Comparison of asteroid and meteorite spectra - Classification by principal component analysis

ICARUS 99, 153--166 (1992)

Comparison of Asteroid and Meteorite Spectra: Classification by Principal Component Analysis

DANIEL T. BRITT

Lunar and Planetary Laboratory, University of Arizona, Tucson, Arizona 85721

DAVID J. THOLEN

Institute for Astronomy, University of Hawaii at Manoa, 2680 Woodlawn Drive, Honolulu, Hawaii 96822

JEFFREY F . BELL

Planetary Geosciences Division, Department of Geology and Geophysics, School of Ocean and Earth Sciences and Technology, University of Hawaii at Manoa, 2525 Correa Road, Honolulu, Hawaii 96822

AND

CARLE M. PIETERS

Department of Geological Sciences, Box 1846, Brown University, Providence, Rhode Island 02912

Received December 18, 1990; revised April 29, 1992

The relationships between asteroid and meteorite spectral types are a subject of much discussion and debate. To provide an overview of spectral similarities between and among asteroids and meteorites, the spectra of 103 meteorites and 411 asteroids were statistically compared using principal component analysis. This analysis produced a "map" of statistically defined spectral similarities in which distance is a measure of relative similarity. There appears to be considerably more spectral variance in the meteorite collection than is seen in the asteroids. Also, many planet-crossing asteroids are spectrally much more similar to the meteorites than most main belt asteroids. The principal components of meteorite spectra are, in general, offset from those of the bulk of the asteroid population. The offset of the carbonaceous chondrite and iron meteorites from their most frequently suggested asteroidal analogues is particularly notable. Finally, the analysis shows no direct meteorite analogues from the outerbelt B-, D-, F-, P-, and T- type asteroids, confirming that meteorites sample only a relatively limited portion of the inner asteroid belt. © 1992 Academic Press. Inc.

INTRODUCTION

The study of minor planets poses a vareity of fascinating questions and challenges for planetary science. Their abundance, mineralogical variety, and potential as geo- chemical windows onto the early Solar System make them

a treasury of data. However, their numbers and relative inaccessibility rule out close spacecraft reconnaissance of all but a handful of these objects for at least the next several decades. Even with spacecraft flybys of a few objects, remote sensing from groundbased photometry and spectrophotometry is likely to remain the major source of information on the composition, distribution, and evolution of most asteroids. The remote study of any object is greatly aided by the provision of "ground truth" samples that provide laboratory calibration of mineralogically diagnostic spectral features. The study of asteroids, despite the disadvantages of distance, inaccessibility, and great numbers, has one major advantage over the studies of most other Solar System objects: we have samples of asteroidal material in the meteorites. Meteorites have been linked by dynamical (e.g., Wetherill 1985, Wisdom 1987, Greenberg and Nolan 1989) and spectral (e.g., Mc- Cord and Gaffey 1974, Chapman 1976, Bell et al. 1989a, Pieters et al. 1976, Pieters 1984, McFadden et al. 1984, Wetherill and Chapman 1988, Cruikshank et al. 1991) arguments to asteroidal parent bodies. Even without knowing which parent bodies individual meteorites come from, the mineralogy and petrology of meteorites provide a vital calibration on the range and mineralogical sources of asteroidal spectral variation.

However, meteorites are imperfect sources of ground

153 0019-1035/92 $5.00 Copyright © 1992 by Academic Press, Inc.

All rights of reproduction in any form reserved.

154 BRITT ET AL.

truth for spectral studies. First, they lack provenance since it is impossible to link any meteorite with absolute certainty to any individual asteroid. What can be done is to use meteorites as guides to the general spectral types found in the asteroid belt. But that raises a second point: not all asteroids seems to have spectral analogues in the meteorite collection. Orbital dynamical, compositional, and Solar System accretion arguments indicate that the meteorite collection represents samples of a relatively limited zone in the asteroid belt (Wetherill 1985, Chapman 1976, Bell et al. 1989a, Gradie and Tedesco 1982). Another important limitation is that remote sensing, using either visible, infrared, or thermal wavelengths, senses only the upper few micrometers to centimeters of an object's surface. That surface is exposed to the space environment and can be altered optically and mineralogically by a variety of regolith processes. Meteorites, however, can and usually do come from any level within a parent body. Finally, meteorites may be altered from their pristine in- space condition by terrestrial weathering. Even with their limitations, meteorites remain the best sources of ground truth for the remote study of asteroidal spectral variation.

The classification of asteroids on the basis of their spectral and photometric properties has been an active and evolving aspect of asteroid science during the past two decades. Because of the length and complexity of the several classification schemes, this work discusses only the Tholen (1984) classification system with the addition of the K-class (Bell 1988, Tedesco et al. 1989). The advantages of using the Tholen (1984) system are that it is widely used and broadly similar to several other recent classifications (Barucci et al. 1987, Tedesco et al. 1989, Chapman 1990). For an excellent overview of this and other recent classification systems see the chapter by Tholen and Barucci (1989) in Astero ids II. Examples of asteroidal spectra for all the spectral types are reviewed in Tholen and Barucci (1989) and Tedesco et al. (1989). The common feature of all asteroidal classifications is that they are optical classifications; similarities between objects are determined by the similarities of their spectral or photometric data. Mineralogical similarities are, of course, only inferred.

Several studies have interpreted asteroid and meteorite spectral data to produce a mineralogy of the asteroid belt (e.g., Gaffey and McCord 1978, Bell et al. 1989a, Chap- man 1976). Other workers have used statistical techniques to classify asteroidal spectral data (Tholen 1984, Barucci et al. 1987, Tedesco et al. 1989, Chapman 1990). The object of this paper is to combine these approaches to provide an "overview" of asteroid and meteorite spectral variation. Using the statistical technique of principal components analysis, the spectra of a representative sample of asteroids and meteorites are analyzed, and statistical similarities and dissimilarities are displayed in statistical

space. Because statistical analysis requires that all data be in the same wavelength format, both the asteroid and the meteorite data are presented in the Eight Color Aster- oid Survey (ECAS) format, using a wavelength range from 0.33 to 1.1/zm.

DATA ANALYSIS

Cavea ts and L imi ta t ions on Spec t ra l C o m p a r i s o n s

To make statistically meaningful comparisons between asteroid and meteorite spectral data the respective data sets must meet some basic criteria. They must be (I) large enough to be statistically meaningful, (2) representative of the spectral variation within and between groups, (3) consistent in their data collection techniques throughout the data set, and, of course, (4) cover the same spectral range. These requirements significantly constrain the possible choices. The ECAS data set is the logial choice to represent asteroid spectral variation because it includes by far the largest number of objects (over 580) and it was undertaken expressly to determine the spectral diversity of the asteroid belt (Zellner et al. 1985). For meteorite data the most consistent and comprehensive data set is the Gaffey (1976) study of meteorite spectra.

However, there are several problems in comparing these data sets. First, the methods used to acquire the asteroid and meteorite spectral data sets are photometri- cally different. Asteroid spectra acquired with telescopic remote sensing are, by their very nature, bidirectional. Simply put, bidirectional reflectance is the measured reflectance of a disk-integrated asteroidal object viewed at a specific direction and illuminated from another direction (Hapke 1981). These data are taken at a variety of solar phase angles and under a variety of conditions. A few of the asteroid spectra were affected by the phenomenon of "phase reddening" that occurs at high solar phase angles and tends to increase the red continuum slope of the spectrum. In general the asteroid spectra are not cor- rected for phase effects. The data were, however, cor- rected for atmospheric transmission using solar-type stan- dard stars. The Gaffey (1976) meteorite spectra were acquired with an integrating sphere spectrometer (Beckman DK2A ratio recording spectroreflectometer) and the spectra are directional-hemispherical reflectance. This method compares the ratio of all the radiant power emitted from a surface to the total incident irradiance (Hapke 1981). Comparisons using spectra taken with these two techniques must be treated with caution. However, the level of comparability between bidirectional and directional-hemispherical reflectance was tested using spectra of the same meteorites taken with both methods. Results of this test indicate that the two techniques yield roughly comparable results at the level of precision of this study.

