LETTERSPUBLISHED ONLINE: 4 JULY 2010 | DOI: 10.1038/NGEO897

Possible mantle origin of olivine around lunarimpact basins detected by SELENESatoru Yamamoto1*, Ryosuke Nakamura2, Tsuneo Matsunaga1, Yoshiko Ogawa3, Yoshiaki Ishihara4,Tomokatsu Morota5, Naru Hirata3, Makiko Ohtake5, Takahiro Hiroi6, Yasuhiro Yokota1

and Junichi Haruyama5

The composition, structure and evolution of the Moon’smantle is poorly constrained. The mineral olivine, one of themain constituents of Earth’s mantle, has been identified byEarth-based telescopic observations at two craters on thenear side of the Moon, Aristarchus and Copernicus1–3. Globalreflectance spectra in five discrete spectral bands producedby the spacecraft Clementine4–6 suggested several possibleolivine-bearing sites, but one of the candidate occurrencesof olivine was later re-classified, on the basis of continuousreflectance spectra over the entire 1µm band, as a mixture ofplagioclase and pyroxene7. Here we present a global surveyof the lunar surface using the Spectral Profiler onboard thelunar explorer SELENE/Kaguya7,8. We found many exposuresof olivine on the Moon, located in concentric regions around theSouth Pole-Aitken, Imbrium and Moscoviense impact basinswhere the crust is relatively thin. We propose that theseexposures of olivine can be attributed either to an excavationof the lunar mantle at the time of the impacts that formed thebasins3, or to magnesium-rich pluton in the Moon’s lower crust.On the basis of radiative transfer modelling4,8–10, we suggestthat at least some of the olivine detected near impact basinsoriginates from upper mantle of the Moon.

The lunar magma ocean (LMO) scenario proposes fractionalcrystallization of LMO-produced mafic cumulates that made upthe mantle, and plagioclase floatation that made up the crust11,12.Several models have been proposed that describe the compositionaland structural evolution of a crystallizing magma ocean, but thereare still uncertainties in the composition and structure. One of thereasons for the uncertainties is the lack of information on olivineexposure on the Moon, a plausible main material for the lunarmantle. Earth-based telescopic observations have reported only twonearside craters, Copernicus and Aristarchus, having olivine-richspectral features1–3. Although Earth-based observations producecontinuous reflectance spectra, the observational points are sparseand limited to the lunar nearside. On the other hand, the UVVIScamera onboard the Clementine spacecraft (hereafter Clementine),which had five discrete bands, provided global data of the Moon4–6.Olivine Hill in the South Pole-Aitken (SPA) basin and the centralpeaks of five craters were identified as possible olivine-bearingsites by Clementine5,6. However, after a re-examination using datataken by the Spectral Profiler (SP) onboard the Japanese explorerKaguya, one of the Clementine candidates, the Tsiolkovsky crater,

1Center for Global Environmental Research, National Institute for Environmental Studies, 16-2 Onogawa, Tsukuba, Ibaraki 305-8506, Japan, 2InformationTechnology Research Institute, National Institute of Advanced Industrial Science and Technology, 1-1-1 Umezato, Tsukuba, Ibaraki 305-8568, Japan,3ARC-Space/CAIST, The University of Aizu, Ikki-machi, Aizuwakamatsu, Fukushima 965-8580, Japan, 4RISE project, National Astronomical Observatoryof Japan, 2-12 Hoshigaoka, Mizusawa, Oshu, Iwate 023-0861, Japan, 5Institute of Space and Astronautical Science, Japan Aerospace Exploration Agency,3-1-1 Yoshinodai, Chuo-ku, Sagamihara, Kanagawa 252-5210, Japan, 6Department of Geological Sciences, Brown University, Providence, Rhode Island02912, USA. *e-mail: yamamoto.satoru@nies.go.jp.

