevrel et al., 2002b; Lawrence et al., 2002). Ourmain idea is to use low resolution GRS data as a‘‘Ground truth’’ to establish relationships linkingoptical data (i.e., mineralogy) and geochemical informa-tion. We search for the relationships with maximalcorrelation coefficients for measured and predicted data.Then we use the relationships to derive elementalabundance maps (Fe, Ti, O, Al, Ca, and Mg maps) ofintermediate spatial resolution by means of availableCSR mosaics. Thus we make an attempt to extractgeochemical information from CSR data fundamentallyusing the GRS data.

We specially note that the same approach can beapplied to D-CIXS/Smart-1 data. For this purpose theCSR mosaics and/or data obtained with AMIE/Smart-1camera (Josset et al., 2002) can be used.

The CSR data are presented with reflectance in fivedifferent spectral bands (415, 750, 900, 950, and1000 nm). At first glance, there are not enoughmeasurements to constrain the elemental abundance ofsix different elements (Fe, Ti, O, Al, Ca, and Mg).The reason why we obtained results that make senseis that not all of these elemental abundances areindependent of each other. The best example is thewell-known anti-correlation between Fe and Al (e.g.,Fischer and Pieters, 1995). We note that Lawrence et al.(2003) have independently used a measured anti-correlation between Lunar Prospector data on Th andClementine Fe data in the Kepler crater region toimprove the spatial resolution determination of Thabundance.

ARTICLE IN PRESSY.G. Shkuratov et al. / Planetary and Space Science 53 (2005) 1287–1301 1289

2. Method of interpolation and maps

2.1. Approach

Using a set of optical characteristics (e.g., albedo,colour indexes, etc.), we formally correlate GRS datawith a combination of the optical characteristics andfind regression parameters providing the maximalcorrelation coefficients. For this we bring CSR data tothe GRS grid (approximately 51� 51) using simplerebinning. Note that this grid is not actual resolutionof the GRS observations; this rather is a pixel resolutionfor display and data processing purpose. Then therelationships between the GRS and optical data are usedto derive elemental abundance maps with an improvedspatial resolution. The organigram presented in Fig. 1explains once again how to exploit CSR and GRS datain order to map the mentioned elements.

It should be noted that even if a correlation betweenthe GRS and optical data are casual (not causal), it mayprovide the application of the technique. This is anadvantage of our approach, since in first approximationit permits an empirical prediction of the elementabundance without a detailed physical grounding ofthe correlation. Next steps related to the use andinterpretation of any empirical relation betweenoptical parameters and elemental composition must bebased on an understanding of the natural physics ofthe system.

Fig. 1. Organigram illustrated the interpolating technique. The CSR

and GRS data are used to obtain the coefficients in Eqs. (1) and (2)

maximizing the correlation coefficients between measured predicted

data (dark arrows). Then from this and CRS data new GRS maps with

amplified spatial resolution can be produced (grey arrows).

The main limitation of the technique is its dependenceon how well the abundance of the elements correlateswith the Clementine UVVIS data.

2.2. Basic equations

Many non-linear combinations of optical parameterscan be chosen to find a rule for interpolation of GRSdata. In principle, the final result depends on the choice.However, if a few different combinations are chosenmore or less adequately the final results should be closeone to another. We use below two empirical combina-tions of different sets of optical characteristics, whichhave been used to predict abundance of chemicalelements in the lunar surface materials. The firstcombination is presented in Lucey et al. (1995, 1998,2000a). The function that we consider here, only as aformal non-linear combination of different albedo is thefollowing:

L ¼ q arctanAðl1Þ=Aðl2Þ � y

Aðl2Þ � x

� �� s

þ p, (1)

where L is a geochemical parameter (we use below Fe orTi abundance in weight%), A(l) is the albedo (%) at agiven wavelength l. At estimates of iron content we usealbedo at l1 ¼ 950 nm and l2 ¼ 750 nm. For titaniumthe wavelengths are l1 ¼ 415 nm and l2 ¼ 750 nm.

For reference, the coefficients q, p, s, x, and y thatcorrespond to Lucey’s CSR iron and titanium maps(Lucey et al., 1998, 2000a; Blewett et al., 1997) are givenin the first two lines of Table 1. Using iron and titaniummaps obtained from GRS data yields the coefficientsgiven, respectively, in the last two lines of Table 1.

The second empirical combination of spectral para-meters has been applied (Shkuratov et al., 2003a, b)to map chemical and mineral composition, using theCSR data at 1 km resolution (Eliason et al., 1999),and data of the Lunar Soils Characterisation Consor-tium (Taylor et al., 2001) for mare lunar soils. Thiscombination is

log P ¼ aAR þ bCBR þ cCIR1 þ hCIR2

þ fCIR3 þ eD þ g, ð2Þ

where P is a geochemical parameter (Fe, Ti, O, Al, Ca,or Mg abundance in weight%), AR ¼ Að750 nmÞ, in %;Cbr ¼ Að415 nmÞ=Að750 nmÞ,CIR1 ¼ Að900 nmÞ=Að750 nmÞ,CIR2 ¼ Að950 nmÞ=Að750 nmÞ,CIR3 ¼ Að1000 nmÞ=Að750 nmÞ,D ¼ Að750 nmÞAð1000 nmÞ=½Að900 nmÞ�2, and the coeffi-cients a, b, c, h, f, e, and g should provide the maximalcorrelation coefficients in correlations with the GRS‘‘Ground truth’’ data.

