Skip to main content
SearchLoginLogin or Signup

A quantitative semi-automated technique to distinguish lunar impacts from shallow moonquakes

Presentation #119.08 in the session Moon & Earth (Poster + Lightning Talk)

Published onOct 23, 2023
A quantitative semi-automated technique to distinguish lunar impacts from shallow moonquakes

During the 7 year operation of the Apollo seismic network, tens of thousands of seismic signals were detected and cataloged (Nakamura et al., 1981), including the signal from impacts and shallow moonquakes. Classifying seismic signals correctly is key for evaluating the moon’s impact and seismicity rates. Yet, 60% of the original Nakamura et al. (1981) catalog events remain unclassified because the highly scattered lunar waveforms appear visually similar. In this work, we develop a rigorous, quantitative method to discriminate shallow moonquakes from impacts. First, we convert short- and long-period spectrograms to smoothed probability density functions and then calculate the K-L divergence between the distributions associated with pairs of events. The K-L divergence is a nonparametric measure of the differences between the two distributions; a K-L divergence of 0 shows an identical pair of signals. We test this new statistical method on classified events in the catalog. Preliminary results show a K-L divergence of >1 between previously identified shallow moonquakes and impacts, but a divergence of <0.5 between pairs of shallow moonquakes; shallow moonquakes are more similar to each other than to impacts. The K-L divergence can discriminate between shallow moonquakes and impacts. In addition, we analyze the short period time series of the seismic signals recorded at station S15 using the Python package tsfresh. This package automatically calculates over 1200 features from a time series, such as autocorrelation, entropy, and Fourier coefficients. We again test this method on classified events in the catalog. Our preliminary results identify 6 features that differ between shallow moonquakes and impacts. By combining these two methods, our preliminary results suggest we may robustly and semi-automatically classify new or previously unclassified events.

Comments
0
comment
No comments here