On the Application of Machine Learning to Planetary Sciences

Today, the most pressing scientific and engineering problems are typically nonlinear, dynamic, and multidimensional in space and time. To capture such complexity, researchers of every discipline are progressively adopting machine learning algorithms which model physical phenomena and interpret observations by learning through data. The aforesaid is provoking a fundamental change in the way we do science. This thesis describes the potentials and limits of the application of machine learning to planetary sciences. In Chapter 1, the state-of-the-art of data-driven applications to planetary sciences is reviewed. In Chapters 2 and 3, machine learning is applied to improve the realism of terrestrial planet formation studies by streamlining high-resolution collision simulations into machine learned response functions. A more realistic treatment of collisions in N-body terrestrial planet formation studies is found to profoundly affect the predicted mass, composition, and internal structure of terrestrial planets with respect to previous studies that assumed perfectly inelastic collisions (perfect merging). In Chapters 4 and 5, a novel approach that combines machine learning and Bayesian statistics to analyze data from remote sensing is presented. In Chapter 4, this method is first used to refine the interpretations of asteroids' properties from measurements of the surface thermal emission, including regolith abundance that eluded the traditional techniques of previous investigations. In Chapter 5, the approach simultaneously inverts remote sensing surface temperature and radar measurements of a terrestrial analog of Jupiter's moon Europa (Lake Vostok, East Antarctica) in order to constrain the temperature profile and the composition of the ice. The expected performance of this new data fusion approach is finally discussed in the context of the forthcoming NASA Clipper and ESA JUICE missions to Europa, which will both carry a radar sounder and a thermal imager.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

doctoral

Discipline thesis:degree_discipline

Graduate College

Grantor dc:publisher

The University of Arizona.

Year dc:date.issued

2020

Author dc:creator

• Cambioni, Saverio

Advisor dc:contributor.advisor

• Asphaug, Erik

Committee members dc:contributor.committeemember

• Carter, Lynn M.
• Delbo, Marco
• Furfaro, Roberto
• Reddy, Vishnu

• Asteroids
• Giant Impacts
• Machine Learning
• Planetary Formation
• Regolith
• Remote Sensing

Statement dc:rights

• Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction, presentation (such as public display or performance) of protected items is prohibited except with permission of the author.

Language dc:language.iso

en

Handle dc:identifier.uri

http://hdl.handle.net/10150/650775

OAI identifier oai:identifier

oai:repository.arizona.edu:10150/650775