Latit

ude

(°)

Longitude (°)

¬90

90

60

30

0

¬30

¬60

(km)

1101009080706050403020100

AB

C

D

E

F

G

H

I

J

M

C1

C6

¬180¬150 ¬120 ¬90 ¬60 ¬30 0 30 60 90 120 150 180

Figure 1 |Global distribution of olivine-rich points on the Moon. Thebackground map is the total lunar crustal thickness (crustal materials andmare basalt fills) based on SELENE gravity and a topographic modelproduced by the Kaguya explorer13,28–30. The red squares indicateolivine-rich points with multiple SP data points showing a clear olivinespectral signature. The small red crosses indicate single SP detections.Note that most of the olivine-rich points are distributed around impactbasins. The SP successfully detected olivine at the Copernicus (C1) andAristarchus (C6) craters, which were identified as olivine-bearing areas byEarth-based observation1–3.

was classified as a mixture of plagioclase and pyroxene, rather thanas pure olivine7. This SP finding demonstrated the importanceof obtaining continuous reflectance spectra over the visible andnear-infrared range covering the entire 1 µm band, which canbe used to as a diagnostic tool for olivine and other silicates inidentifying olivine exposure sites on theMoon.

The SP has obtained continuous spectral reflectance data forabout seventy million points (a 0.2–0.5 km by 0.5 km footprint)on the Moon over the 0.5–2.6 µm wavelength range (λ) witha spectral resolution of 6–8 nm during its mission period fromNovember 2007 to June 2009 (refs 7,8). Analysing all of the spectraldata, we identified 245 olivine-rich points by picking up spectrahaving absorption band minima within the wavelength range ofλ= 1.05± 0.03 µm after removing a linear tangential continuum.Most of the spectra for the selected points (hereafter referred to asolivine-rich points) show clear olivine bands with λ=0.85, 1.05 and1.25 µm as shown in Supplementary Figs S1–S3, although some ofthe spectra show less clear olivine bands, which may be due to thepresence of minor amounts of high-Ca pyroxene or other geologicunits in the SP field of view.

NATURE GEOSCIENCE | VOL 3 | AUGUST 2010 | www.nature.com/naturegeoscience 533© 2010 Macmillan Publishers Limited. All rights reserved.

LETTERS NATURE GEOSCIENCE DOI: 10.1038/NGEO897

Moscoviense

A1

100 km

A2

Crisium

B2

100 km

Imbrium

C2

200 kmC1

C3

C6

C4 and C5

Schrödinger

E1

Zeeman

SPA

200 km

Humorum

D1

100 km

Nectaris

Theophilus

100 km

Frigoris F1

100 km

Serenitatis

200 km

E1

ComptonHumboldtianum

100 km

Rc

R c Rc Rc

Rc

R c

Wavelength (µm)

Wavelength (µm) Wavelength (µm) Wavelength (µm)

Wavelength (µm) Wavelength (µm)

1.61.41.21.00.80.61.61.41.21.00.80.6

1.61.41.21.00.80.6

1.021.000.980.960.940.920.900.88

1.61.41.21.00.80.6 1.61.41.21.00.80.6

1.021.000.980.960.940.920.900.880.86

1.00

0.95

0.90

0.85

0.80

1.00

0.95

0.90

0.85

1.61.41.21.00.80.6

A1 C1C1A1 B2B2

E1D1 F1D1 F1

a

d

g h i

e f

b c

1.021.000.980.960.940.920.900.880.86

9

0

¬9

3

¬3

6

¬6

(km)

1.021.000.980.960.940.920.90

Figure 2 | Local distribution of olivine-rich sites around various basins or maria. The background map is the surface topography obtained by the Kaguyamission28. The larger rectangles and smaller circles indicate olivine-rich points with multiple and single SP detections, respectively. Photos A1, B2, C1, D1,E1 and F1 are close-up images of olivine-rich sites taken by the MI or the TC onboard Kaguya14,15. On the close-up images, olivine-rich points are plotted asred rectangles with white 5 km scale bars. The accompanying plots show the continuum-removed reflectance spectra Rc at the location of the yellowsquare marked on each close-up image.