We use the GRS maps of Fe, Ti, O, Al, Ca, and Mgacquired during the high-altitude portions of the Lunar

ARTICLE IN PRESS

Table 1

Coefficients of Eq. (1) found with the optical (Lucey’s approach) and GRS-interpolating techniques

q p x y s k

For FeO from Lucey et al. (2000a) �17.43 �7.56 0.08 1.19 1 0.82

For TiO2 from Lucey et al. (2000a) 3.71 0 0 0.42 5.98 0.81

For Fe found from GRS data �11.23 �2.41 0.08 1.16 1 0.92

For Ti found from GRS data 2.77 �0.69 0 0.42 3 0.90

k is the correlation coefficient.

Table 2

Coefficients of Eq. (2) found with our technique using CSR and GRS data

a b c h e f g k

Fe �0.039 1.354 �38.321 0.314 16.251 �18.2 40.819 0.96

Ti �0.090 8.303 �95.841 0.72 43.169 �44.783 93.31 0.86

O 0.0015 �0.069 7.910 0.148 �3.676 4.217 �6.966 0.75

Al 0.020 �1.029 34.5 �0.174 �15.217 17.34 �35.26 0.76

Ca 0.020 �1.581 32.747 �1.425 �14.613 14.369 �29.523 0.67

Mg �0.031 1.217 �25.209 2.161 10.101 �5.940 19.47 0.73

k is the correlation coefficient.

Y.G. Shkuratov et al. / Planetary and Space Science 53 (2005) 1287–13011290

Prospector mission (the versions of June 15, 2002,http://pds-geosciences.wustl.edu/missions/lunarp/reduced.html). The map bin size is approximately 51 by 51 (i.e.,150 km by 150 km). We treat these maps as a ‘‘Groundtruth’’ to obtain coefficients q, p, s, x, y and a, b, c, h, f,e, g. Using the least-square method we search for themaximal correlation coefficients k for relationshipsbetween the ‘‘Ground truth’’ data and the parametersL and P. The values of the coefficients are given inTables 1 and 2 for L and P, respectively. One cancompare correlation coefficients k for Fe and Ti whenwe use the values of the coefficients q, p, s, x, and y

derived with the optical algorithm (Lucey et al., 2000a;Blewett et al., 1997) and with the GRS data. Highercorrelation coefficients k for GRS data (compare lines 1and 2, respectively, with lines 3 and 4 in Table 1) justifythe modified coefficients q, p, s, x, and y. Initially weinvolved in the analysis the element Si. However, in thiscase the correlation coefficient is as low as 0.6 and finallywe rejected this parameter.

Figs. 2a,b,3a,b, and 4a–d show correlation diagramsfor each mentioned chemical element. The diagramspresent the predicted values via GRS-measured data.Figs. 2a, 2b, 3a and 3b correspond to using Eqs. (1) and(2), respectively. Although the scatterplots seem to bedispersed, the correlation coefficients corresponding tothe central linear regression are fairly high (see Tables 1and 2), especially for iron (k ¼ 0:96). The worstcorrelation is observed for Ca (k ¼ 0:67). As can beseen the dependence in Fig. 2a is not linear; that is whywe use below only Eq. (2) for Fe prognosis; the same isobserved for O, Al, Ca, and Mg. In contrast, Eq. (1)

provides the linearity and higher correlation coefficientfor Ti.

In spite of relatively high correlation coefficients,uncertainties in the determination of element abun-dances are fairly high. For instance, they are +1.4 and�0.6wt% at the average 2wt% for Ti. We note,however, that the values are typical for opticaldeterminations of Ti (e.g., Charette et al., 1974; Gilliset al., 2003; Blewett et al., 1997).

The idea to study correlations of other elements, likethorium, uranium, or hydrogen in the same manner anduse this for the interpolation seems to be very attractive.Unfortunately, carrying out this, we have found veryweak correlations between measured and predictedvalues. Therefore, we exploited further only Fe, Ti, O,Al, Ca, and Mg data. On the other hand, for a smallregion rather close correlation between Fe and Th hasbeen found (Lawrence et al., 2003) and, therefore, ourtechnique can probably be extended for some lunarregions for thorium, uranium, or even hydrogenprognoses.

Another important issue is that the ClementineUVVIS data are strongly influenced with variations ofthe maturity degree of the lunar soils (e.g., Lucey et al.,2000b), while this important parameter does not affectthe GRS data. However, the Lunar Prospector ‘‘Groundtruth’’ data may discriminate regions with the immatureregolith that is characteristic of young craters, their raysystems, and perhaps swirls (e.g, Pinet et al., 2000;Starukhina and Shkuratov, 2004a), if maturity degreevariations accompany chemical composition changes(see below).

http://pds-geosciences.wustl.edu/missions/lunarp/reduced.html

http://pds-geosciences.wustl.edu/missions/lunarp/reduced.html

ARTICLE IN PRESS

Fig. 2. (a,b) Predicted Fe vs. GRS-measured Fe (wt%). Diagrams (a)

and (b) present data obtained with Eqs. (1) and (2), respectively.

Predicted and measured data correspond to the pixel size 150 km.

Fig. 3. (a,b) Predicted Ti vs. GRS-measured Ti (wt%). Diagrams (a)

and (b) present data obtained with Eqs. (1) and (2), respectively.