ASTEROID AND METEORITE SPECTRA 155

Data from these tests are discussed in detail in later sections.

Second, laboratory spectral data of meteorites were taken at spectral resolution and signal-to-noise ratios much higher than those possible to achieve with telescopi- cally obtained asteroid spectra. Before comparisons could be made the laboratory meteorite spectra had to be con- verted to the same low-resolution ECAS spectral system as the asteroids. Digital versions of ECAS filter transmission and detector sensitivity curves were used to resample the meteorite spectra into the ECAS system. A problem with resampling is that the UV-wavelength s-filter of the ECAS system is centered at 0.337 /zm and the Gaffey meteorite spectra all start at the longer wavelength of 0.35 /.~m. The result is that the s-filter is undersampled for the meteorites and is potentially meaningless. This is probably not a serious problem since Tholen (1984) noted a high degree of covariance (0.982) between the s- and u-filters, indicating that most of the information sensed by the s- filter is replicated by the u-filter. For this analysis, the s- filter is not used because of the spectral limitations of the meteorite data and the analysis is confined to what is effectively seven-color data.

The third caveat concerns the nature of meteorite samples and asteroidal surfaces. Laboratory spectra of meteorite samples are typically taken from crushed powders or freshly broken surfaces. In the Gaffey (1976) study, meteorites were, with a few exceptions, crushed to particle sizes in the 20- to 300-/zm range. Spectra were usually taken of bulk powder samples, but in some cases particle size separates were taken using dry sieving into several size bins. It is known that the particle size of a sample can affect several spectral parameters, including the slope of the spectra, the spectral contrast between absorption features, and the depth of absorption bands (Adams and Filice 1967). To be directly comparable to asteroid spectra, laboratory meteorite spectra must simulate, as much as possible, the conditions and particle size distributions found on asteroidal surfaces. However, asteroidal surface conditions are poorly understood and what we do know comes from a mixture of theoretical studies, remote sensing observations, and meteorite petrology. Theoretical studies suggest that asteroids greater than 10 km in diame- ter will be able to retain some regolith varying from meters thick in small objects to perhaps kilometers thick in the largest asteroids (McKay et al. 1989). Optical photopolarimetry observations indicate that asteroids are covered by a layer of loose regolith that tends to be coarser than lunar regoliths. Examples of particle size ranges are 50-200 ~m for S-type asteroids (Dollfus et al. 1989) and >50/xm with a partial coating of small (< 10/~m) particles for Vesta (Le Betre and Zellner 1980). Asteroids are probably depleted in the smallest size fractions relative to lunar regolith soil because of their low escape velocity (McKay

et al. 1989). Petrological studies of meteorites containing material that was probably on the surface of asteroids show asteroidal regoliths to be less mature relative to lunar regoliths (McKay et al. 1989). However, the same meteorites also often show alteration of their spectral characteristics, probably due to shock blacking while exposed on the surface (Bell and Keil 1988, Britt and Pieters 1990).

Regolith formation processes probably have some effect on the spectral properties of asteroidal surface material, either in changing particle sizes or in spectral alteration. The crushing of meteorites under laboratory conditions is probably not a good approximation of these regolith-forming processes, particularly for metal-bearing meteorites. However, with our level of understanding of these processes, crushing is still the best available alterna- tive for examination of particulate materials. In subsequent sections the effect of particle size variations on meteorite/asteroid spectral similarities is examined.

Asteroid and Meteori te Spectral Data

The data selected for this study are restricted to subsets of the ECAS asteroid survey (Zellner et al. 1985) and the Gaffey (1976) study of meteorite spectral reflectance. In the asteroid data set, 411 objects with single-letter classifications are used to sharply define the statistical space occupied by each asteroid type. There are a number of additional spectra of asteroids with multiple-letter classifications available that were not included because the object of this study is to compare meteorite spectral characteristics with the best classified and characterized asteroids. The inclusion of these "transitional" asteroids would tend to blur the distinctions between well-defined asteroid groups and clutter already very " b u s y " figures.

The meteorite data set was first examined on the basis of Gaffey's "quality" criterion and all spectra with a quality rating of greater than 2 (i.e., minor rust) were eliminated. An exception to this criterion was made for the C30 carbonaceous chondrite Lance, which had a quality rating of 3 and is a member of a relatively rare meteorite group. All average spectra and most duplicate spectra were also eliminated to avoid giving undue statistical weight to single samples. In four instances, multiple spectral runs of the same meteorite sample were included for meteorite types that would otherwise be greatly underrepresented. In three additional cases, particle size separates of the same meteorite were included in the analysis to gauge the effects of particle size on spectral characteristics and the resulting principal components. In all, 116 spectra of 103 individual meteorites were analyzed. All spectra were scaled to unity at 0.55/.~m to remove albedo as a factor in the analysis. Of the 527 total spectra used in this analysis, over 70% are asteroid spectra. In this form of statistical

156 BRITT ET AL.

analysis the space is defined by the data and the objective is to allow the already well-defined asteroid spectral classification system to define the space in order to classify the meteorite spectra relative to the asteroids.

Applications of Principal Component Analysis

To compare and classify the 527 asteroid and meteorite spectra in a quantitative and consistent manner, the multivariate statistical technique of principal component analysis (PCA) was used. Because the technique is described in depth elsewhere (Davis 1986, Tholen 1984) this discussion is limited to a brief review of the major features, advantages, and disadvantages of PCA. The goal of principal component analysis is to reduce the dimensionality of multivariate data sets so that the major sources of variance can be displayed in a two- or three-dimensional plot. The variation within the data set is quantified by statistical distance measures and then transformed into a set of statistically independent vectors. Each vector is rotated in statistical space to be orthogonal to and statistically independent of all other vectors. The first vector defines the largest sources of variation, the second vector the second major source of variation, and so on. Each sample is assigned scores along these vectors, or principal component axes, that quantify the sample's degree of similarity with all other samples for each principal component of variation. In a two-dimensional plot of two principal components, distance between samples becomes a measure of similarity or dissimilarity within the plane defined by those two statistically determined axes of variation.

The major advantage of using principal components to display relationships between samples is that, for statistically defined sources of variation, the similarity relationships are quantified and are free from experimenter preconceptions and assumptions. Not only are the relationships between samples quantified, but this technique also quantifies the major sources of statistical variation. This is accomplished by putting each statistically independent source of variation into a separate principal component. Analysis of the factor loadings of the components can provide substantial insight into spectrally diagnostic features. PCA can also be an excellent tool for rough classification of large data sets, particularly sets like the 527 asteroid and meteorite spectra used in this study. Finally PCA is useful for analysis and suppression of noise in spectral data sets. In the case of random noise, this variation would be distributed relatively evenly among the samples so PCA would tend to concentrate it into the minor components. Noise from consistent features like atmospheric water bands or instrument error would show up as a single component and can be eliminated from the analysis.

