Machine Learning for Planetary Science

Kurzfassung

Humans have used robots to explore other planets in our solar system since the beginning of planetary exploration. In the Apollo era, robotic spacecraft were sent to the Moon ahead of humans to collect critical information about lunar surface properties and landing sites. Today, scientists perform detailed scientific analyses and field geology on the surface of Mars using the Perseverance and Curiosity rovers, robotic vehicles equipped with cameras, spectrometers, and other instruments. Today, humans are actively exploring planets and other bodies in the solar system with more than 20 robotic spacecraft, with more to come in the next decade [1]. As humans send more robotic explorers into the solar system carrying increasingly sophisticated instruments, each observation contains even more information for scientists to analyze and increases the volume of archived data from planetary exploration missions. This is evidenced in Fig. 0.1, which shows the first close-up image of Mars ever taken in 1965 alongside an image of Mars taken nearly 50 years later in 2014. Analyzing the enormous volumes of data returned by past, present, and future planetary exploration missions will require scientists to adopt a different kind of robot—machine learning, a subfield of artificial intelligence which learns patterns, perceptions, and predictions from data in an automated way. Machine learning is a topic that spans a broad range of methods, models, learning types, and machine behaviors. In planetary science, machine learning can be used to facilitate scientific discovery and analysis in multiple ways: by uncovering patterns or features of interest in large, complex datasets that are difficult for humans to analyze; by inspiring new hypotheses based on structure and patterns revealed in data; or by automating tedious or time-consuming tasks. The goal of this book is to provide a bridge between the communities of machine learning and planetary science to enable increased uptake of machine learning methods in planetary science, and improve the accessibility of planetary science data for the machine learning community. In the first chapter, we will cover the basics of machine learning, special considerations for applying machine learning to planetary science datasets, guidelines for implementing machine learning models, and resources for the reader to find a deeper understanding of machine learning methods if desired. In the second chapter, we will cover the types of data and challenges that are encountered in planetary science. The third chapter provides tutorials for accessing and preparing planetary science data sets for machine learning applications. The final chapter presents several case studies detailing how machine learning has been implemented for a variety of planetary science applications and data types.