Predicted and measured data correspond to the pixel size 150 km.

Y.G. Shkuratov et al. / Planetary and Space Science 53 (2005) 1287–1301 1291

2.3. Maps

Fig. 5 presents maps regarding the iron distribution.The upper image is the initial GRS data. The lowerimage is a prediction for iron distribution obtained withEq. (2) at the spatial resolution of 15 km. Our Fe maplooks very similar to published maps with analogouscontent (e.g., Lucey et al., 2000a; Chevrel et al.,2002a, b; Lawrence et al., 2002). Highlands have typicalvalues of Fe content near 4%. Mare regions, such as

Procellarum Ocean, Mare Tranquillitatis, Mare Sereni-tatis, and Mare Imbrium, show significantly highercontent of Fe than highlands; the same is observed forthe South Pole—Aitken Basin. We may anticipate falsedetails associated with young craters, since they containimmature soils. Indeed, the Fe map reveals abnormal-ities related to large young craters. Owing to that themare regions are inhomogeneous.

Fig. 6 shows analogous maps for titanium. Nearsideportion of the map presented in lower panel looks

ARTICLE IN PRESS

Fig. 4. (a,d) Predicted vs. GRS-measured element contents. The diagrams (a)–(d) present O, Al, Ca, and Mg, respectively. Eq. (2) was used for the

prediction. Predicted and measured data correspond to the pixel size 150 km.


qualitatively close to published ones (e.g., Johnson et al.,1991; Blewett et al. 1997; Shkuratov et al., 1999a). Inparticular, well-known titanium difference betweenMare Serenitatis and Mare Tranquillitatis is clearlyseen. The highland regions are fairly homogeneous withtypical values of titanium content near 0.3%.

We study correlations between the pure optical(Lucey’s technique) and interpolated GRS distributionsof iron and titanium taking into account the differencebetween elemental (e.g., Fe and Ti) and oxide (FeO andTiO2) abundances. Fig. 7 shows that the correlation forFe is non-linear as has been shown by Lawrence et al.(2002). We also note that Lucey’s optical technique hasa limitation in the domain of high values of iron content,whereas the interpolated GRS data allow the valueshigher than 20% that is in agreement with conclusionsby Lawrence et al. (2002).

Analogous results obtained for titanium are shown inFig. 8. Depending on the choice of the values of thecoefficients given in Table 1 (the second line or thefourth line), we may use Eq. (1) in two versions, as a toolfor the optical prognosis (Lucey’s approach) and as aninterpolator of the GRS data (our approach). Thisallows us to study the correlation between these twopredictions for titanium (see Fig. 8). Since we use almostthe same formula to calculate the titanium distributions,the dependence is functional demonstrating a non-linearrelation between the pure optical and interpolated GRSpredictions. The interpolated GRS data show less valuesof Ti content in the high titanium domain as comparedto Lucey’s optical determinations. The same differencewas obtained by Elphic et al. (2002).

As has been mentioned, Eq. (2) provides the linearityand higher correlation coefficients for the elements O,

ARTICLE IN PRESS

Fig. 5. Iron distributions over the lunar surface. Maps correspond to

the initial GRS data (upper panel, pixel size 150 km) and applying

Eq. (2) (lower panel, pixel size 15 km); one scale is given for the two

maps. The map projection is simple-cylindrical.

Fig. 6. Titanium distributions over the lunar surface. Maps corre-

spond to the initial GRS data (upper panel, pixel size 150 km) and

applying Eq. (2) (lower panel, pixel size 15 km); one scale is given for

the two maps. The map projection is simple-cylindrical.

Fig. 7. Correlation diagram for Fe distributions obtained with Eq. (2)

and using Lucey’s approach (coefficients from first line of Table 1).

The latter is indicated as Lucey et al. (2000a, b).

Fig. 8. Correlation diagram for Ti distributions obtained with Eq. (1)

using coefficients corresponding to the second and fourth lines of

Table 1. The first is indicated as Lucey et al. (2000a).


Al, Ca, and Mg than Eq. (1). Therefore, we use Eq. (2)for further determinations. Figs. 9–12 present maps ofthe elements O, Al, Ca, Mg, respectively. The upperpanels are the initial Lunar Prospector data. The lowerpanels present results of our interpolations. One canexpect false details in areas with young craters as theycontain immature soils. However, the obtained maps arefairly smooth and do not reveal big abnormalitiesrelated to young craters. An exception is the Mgdistribution.

We emphasise once more that though the listedelements are not chomophores, they correlate withoptical parameters through iron and titanium.Additional arguments in favour of the possibilityto map such elements as Ca, Mg, and others are asfollows. Spectral properties of the lunar regolith areformed with minerals, not with elements. An important

ARTICLE IN PRESS

Fig. 9. Oxygen distributions over the lunar surface. Maps correspond

to the initial GRS data (upper panel, pixel size 150 km) and applying

Eq. (2) (lower panel, pixel size 15 km); one scale is given for the two

maps. The map projection is simple-cylindrical.

Fig. 10. Aluminium distributions over the lunar surface. Maps

correspond to the initial GRS data (upper panel, pixel size 150 km)

and applying Eq. (2) (lower panel, pixel size 15 km); one scale is given

for the two maps. The map projection is simple-cylindrical.

