<<

Lunar and Planetary Science XXXVI (2005) 2174.pdf

Factor Analysis and Target Transformation of Mini-TES Spectra: Recovery of Scene Endmembers at Merid- iani Planum. T. D. Glotch1, J. L. Bandfield1, and P. R. Christensen1, 1Department of Geological Sciences, Arizona State University, Tempe, AZ 85287-6305

Introduction: A objective of sending the which transforms, or rotates, the coordinate axes of the MER Rover to Meridiani Planum was to abstract matrices, and aligns them in a fashion that investigate the occurrence of gray, crystalline yields a physically meaningful result. From a practical detected by the Global Surveyor Thermal Emis- standpoint, this is accomplished by performing a linear sion Spectrometer (MGS-TES) instrument. Among the least squares fit of the eigenvectors onto a test vector instruments on the rover payload is the Mini-TES [1], that is thought to be a component of the system. If the which serves as a complement to the orbital TES in- best-fit spectrum is an acceptable match to the test strument and mid-infrared laboratory studies. The gray spectrum, then it is a physically meaningful component hematite at Meridiani Planum is present in the form of in the data being analyzed. The Mini-TES data that spherules that have been interpreted as were analyzed for this study were a series of system- weathering out of a -rich outcrop [2]. The spher- atic “vertical scans” and low-emission angle “fore- ules are also abundant on the plains of Meridiani and ground rasters” over the Meridiani Plains that provided may be present as a lag. The shapes of the hematite spectra with variable abundances of hematite and other features in Mini-TES spectra have not varied sub- soil components. stantially over the course of the Opportunity mission, Results and Discussion: Figure 1 shows the implying that the composition of the hematite spher- hematite spectrum recovered from the Mini-TES data ules is relatively uniform. using a goethite-derived hematite target vector. There Factor analysis and target transformation tech- is a particularly good fit in terms of band shapes and niques have been applied to Mini-TES spectra to iso- positions, although the relative depths of the two major late the surface component spectral shapes at Merid- absorption bands are slightly different from the test iani Planum, including that of hematite. These tech- vector. This result also indicates that the spectra of the niques have previously been applied to TES spectra to hematite spherules at Meridiani Planum do not exhibit recover atmospheric and surface spectral shapes [3]. a 390 cm-1 feature, and the best match to the The recovered hematite spectrum is similar to the hematite spectrum is that of a goethite-derived hema- TES hematite spectrum derived in the same way, and tite [4]. is best matched by a spectrum of hematite derived by Another main component of the Meridiani Planum dehydroxylation of goethite [4]. Three other spectral surface is basaltic (Figure 2). This sand has a shapes were recovered from the Mini-TES data. They spectral shape that is similar to the TES Surface Type include a relatively good match to TES Surface Type 1 1 () as seen from orbit, although there is some (basalt) [5], a spectrum similar to the TES globally slight variation in the spectral shape between 8 and 12 homogenous dust spectrum [6], and a final surface microns These differences can be explained by minor component that appears to be composed largely of variations in plagioclase or pyroxene abundances com- sulfate and amorphous silica. An acceptable match to pared to the global average basalt. It should be noted TES Surface Type 2 was not found. that we were not able to recover an acceptable match Endmember Recovery: The factor analysis tech- to the TES Surface Type 2 spectrum, which has been nique can be applied to any data set in which the com- interpreted as either a basaltic andesite [5] or a ponents of the data set vary independently, and the weathered basalt [8]. This indicates that the TES Type measurements within the data set can be expressed as a 2 composition is not present in appreciable concentra- linear sum of the components. Because thermal infra- tions at Meridiani Planum. red spectra combine linearly [7], the TES and Mini- Although Meridiani Planum is a classic dark al- TES data can be analyzed with these techniques. The bedo region of Mars, the Mini-TES instrument has first step in factor analysis is transforming the data detected the signature of fine particulates mixed with matrix into two abstract row and column matrices, or the soil (Figure 3). The recovered dust spectrum is eigenvector matrices with associated eigenvalues. The similar to the TES-derived global average dust spec- eigenvalues associated with each eigenvector indicate trum. The match is particularly good at the shortest the relative importance of the eigenvectors with re- wavelengths, where the signatures of bound and spect to the total spectral variation in the scene. are present [9]. The main framework silicate The second step is to transform the abstract eigen- transparency feature between 800 and 900 cm-1 is vectors into physically meaningful components. This is shifted to shorter wavelengths compared to the TES accomplished by the target transformation process, global average dust spectrum. If this is a real result, it Lunar and Planetary Science XXXVI (2005) 2174.pdf

would indicate that the fines at Meridiani Planum have a more sodic plagioclase composition that the globally homogenous dust, or there may be an additional fine sulfate component. An alternate hypothesis is that high emission angle spectra included in the analysis may have introduced a phase function effect. Future work will include analyzing only low emission angle meas- urements. We have somewhat less confidence in the final de- rived spectral shape (Figure 4) than in the other shapes that have been recovered. This shape, which is proba- bly representative of the sulfate-rich outcrop rock, can be modeled largely with a combination of sulfate and a Figure 2. Mini-TES spectrum of basalt recovered with silica phase. It was not accurately matched to any rock the target transformation and factor analysis technique. or mineral test vector, but because this shape was con- The Mini-TES spectrum is a good match to the TES sistently recovered over a range of test vectors, it may Surface Type 1 spectrum, and is a major component of be a real component in the scene. Additional refine- the surface of Meridiani Planum. ment of the atmospheric correction (including down- welling radiance) for Mini-TES spectra and the intro- duction of additional spectra to the factor analysis and target transformation process may confirm the pres- ence of this spectral shape.

References: [1] Christensen, P. R., et al., Science, 306, 1733-1739, 2004. [2] Squyres, S. W. et al., Sci- ence, 306, 1709-1714, 2004. [3] Bandfield et al., J. Geophys. Res., 105., 9573-9587, 2000. [4] Glotch, T. D., et al., J. Geophys. Res., 109, E07003, 2004. [5] Bandfield, J. L., et al., Science, 287, 1626-1630, 2000.

[6] Bandfield, J. L., and M. D. , Icarus, 161, 47- Figure 3. Mini-TES dust spectrum recovered with the 65, 2003. [7] Ramsey, M. S. and P. R. Christensen, J. target transformation and factor analysis technique. Geophys Res., 103, 577-596, 1998. [8] Wyatt, M. B. The spectrum is a good match to the TES global dust and H. Y. McSween Jr., Nature, 417, 263-266, 2002 spectrum at short wavelengths, but the transparency [9] Bandfield, J. L. et al., Science, 301, 1084-1087, feature at ~900 cm-1 is shifted to shorter wavelengths 2003. compared to the TES global dust.

Figure 1. Mini-TES hematite spectrum recovered using Figure 4. spectral shape recovered a goethite-derived hematite as the test vector. The re- from the Mini-TES data with the factor analysis and -1 covered spectrum lacks a 390 cm feature and the po- target transformation methods. The main spectral fea- sitions and shapes of the major absorption bands are tures can be largely modeled with a simple combina- good matches to the test vector. tion of sulfate and silica glass.