DETERMINATION OF PETROGENETIC ASSOCIATION OF DETRITAL TOURMALINE: A TOOL FOR RESOURCE EXPLORATION
We analyzed 164 tourmalines from five petrogenetic associations: calcareous metamorphic rocks (34 samples, 1 locality), pelitic metamorphic rocks (18 samples,3 localities), hydrothermal deposits (33 samples,7 localities), silicic igneous rocks including Li-poor pegmatites (49 samples, 6 localities), and Li-rich pegmatites (30 samples, 13 localities). Each sample was analyzed by Laser-Induced Breakdown Spectroscopy in Ar. Most samples were analyzed 150 times; smaller samples were analyzed 50-100 times. All spectra from a single specimen were averaged into one spectrum. In the case of zoned tourmalines, cores and rims were analyzed and modeled separately.
Multivariate modeling of the spectra used a series of Partial Least Squares Regression models. In each model, one petrogenetic association was investigated; samples were modeled as either belonging to that association or belonging to the group of all other associations. Each investigated association was eliminated from subsequent models. Spectra from half of the locations for each petrogenetic association were used to train the models; the remaining half of the spectra were used to validate the models and calculate success rates. The first model recognizes calcareous metamorphic tourmalines (92% success). Next, tourmalines from pelitic metamorphic rocks are recognized (96% success). Tourmalines from hydrothermal systems are next recognized (94% success). Finally, silicic igneous tourmalines are distinguished from those in Li-rich pegmatites (92% success). Overall, 74 of the 79 samples in the validation set were correctly identified (93.7% success rate).