Fig. 11. Calcium distributions over the lunar surface. Maps corre-

spond to the initial GRS data (upper panel, pixel size 150 km) and

applying Eq. (2) (lower panel, pixel size 15 km); one scale is given for

the two maps. The map projection is simple-cylindrical.

Fig. 12. Magnesium distributions over the lunar surface. Maps

correspond to the initial GRS data (upper panel, pixel size 150 km)

and applying Eq. (2) (lower panel, pixel size 15 km); one scale is given

for the two maps. The map projection is simple-cylindrical.


factor influencing the intensity and position of absorp-tion bands is the distances between ions in the crystallinelattice of minerals. Non-transition elements changethe lattice distances affecting in that way the opticalspectra. For example, Ca in pyroxenes noticeablychanges the positions of the crystal-field bands near 1and 2 mm (Adams, 1974), though it is not transitionelement.

All the maps with the high resolution are veryattractive. However, in what measure are they true?We can suggest that two points supported theirreliability. Note, first of all, that there are no principlerestrictions on our technique, as all the studied elementscorrelate with optical data and there are no reasons toexpect that the extrapolation should destroy thecorrelations. This is confirmed with the fact that thesame analysis gives reasonable results when we use lunar

ARTICLE IN PRESS

Fig. 13. Correlation between Fe distributions obtained with Eq. (2)

and low-altitude GRS data with 15 km pixels.

Fig. 14. (a,b) Optical/GRS ratios for iron (a) and titanium (b)

distributions. The map projection is simple-cylindrical.


samples as the ‘‘Ground truth’’ data (Shkuratov et al.,2003a, b). Second, we have a unique opportunityto check our interpolation technique with real datafor iron distribution. During the low-altitude portionof the Lunar Prospector mission, an iron distributionwith higher spatial resolution was obtained. Theresolution allows the pixel size near 15 km. Weemphasise that the 15-km resolution is conditional,since gamma-ray detectors gather information from thefull hemisphere when looking at a planet, though it ismore sensitive to photons coming from the nadir. Thereis some signal mixing when the instrument measuressimultaneously two different compositions that are inthe same field of view. Consequently, a value in a givenpixel does not exactly represent the abundance intothat pixel. Nevertheless, this is a case that provideshigher resolution.

Fig. 13 shows a correlation diagram for interpolatedand measured data both are with 15-km pixels. As onecan see the correlation is strong (k ¼ 0:93), though pointscatter is also observed. The reasons of the scatter can benot only shortcomings of our technique, but also actualdifference between the 150 and 15-km GRS data. Toverify this, we compared the 150-km map to 15-km databringing formally the latter to the 150-km resolution.The point dispersion in this case is even slightly higher(k ¼ 0:92) in comparison with the diagram shown inFig. 13. Thus we may conclude that the intermediateresolution maps of the elements Fe, Ti, O, Al, Ca, andMg presented in Figs. 5, 6, 9–12, being preliminary, canbe used for further analyses.

3. Discussion

3.1. Fe and Ti content

Maps of global distributions of iron and titaniumhave been described in the literature many times (e.g.,Lucey et al., 1998, 2000a; Blewett et al., 1997; Gillis etal., 2003; Lawrence et al., 2002, 2003). We may give tothose descriptions additional information concerningdifference in distributions of optically determined(Lucey et al., 2000a; Blewett et al., 1997) andinterpolated GRS iron (titanium) with 15-km pixels.Fig. 14a and b presents the optical/GRS ratios. Anumber of features in Fig. 14a can be immediatelynoted. Maria of the lunar nearside have somewhat lowerabundance of iron determined with the optical algo-rithm. The same is observed for the north highland ofthe farside. In contrast, the South Pole—Aitken Basin,cryptomaria (e.g., the Schickard—Schiller region), andhighlands surrounding the nearside maria are charac-terised with relatively high values of iron concentrationestimated with the optical method. The discrepanciesfound for the South Pole—Aitken Basin are in agree-ment with earlier studies by Chevrel et al. (1999, 2002a)and Lawrence et al. (2002). Typical variations of theratios are fairly small, with the RMS deviation nearby10 relative %.

The optical/GRS ratio for titanium (see Fig. 14b)reveals a prominent latitude variation that is addressedto photometric quality of the Clementine multispectralmosaics. Phase angles of the Clementine imaging arerelated to the latitude of observation. Although the

ARTICLE IN PRESS

Fig. 15. (a–d) Correlation diagrams for different combinations of elements using prognosis maps with 15 km pixels (grey colour) and initial GRS

data with 150 km pixels (circles): (a) Ti–Fe, (b) Fe–Al, (c) O–Fe, and (d) Mg–Ca.


phase-angle effect have been removed from the Clem-entine UVVIS data used here, the data might still beaffected by residual latitude effects (e.g., Le Mouelic etal., 2000; Shkuratov et al., 2003b). Mare regions withvery high concentrations of Ti (e.g., Mare Traquillitatis)show up on the ratio image. This means that the scale ofGRS titanium is stretched in comparison with opticalestimates. Typical variations of the ratios for Ti are ratherhigh, with the RMS deviation nearby 30 relative %.

Thus, at high concentrations of Fe and Ti the opticalprognosis underestimates iron and overestimates tita-nium concentrations in comparison with the GRStechnique.