In Fig. 1 we plot 245 olivine-rich points on a lunar crustalthicknessmap obtained byKaguya13.Most of the olivine-rich pointsare grouped into several local sites. For example, SP detected 59 inthe Copernicus crater and 4 in the Aristarchus crater. Taking intoaccount the local geologic context, based on images obtained byKaguya’s Multiband Imager (MI) or Terrain Camera (TC) duringthe SP observation14,15, we found that most of the localities havingmultiple olivine-rich points are associated with small fresh, craters(Fig. 2). Therefore, we divided and assigned the 245 olivine-richpoints to 34 olivine-rich sites (Supplementary Table S1). Therepresentative spectra for the individual olivine-rich sites are shownin Supplementary Figs S1–S3.

Figure 1 shows that most of the olivine-rich sites are locatedaround impact basins: that is, (A) Mare Moscoviense, (B) Crisium,(C) Imbrium, (D) Humorum, the SPA basin ((E) Schrödingerand (M) Zeeman craters), (G) Nectaris, (H) Serenitatis, (I)Humboldtianum and (J) Australe. These basins are located on

thinner crusts with a thickness of about 30–50 km. Most of theolivine-rich sites are concentrated on the lunar nearside. Whereasthere is no olivine-rich site in the Feldspathic Highlands Terrane16,olivine-rich sites are found in the SPA and Moscoviense, on the farside, in locations where the crust is thin.

Furthermore, in the vicinity of each basin, olivine-rich sites aredistributed along the concentric region of the basin. For example,Fig. 2a shows that olivine-rich sites are distributed along the rimof Moscoviense, whereas there is no olivine-rich site in the centralregion of the mare or regions far from the outer ring. Anotherconspicuous example is Crisium, where the olivine-rich sites arelimited to a narrow concentric region around the mare (Fig. 2b).Around Imbrium (Fig. 2c) we found olivine-rich sites in theCopernicus (C1), Eratosthenes (C3), Aristarchus (C6), the MontesAlpes (C2) and the terrace in the Sinus Iridum (C4 and C5). Theirlocations seem to correspond to the prominent rings of Imbrium17.In the SPA (Fig. 2e), there are two craters with olivine-rich sites

534 NATURE GEOSCIENCE | VOL 3 | AUGUST 2010 | www.nature.com/naturegeoscience

© 2010 Macmillan Publishers Limited. All rights reserved.

NATURE GEOSCIENCE DOI: 10.1038/NGEO897 LETTERS

D

E

F

E1

1.000.950.900.850.80 D E F

A

B

C

D1 1.00

0.96

0.92

0.88

1.41.21.00.8

1.41.21.00.8

A B C

REFF

1.41.21.00.80.6

A B C

REFF

D E F

0.300.250.200.150.100.05

0.350.300.250.200.150.10

1.41.21.00.80.6

R cR c

Wavelength (µm) Wavelength (µm)

Wavelength (µm) Wavelength (µm)

Figure 3 | Colour-composite image maps of olivine-rich sites D1 and E1 taken by the MI. (See ref. 14 for more detail on the MI images.) Blue, green and redare assigned to reflectances of 900, 1,050 and 1,250 nm, respectively. The continuum-removed reflectance spectra Rc and reflectance factor9 (REFF) at thesix locations marked A–F in the images are also plotted. The locations of A and E show olivine-rich spectra. All reflectance spectra are given as the averageof a 500 m×500 m area to remove spatial variation. Saturated data areas are masked (black filled areas).

(Schrödinger and Zeeman). They are located near the edge of theSPA, whereas there is no olivine-rich site in its central region.Although the number of olivine-rich sites was limited, the samedistribution patternwas observed at other basins (Fig. 2d–i).

