Numerical modelling and observations of nuclear-explosion coda wavefields

Abstract

Frequency-dependent earthquake coda attenuation values are often reported; however such measurements usually depend on the types of the attenuation models employed. In this thesis, I use numerical modeling of Peaceful Nuclear Explosion (PNE) codas at far regional to teleseismic distances to compare two of such models, namely the conventional frequency-dependent attenuation with parameters (Q₀, η) defined by Qcoda(f) = Q₀f¹ and frequency-independent effective attenuation (Qₑ) with geometrical attenuation (γ). The results favour strongly the (γ, Qₑ) model and illustrate the mechanisms leading to apparent Qcoda(f) dependencies. Tests for variations of the crustal velocity structures show that the values of γ are stable and related to lithospheric structural types, and the inverted Qₑ values can be systematically mapped into the true S-wave attenuation factors within the crust. Modeling also shows that γ could increase in areas where relatively thin attenuating layers are present within the crust; such areas could likely be related to younger and active tectonics. By contrast, when interpreted by using the traditional (Q₀, η) approach, the synthetic coda shows a strong and spurious frequency dependence with η ≈ 0.5, which is also similar to many published observations. Observed Lg codas from two Peaceful Nuclear Explosions located in different areas in Russia show similar values of γ ≈ 0.75·10ˉ² sˉ¹ , which are also remarkably close to the independent numerical predictions in this thesis. At the same time, coda Qₑ values vary strongly, from 850 in the East European Platform to 2500 within the Siberian Craton. This suggests that parameters γ and Qₑ could provide stable and transportable discriminants for differentiating between the lithospheric tectonic types and ages, and also for seismic coda regionalization in nuclear-test monitoring research.