We study also correlation between iron and titanium.As can be seen in Fig. 15a, there is a close non-linearcorrelation between these elements in agreement withour previous results (e.g., Shkuratov et al., 1999a;

Pieters et al., 2002). The Ti–Fe diagram shown inFig. 15a contains data including both our prognosiswith 15 km pixels (grey colour) and initial LunarProspector data with 150 km pixels (circles). Thesedifferent kinds of data demonstrate good coincidence.

3.2. Other elements

The distribution of aluminium abundance (see Fig. 10)resembles an albedo image with noticeable suppressingof bright craters and their ray systems. Thus thealuminium distribution strongly anti-correlates withthe iron content. This is confirmed with the correlationdiagram Fe–Al presented in Fig. 15b. This result isconsistent with numerous laboratory studies of lunarsamples (e.g., Fischer and Pieters, 1995; Taylor et al.,2001; Pieters et al., 2002).


The oxygen distribution is generally correlated withabundance of iron and titanium (cf. Fig. 9,5 and 6), asultrabasic rocks contain usually more Fe and Ti thanbasic ones. Fig. 15c confirms this showing the propercorrelation diagram O–Fe that, as in the case of theTi–Fe correlation, presents two sorts of data, ourprognosis (grey colour) and initial GRS maps (circles).Note several interesting anomalies of the oxygendistribution. They are associated with the SouthPole—Aitken Basin, the Schickard—Schiller region,and the crater Tycho. We note also a subtle regionalstructure in Procellarum Ocean that corresponds to lavaflows of different ages (Hiesinger et al., 2000).

The distribution of calcium (see Fig. 11) showsrelatively low contrast mare/highland, though all mariaare clearly seen. Higher content of the element isobserved for the equatorial regions of the farside. Fig.12 presents magnesium abundance. Mare regions areinhomogeneous for this parameter. Some bright marecraters can be seen in the map. That is perhaps ademonstration of shortcomings of our technique thatcannot be always applicable for areas with immatureregolith. Concentrations of calcium and magnesium areglobally anti-correlated. Fig. 15d illustrates this, reveal-ing also noticeable point scatter. More detailed con-sideration of the correlation diagram Mg–Ca shows twostrongly overlapped branches.

3.3. Histograms

We produce also histograms of distributions of all themapped elements for initial and amplified resolutions(see Fig. 16a and b). As can be seen, the histogramscorresponding to low (a) and high (b) resolutions havealmost the same width and shape for each element. Theshapes of the histograms vary for various elements. Thehistograms of iron, titanium, and oxygen distributionsare very asymmetric, revealing maximum correspondingto highland values of the parameter. The histogram ofcalcium distribution is the most symmetric, reflecting thesmall mare/highland contrast of the parameter. We donot find clear bi-modality of the distributions of thestudied chemical elements. That means strong over-lapping of mare and highland distributions of thestudied parameters.

3.4. Mixing of lunar surface materials

Vertical and horizontal mixing of mare and highlandmaterials due to meteorite impacts is a very importantprocess influencing the lunar surface composition. Inmare regions the highland component is mainlyexcavated from the highland substrate of maria due torather powerful impacts (vertical mixing). Contamina-tion of highland regions by mare material is weaker. The

contamination is basically due to the horizontal trans-port at impact events.

We note that highland young craters and their brightray systems are clearly seen on the maps of Fe and Al(see Figs. 5 and 10); the Fe content is low (and the Alcontent is higher) for these regions than for thesurroundings. This means the Fe and Al abundance tocorrelate with the exposition age of the lunar surfaces.At first glance, such a correlation is not obvious and onecan think that the showing up of the young crates andtheir ray systems is a manifestation of shortcomings ofour technique. We show, however, that this correlationis probably real and relates perhaps to the horizontaltransport of mare materials.

To verify the correlations, we use data for lunarsamples. The exposition age of the lunar surface can becharacterised with several parameters. Among them theregolith maturity parameter Is/FeO that is the ratio offerromagnetic resonance intensity Is (which is propor-tional to metallic iron content) to the total Fe content(Morris, 1976, 1977, 1978, 1980) is of the most reliable.We studied relationship between Is/FeO and FeO usingdata for highland lunar samples from Apollo-16 landingsites (Morris, 1976, 1977, 1978, 1980). Fig. 17 clearlyshows the correlation: the higher the maturity degreeIs/FeO, the higher the FeO content.

Another parameter characterising the exposition ageof the lunar surface is abundance S of the crystallinecomponent in the lunar regolith. The abundance ishigher for immature lunar soils (small exposition age).We studied correlation between S and Al2O3 contentusing data for highland lunar samples from Apollo-16landing sites (Taylor et al., 1991). Fig. 18 clearlydemonstrates the correlation: the higher the parameterS, the higher the Al2O3 content. Thus both Figs. 17 and18 are in qualitative agreement with the maps presentedin Figs. 5 and 10.

As has been mentioned, the reason of the correlationbetween chemical composition and exposition age of thelunar surface can be global transport of the lunarsurface material over the Moon in rather large impactevents. This transport leads to contamination of thehighland surface with mare materials, which depends ontime. This dependence may provide the correlation; seealso relative works (Pieters and Taylor, 2003; Starukhi-na and Shkuratov, 2004b).

