Stacking autocorrelograms to map Moho depth with high spatial resolution in southeastern Australia

Abstract

Current estimates of Moho depth in southeastern Australia are based on sparse sampling. The results are augmented with 180 new Moho estimates constructed from spatial stacks of crustal P wave reflectivity derived from autocorrelograms at over 750 stations. The spatial stacks of reflectivity are constructed using a Gaussian with half width 0.5 ∘. Picks of the base of crustal reflectivity are made with the aid of the previous Moho model, based on the sparser data, and knowledge of the variation in the character of the crust-mantle boundary across the region. Good ties can be made to previous results from deep reflection profiling. The new information fills in many holes in coverage and provides a Moho map with closer ties to geological provinces. The procedure exploits the continuous records at the stations and just the vertical components and so can be applied to older data for which receiver function techniques cannot be used.

Crustal Imaging With Bayesian Inversion of Teleseismic P Wave Coda Autocorrelation

“…It is noteworthy that we find deeper Moho in the southeast of the Eromanga Basin and underneath the Murray Basin and also in an area with a relatively thick crust near 33°S and 140°N, as seen in deep reflection profiling (Kennett, ) and receiver functions (Fontaine et al, ) images. These results are also consistent with those obtained by Kennett et al () using spatial stacking of high‐frequency autocorrelograms. Shallower Moho depths imaged here around 37°S and 148°N also matches the results of Kennett et al ().…”

“…These results are also consistent with those obtained by Kennett et al () using spatial stacking of high‐frequency autocorrelograms. Shallower Moho depths imaged here around 37°S and 148°N also matches the results of Kennett et al (). In summary, the dissimilarity between our Moho model and the AuSREM Moho model may be generally due to different approaches and/or data sets with disparate sensitivities and coverage being used.…”

“…The 1 σ uncertainties are ≤10% of the maximum Moho depth (≤6.0 km). Since the autocorrelograms are more sensitive to velocity/density discontinuities (similar to receiver functions), one of the possible reasons for the high uncertainties (e.g., 5 km) is the lack of strong impedance contrast at the crust‐mantle boundary at some locations, especially in east and southeast of Australia (Kennett et al, , ). The wide prior ranges used for crustal thickness variations (Table ) could also contribute to the higher uncertainties.…”

“…An important point about the AuSREM model is that it will evolve over time when new data and methods become available. As an example, Kennett et al () updated the Moho depth of southeast Australia by spatially stacking of the crustal P wave reflectivity estimated from autocorrelograms. Recently, the Moho structure of Australia has been updated by Kennett et al () using the full range of available information by the end of 2016, including those extracted from many data points (seismic stations) that were not available for the model produced by Salmon et al ().…”

“…Here we should emphasize that our approach does not depend on conversions between wave types and allows access to a wide range of frequencies. Also, except for very complex near‐surface structures, the reflections extracted from autocorrelations are less contaminated by the free surface multiples compared to the receiver functions (Kennett et al, ; Kennett & Sippl, ; Sun & Kennett, ). Therefore, the inversion of these data can be more feasible than the inversion of receiver functions.…”