Optimization of Deep Moonquake Event Stacks in the Apollo Lunar Seismic Data and Applications to Lunar Structure

Our studies of the Apollo lunar seismic data have focused on the deep moonquakes, which comprise 46% of catalogued events and appear mostly on the long-period channels. Thought to be tidally triggered, deep moonquakes occur periodically at specific source locations, each of which produces its own characteristic waveform. Using cross-correlation techniques, deep moonquakes can be sorted into separate groups, stacked, and analyzed for information on lunar interior structure. An accurate catalog of events is crucial to understanding deep event occurrence characteristics and source locations. This requires identification of all deep events associated with known groups, both in existing catalogs and in the continuous time series. Previous studies of lunar deep events have focused on the event catalog, not the complete continuous data. We search for previously unidentified events by cross-correlating preliminary waveform stacks representing a known group with the continuous data. Our analysis so far has concentrated on deep event group A1, which has the largest number of events (323) of all identified source regions. To generate the preliminary stack every event within the A1 group is cross-correlated with every other event in the group. Those event pairs with cross-correlation coefficients greater than some threshold value (i.e., events that are clearly members of the A1 group) are retained. By combining the cross-correlation approach with a robust median despiking algorithm, we have produced improved differential times and amplitudes, enabling us to construct cleaner stacks. Each deep event group stack is then cross-correlated with the continuous time series. Using this method, we have successfully identified 34 new A1 events on stations 12 and 16. The extension of our methods to data from stations 14 and 15, as well as to target stacks from other deep event groups, will provide an improved data set for investigating the temporal occurrence of deep moonquakes. After identifying new events, we optimize stacks from A1 and other groups to obtain the cleanest possible waveforms from which to pick P- and S-wave arrival times. The initial stack is cross-correlated with all members of a given deep event group, a new stack formed by weighting event group members, and the process repeated. We are investigating different weighting schemes, including using the cross-correlation coefficient and measures of signal-to-noise. Convergence is achieved after only a few iterations, and picks can be made from the resulting optimized stack. These picks can be used to test and refine lunar seismic velocity models and location estimates for the deep event groups.