4. Conclusion

Thus we suggest an empirical approach that allowsone to derive elemental abundance maps from LunarProspector data for such elements as Fe, Ti, O, Al, Ca,and Mg at an improved resolution, intermediatebetween those of GRS and CSR data (with a potentialgain of a factor of 10) permitting regional geologic

ARTICLE IN PRESS

Fig. 16. Histograms for distributions of all the studied elements (Fe, Ti, O, Al, Ca, and Mg) at different sizes of pixels. Plots (a) and (b) correspond to

the pixels 150 km (initial data) and 15 km (interpolated data), respectively.


investigations. The main limitation of the technique is itsdependence on how well the abundance of the elementscorrelates with the Clementine UVVIS data.

Our new iron and titanium maps were produced usingCSR data, which were calibrated with GRS data, exhibitnon-linear relationships with corresponding distribu-tions obtained with Lucey’s technique. That is consistent

with previous works (e.g., Chevrel et al., 2002a, b;Lawrence et al., 2002). We confirmed reliability of ourapproach with the correlation between the predictediron distribution and measured one using in both thesecases the maps with 15-km pixels. The maps of non-chromophore elements, O, Al, Ca, and Mg appear tobe informative. For example, the map of oxygen

ARTICLE IN PRESS

Fig. 17. Correlation between maturity degree Is/FeO and FeO

abundance for highland samples (Morris, 1976, 1978, 1977, 1980).

Fig. 18. Correlation between crystalline component S and Al2O3

abundances for highland samples (Taylor et al., 1991).


demonstrates an anomaly in the crater Tycho, and themap of calcium shows low mare/highland contrast.

The maps of Fe and Al contents show for highlandregions slight variations related to young craters andtheir ray systems. The correlation between the bulkchemistry and maturity degree of the lunar regolith isconfirmed with data for lunar samples from the Apollo-16 landing site. The reason of the correlation can beglobal transport of the lunar surface material due tometeorite impacts.

Our future work will be devoted to investigation ofcorrelations between the chemical elements and toclassification of the lunar surface with these correlations(cluster analyses). We will refine our technique (Shkur-atov et al., 2003a, b) based on the Lunar SoilsCharacterisation Consortium data (Taylor et al., 2001)to produce maps of the same chemical elements and tocompare the maps with obtained in the present work. Inaddition, we plan to present a detailed description of thenew maps. The technique described above can be easilyapplied to D-CIXS/Smart-1 data in order to increasespatial resolution of lunar compositional maps.

Acknowledgement

This work is partially supported by INTAS Grant#2000-0792, CRDF Grant #UKP2-2614-KH-04, theFrench Space Agency CNES, and the Paul SabatierUniversity of Toulouse, with the attribution of visitingpositions to YGS and DGS. The authors thank Drs. D.Lawrence and O. Gasnault for the thoughtful discus-sions and many pieces of advice.

References

Adams, J.B., 1974. Visible and near-infrared diffuse reflectance spectra

of pyroxenes as applied to remote sensing of solid objects in the

solar system. J. Geophys. Res. 79, 4829–4836.

Arnold, I., Metzger, A.E., Reedy, R.C., 1977. Computer-generated

maps of lunar composition from gamma-ray data. In: Proceedings

of the Lunar Science Conference, Eighth. Pergamon Press, NY, pp.

945–948.

Bielefeld, M.J., 1977. Lunar surface chemistry of regions common to

the orbital X-ray and gamma-ray experiments. In: Proceedings of

the Lunar Science Conference, Eighth. Pergamon Press, NY, pp.

1131–1148.

Bielefeld, M.J., Andre, C.G., Eliason, E.M., Clark, P.E., Adler, I.,

Trombka, J.I., 1977. Imaging of lunar surface chemistry from

orbital X-ray data. In: Proceedings of the Lunar Science

Conference, Eighth. Pergamon Press, NY, pp. 901–908.

Blewett, D.T., Lucey, P.G., Hawke, B.R., Jolliff, B.L., 1997.

Clementine images of the lunar sample-return stations: refinement

of FeO and TiO2 mapping techniques. J. Geophys. Res. 102,

16,319–16,326.

Burns, R.G., 1993. Mineralogical Application of Crystal Field Theory.

Cambridge University Press, Cambridge, MA, p. 551.

Charette, M., McCord, T., Pieters, C., Adams, J., 1974. Application of

remote spectral reflectance measurements to lunar geology

classification and determination of titanium content of lunar soils.

J. Geophys. Res. 79, 1605–1613.

Chevrel, S., Pinet, P., Barreau, G., Daydou, Y., Richard, G., Maurice,

S., Feldman, W., 1999. Integration of the UV-VIS spectral

Clementine data and the gamma-ray Lunar Prospector data:

preliminary results concerning FeO, TiO2, and Th abundances of

the lunar surfaces at global scale. In: Workshop on the New View

of the Moon II: Understanding the Moon Through the Integration

of Diverse Datasets, LPI Contrib. 980, LPI Houston.

Chevrel, S., Pinet, P., Daydou, Y., Maurice, S., Lawrence, D.,

Feldman, W., Lucey, P., 2002a. Integration of the Clementine

UV-VIS spectral reflectance data and the Lunar Prospector

ARTICLE IN PRESSY.G. Shkuratov et al. / Planetary and Space Science 53 (2005) 1287–13011300

gamma-ray spectrometer data: a global-scale multi-element analy-

sis of the lunar surface using iron, titanium, and thorium

abundance. J. Geophys. Res. 107 (E12).

Chevrel, S.D., Pinet, P., Daydou, Y., Feldman, W., 2002b. Integration

and comparison of the Clementine and the Lunar Prospector data:

Global-scale multi-element analysis (Fe, Ti and Th) of the lunar

surface. Solar Syst. Res. 36, 458–465.