At each olivine-rich site, most of the olivine exposure wasdetected at several consecutive SP footprints. This indicates that theolivine-rich exposures extend over several footprint sizes spanningseveral kilometres. They are found on crater walls (for example,B2, D1 and E1) and on continuous ejecta (for example, A1 andF1). Figure 3 shows the MI images for sites D1 and E1, where theolivine-rich spectra appear in the landslide features on the craterwall. At site E1, there is also an area that has a clear plagioclasespectrum showing a strong 1.25 µm band on the crater wall(marked ‘F’). On the other hand, spectral features for areas outsideolivine (or plagioclase) exposures are too unclear to allow correctinterpretation of their mineral compositions. This is because mostof the lunar surface is covered with mixtures of various minerals.Space weathering also obscures spectral features. The olivine-richexposures, however, are found in fresh areas such as landslidefeatures on craterwalls or recently formed craters (for example, F1).

Figure 1 does not includes the Olivine Hill, Langrenus, Keeler,Crookes and Tsiolkovsky craters, which were suggested as olivine-bearing areas by Clementine5,6. This is because the Clementineanalysis was based on discrete spectral data with a limitedwavelength coverage of λ≤ 1 µm, whereas the SP has continuousspectral data with λ=0.5–1.6 µm (ref. 7). However, the Theophilussuggested by Clementine is identified as an olivine-rich site by theSP. Figure 2g shows that this crater is located in the concentricregion around Nectaris.

In summary, olivine exposures on the Moon are limited to con-centric regions around the impact basins that have thinner crusts.On a local scale, they are found mainly on small, fresh crater wallsor continuous ejecta. What mechanism produced this distribution?We propose that basin formation is responsible for the observeddistribution of the olivine exposures. Each basin formation couldhave blasted away the upper crust, excavating and redistributingdeep-seated olivine-rich material to the rim. Whereas the centralregion of the basin would be covered with basaltic lava that eruptedlater in the cases of nearside basins andMoscoviense, the rim regionwould not. For the SPA the impact resulted in the production of alarge amount of melted material, which puddled on the floor ofthe excavated cavity as a melt sheet. Local differentiation occurredin these melt layers, forming an orthopyroxene layer that overliesthe olivine-rich layer18. Indeed, a recent SP survey8 revealed theexistence of an extensive layer of differentiated orthopyroxene in thecentral region of the SPA; the central peaks of the Finsen, Antoniadi,Bhabha and Lyman craters show clear orthopyroxene spectra. Thus,

the deep-seated olivine-rich layers in the central region of theSPA would be hidden by the differentiated impact melt. Althougholivine in the rim regions would have been covered with ejecta fromthe surroundings, later impacts could have excavated the olivine,exposing it to the surface. As a result, olivine-rich sites are observedonly at fresh craters in the concentric regions around large basins.

Although most main thin-crust basins (for example,Moscoviense, Crisium, Humboldtianum) have olivine-rich sites,some basins (for example, Mare Smythii) do not. This may be dueto the incomplete coverage of our survey. Some basins located inthin-crust regions may have olivine exposures in their concentricregions that the SP survey did not discover.

Where did the olivine-rich material originate? This is an impor-tant question for increasing our understanding of the structure andevolution of theMoon.Herewe propose twopossible scenarios. Thefirst scenario is that the olivine-rich exposures originated in the up-per lunar mantle. The basins with olivine-rich sites are located onlyin regions where the crust is relatively thin (Fig. 1). For example, ifthe general impact cratering theory19 is applied to themare Crisium(∼1,000 km diameter), the depth of the excavation is >∼ 100 km.The original crust thickness at Crisium could have been thinnerthan themaximum thickness of the current feldspathic crust, whichis about 100 km (Fig. 1). Thus, basin formation impacts couldplausibly have penetrated to the crust–mantle boundary.

