Spectral reflectance studies of plutonic rocks in the 0.45 to 2.45 micron region

Field spectral reflectance measurements in the 0.45 to 2.45 micron region have been made on a wide variety of plutonic rock types using the Portable Field Reflectance Spectrometer (PFRS) of the Jet Propulsion Laboratory (JPL). Detailed mineralogical information for 135 of the 182 sites sampled was obtained by laboratory analyses of rock samples taken from the field of view of the PFRS after spectral sampling. Data analysis was done on the IBM 360/60 of the Image Processing Laboratory at JPL. Regression analysis on the mineralogical and spectral data did not reveal strong correlation between any single mineral and spectral reflectance values. Good correllation was obtained between total iron bearing minerals and spectral reflectance values. Good correlation was also obtained between the minerals orthoclase and quartz and spectral reflectance values. This indicates that the general composition of a plutonic rock can be estimated from spectral reflectance measurements. The ranges of correlations indicated the importance of the 2.2 micron region in geological remote sensing. In order to test the feasibility of using spectral reflectance to discriminate plutonic rock types, the 135 samples were divided into eleven groups according to mineralogy. These groups approximately correspond to traditional rock classifications and include granite, quartz monzonite, monzonite, quartz diorite, syenite, anorthosite, leucogabbro, melanogabbro, pyroxenite, dumite and olivine and pyroxene bearing rocks. A linear discriminant analysis program was used to analyze and classify these groups by spectra with excellent results. Identification of rock type by spectra averaged 83% correct. If minor errors in classification are ignored, accuracy is 93%. The 47 spectra not used in generating the classification functions were classified with 85% accuracy. The PFRS spectral data were also resampled to simulate the data that would be returned by the proposed Landsat D satellite. These data were analyzed using the same groups. Accuracy decreased to 71% or 85% ignoring minor errors. The 47 spectra not used in generating the classification functions were classified with 70% accuracy. In several cases minor compositional variations could be detected on related rocks sampled 1n the same general areas. These results indicate that it should be possible to reliably discriminate plutonic rock types with good resolution based on spectral reflectance measurements from aircraft or satellites scanning the 0.45 to 2.45 micron wavelength region. Accuracy would be considerably enhanced if some ground control was available in the proposed study area to guide development of the discriminant functions and groups.