Clark, P., McFadden, L., 2000. New results and implications for lunar

crustal iron distribution using sensor data fusion techniques. J.

Geophys. Res. 105, 4291–4316.

Dunkin, S.K., Grande, M., Casanova, I., Fernandes, V., Heather,

D.J., Kellett, B., Muinonen, K., Russell, S.S., Browning, R.,

Waltham, N., Parker, D., Kent, B., Swinyard, B., Perry, A.,

Feraday, J., Howe, C., Phillips, K., Huovelin, J., Muhli, P.,

Hakala, P.J., Vilhu, O., Thomas, N., Hughes, D., Alleyne, H.,

Grady, M., Lundin, R., Barabash, S., Baker, D., Clarke, P.E.,

Murray, C.D., Guest, J., Casanova, I., d’Uston, L., Maurice, S.,

Foing, B., 2003. Scientific rationale for the D-CIXS X-ray

spectrometer on board ESA’s SMART-1 mission to the Moon.

Planet. Space Sci. 51, 435–442.

Eliason, E.M., Isbel, C., Lee, E., Becker, T., Gaddis, L., McEwen, A.,

Masson, R., 1999. To the Moon: the Clementine UVVIS global

mosaic. US Geological Survey, Flagstaff, Ariz.

Elphic, R.C., Lawrence, D.J., Feldman, W.C., Barraclough, B.L.,

Maurice, S., Lucey, P.G., Blewett, D.T., Binder, A.B., 2002. Lunar

prospector neutrons spectrometer constraints on TiO2. J. Geophys.

Res. 107 (#E4).

Feldman, W.C., Barraclough, B.L., Fuller, K.R., Lawrence, D.J.,

Maurice, S., Miller, M.C., Prettyman, T.H., Binder, A.B., 1999.

The lunar prospector gamma-ray and neutron spectrometers. Nucl.

Instrum. Methods Phys. Res. A 422, 562.

Fischer, E.M., Pieters, C.M., 1995. Lunar surface aluminum and iron

concentration from Galileo solid state imaging data, and the

mixing of mare and highland materials. J. Geophys. Res. 100,

23,279–23,290.

Gillis, J., Jollif, B., Elphic, R., 2003. A revised algorithm for

calculating TiO2 from Clementine UVVIS data: A synthesis of

rock, soil, and remotely sensed TiO2 concentrations. J. Geophys.

Res. 108 (E2).

Grande, M., Dunkin, S., Heather, D., Kellett, B., Perry, C.H.,

Browning, R., Waltham, N., Parker, D., Kent, B., Swinyard, B.,

Fereday, J., Howe, C., Huovelin, J., Muhli, P., Hakala, P.J., Vilhu,

O., Thomas, N., Hughes, D., Alleyne, H., Grady, M., Russell, S.,

Lundin, R., Barabash, S., Baker, D., Clark, P.E., Murray, C.D.,

Christou, A., Guest, J., Casanova, I., D’Uston, L.C., Maurice, S.,

Foing, B., Kato, M., 2002. The D-CIXS X-ray spectrometer, and

its capabilities for lunar science. Adv. Space Res. 30, 1901–1907.

Hiesinger, H., Jaumann, R., Neukum, G., Head, J.W., 2000. J.

Geophys. Res. 105 (E12), 29,239–29,275.

Johnson, J.R., Larson, S.M., Singer, R.B., 1991. Remote sensing of

potential lunar resource: 1. Near-side compositional properties. J.

Geophys. Res. 96, 18,821–18,861.

Josset, J.-L., Heather, D.J., Dunkin, S.K., Roussel, F., Beauvivre, S.,

Kraenhenbueh, D., Plancke, P., Langevin, Y., Pinet, P., Chevrel,

S., Cerroni, P., De Sanctis, M.-C., Dillelis, A., Sodnik, Z.,

Koschny, D., Barucci, A., Hofmann, B., Josset, M., Muinonen,

K., Piironen, J., Ehrenfreud, P., Shkuratov, Yu., Shevchenko, V.,

2002. Asteroid Moon micro-Imager Experiment (AMIE) for

SMART-1 Mission, Science Objectives and Development Status.

EGS XXVII General Assembly, Nice, France, April 2002, abstract

#EGS02-A-06862.

Lawrence, D.J., Feldman, W.C., Elphic, R.C., Little, R.C., Prettyman,

T.H., Maurice, S., Lucey, P.G., Binder, A.B., 2002. Iron

abundances on the lunar surface as measured by the lunar

prospector gamma-ray and neutron spectrometers. J. Geophys.

Res. 107 (E12).

Lawrence, D.J., Elphic, R.C., Feldman, W.C., Prettyman, T.H.,

Gasnault, O., Maurice, S., 2003. Small-area thorium features on

the lunar surface. J. Geophys. Res. 108 (E9) 6-1–6-25.

Le Mouelic, S., Langevin, Y., Erard, S., Pinet, P., Daydou, Y.,

Chevrel, S., 2000. Discrimination between maturity and composi-

tion of lunar soils from integrated Clementine UVVIS-NIR data.

Application to Aristarchus Plateau. J. Geophys. Res. 105 (E4),

9445–9455.

Lucey, P., Taylor, G., Malaret, E., 1995. Abundance and distribution

of iron on the Moon. Science 268, 1150–1153.