Principal component analysis, however, is not without

its disadvantages. It is just a tool to aid intelligent observa- tion and skilled data analysis. PCA's major flaw is that it uses a statistical definition of variance, and that may or may not correspond with the actual physical processes relfected in the data. PCA can also ignore or trivialize subtle spectral features, which may be diagnostic of mineralogy, in favor of the major sources of variation. In spectral data, inflections, shoulders, rare features, and very shallow bands will not be well represented and tend to be put into the minor components. PCA cannot differ- entiate between two spectral features that combine to produce a single apparent variation. A spectrum with a strong UV drop-off and an IR red slope will look like it has an overall red slope to the program and will be classified accordingly.

Correspondence analysis, the statistical technique used in this work, has several properties which make it a power- ful tool for the study of spectral data sets. First, the analysis rescales the data set by dividing the data matrix by the covariance matrix. This, in effect, moves the centroid of the data matrix to the center of the statistical space and allows the data to define the statistical space. In turn the data become a matrix of marginal probabilities and can be treated as multinomial probability distributions. Since variance in spectral data sets is subtle in statistical terms, this transformation has the effect of focusing the variation in the data set. The second advantage of correspondence analysis is that it produces both R- and Q-mode component scores simultaneously and on equivalent scales. Q- mode statistics analyzes samples in variable space while R-mode analyzes variables in sample space. For the pur- poses of this work, the R-mode analysis produces what is, in effect, an average spectrum for each principal component. These scores are used to determine what physical relationships are highlighted in each component. Q-mode scores, in turn, show the similarity relationships between the samples for each component. After the data matrix has been processed, a similarity measure is applied to produce a matrix of similarities. In this case Gower 's distance was used because it is a scaled measure and it is a good measure for discontinuous data such as spectral data points. R- and Q-mode component scores are then calculated.

STATISTICAL RESULTS AND DISCUSSION

Classification and Components

Shown in Fig. 1 are the first two principal components of the full asteroid/meteorite spectral data set. In this and all subsequent plots asteroids are represented by their Tholen (1984) type letter designation (except for the spectrally similar E-, M-, and P-type asteroids which, for sim- plicity, all have the designation "X" ) and meteorites are represented by the Symbols shown in Table I. The first


two components account for 59.6 and 31.9%, respectively, of the statistically recognized variance in the spectral data set. That 91.5% of the spectral information in seven filters can be distilled into two components is reasonable given the major sources of spectral variation in this wavelength range. From 0.33 to 1.1 p,m there are three major features in materials common in meteorites: (1) the silicate Fe2+-Fe 3÷ charge transfer absorption in the UV that produces the UV turndown, (2) the Fe 2÷ crystal field absorption near 1.0/~m that is diagnostic of pyroxenes and olivines, and (3) the red continuum slope seen in pure FeNi meteorites and iron-free enstatites (Gaffey 1976).

The first principal component can be seen as a rough measure of the continuum slope. However, high scores in this principal component can be due to two causes, both of which can fall under the statistical definition of a continuum slope. The first case consists of the strong red slopes of the iron-free enstatites and pure iron-nickel meteorites. The second possibility is a combination of a strong UV turndown and a red slope in the IR region. In each case the reflectance of the spectrum increases strongly from lower to higher wavelengths, but for different reasons. An

A V [] +

0 <>

TABLE I Meteorite PCA Symbols

Eucrites, SNCs, ADOR CI carbonaceous chondrites CM carbonaceous chondrites CO and CV carbonaceous chondrites Black chondrites Ordinary chondrites Irons Aubrites and enstatite chondrites

example of the first possibility is shown in the plot of Alais (CI) in Fig. 2a. In general, the red slope increases with component scores, so high positive values in the first component represent very red spectra. At the opposite end of the first component is the fiat and somewhat blue sloped spectrum of the F-type asteroid 1076 Viola, which produces very negative scores.

The second principal component is a general measure of the curve of the spectrum produced by a combination of the UV drop-off and the 1.0-p,m band. This component corresponds roughly to the "bend" parameter of

0 . 1 0 - Q c- O Cb

E o-0.1 0

O

f~3 O_

. ~

o-0 .30 c-

EL_

-C3 -0.5 0 * 0 0 (b

co

-0.70 -0.30

, i I ' I ' I ' I ' I D o DD, D ~ DD I~ D D

X " X~ vX'~'~J~ ^ I ~ D T < d~

~A

+

, .

*

*

, +

+ + + + , O +

* ~ ~ + v + + + + +

, + + ++ + , +, **+

÷ * + +

A + + A

+ + + R A

q+

i

I A I , , I i I ~ I , I I

-0.20 -0.1 0 0.00 0.1 0 0.20 0.30

First Principal Component FIG. 1. Plot of first two principal components of the combined asteroid and meteorite ECAS spectral data set. Asteroids are represented by

the letter of their Tholen (1984) spectral type and meteorites are represented by symbols explained in Table I.

158 BRITT ET AL.

a 1.4

1.2

o ~ "d

y, .8

.4

,2 .3

' ' I ' ' I ' ' I ' ' I ' ' I

i r I i I I i f I i , I i i I .45

1 4 f

L 1 ,6B .....

> .B

r

,4

.2 / d~ , I ~ ~ I ~ ~ I , ~ I , , I .6 .75 .9 1.05 .3 .45 .6 .75 .9 1.05 Wavelength (/~m) Wavelength (~m)

FIG. 2. (a) Spectra of objects with extreme values in the first principal component. The asteroid 1076 Viola has a very negative value and the carbonaceous chondrite Alais (CI) has a very positive value in the first component. The shapes of these spectra indicate that the first component is a measure of slope with high positive values corresponding to an increasing or red continuum and high negative values indicating a decreasing or blue slope. Spectral data are from Gaffey (1976) and Zellner et al. (1985). (b) Spectra of objects with extreme values in the second principal component. The asteroid 1746 Brouwer has a very positive value and the SNC meteorite Chassigny has a very negative value in the second component. The second component is a rough measure of the strength of the 1.0-/zm band with low values corresponding to a very strong, deep band and high values indicating no discernible band. Spectral data are from Gaffey (1976) and Zeliner et al. 1985).

Chapman et al. (1975). As shown in Fig. 2b the extreme possibilities are a very steeply curved spectrum as shown by the olivine-rich SNC meteori te Chassigny or the absence of bands as shown by the D-type asteroid 1746 Brouwer. In this case high positive values of the second component represent no bands while large negative values indicate strong UV and 1.0-/xm absorption bands.

As shown in Fig. 1 the statistical analysis is successful at recognizing spectral similarities. Correspondence analysis sorts the asteroids into sharply defined areas of statistical space and the membership of these zones corresponds to previous spectral classifications (Tholen 1984, Barucci et al. 1987, Chapman 1990, Tedesco et al. 1989). This analysis is based on the Tholen (1984) classification, but with only seven instead of eight colors for classification. Even without the ECAS s-filter, the program replicates Tholen 's statistical results almost exactly. The analysis also tends to put similar meteorite types together. The meteorites, in general, seem to occupy one broad spectral trend starting in the upper right with the irons and going toward the lower left with successive apparent bands of meteori te types. The irons are followed by the enstatite chondrites, then the carbonaceous chondrites, the ordinary chondrites, and finally the eucrites. The spectral variance shown by the meteorite collection is similar to that of the asteroids in the first (continuum slope) component . The suite of asteroids occupies only about 40% of the statistical space defined by the meteorites. Much of the extra variance in the meteorites, however , is produced by the relatively rare

and highly differentiated eucrites and SNCs with very strong features in the UV and IR that would produce high values in the second component . The spectral/ statistical relationships between asteroids and meteorites highlighted by PCA tend to confirm some concepts about the asteroid/meteori te connection, but it has also produced a few surprises. In the following sections these relationships are discussed in detail using a series of enlarged plots of the PCA space.