The second scenario is that the olivine-rich exposure originatesfrom the mafic-rich lower crust. In other words, the basin forma-tions excavated the Mg-rich pluton intruding into the lunar lowercrust20–22. Note that some of the olivine-rich sites are associatedwithplagioclase; the Schrödinger and Aristarchus craters were reportedas the purest-anorthosite-bearing regions14. In addition, site E1 inthe Schrödinger crater (Fig. 3) has areas that exhibit the 1.25 µmplagioclase absorption band adjacent to areas showing olivine-richspectra. Thismay suggest that the basin formations excavated intru-sions with spatially inhomogeneous plagioclase/olivine ratios in thelower crust, although there is the possibility that the excavation ofthe crust–mantle boundary resulted in mixtures of mantle olivineand anorthosite during the excavation process. If this scenario istrue, the spatial distribution of olivine exposures (Fig. 1) gives im-portant insights into constraints on early lunar basaltic magmatismand crustal growth after the crystallization of the LMO(refs 23,24).

Which of the above scenarios is more plausible? If the olivine-rich exposure originates from the upper mantle, the compositionshould be similar to dunite rather than troctolite in the lowerlunar crust. To confirm whether this is the case, we examinedthe spectral data for some of the olivine-rich sites using radiativetransfer modelling based on an intimate mixture model4,8–10 (seeSupplementary Information). Supplementary Fig. S4 shows that the

NATURE GEOSCIENCE | VOL 3 | AUGUST 2010 | www.nature.com/naturegeoscience 535© 2010 Macmillan Publishers Limited. All rights reserved.

LETTERS NATURE GEOSCIENCE DOI: 10.1038/NGEO897

representative spectra are more consistent with a dunite-dominantmodel than with troctolite. This is mainly because most of thespectral data for the olivine-rich sites in Supplementary Fig. S4 havelower absolute reflectance (<0.23 at λ∼ 0.7 µm). In other words,it is difficult to fit the dark spectra of the olivine-rich exposureswith bright components such as plagioclase (see SupplementaryInformation). Therefore, radiative transfer modelling supports theconcept that materials in the olivine-rich sites originated in theupper mantle. We cannot, however, rule out the possibility thatplagioclase is present in olivine exposures, but its diagnostic bandwas completely erased by shock25. Although furtherwork is requiredto resolve the origin issue, we believe that at least some of the largerbasin formations could have excavated the lunar upper mantle toproduce the basin-related olivine. In this case, the distribution ofolivine-rich sites gives us valuable new constraints for the Moon’sevolution. In the mantle cumulate crystallization sequence11,12, thelowermost layer would be Mg-rich olivine and the residual KREEP(potassium, rare-earth elements and phosphorus) layer wouldunderlie the anorthositic lunar crust. Therefore, basin formationwould have excavated not only the lunar mantle but also theKREEP layer as well. However, although neither Moscoviense norCrisium, which have the thinnest crust on the Moon, shows a highthorium concentration16, their formations excavated the Mg-richolivine layer. A large-scale overturn of the mantle cumulate26may have occurred before the two basins were formed. Althoughfurther work is required to resolve this issue, the present dataprovide important modelling constraints for the evolution of thelunar crust and mantle.

MethodData in the SP Level 2A (L2A) product processed by the SELENE Operationand Analysis Center at the Institute of Space and Astronautical Science wereradiometrically calibrated. (Dark current removal, non-linearity correction andminor wavelength correction are included in the calibration procedures.) The L2Adata were then further corrected using the reflectance spectra of the Apollo 16landing site and the spectral reflectance of the Apollo 16 soil 62231 as measured inthe laboratory27. We also corrected small gaps between data measured by differentSP detectors resulting from differences in exposure time and sampling timing. Asthe radiometric calibration for wavelengths longer than 1.6 µmremains incomplete,we use only data at wavelengths shorter than 1.6 µm.

