USING SURFICIAL REFLECTANCE PROPERTIES TO ANALYZE CARBONACEOUS-LIKE ASTEROID SUBCLASSES

In our research we applied AutoClass, a data mining technique based upon Bayesian Classification, to the reflectance properties of asteroids most similar to carbonaceous meteorites. Previous reflectance studies relied mostly on Principal Component Analysis (PCA) to differentiate subgroups of asteroids within the carbonaceous-like group (e.g. B, G, F, Ch, Cg and Cb). The advantage of AutoClass is that it calculates the most probable classification automatically, removing the human factor from this part of the analysis.

Using the Sloan Digital Sky Survey (SDSS) surficial reflectance data, AutoClass divided the carbonaceous-like asteroids into two large classes and six smaller classes. The two large classes (n=4974 and 2033, respectively) displayed distinct regions with some overlap in color-vs-color plots. Each cluster's average spectrum was compared to 'typical' spectra of the carbonaceous-like group subtypes as defined by Tholen (1989) and each cluster's members were evaluated for consistency with previous taxonomies.

Of the 117 asteroids classified as B-subtype in previous taxonomies, only 12 were found with SDSS colors that matched our criteria of having less than 0.1 magnitude error in u(wavelength 0.35 microns) and 0.05 magnitude error in wavelengths 0.48, 0.62, 0.76, and 0.91 microns. Although this was a relatively small group, 11 of the 12 B-types were placed by AutoClass in the same cluster. By determining the carbonaceous-like group sub-classifications in the large SDSS database, this research furthers our understanding of asteroid surficial composition variability.