Eucri tes and Ordinary Chondri tes

Shown in Fig. 3 is the area of principal component space occupied by the eucrite and ordinary chondrite meteorites. The location of this area in PCA space suggests a great deal about the general spectral characterist ics of these objects. The moderate first-principal-component scores indicate very modest IR red slopes but strong UV drop-offs. The low to ext remely low second-component values reflect the characteristic deep absorptions in the UV and near 1.0/zm of the or thopyroxene-r ich mineralogy of eucrites. The overlap between the ordinary chondrites and the eucrites indicates the similarities and differences in the mineralogy of these two groups. The ordinary chondrites tend to have more olivine and smaller amounts of or thopyroxene than eucrites and the larger scatter in the first-principal-component scores reflects their more complex mineralogy.

The distribution of these objects in statistical space corresponds very well with previous ideas on asteroid/ meteorite spectral relationships. Several workers have

A S T E R O I D A N D M E T E O R I T E S P E C T R A 159

0.00

c--

~ - 0 . 1 0

0 C)

E-0.20 0

0 - - -0.30

. ~

~-0.40

El_ - 0 . 5 0 - *

c - O o - 0 . 6 0 - (1)

CO

-0.70 -0.30

i I ,.i ~z~l Q B a

a v v

* + 4-

+

+

+ +

* o

g + v

~- + + + +

4- 4 - , 4 - , +

"11" * +

~1~ + *

8 8 +

+ +

4- 4- 4- +

4- R 4-+

I a I, , I ,,, I

-0.20 -0.1 0 0.00 0.1 0 F rst Principal Component

FIG. 3. The area o f principal componen t space occupied primarily by eucri tes and ordinary chondri tes . This plot is an en la rgement of the lower left port ion o f Fig. 1 and the scales are significantly changed f rom those of Fig. 1.

' ' I ' ' I ' ' I ' ' I ' ' I /

1.2

1.2 •

. 6 - -

. 4 , , I ~ , I , , I , , I , , I • 3 .45 .6 .75 .9 1.05

Wavelength O~m)

FIG. 4. Examples of asteroid and meteori te spectra. Data points on asteroid spect ra have error bars while laboratory meteori te spect ra do not. Spectral data are f rom Gaffey (1976) and ZeUner e t al. (1985).

linked the spectral properties of eucrites to Vesta or Vesta-like differentiated asteroids (McCord et al. 1970, Gaffey and McCord 1978, Cruikshank et al. 1991). Eu- crites in Fig. 3 show substantial variance, but tend to cluster around Vesta, their suggested spectral analogue. Shown in Fig. 4 are the spectra of asteroid 4 Vesta and the eucrite Juvinas. Ordinary chondrites also show substantial variance and overlap the distribution of the eucrites. They also plot around their suggested asteroidal analogue, the Q-type asteroid 1862 Apollo (McFadden 1983). Shown in Fig. 4 are the spectra of 1862 Apollo and the LL5 ordinary chondrite Olivenza.

Ordinary chondrites and the S-type asteorids. One of the major problems in the asteroid/meteorite connection is the lack of spectrally identified main asteroid belt parent bodies for ordinary chondrites. The only pure spectral Q-type asteroid is 1862 Apollo and it is a relatively small Earth crosser. The ordinary chondrites, on the other hand, are by far the most common meteorite group, ac- counting for 79.4% of meteorite falls (Graham et al. 1985). Based on the fall statistics, many workers suggest that there should be at least a few main asteroid belt ordinary chondrite parent bodies (Pellas 1988). This problem is

160 BRITT ET AL.

replicated in statistical space by the lack of other asteroids plotting in the area defined by ordinary chondrite spectra. A number of theories have been advanced to account for this discrepancy. One idea is that the numerous S-type asteroids are the parent bodies of ordinary chondrites. Spectrally, S-type asteroids tend to show a much weaker and broader 1.0-/xm band than ordinary chondrites, indicating a mineralogy richer in olivine (Feierberg et al. 1982). In addition, S-types have a characteristic spectral red slope that is thought to indicate a high content of elemental iron. These characteristics are emphasized in statistical space by the offset of the S-type area from the area occupied by ordinary chondrites. The greater red slope forces the bulk of the S-types farther to the right than the ordinary chondrites and the weaker and broader 1.0-/~m absorption moves the group closer to the area occupied by the iron meteorites. Suggestions to account for the spectral differences between ordinary chondrites and S-type asteroids have included "unknown" regolith processes that have altered the surface material on ordinary chondrite parent bodies, increasing the apparent surface abundance of olivine and metal (Wetherill and Chap- man 1988, Gaffey et al. 1989). Other workers propose that the spectral indications giving the S-types a much more olivine and metal-rich composition than the ordinary chondrites reflect the basic mineralogy of this group (Gaffey 1984, 1986). They hypothesize that the stony- iron meteorites such as pallasites and mesosiderites are spectral analogues for the S-types.

A- and R-type asteroids. The rare A- and R-type asteroids are also offset from the eucrite and ordinary chondrite field. In statistical space, these asteroids plot along a trend from left to right of decreasing orthopyroxene and increasing olivine content (Bell et al. 1989a). The one R-type asteroid (a single-member class, consisting of only 349 Dembowska) is thought to represent a differentiated parent body with an olivine/orthopyroxene ratio much higher than that of the V-type asteroids (Gaffey et al. 1989). As yet no meteorites that correspond to this hypothesized mineralogy have been found. The A-type asteorids are thought to be the other end member of the orthopyroxene-olivine trend with a composition of almost pure olivine (Cruikshank and Hart- mann 1984, Gaffey et al. 1989). The very rare Brachinites are expected to be reasonable spectral analogues for the A-type asteroids, but unfortunately no spectrum from a meteorite in this group was available for inclusion in this analysis. The olivine-rich SNC Chassigny is a possible meteoritic analog for the A-types, but in PCA space Chassigny plots in the extreme lower left of the space, substantially away from the A-type asteroids. In this case Chassigny's very large spectral contrast in the UV/ IR absorptions and low overall red slope set it apart from the A-type asteroids.

Carbonaceous Chondrites and Black Chondrites

Carbonaceous chondrites and black ordinary chondrites represent a suite of low-reflectance meteorites that are rich in opaque materials such as carbon or micrometer- sized metal and troilite grains. The opaque component in these meteorites often strongly attenuates reflectance, producing dark spectra with very subdued absorption features. The distribution of low-reflectance meteorites in statistical space is shown in Fig. 5. The striking characteristics of these data are the very large variance in the first principal component, which tends to reflect spectral slope, and the relatively small variance in the second component, which measures the absorption features in the UV and IR. The varieties of spectrally dark carbonaceous chondrites, the CI, CM, CO, and CV, are spread in a swath across the distribution of spectrally light S-type asteroids rather than with their suggested analogues, the C-type asteroids (Johnson and Fanale 1973, Bell et al. 1989a). Some of the spectrally dark black ordinary chondrites plot with the C-type asteroids. The offset of the carbonaceous chondrites from the dark asteroids and their plotting in the same statistical area as the bright S-type asteroids seem counterintuitive.

The first step in unraveling this puzzle is to examine the spectral similarities indicated by the statistical analysis. Shown in Fig. 6 are the spectra of the S-type asteroid 7 Iris and the CI chondrite Orgueil. The two spectra are different in important details, especially in the center of the IR absorptions, but the general shapes of the spectra are very similar and these similarities are what the PCA identifies. Other examples are 115 Thyra (S) and Felix (C30). Once again, the shape of the UV drop-off and the shape and location of the 1.0-~m band are very similar. This is not to say that 7 Iris is the CI parent body; the very large albedo differences between these two objects rule that out. However, there are systematic offsets apparent in these data that have the dark carbonaceous meteorites plotting away from the statistical space where the dark B, C, F, G, and P asteroids are found.

