Comprehensive Processing of the Apollo Lunar Seismic Data

The Apollo Passive Seismic Experiment consisted of four seismic stations deployed on the lunar surface between 1969 and 1972. We have successfully read and processed all of the lunar seismic event tapes available from the IRIS Data Management Center. Our studies have focussed on the long-period data, and have included filtering and despiking noisy data, event classification, cluster identification, and robust methods for amplitude estimation. We have completed cross-correlation analyses for known groups of deep events. In these analyses, every event in a previously-classified group was cross-correlated with every other event in that group. We find good correlation between events, confirming earlier visual classifications. By combining the cross-correlation approach with a robust median despiking algorithm, we have produced greatly improved differential times and amplitudes, enabling us to construct much cleaner stacks. Because of the inferior signal-to-noise ratio exhibited by a majority of the traces, we have tested several methods of calculating the amplitude of an event. These include relative amplitudes calculated through cross-correlation and several different estimates of absolute amplitude. Our preferred amplitude estimate (the 75th percentile of the absolute deviation about the median) produces the best x-component vs. y-component correlation when plotting events from a single deep event group. Because of the small geographical area spanned by the stations, deep events from the same source region should produce similar amplitudes on the horizontal channels at different stations. Both our previous "grading" technique (in which each event is assigned a grade of A, B, or C based on such factors as signal-to-noise and initial impulse coherence) and our new amplitude estimates indicate that stations 12 and 16 have the greatest number of "visible" events with the most consistent amplitudes. We will begin our search for more events using the continuous data from these stations. Each deep event group can be represented by a stack of its constituent members, appropriately despiked and time-shifted. This will permit new waveform cross-correlations. Instead of using the time-consuming event-by-event method employed earlier, each event group, represented by a single waveform stack, can be correlated with the continuous time series. Using this approach, we hope to identify additional events that can be associated with previously-defined deep clusters.