Observational and theoretical constraints on crustal and upper mantle heterogeneity

Abstract

The effects of crustal and mantle heterogeneity are reflected in the relative complexity of short-period seismograms at regional to teleseismic distances. Observations of events at regional ranges from Indonesia and New Guinea at the Warramunga array in Northern Australia allow a separation of the propagation through the mantle by concentrating on the coherent signal across the medium aperture array. Most of the incoherent features on the seismograms arise from signal generated noise in the general vicinity of the array with scale lengths of the order of 1–2 km or smaller. Partially coherent phases with relatively rapid changes in waveform across the array are to be associated with organised heterogeneity. This is most likely to arise from features in the crust or at the crust-mantle boundary near the receiver, but could in some cases be due to structural effects near the source. The coherent signals show considerable complexity for phases travelling large portions of their path above 80 km. As the depth of sources increases this coherent portion of the seismic field tends to become simpler, which suggests that the scale-length of heterogeneity is larger at depth. The use of horizontally stratified models has provided a convenient framework for the understanding of many seismic wave phenomena via the calculation of synthetic seismograms. However, further development is needed for calculating the response of the laterally heterogeneous Earth. The convenience of physical interpretation provided by working with the reflection and transmission properties of the medium can be retained if the structure can be split into a stratified reference model with superimposed heterogeneity. With a plane-wave decomposition of the wavefield, the effect of lateral variations can be introduced by forcing mixing of wavenumber components. A wide variety of phenomena can be simulated by varying the statistical description of the heterogeneity field in the wavenumber domain. This varies the cross-coupling between wavenumber components and so changes the nature of the resulting theoretical seismograms. Such an approach allows a quantitative study of the degradation of the major seismic phases compared to the predictions for stratified models, and also the characterisation of the codas of these phases.