The red slope in carbonaceous chondrites. There are several possible explanations for this apparent offset. The first possibility is the nature of the meteorite data them- selves. The meteorite spectra were taken with an integrating sphere spectrometer collecting data from crushed samples sieved to specific particle sizes, while the asteroid spectra are bidirectional and of objects with surface material of an unknown particle size. The effects of these two factors, particle size and photometric data collection, can be tested to determine the direction and magnitude of any possible offsets in PCA space.

The easiest to test is particle size. Previous work has shown that the spectra of samples with smaller particle sizes tend have a redder continuum slope than samples


0.1 0

c - (1) c - O

d:z. 0.05 E 0

0

( U

0.00 0 c--

. ~ M_

O.__

-c~_0.05 ( -

0 0 (1)

co

-0.1 0 -0 .30

F

F

F

¢ F F

F

I ' j< x x~xXx x I ' I "r ' ol ' ~x x °

F X X X x x X 0 ;~ T T o

F X X X ~ X

F F ( : ~ C iv "C~ ~ ( L~CI~L t ' c ¢C" P" x X S F F - C o ~ C(~'C S 8 8 8 .

FI~ C ~ Oz~ C C ,~ _~ ,~ = ~ '~B

0

"V C V SSSSS S ~88 S-

Cold B o k i l e d / ~ t ( ~ ) Q S - B . ~ ~-- 3 S S 8 8 [] S VS S SS~ S S S

v v s ~ s s S ~ s ss

°

• S A

+

+ s I~ +

t I I ÷ , l , I = I t

-o.2o -o. i o o.oo o.i o o.2o o.3o

F rst Principal Component FIG. 5. The area of principal component space occupied primarily by low-albedo asteroids and low-reflectance meteorites. This plot is an

enlargement of the central portion in Fig. I and the scales are significantly changed from those in Fig. 1. Lines connect the component scores of spectra of the particle size separates from Cold Bokkevelt (CM), Warrenton (C30), and Grosnaja (C3V).

1.2

=o P L 2

1

~ - 329 SVEA (C) AND SEVRU'KOVO (BLACK 1,5) .

, , l , , I , , I ~ , I , , I .45 .6 .75 .9 1.0S

W a v e l e n g t h ( /~m)

FIG. 6. Examples of asteroid and meteorite spectra. Data points in the asteroid spectra are marked by their error bars. Spectral data are from Gaffey (1976) and Zellner e t al. (1985).

with larger particle sizes (Johnson and Fanale 1973, Ad- ams and Filice 1967). This slope effect is reproduced in the statistical results. With larger particle sizes the values of the first principal component drop, shifting the samples progressively to the left in PCA space. The second principal component is largely unaffected by particle size changes because the 1.0-tzm absorption band in carbonaceous chondrites is either missing or very modest. In- cluded in the analysis were data from three carbonaceous chondrites, Cold Bokkevelt (CM), Grosnaja (C3V), and Warrenton (C30), that each had spectra available of three different particle sizes. The results are shown in Fig. 5 as lines connecting data points for individual meteorites. This shift in the first component is consistent with a pro- gressive spectral reddening with finer particle sizes.

Testing the effect of using bidirectional meteorite spectra rather than directional-hemispherical spectra is also relatively simple. Bidirectional reflectance spectra of several meteorite samples that were included in the analysis are available and PCA scores for these spectra can be calculated. The tested meteorites were Farmington (L5), Paragould (LL5), Orguiel (CI), Murry (CM), Allende (CV), Meghei (CM), Grosnaja (CO), and Murchison (CM).

162 BRITT ET AL.

Although these samples are not the identical material used for the Gaffey (1976) study, they do come from the same meteorites and do represent the same mineralogy. Results show that the PCA scores of spectra taken with the directional-hemispherical technique are nearly identical to the PCA scores of spectra of the same meteorite taken with the bidirectional technique. For example, the directional-hemispherical spectrum of Grosnaja produced first- and second-component scores of 0.122 and -0.057, respectively. The bidirectional spectrum had scores of 0.122 and - 0.102, respectively. The results for Meghei were directional-hemispherical, 0.030 and 0.010, and bidirectional 0.079 and 0.010. Assuming the samples are comparable, the offset due to spectral data collection techniques is about an order of magnitude smaller than the particle size offset.

Although there is a substantial offset that could be at- tributed to particle size effects, the magnitude of the observed offset is not great enough to move the carbonaceous laboratory spectra into the area occupied by the C-type asteroids. Some additional factors may be required to shift the spectra into the C-type region and there are a number of possibilities. First, if the surface materials of C-type asteroids are characterized by large particle sizes, the size ranges tested in the laboratory may still be too fine to simulate C-type soils. Second, the microstructure of the asteroidal regolith soil might be similar to the complex structures observed in lunar soils. The spectral effects of such structures are hard to predict and this structure would be difficult to simulate in the laboratory. Regolith processes may be another factor that could produce subtle alteration effects on the surface material of C-type asteroids. Some of these effects can be investigated by studying the spectra of the portions of carbonaceous chondrites that are rich in solar wind-implanted gases. This material would have been present on the surface layer of an asteroidal parent body and can be used as a source of spectral information to address the optical effects of surface processes on C-type asteroids. A fourth effect could be terrestrial weathering in the meteorite collection. A major weathering effect in meteorites turns metal and troilite into iron oxides such as geothite. This process has a reddening effect on the spectrum. The meteorites in the Gaffey (1976) study were all examined for terrestrial weathering and selected for minimum apparent weathering, but the terrestrial environment is corrosive to meteorites and a weathering effect is certainly possible. For example, the very strong reddening in Orgueil may be due to terrestrial weathering. A fifth possibility is that the laboratory measurements introduced a phase reddening. This is not very likely because the geometry of directional-hemispherical spectral data collection would tend to minimize phase effects. In addition the effects of phase were partially tested by mesuring a number of meteorites by

L,2

1.2

.8

.6 .3

2~i zos (r)

I

)

i , I i i I ~ ~ f ~ i I f , f .45 .6 ,75 .9 1,05

Wavelength OJrn)

FIG. 7. Examples of asteroid and meteorite spectra. Data points in the asteroid spectra are marked by their error bars. Spectral data are from Gaffey (1976) and Zellner e t al. (1985).

two different laboratory techniques. These results did not shown any apparent phase reddening; however, this effect cannot be completely ruled out. Finally, it is possible that the telescopic data have an average "blue" bias relative to the laboratory data.

Black chondrites. The next question to explore is the spectral similarity between the shock-blackened ordinary chondrites and the C-type asteroids. The principal component analysis tends to plot the black chondrites very near the dark C-type asteroids. As shown in Fig. 6 there are spectral similarities between black chondrites and some C-type asteroids. Britt and Pieters (1990) have suggested that shock and other regolith processes on some ordinary chondrite parent bodies may have altered the spectral signature of their surface material, producing a dark, flat spectrum similar to that of C-type asteorids. Some of the parent bodies of ordinary chondrites may be inner asteroid belt objects that have been classified as C-type because of their low albedo and relatively featureless spectra. Al- though black chondrites are not reasonable analogues for the 60% of C-type asteroids that have been shown to have water of hydration bands (Lebofsky et al. 1990), the other 40% of C-type asteorids that are anhydrous may contain some objects that are candidates for ordinary chondrite parent bodies.

