Evaluation of one-dimensional seismic models of the lunar interior

Continuous data from the Apollo Passive Seismic Experiment (PSE) were recorded from 1969 to 1977 at four sites on the lunar surface. These data have subsequently been used to generate nine unique one-dimensional seismic velocity models for the moon. In spite of the fact that these models were generated from the same data set, significant differences exist between them. We analyzed travel-time and waveform effects of these previously published seismic velocity models. In order to examine the predictive power of lunar seismic models currently in existence, we calculated the mean squared error between the predicted travel times of direct P- and S-waves for each and four published catalogs of seismic arrival time readings. The mean squared error of P-wave arrival times are smaller than that of S-wave arrival times for each model. Models typically fit artificial impacts, meteoroid impacts, and deep moonquakes better then shallow moonquakes. We found no systematic variation in travel times based on location of seismic station. We constructed a new model (referred to as YY13) of the lunar interior based on a weighted average of the other models with respect to how well they predict measured travel times. The YY13 model provides the least misfit based on all available lunar seismic data. In addition, we reprocessed a total of 85,466 moonquake traces recorded by the PSE and calculated the coda decay rate separately for shallow and deep events. Shallow events show relatively longer codas than deep events. We computed synthetic seismograms for the YY13 mode and alternative versions of this with a varying thickness low velocity zone at the surface. By comparing synthetic seismograms with original data, it is challenging to reproduce the long seismic coda inherent in all lunar seismograms with a single low velocity or thin scattering layer as suggested by previous efforts.