A COMPARISON OF METHODOLOGIES FOR REPRESENTING PATH EFFECTS ON REGIONAL P/S DISCRIMINANTS

Abstract

Short-period regional P/S amplitude ratios hold much promise for dis-criminating low-magnitude explosions from earthquakes in a Comprehensive Test Ban Treaty (CTBT) monitoring context. However, propagation path effects lead to variability in regional-phase amplitudes that, if not accounted for, can reduce or eliminate the ability of P/S ratios to identify the seismic source. Here we compare four different methodologies that account for the effect of heterogeneous structure on P/S amplitude variance: (1) distance corrections, (2) path-specific crustal wave-guide parameter regressions, (3) cap averaging (running mean smoothing), and (4) kriging. The predictability of each method is established by cross-validation (leave-one-out) analysis. We apply these techniques to Pn/Lg, Pg/Lg, and Pn/Sn observa-tions in three frequency bands from 0.75 to 6.0 Hz at station ABKT (Alibek, Turk-menistan), site of a primary station of the International Monitoring System (IMS). Paths to ABKT sample diverse crustal structures (e.g., various topographic, sedimen-tary, and geologic structures), leading to great variability in the observed P/S am-plitude ratios. For these data to be useful for isolating source characteristics, the scatter needs to be reduced by accounting for the path effects, and the resulting distribution needs to be Gaussian for most existing spatial interpolation and discrimi-nation strategies to have valid application. Each method reduces the scatter of the P! S amplitude measurements with varying degrees of success; however, kriging has the distinct advantages of providing the greatest variance reduction and a continuous correction surface with an estimate of the model uncertainty. The largest reductions in scatter are found for the lowest frequency P/S ratios (<3.0 Hz).