The K-type asteroids. The region in statistical space between the S- and C-type asteroids is occupied by the proposed spectral type K asteroids (Bell 1988, Tedesco et al. 1989). The K-type asteroids are objects that were formerly classified as S-types but have spectral and albedo similarities to the anhydrous CV and CO carbonaceous chondrites. So far, most asteroids classified as K-types are members of the Eos family. These objects have orbits with semimajor axes of approximately 3.0 AU and may be near a dynamical space that would make it relatively


easy to perturb fragments into Earth-crossing orbits (Bell et al. 1989). Shown in Fig. 7 are the spectra of the K-type asteroid 221 Eos and the C3V chondrite Allende. Both objects show strong spectral similarities, particularly in the UV turndown region. In statistical space the CV and CO chondrites tend to plot in a swath across the region suggested for the K class.

The mineralogy of C- and G-type asteroids. The low albedo and relatively featureless spectra of C- and G-type asteroids suggest comparisons with the dark carbonaceous chondrites. However, analysis has shown that the spectra of carbonaceous chondrites are systematically offset in statistical space from the spectra of these asteroids. The possible causes of this apparent offset have been discussed at length in previous sections. Whatever the cause of the offset, CI and CM carbonaceous chondrites remain the only samples of hydrated materials in the meteorite collection (Dodd 1981). As such, they are reasonable analogues for the G-type asteroids that are all hydrated and the 60% of the C-type asteroids which show hydration features (Lebofsky et al. 1990). The remaining 40% of the C-types lack hydration features and may be spectrally analogous to some of the anhydrous CO and CV carbonaceous chondrites and the black ordinary chondrites (Britt and Pieters 1990). Either or both of these groups exhibit spectral features consistent with the spectral characteristics of anhydrous C-type asteroids. However, most CO and CV chondrites show red slopes and albedos that put them into the K-class area and only a handful of these meteorites plot with the C-type asteroids. Another possibility is the proposed "dehydrated" CM3 meteorites. CM3 meteorites are thought to be the precursor of the aqueously altered CM material; however, no samples of this material are found in the meteorite population (Bell et al. 1989a).

Irons and Enstatite Meteorites

The final group of meteorites to consider are the irons, enstatite achondrites, and enstatite chondrites. These objects tend to be spectrally similar and are characterized generally by red sloped and relatively featureless spectra. The statistical results confirm these characteristics by as- signing high values in both principal components as shown in Fig. 5. Once again, however, the iron meteorites are systematically offset from their usual asteroidal analogues, the M-type asteroids (Ostro et al. 1985, Bell et al. 1989a). Examples of this spectral offset are shown in Fig. 7 with the spectra of the iron meteorite Chulafinne, the T-type asteroid 1236 Thais, and the M-type asteroid 16 Psyche. The iron meteorite in this plot is systematically redder than its usual analogue. The T-type asteroids, even though they do have a rough spectral similarity to these meteorites, are too dark to be reasonable analogues. As

in the case of the carbonaceous chondrites, a factor in this spectral offset may be particle size. Since iron is very tough and ductile at ambient terrestrial temperatures, crushing iron meteorites to produce powders that can simulate a metallic asteroid regolith spectrally is ex- tremely difficult. The iron meteorite spectra used in this analysis were obtained from cut slabs rather than powders. Laboratory work with slabs of iron meteorites has shown that the spectral red slope is very sensitive to the micrometer-scale roughness of the slab (Britt and Pieters 1988). With finer-scale roughness, the iron red slope tends to flatten. This is opposite to the particle-size-related red slope seen in silicates, but conducting metals and noncon- ducting silicates are electromagnetically very different objects. The net result may be that M-type parent bodies have a very fine grained metallic regolith.

The suggested compositional analogues based on spectra for the E-type asteroids are the enstatite achondrite meteorites or aubrites (Bell et al. 1989a, 1989b). Cumber- land Falls was the only aubrite spectrum chosen for inclusion in the analysis. Its very fiat spectrum resulted in scores of - 0.117 for the first component and 0.029 for the second, which place it in the same area of the C- and F-type asteroids. The first-component score is similar to that of the E-type asteroids, but Cumberland Falls is offset by a lower second-component score. This offset is a result of its fiat spectrum in the 1.0-/xm band, while the E-type asteroids tend to have a positive spectral slope in the 1.0-/~m region. The mineralogically similar enstatite chondrites also are offset from the E-type asteorids. They tend to be systemically redder than the aubrites (Gaffey 1976) and are spectrally similar to the irons. The more moderate red slope of the aubrites tends to move their spectra closer to those of the E-type asteroids, making them reasonable analogues.

The mineralogy of M- and E-type asteroids. With the strong (and somewhat perverse) spectral similarity between iron-poor enstatite and FeNi meterorites, the principal diagnostic factor for these objects is albedo at visible and radar wavelengths. The aubrites are systematically brighter than the irons in visible wavelengths (Gaffey 1976), while the irons are much more reflective of radar (Ostro 1989). Viewed as a function of visible albedo the X-group (i.e., the E-, M-, and P-types, which are spectrally similar) neatly separates into three groups of high-, moderate-, and low-albedo objects. The M-types, with moderate visible and high radar albedos, are probably very metal-rich objects. The E-types, characterized by high visible albedos, are probably stony objects composed primarily of the iron-poor pyroxene enstatite.

Missing Meteorites: The B, D, F, P, and T Asteroids

One of the most interesting results of this analysis is a negative one; the lack of meteoritic spectral analogues for

164 BRITT ET AL.

the B-, D-, F-, P-, and T-type asteroids. This is hardly a new idea; several workers have pointed out that these asteroids lack direct analogues (e.g., Bell et al. 1989a, Gaffey and McCord 1978), but the analysis shows graphi- cally that significant populations in the asteroid belt di- verge from the spectral trend of material delivered to Earth. Probably two factors combine to produce this result. First, orbital dynamics arguments suggest that most meteoritic materials are produced from relatively limited zones in the inner asteroid belt (Wetherill 1985). The B-, D-, F-, P-, and T-type asteroids are low-albedo objects that are poorly represented in these zones. The bulk of the possible parent bodies for the missing meteorites are asteroids which reside primarily in the most distant zones of the outer asteroid belt or with the Trojan asteroids in the Lagrangian points of Jupiter's orbit (D- and P-types). Dynamical mechanisms to move this material into the inner Solar System probably do not exist (Wetherill 1985). The remaining B-, F-, and T-type asteroids are evenly distributed throughout the asteroid belt, but they are relatively rare objects and may not contribute significantly to the meteorite flux.

The second major factor may be the mineralogy of these objects. The B-, D-, F-, P-, and T-type asteroids are low- albedo objects with an inferred mineralogy rich in clays, organics, and opaques. Such material may not be as strong as the ordinary chondrites or the irons and thus would be less likely to be able to withstand the stresses produced by high-velocity entry into the Earth's atmosphere. In this case, the Earth's atmosphere would act as a filter, eliminating the relatively weak unmetamorphosed materials in favor of material that has been recrystallized. How- ever, a word of caution is appropriate for this discussion. Since direct meteorite analogues are not available, the mineralogy of these "missing" asteroids is open for spec- ulation. The primary tool for mineralogical analysis must be the interpretation of reflectance spectra, but these asteroids are characterized by spectral features that are simply not seen in the meteorite collection. Inferring mineralogy based on extrapolation from poorly understood spectral features is not a recipe for consistent success.