Lucey, P., Blewett, D., Hawke, B., 1998. Mapping the FeO and TiO2

content of the lunar surface with multispectral imagery. J.

Geophys. Res. 103, 3679–3700.

Lucey, P.G., Blewett, D.T., Jolliff, B.L., 2000a. Lunar iron and

titanium abundance algorithms based on final processing of

Clementine ultraviolet-visible images. J. Geophys. Res. 105,

20,297–20,306.

Lucey, P.G., Blewett, D.T., Taylor, G.J., Hawke, B.R., 2000b.

Imaging of lunar surface maturity. J. Geophys. Res. 105,

20,377–20,386.

Morris, R., 1976. Surface exposure indices of lunar rocks: a

comparative FMR study. In: Proceedings of the Lunar Planetary

Science Seventh, LPI Houston, pp. 315–335.

Morris, R., 1977. Origin and evolution of the grain-size dependence of

the concentration of fine-grained metal in lunar soils: the

maturation of lunar soils to a steady-state stage. In: Proceedings

of the Lunar Science Conference Eighth, LPI Houston,

pp. 3719–3747.

Morris, R., 1978. The surface exposure (maturity) of lunar soils: some

concepts and Is/FeO compilation. In: Proceedings of the

Lunar Planetary Science Conference Nineth, LPI Houston,

pp. 2287–2297.

Morris, R., 1980. Origin and size distribution of metallic iron particles

in the lunar regolith. In: Proceedings of the Lunar Science

Conference 11th, LPI Houston, pp. 1697–1712.

Pieters, C., Taylor, L., 2003. Systematic global mixing and melting in

lunar soil evolution. Geophys. Res. Lett. 30 (20), 2048.

Pieters, C.M., Stankevich, D.G., Shkuratov, Yu.G., Taylor, L.A.,

2002. Statistical analysis of the links between lunar mare

soil mineralogy, chemistry and reflectance spectra. Icarus 155,

285–298.

Pinet, P., Shevchenko, V., Chevrel, S., Daydou, Y., Rosemberg, C.,

2000. Local and regional lunar regolith characteristics at Reiner

Gamma formation: optical and spectroscopic properties from

Clementine and Earth-based data. J. Geophys. Res. 105,

9457–9476.

Prettyman, T.H., Feldman, W.C., Lawrence, D.J., McKinney, G.W.,

Binder, A.B., Elphic, R.C., Gasnault, O.M., Maurice, S., Moore,

K.R., 2002. Library least squares analysis of Lunar Prospector

gamma-ray spectra. Lunar and Planetary Science Conference 33rd,

LPI Houston, Abstract #2012.

Shkuratov, Yu.G., Kaydash, V.G., Opanasenko, N.V., 1999a. Iron

and titanium abundance and maturity degree distribution on lunar

nearside. Icarus 137, 222–234.

Shkuratov, Yu., Starukhina, L., Hoffmann, H., Arnold, G., 1999b. A

model of spectral albedo of particulate surfaces: implication to

optical properties of the Moon. Icarus 137, 235–246.

Shkuratov, Yu., Pieters, C., Omelchenko, V., Stankevich, D.,

Kaydash, V., Taylor, L., 2003a. Estimates of the lunar surface

composition with Clementine images and LSCC data. Lunar

Planetary Science Conference 34th, LPI Houston, Abstract #1258.

Shkuratov, Yu., Stankevich, D., Kaydash, V., Omelchenko, V.,

Pieters, C., Pinet, C., Chevrel, S., Daydou, Y., Foing, B., Sodnik,

Z., Josset, J.-L., Taylor, L., Shevchenko, V., 2003b. Composition

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

using Clementine data. J. Geophys. Res. 108 (E4).


Shkuratov, Yu., Pinet, P., Omelchenko, V., Kaydash, V., Stankevich,

D., Chevrel, S., Daydou, Y., 2004. Derivation of elemental

abundance maps at 15-km spatial resolution from the

merging of Clementine optical and Lunar Prospector geo-

chemical data. Lunar Planetary Science 35th, LPI Houston,

Abstract #1162.

Starukhina, L.V., Shkuratov, Yu.G., 2001. A theoretical model of

lunar optical maturation: effects of submicroscopic reduced iron

and particle size variations. Icarus 152, 275–281.

Starukhina, L.V., Shkuratov, Yu.G., 2004a. Swirls on the Moon and

Mercury: meteoroid swarm encounters as a formation mechanism.

Icarus 167, 136–147.

Starukhina, L.V., Shkuratov, Yu.G., 2004b. Global mixing as a

mechanism for compositional anomalies of agglutinitic glasses.

Lunar Planetary Science Conference 35th, LPI Houston, Abstract

#1497.

Taylor, J., Warren, P., Ryder, G., Delano, J., Pieters, C., Lofgren, G.,

1991. Lunar rocks. In: Heiken, G., Vaniman, D., French, B. (Eds.),

Lunar Source Book. Cambridge University Press, Cambridge, pp.

183–284.

Taylor, L.A., Pieters, C.M., Morris, R.V., Keller, L.P., McKay, D.S.,

2001. Lunar mare soils: space weathering and the major effects of

surface-correlated nanophase Fe. J. Geophys. Res. 106,

27,985–28,000.

derivation of elemental abundance maps at intermediate resolution from optical interpolation of...

Documents