The SP obtained continuous spectral reflectance data for more than 68millionpoints during its mission period from November 2007 to June 2009. (Thetotal SP coverage area is about 20% of the lunar surface at the equator.) Wediscovered olivine-rich points using the following procedure. First, we rejected thefollowing data with low signal-to-noise ratios: (1) data in which the radiance atwavelength λ= 0.5126 µm is less than 23.3Wm−2 µm−1 sr−1 and (2) data in whichcontinuum-removed reflectance Rc (divided by a linear tangential continuum)does not show an absorption band minimum with Rc< 0.95 within the wavelengthrange from 0.7 to 1.6 µm. This is because such data cannot be used to examinewhether the characteristic olivine bands exist at λ=0.85, 1.05 and 1.25 µm. Second,we searched for the wavelengths at which Rc with λ= 0.7 and 1.6 µm are thefirst-, second- and third-lowest values, and selected the data in which these threewavelengths were within the range of 1.05±0.03 µm. Third, we rejected the jaggedspectral data in which the difference in Rc between λ= 1.04 and 1.07 µmwas largerthan 0.02. After applying this algorithm, only 266 candidate spectra remainedout of more than 68million original spectra. Then, to remove spectra that had aclear plagioclase absorption band, we also rejected those for which the differencebetween the mean Rc over the wavelength range from 1.04 to 1.07 µm and the meanRc over the wavelength range from 1.17 to 1.20 µm was less than 0.01. This left 245observational points, which we designated olivine-rich points.

Received 24 December 2009; accepted 24 May 2010;published online 4 July 2010

References1. Pieters, C. M. Conspicuous crater central peak: Lunar mountain of unique

composition. Science 215, 59–61 (1982).2. Lucey, P. G., Hawke, B. R., McCord, T. B., Pieters, C. M. & Head, J. W. A

compositional study of the Aristarchus region of the Moon using near-infraredreflectance spectroscopy. J. Geophys. Res. 91, D344–D354 (1986).

3. Pinet, P. C., Chevrel, S. D. & Martin, P. Copernicus: A regional probe of thelunar interior. Science 260, 797–801 (1993).

4. Lucey, P. G. Mineral maps of the Moon. Geophys. Res. Lett. 31, L08701 (2004).5. Tompkins, S. & Pieters, C. M. Mineralogy of the lunar crust: Results from

Clementine.Meteorit. Planet. Sci. 34, 25–41 (1999).6. Pieters, C. M., Head, J. W. III, Gaddis, L., Jolliff, B. & Duke, M. Rock types of

South Pole-Aitken basin and extent of basaltic volcanism. J. Geophys. Res. 106,28001–28022 (2001).

7. Matsunaga, T. et al. Discoveries on the lithology of lunar crater central peaksby SELENE Spectral Profiler. Geophys. Res. Lett. 35, L23201 (2008).

8. Nakamura, R. et al. Ultramafic impactmelt sheet beneath the South Pole-Aitkenbasin on the Moon. Geophys. Res. Lett. 36, L22202 (2009).

9. Hapke, B. Space weathering from Mercury to the asteroid belt. J. Geophys. Res.E106, 10039–10073 (2001).

10. Cahill, J. T. S., Lucey, P. G. &Wieczorek, M. A. Compositional variations of thelunar crust: Results from radiative transfer modelling of central peak spectra.J. Geophys. Res. E114, E09001 (2009).

11. Snyder, G. A., Lawrence, T. A. & Clive, N. R. A chemical model forgenerating the sources of mare basalts: Combined equilibrium and fractionalcrystallization of the lunar magmasphere. Geochim. Cosmochim. Acta 56,3809–3823 (1992).

12. Wieczorek, M. A. et al. in New Views of the Moon vol. 60 (eds Jolliff, B. L. et al.)Ch. 3, 221–364 (The Mineralogical Society of America, 2006).

13. Ishihara, Y. et al. Crustal thickness of the Moon: Implications for farside basinstructures. Geophys. Res. Lett. 36, L19202 (2009).