Planet-Crossing Asteroids and the Meteorites

One of the immediate reservoirs for meteorites is probably the planet-crossing Aten, Apollo, and Amor asteroids. These objects have dynamical lifetimes on the order of 107-108 years, indicating a consistent replenishment from the asteroid belt and a consistent loss due to ejection from the Solar System or planetary collision (Wetherill 1985, McFadden 1989). The orbits of the Earth-crossing Apollo asteroids, in particular, are similar to orbits inferred from meteor fireballs and are probably a source of at least some of the meteorite flux. Shown in Fig. 8 are the first two principal components of the meteorites and the 13 planet- crossing asteroids. Although the sample of 13 asteroids is

too small and probably strongly bised, a few tentative inferences can be drawn from this exercise. First, the asteroid spectral characteristics are surprisingly consistent. All of the S-type planet crossers occupy the portion of the S-type field closest to the ordinary chondrite meteorites and a few overlap the ordinary chondrite field. The relationship suggests that the planet-crossing population does not represent the spectral variance in the S-type population and that the planet-crossing S-types are, on average, more like the ordinary chondrites than the main belt S-types. The single Q-type asteroid, 1862 Apollo, plots well within the field of the ordinary chondrite meteorites and represents the most common meteorite fall type. This object is joined in PCA space by the QRS-type asteroid 1981QA and the QU-type 1980WF. Finally it is interesting, but perhaps not significant, that most of the limited population of planet-crossing asteroids plots in PCA space within the general trend of the meterorites while most of the main belt objects are offset in statistical space from the meteorites. The only C-type planet crosser is the notable exception to this generalization and indicates that this limited observed population is probably biased against low-albedo objects and much more observational work needs to be done.

There are several factors that may contribute to this apparent spectral offset between planet-crossing and main belt asteroids. First, there is considerable evidence that planet crossers tend to have much shorter lifetimes and much more immature regoliths than the main belt asteroids (Ostro 1989, McFadden 1989). Planet crossers, because of their small size, short lifetimes, and probable origin as collisional fragments, are much more likely to have relatively fresh exposed bedrock on their surfaces. A second factor is phase effects. Planet-crossing asteroids are often observed at much different phase angles than are common for observations of the main belt and strong phase reddening has been observed in at least one planet- crossing asteroid (Tholen 1984).

CONCLUSIONS

Principal component analysis provides a description of similarities between asteroid and meteorite spectra. It has the advantage of a consistent treatment of the major sources of spectral variance. Its disadvantage is in ignor- ing minor, but potentially diagnostic, spectral features in favor of the major trends. The results of the statistical analysis of asteroid and meteorite spectral data indicate the following:

(1) Principal component analysis is successful at classi- fying and characterizing the primary spectral variance in the asteroid and meteorite populations.

(2) There is more statistical variance in the spectra of the meteorite collection than is observed in the asteroid population.

(3) Some meteorite groups plot in the same area of


0.20

c-

0.05 0 c]._ E 0--0.1 0

(D

cu C:z--0.25 (.3 (-

0_-0.40

° - 0 . 5 5 dp

CO

-0 .70 -0 .30

~t 4-

I ' I ' I ' I ' i ' i '

o 0 0

0 o

CF 0 c7 C a o 0

V ~A V~O , < > S ~ o O

o ~S~V~vVVs+,Sv a V V SM

+ v++ + S S + + + + q ~ S +

+ , Q + + * ~ g + + + +

+ + + 4 - + ' ~ + +

++ +, ~., **~

4- * +

:¢

+

:¢

o ¢ O <> *

s V

i I , I , l I , I , I , I i

-0 .20 -0 .10 0.00 0.10 0.20 0.30

First Principal Component FIG. 8. Plot of the full area of the principal component space with the meteorite spectral data set and only those asteroids that have planet-

crossing orbits. The planet-crossing asteroids are denoted by the letter(s) of their Tholen (1984) classification.

PCA space as their suggested asteroidal analogues. These include Vesta and the eucrites, Apollo and the ordinary chondrites, some black chondrites and the C-type asteroids, some CO and CV carbonaceous chondri tes , and the K-type asteroids.

(4) A large number of meteori tes plot in a general trend that is offset f rom the bulk of the asteroid population. This is particularly evident for the low-albedo C-, B-, D-, F-, P-, and T-type asteroids as well as for the brighter E- and M-types.

(5) Several meteori te groups are systematically offset from their suggested analogues. These include the carbonaceous chondrites and the C- and G~type asteroids and the irons and the M-type asteroids. These spectral offsets are primarily seen in the first principal component and are the result of stronger red cont inuum slopes and/or stronger UV absorptions in the meteorites. Surface processes on asteroidal parent bodies, particularly particle size and regolith alteration effects, may be the cause of these apparent offsets.

(6) No meteori tes plot with the B-, D-, F-, P-, and T-type asteroids. These asteroid types are either relatively rare or common only in the outer asteroid belt, far from dynamical regions that could perturb these objects into Earth-crossing orbits. This result tends to confirm work implying that the meteori te collections sample only a relatively limited portion of the inner asteroid belt.

(7) The spectral characteristics of the small sample of planet-crossing asteroids tend to be more similar to the general spectral trend of the meteori te collection than the main-belt asteroids. This may confirm the heliocentric

bias of the meteori te collection. If true it implies that the meteori te collection may be a relatively good sample of the planet-crossing asteroid population, but much more observational work needs to be done.

Although principal component analysis is no substitute for detailed interpretation of spectroscopic measurements, the results of this analysis show that the technique is useful for classification of spectral types. It also has the advantage of presenting a spectral " o v e r v i e w " of the variation and relationships within a spectral data set. This overview of asteroids and meteori tes highlights both the great progress that has been made in asteroid science and the major gaps that still exist in our knowledge of asteroid mineralogy, processes, and evolution.

ACKNOWLEDGMENTS

The authors express their thanks to M. J. Gaffey for providing exten- sive spectral data, to J, A. Burns for his editorial assistance, and to C. R. Chapman and an anonymous colleague for helpful and constructive reviews. This research was supported by NASA Grant NAGW-28 and by a NASA graduate student research fellowship to D. T. Britt. RELAB is supported as a NASA-multiuser facility under Grant NAGW-748.

R E F E R E N C E S

ADAMS, J. B., AND A. T. FILICE 1967. Spectral reflectance 0.4 to 2.0 microns of silicate rock powders. J. Geophys. Res. 72, 5705-5717.

BARUCCI, M. A., M. T. C A P R I A , A. CORADINI, AND M. FULCHIGNONI 1987. Classification of asteroids using G-mode analysis. Icarus 72, 304-324.

BELL, J. F. 1988. A probable asteroidal parent body for the CV and CO chondrites. Meteoritics 23, 256-257.

166 BRITT ET AL.

BELL, J. F., AND K. KEIL 1988. Spectral alteration effects in chondritic gas-rich breccias: Implications of S-class and Q-class asteroids. In Proceedings, Lunar Planet. Sci. Conf. 18th (G. Ryder, Ed.), pp. 573-580. Cambridge Univ. Press, New York.

BELL, J. F., D. R. DAVIS, W. K. HARMANN, AND M. J. GAFFEY 1989a. Asteroids: The big picture. In Asteroids H (R. P. Binzel, T. Gehrels, and M. S. Matthews, Eds.), pp. 921-945. Univ. of Arizona Press, Tucson.

BELL, J. F., B. R. HAWKE, P. D. OWENSBY, AND M. J. GAFFEY 1989b. Atlas of asteroid infrared reflection spectra. Preprint.

BRITT, D. T., AND C. M. PIETERS 1988. Bidirectional reflectance properties of iron-nickel meteorites. In Proceedings Lunar Planet. Sci. Conf. 18th (G. Ryder, Ed.), pp. 503-512. Cambridge Univ. Press, New York.

BRITT, D. T., AND C. M. PIETERS 1990. The spectral effects of opaque phases in optically altered ordinary chondrites. Lunar Planet. Sci. XXI, 127-128.

