Predicting Deep Moonquake Source Regions Using Their Temporal and Spatial Patterns and Machine Learning

• Published in: Journal of geophysical research. Machine learning and computation Vol. 1; no. 2
• Main Authors: Majstorović, Josipa, Lognonné, Philippe, Kawamura, Taichi, Panning, Mark P.
• Format: Journal Article
• Language: English
• Published: American Geophysical Union/Wiley 01.06.2024
• Subjects: classification; deep moonquakes; farside seismic suite; machine learning; Moon; Sciences of the Universe
• ISSN: 2993-5210; 2993-5210
• DOI: 10.1029/2023JH000117

In the near future, lunar exploration should be enhanced by deploying a new seismic station on the farside of the Moon, the Farside Seismic Suite (FSS), a package of two seismometers recently selected by NASA to fly on a commercial lander. The new data should provide us with new insights into Moon's seismic activity, its interior, and composition. To fully benefit from the new data, we need to take advantage of the data acquired during the Apollo missions. The problem of relating new and old data is complex due to the single‐station nature of the future deployment. In this study, we tackle this issue by developing a machine learning model in the context of the deep moonquake (DMQ) classification problem. The DMQs form the largest group of detected events from Apollo data, and their source regions have been located and are known to exhibit temporal and spatial patterns connected with the monthly lunar tidal periods. Therefore, we propose to utilize a machine learning (ML) algorithm named random forest to identify DMQs source regions without using waveform information and only using the lunar orbital parameters related to DMQs time occurrences. We show that ML models perform well (with an accuracy >70%) when they are trained to classify 4 or fewer different source regions. This approach gives us a good first location approximation of the DMQs source regions and opens up a new approach to their location estimate when captured by the future FSS single‐station seismometers. Plain Language Summary Future space missions will provide us with new in situ ground vibration measurements of the Moon. From the Apollo missions in 1970, we know that the Moon can produce various seismic events. The most numerous ones are deep moonquakes (DMQs), which are events associated with the displacement deep within the lunar interior. These events occur in specific source regions, and their occurrence is related to the monthly motion of the Moon around the Earth. To further study lunar interior and DMQs with future missions, we need to define a way to relate the new data with the existing Apollo data. Due to the single‐station nature of the future deployment, the mapping between these two data sets is challenging with classical techniques. Therefore, in this study, we tackle this issue using machine learning (ML). We propose a model that is able to classify DMQs source regions using only the time of the occurrence of quakes. We show that models perform well (with an accuracy >70%) when they are trained to classify 4 or fewer source regions. This gives us a good preliminary location of the DMQs source regions and an opportunity to detect them with future lunar missions. Key Points We study the seismic data acquired during the Apollo missions for the future single‐station deployment on the far side of the Moon We propose the classification model of deep moonquake source regions using only their occurrence times The models perform with accuracy greater than 70% when trained to classify combinations of four or fewer source regions