| Summary: | LLNL identification research is focused on the problem of correctly discriminating small-magnitude explosions from a background of earthquakes, mining tremors, and other events. The goal is to reduce the variance within the population of each type of event, while increasing the separation between the explosions and the other event types. We address this problem for both broad categories of seismic waves: body waves and surface waves. First, we map out the effects of propagation and source size in advance so that they can be accounted for and removed from observed events. This can dramatically reduce the population variance. Second, we try to optimize the measurement process to improve the separation between population types. For body waves we focus on the identification power of the short-period regional phases Pn, Pg, Sn and Lg, which can often be detected down to very small magnitudes. Many studies have shown that particular ratios of these phases, such as 6-to 8-Hz Pn/Lg, can effectively discriminate between closely located explosions and earthquakes. To extend this discrimination power over broad areas, we use our revised Magnitude and Distance Amplitude Correction (MDAC2) procedure. This joint source and path model fits the observed spectra and removes magnitude and distance trends from the data. The MDAC residuals are kriged to provide full 2-D path corrections by phase and frequency band. The MDAC residuals allow the exploration of all possible ratios and multivariate combinations of ratios for their discrimination power. We also make use of the MDAC spectra and the noise spectra to determine the expected detectability of each phase and use that to optimize the multivariate discriminants as a function of location. We quantify the discrimination power using the misidentified event trade-off curves and an equi-probable measure. We evaluate the correction surfaces using a cross-validation technique. The result is an end-to-end validation and discrimination performance measure for each station analyzed, and a set of measurement algorithms and correction surfaces that greatly improves discrimination. For surface waves we continue to make improvements in our regional group velocity tomography models. We have recently expanded our Rayleigh and Love wave dispersion measurements from the Middle East and North Africa into the European Arctic. The resulting tomography models now cover all of western Eurasia and provide high-resolution maps of group velocity from 10-to 100-seconds period. The maps also provide estimates of the expected phase spectra of new events that can be used in phase-match filters to compress the expected signals and improve the signal-to-noise ratio on surface wave magnitude (Ms) estimates. Phase match filters in combination with regional Ms formulas can significantly lower the threshold at which Ms can be measured, extending the Ms-mb discriminant. We have measured Ms in western Eurasia for thousands of events at tens of stations, with and without phase match filtering, and found a marked improvement in discrimination. Finally we are exploring the Ms-yield relation for the explosions in the dataset with announced yields.
|
|---|