CHAPMAN, C. R., D. MORRISON, AND B. ZELLNER 1975. Surface properties of asteroids: A synthesis of polarimetry, radiometry, and spectre- photometry. Icarus 25, 104-130.

CHAPMAN, C. R. 1976. Asteroids as meteorite parent-bodies: The astro- nomical perspective. Geochim. Cosmochim. Acta 40, 701-719.

CHAPMAN, C. R. 1990. Compositional structure of the asteroid belt and it families. In preparation.

CRUIKSHANK, D. P., AND W. K. HARTMANN 1984. The meteorite-asteroid connection: Two olivine-rich asteroids. Science 223, 281-283.

CRUIKSHANK, D. P., D. J. THOLEN, W. K. HARTMANN, J. F. BELL, AND R. H. BROWN 1991. Three basaltic Earth-approaching asteroids and the source of the basaltic meteorites. Icarus, $9, 1-13.

DAVIS, J. C. 1986. Statistics andData Analysis in Geology. Wiley, New York.

DODD, R. T. 1981. Meteorites: A Petrologic-Chemical Synthesis. Cam- bridge, New York.

DOLLFUS, A., M. WOLFF, J. E. GEAKE, D. F. LUPISHKO, AND L. M. DOUGHERTY 1989. Photopolarimetry of asteroids. In Asteroids II (R. P. Binzel, T. Gehrels, and M. S. Matthews, Eds.), pp. 594-616. Univ. of Arizona Press, Tucson.

FEIERBERG, M. A., H. P. LARSON, AND C. R. CHAPMAN 1982. Spectro- scopic evidence for undifferentiated S-type asteroids. Astrophys. J. 257~ 361-372.

GAFFEY, M. J. 1976. Spectral reflectance characteristics of the meteorite classes. J. Geophys. Res. 81, 905-920.

GAFFEY, M. J. 1984. Rotational spectral variations of asteroid (8) Flora: Implications for the nature of the S-type asteroids and for the parent bodies of the ordinary chondrites. Icarus 60, 83-114.

GAFFEY, M. J. 1986. The spectral and physical properties of metal in meteorite assemblages: Implications for asteroid surface materials. Icarus 66~ 468-486.

GAFFEY, M. J., J. F. BELL, AND D. P. CRUIKSHANK 1989. Reflectance spectroscopy and asteroid surface mineralogy. In Asteroids H (R. P. Binzel, T. Gehrels, and M. S. Matthews, Eds.), pp. 98-127. Univ. of Arizona Press, Tucson.

GAFFEY, M. J., AND T. B. McCoRo 1978. Asteroid surface materials: Mineralogical characterizations from reflectance spectra. Space Sci. Rev. 21, 555-628.

GRADIE, J. C., AND E. F. TEDESCO 1982. Compositional structure of the asteroid belt. Science 216, 1405-1407.

GRAHAM, A. L., A. W. R. BEVAN, AND R. HUTCHISON 1985. Catalogue of Meteorites. Univ. of Arizona Press, Tucson.

GREENBERG, R., AND M. C. NOLAN 1989. Delivery of asteroids and meteorites to the inner solar system. In Asteroids H (R. P. Binzel, T.

Gehrels, and M. S. Matthews, Eds.), pp. 778-804. Univ. of Arizona Press, Tucson.

HAPKE, B. 1981. Bidirectional reflectance spectroscopy. 1. Theory. J. Geophys. Res. 86, 3039-3054.

JOHNSON, T. V., AND F. P. FANALE 1973. Optical properties of carbonaceous chondrites and their relationship to asteroids. Geophys. Res. 78, 8507-8518.

LE BERTRE, T., AND B. ZELLNER 1980. Surface texture of Vesta from optical polarimetry. Icarus 43, 172-180.

LEBOFSKY, L. A., T. D. JONES, P. D. OWENSBY, M. A. FEIERBERG, AND G. J. CONSOLMAGNO 1990. The nature of low-albedo asteroids from 3 micron multi-color photometry. Icarus 83, 16-26.

McCORD, T. B., AND i . J. GAFFEY 1974. Asteroids: Surface composition from reflection spectroscopy. Science 186, 352-355.

McCORD, T. B., J. B. ADAMS, AND T. V. JOHNSON 1970. Asteroid Vesta: Spectral reflectivity and compositional implication. Science 168, 1445-1447.

MCFADDEN, L. A. 1983. Spectral reflectance of near-earth asteroids: Implications for composition, origin, and evolution. PhD thesis, Uni- versity of Hawaii.

MCFADDEN, L. A., M. J. GAFFEY, AND T. B. MCCoRO 1984. Mineralog- ical-petrological characterization of near-Earth asteroids. Icarus 59, 25-40.

MCFADDEN, L. A. 1989. Physical properties of Aten, Apollo, and Amor asteroids. In Asteroids H (R. P. Binzel, T. Gehrels, and M. S. Mat- thews, Eds.), pp. 442-467. Univ. of Arizona Press, Tucson.

McKAY, D. S., T. D. SWINDLE, AND R. GREENBERG 1989. Asteroidal regoliths: What we do not know. In Asteroids H (R. P. Binzel, T. Gehrels, and M. S. Matthews, Eds.), pp. 617-642. Univ. of Arizona Press, Tucson.

OSTRO, S. J., D. B. CAMPBELL, AND I. I. SHAPIRO 1985. Mainbelt asteroids: Dual-polarization radar observations. Science 229, 442-446.

OSTRO, S. J. 1989. Radar observations of asteroids. In Asteroids H (R. P. Binzel, T. Gehrels, and M. S. Matthews, Eds.), pp. 192-212. Univ. of Arizona Press, Tucson.

PELLAS, P. 1988. Ordinary chondrite-type asteroids in the main belt: Can we believe the zero abundance found by spectral studies from earth-based telescopes? Chem. Geol. 70, 32.

PIETERS, D. 1984. Asteroid-meteorite connection: Regolith effects im- plied by lunar reflectance spectra. Meteoritics 19, 290-291.

PIETERS, C. M., M. J. GAFFEY, C. R. CHAPMAN, AND T. B. McCORD 1976. Spectrophotometry (0.33 to 1.07/~m) of 433 Eros and compositional implications. Icarus 28, 105-115.

TEDESCO, E. F., J. G. WILLIAMS, D. L. MATSON, G. L. VEEDER, J. C. GRADIE, AND L. A. LEBOFSKY 1989. A three-parameter asteroid taxonomy. Astron. J. 97, 580-606.

THOLEN, D. J. 1984. Asteroid taxonomy from cluster analysis of photometry. PhD thesis, University of Arizona, Tucson.

THOLEN, D. J., AND M. A. BARUCCI 1989. Asteroid taxonomy. In Asteroids II (R. P. Binzel, T. Gehrels, and M. S. Matthews, Eds.), pp. 298-315. Univ. of Arizona Press, Tucson.

WETHERILL, G. W. 1985. Asteroidal source of ordinary chondrites. Meteoritics 20, 1-22.

WETHERILL, G. W., AND C. R. CHAPMAN 1988. In Meteorites and the Early Solar System (J. F. Kerridge and M. S. Matthews, Eds.). Univ. of Arizona Press, Tucson.

WISDOM, J. 1987. Chaotic dynamics in the solar system. Icarus 72, 241-275.

ZELLNER, B., D. J. THOLEN, AND E. F. TEDESCO 1985. The eight-color asteroid survey: Results for 589 minor planets. Icarus 61, 355-416.

Comparison of asteroid and meteorite spectra - Classification by principal component analysis

Documents

Transcript of Comparison of asteroid and meteorite spectra - Classification by principal component analysis