GSA Connects 2021 in Portland, Oregon

Paper No. 4-2
Presentation Time: 8:20 AM


LAMM, Sarah1, RODRIGUEZ, Laura E.2, SHEPARD, Rachel Y.2, PERL, Scott M.2, CELESTIAN, Aaron J.3 and BARGE, Laura M.2, (1)NASA Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Drive, Pasadena, CA 91109; Department of Geology, Kansas State University, Manhattan, KS 66502, (2)NASA Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Drive, Pasadena, CA 91109, (3)Natural History Museum of Los Angeles, Los Angeles, CA 90007

Here, we explored the feasibility of using mission ready techniques (infrared spectroscopy (IR), Raman spectroscopy, and Laser-Induced Breakdown Spectroscopy (LIBS)), alone and together, as a means of identifying iron (oxy)hydroxides and iron sulfides in planetary samples. Iron (oxy)hydroxide and sulfide minerals are widespread in hydrothermal environments and are thought to have been common in the ancient surface waters of both early Earth and Mars, as well as potentially the subsurface of ocean worlds such as Europa and Enceladus. In addition to their apparent abundance in environments on wet-rocky worlds that may have been habitable, iron minerals, particularly those that are metastable, may have had a significant impact on facilitating chemistry relevant to the origins of life and/or sustaining microbial communities. The geochemical conditions (pH, Fe2+:Fe3+ ratio, salinity, temperature, redox potential) at which iron minerals precipitate determines which mineralogical species of (oxy)hydroxides and/or sulfides form. In particular, iron (oxy)hydroxides and iron sulfides can help facilitate thermodynamically unfavorable reactions. These minerals can also adsorb organics or anions, providing a favorable setting for concentrating and preserving these materials. Thus, identifying which species are present may help constrain the conditions under which the minerals formed, the conditions they were exposed to after formation, and the specific abiotic organic reactions that took place.

We developed classification models from the IR, Raman, and LIBS spectral datasets of these iron minerals using partial least squares discriminant analysis, random forest, convolutional neural networks, support vector machines, and gradient boosting. Developing machine learning models using multiple samples representative of each mineral species is critical for future planetary work, especially on ocean worlds where communication between Earth and the instrument will have a long delay and where the rover/lander/submersible may have to autonomously decide which spectral datasets to send back. The results of this work will contribute to an expanding spectral library we are building in our lab at JPL to facilitate mineral identification and geochemical exploration of ocean world planetary analogs.