Australia's Moho: A test of the usefulness of gravity modelling for the determination of Moho depth

In general, seismic methods provide a reliable way to image the crust-mantle interface, which is marked by a rapid increase in seismic velocity (the Moho). However, the coverage provided by seismic networks is necessarily limited due to access difficulties, and the cost and labour involved in collecting data. Gravity data provide an alternative way to model the depth to the Moho, and provide more consistent and broader coverage. We discuss the usefulness of gravity data to model Moho depth, and the advantages and disadvantages of several gravity modelling methods. As an example, a model of Australia's Moho is generated through seismically constrained gravity inversion, including an estimate of modelling uncertainty. The inversion results demonstrate that gravity inversion is generally useful, but that its usefulness is subject to the following limitations: 1 - gravity inversion cannot spontaneously generate thick, high-density crust, nor thin, low-density crust, and, unless constrained, will not generate a correct Moho where such crust exists. 2 - major errors in the definition of the a-priori density structure, in particular features that are fixed during inversion, will influence the Moho results. 3 - applying a broad range of inversion parameters is necessary to characterise uncertainty. Model variability maps for Australia show that the average error is less than 5. km. There is a general relationship with seismic coverage, but the areas of highest uncertainty are not necessarily those with the lowest seismic estimate density. Comparison with previous seismic, and seismic-gravity models of Australia's Moho indicates that low seismic data density limits usefulness due to higher uncertainty in the gravity inversion. High-seismic data density also limits usefulness because Moho depth is largely known, and there is little scope for change. The usefulness of gravity inversion is maximum under conditions where seismic coverage is moderately dense, but estimates are well distributed. © 2012 Elsevier B.V.