14. Ohtake, M. et al. The global distribution of pure anorthosite on the Moon.Nature 461, 236–240 (2009).

15. Haruyama, J. et al. Global lunar-surface mapping experiment by the lunarimager/spectrometer on SELENE. Earth Planet. Space 60, 243–256 (2008).

16. Jolliff, B. L., Gillis, J. J, Haskin, L. A., Korotev, R. D. & Wieczorek, M. A.Major lunar crustal terranes: Surface expressions and crust–mantle origins.J. Geophys. Res. E105, 4197–4216 (2000).

17. Melosh, H. J. Impact Cratering (Oxford University Press, 1989).18. Grieve, R. A. F., Stöffler, D. & Deutsch, A. The Sudbury

structure—controversial or misunderstood? J. Geophys. Res. 96,22753–22764 (1991).

19. Croft, S. K. Cratering flow fields: Implications for the excavation and transientexpansion stages of crater formation. Proc. Lunar Planet. Sci. Conf. 11,2347–2378 (1980).

20. Pieters, C. M.&Wilhelms, D. E. Origin of olivine at Copernicus. J. Geophys. Res.90, C415–C420 (1985).

21. Pieters, C. M. & Tompkins, S. Tsiolkovsky crater: A window into crustalprocesses on the lunar farside. J. Geophys. Res. E104, 21935–21949 (1999).

22. Hawke, B. R. et al. Distribution and modes of occurrence of lunar anorthosite.J. Geophys. Res. 108, 5050 (2003).

23. Shearer, C. K. & Papike, J. J. Early crustal building processes on the moon:Models for the petrogenesis of the magnesian suite. Geochem. Cosmochim. Acta69, 3445–3461 (2005).

24. Longhi, J. A new view of lunar ferroan anorthosites: Postmagma oceanpetrogenesis. J. Geophys. Res. 108, 5083 (2003).

25. Bruckenthal, E. A. & Pieters, C. M. Spectral effects of natural shock onplagioclase feldspar. Abstr. Lunar Planet. Sci. Conf. 15, 96–97 (1984).

26. Hess, P. C. & Parmentier, E. M. A model for the thermal and chemicalevolution of the Moon’s interior: Implications for the onset of mare volcanism.Earth Planet. Sci. Lett. 134, 501–514 (1995).

27. Pieters, C. M. The moon as a spectral calibration standard enabled by lunarsamples: The Clementine example. New Views Moon 2, 8025–8026 (1999).

28. Araki, H. et al. Lunar global shape and polar topography derived fromKaguya-LALT laser altimetry. Science 323, 897–900 (2009).

29. Namiki, N. et al. Farside gravity field of the Moon from four-way Dopplermeasurements of SELENE (Kaguya). Science 323, 900–905 (2009).

30. Wessel, P. & Smith, W. H. F. Free software helps map and display data.EOS Trans. AGU 72, 441 (1991).

AcknowledgementsThis research was partly supported by the Grant-in-Aid for Young Scientists (B) fromJapan Society for the Promotion of Science (20740249).

Author contributionsData analyses were conducted by S.Y., R.N., T. Matsunaga, Y.O. and M.O. Themanuscript was produced by significant contributions from S.Y., R.N. and T. Matsunaga.T.H. contributed to the assessments of spectral features in the survey programme. Y.I.contributed to the production of the base maps of Figs 1 and 2 and discussion on the crustthickness. All of the authors, including T. Morota, N.H., J.H. and Y.Y., discussed andprovided significant comments on the results and themanuscript.

Additional informationThe authors declare no competing financial interests. Supplementary informationaccompanies this paper on www.nature.com/naturegeoscience. Reprints and permissionsinformation is available online at http://npg.nature.com/reprintsandpermissions.Correspondence and requests formaterials should be addressed to S.Y.

536 NATURE GEOSCIENCE | VOL 3 | AUGUST 2010 | www.nature.com/naturegeoscience

© 2010 Macmillan Publishers Limited. All rights reserved.