Deep Learning classifier for planetary seismicity detection

Abstract:

Research in planetary seismology is fundamentally constrained by a lack of data, compared to terrestrial seismology. Seismological science observations for future missions can typically only be informed by theoretical signal/noise characteristics of the environment or likely Earth-analogues. Although objectives can be re-assessed after some initial data-collection upon lander arrival, transfer of high-resolution data back to Earth is costly on lander power usage and communications bandwidth, especially for missions to the outer Solar System. Over the last several years, development of GPU computing techniques and open-source high-level APIs have led to rapid advances in deep learning within the fields of computer vision, natural language processing, and collaborative filtering. These techniques are actively being adapted in seismology for a variety of tasks, including: earthquake detection, seismic phase discrimination, and ground-motion prediction. Our objective is to develop a seismicity detection algorithm for use on a future lander that is able to catalog seismic activity without the requirement of local training data. We build a deep learning classifier trained using spectrograms from Earth seismic data and test it on records of the Apollo Passive Seismic Experiment and the Apollo 17 Lunar Seismic Profiling Experiment (LSPE). We find that the algorithm is able to produce detections on the LSPE dataset with greater accuracy than a recent study using Hidden Markov Models. Additionally, we assess the accuracy tradeoff between our original Earth-trained classifier and one built using a training set of lunar seismicity.