Description
- Abstract:
- The increasing size and resolution of planetary datasets calls for analytical approaches that can resolve subtle patterns within complex, multidimensional data. Dimensionality-reduction methods, a branch of unsupervised machine learning, meet this challenge by transforming high-dimensional data into a compact set of axes that capture the dominant modes of variation. Although widely used in planetary spectroscopy (Chapter 1), such approaches remain underutilized in other domains of planetary science. This dissertation introduces Guided Endmember Extraction (GEEX), a framework that extends Principal Component Analysis (PCA) through treatment of correlated data and strategic subsetting of geophysical data. We validate GEEX using observations from NASA’s Gravity Recovery and Interior Laboratory (GRAIL) mission to analyze impact-induced signatures within the lunar crust. In Chapter 2, we establish the foundations of GEEX and show that it reproduces key results from prior studies with an order-of-magnitude greater precision, including the transition associated with the lunar depth of pore closure, previously detectable only through Bayesian modeling. GEEX also reveals distinct eigenvector “shapes” for positive and negative gravity anomalies and identifies a critical porosity of ~16.5%, where craters are equally likely to exhibit positive or negative anomalies, indicating that the porosity of the shallow lunar crust is not in equilibrium. In Chapter 3, we apply GEEX to explore, for the first time, how crustal thickness influences crater gravity signatures. We find that thicker crusts produce a wider range of gravity profile shapes favoring negative anomalies, particularly for craters ≥70±10 km in diameter, and identify a ~30 million km³ domain of low-porosity (10-16%) crust at depths of ~5-20 km within intermediate-thickness (40-48 km) regions. In Chapter 4, GEEX enables the first high-resolution analysis of deep porosity around seven farside impact basins. Using the fraction of low-Dot Product Value (DPV < -0.5) complex craters as a proxy for deep crustal porosity, we find that basin rims preserve deep porosity, while basin centers compact over time from Imbrium-aged to Pre-Nectarian basins. Collectively, these findings inform our understanding of lunar crustal evolution and demonstrate GEEX’s potential as a generalizable framework for revealing meaningful structure in large, multidimensional planetary datasets.
- Notes:
- Thesis (Ph. D.)--Brown University, 2026
Citation
Hundal, Carol Beate,
"Composition and Impact-Generated Structure of Planetary Crusts Revealed by Dimensionality Reduction: Moon & Mars"
(2026).
Earth, Environmental and Planetary Sciences Theses and Dissertations.
Brown Digital Repository. Brown University Library.
https://repository.library.brown.edu/studio/item/bdr:ksh55zcj/
Relations
Collection:
-
Earth, Environmental and Planetary Sciences Theses and Dissertations
Theses and Dissertations for the Earth, Environmental and Planetary Sciences department....