Australia's Moho: A test of the usefulness of gravity ...rses.anu.edu.au/~brian/PDF-reprints/Proofs/TECTO-125531.pdf · 2011), resource exploration (e.g. Begg et al., 2010; Bierlein

Tectonophysics xxx (2012) xxx–xxx

TECTO-125531; No of Pages 12

Contents lists available at SciVerse ScienceDirect

Tectonophysics

j ourna l homepage: www.e lsev ie r .com/ locate / tecto

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

A.R.A. Aitken a,⁎, M.L. Salmon b, B.L.N. Kennett b

a School of Earth and Environment, University of Western Australia, Perth, Western Australia, Australiab Research School of Earth Sciences, Australian National University, Canberra, Australia

⁎ Corresponding author. Tel.: +61 8 6488 5804.E-mail address: [email protected] (A.R.A. Aitk

0040-1951/$ – see front matter © 2012 Elsevier B.V. Alldoi:10.1016/j.tecto.2012.06.049

Please cite this article as: Aitken, A.R.A., et adepth, Tectonophysics (2012), doi:10.1016

a b s t r a c t
a r t i c l e i n f o
Article history:Received 1 March 2012Received in revised form 11 June 2012Accepted 24 June 2012Available online xxxx

Keywords:MohoGravity modellingInversionAustralia

In general, seismic methods provide a reliable way to image the crust–mantle interface, which is marked by arapid increase in seismic velocity (theMoho). However, the coverage provided by seismic networks is necessar-ily limited due to access difficulties, and the cost and labour involved in collecting data. Gravity data provide analternative way tomodel the depth to theMoho, and providemore consistent and broader coverage. We discussthe usefulness of gravity data to model Moho depth, and the advantages and disadvantages of several gravitymodelling methods. As an example, a model of Australia's Moho is generated through seismically constrainedgravity inversion, including an estimate of modelling uncertainty. The inversion results demonstrate that gravityinversion is generally useful, but that its usefulness is subject to the following limitations: 1 — gravity inversioncannot spontaneously generate thick, high-density crust, nor thin, low-density crust, and, unless constrained,will not generate a correctMohowhere such crust exists. 2—major errors in the definition of the a-priori densitystructure, in particular features that are fixed during inversion, will influence the Moho results. 3 — applying abroad range of inversion parameters is necessary to characterise uncertainty.Model variabilitymaps for Australiashow that the average error is less than 5 km. There is a general relationshipwith seismic coverage, but the areasof highest uncertainty are not necessarily those with the lowest seismic estimate density. Comparison with pre-vious seismic, and seismic-gravity models of Australia's Moho indicates that low seismic data density limits use-fulness due to higher uncertainty in the gravity inversion. High-seismic data density also limits usefulnessbecause Moho depth is largely known, and there is little scope for change. The usefulness of gravity inversionis maximum under conditions where seismic coverage is moderately dense, but estimates are well distributed.

© 2012 Elsevier B.V. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 02. Moho modelling using gravity data — a review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0

2.1. Forward modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 02.2. Inverse modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 02.3. Process oriented approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 02.4. Methods used in this study. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0

3. Modelling results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 04. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0

4.1. Australia's Moho . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 04.2. The capabilities and limitations of gravity inversion for Moho modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 04.3. The influence of seismic constraint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0

5. Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0

en).

rights reserved.

l., Australia's Moho: A test o/j.tecto.2012.06.049

1. Introduction

A knowledge of the thickness of the Earth's crust and the mor-phology and character of its interface with the mantle is fundamental

f the usefulness of gravity modelling for the determination of Moho

http://dx.doi.org/10.1016/j.tecto.2012.06.049

mailto:[email protected]

Unlabelled image


Unlabelled image

http://www.sciencedirect.com/science/journal/00401951


2 A.R.A. Aitken et al. / Tectonophysics xxx (2012) xxx–xxx

to understanding the solid earth. Furthermore, crustal thickness is acrucial constraint on many geoscientific endeavours, including the in-terpretation and modelling of tectonic systems (e.g. Aitken et al.,2009; Brandmayr et al., 2011; Prezzi et al., 2009; Salmon et al.,2011), geodynamic modelling of earth processes (e.g. Beaumont etal., 2000; Grobys et al., 2008; Pfiffner et al., 2000; Valera et al.,2011), resource exploration (e.g. Begg et al., 2010; Bierlein et al.,2006) and analyses of crustal stress fields, and their influence onearthquake risk (e.g. Flesch et al., 2001; Naliboff et al., 2012).

The crust–mantle interface (hereafter considered equivalent tothe Moho) is most commonly imaged using seismic methods, mostnotably seismic refraction studies and receiver function analyses,but also seismic reflection profiling and seismic tomography. Eachof these methods is capable of generating an estimate of the localMoho depth. In refraction, the true Moho, i.e. the increase in P-wavevelocity from b7.2 km s−1 to >7.6 km s−1, can be determined. In re-ceiver function analysis, the discontinuity can cause the partial con-version of P-wave energy to an S-wave (or alternatively S-waveenergy to a P-wave), which results in a delayed Ps-wave arrival (oraccelerated Sp arrival). In this case, the crustal velocity structure canbe modelled to match the waveform, so as to achieve an estimate ofMoho depth (Langston, 1979). In seismic reflection, the Moho is de-fined not by its velocity contrast, but by its reflectivity. Moho reflec-tivity can be highly variable and the Moho is not necessarilyreflective, in which case its depth can be estimated from the base oflower crustal reflectivity (Kennett et al., 2011). Seismic tomography,in particular ambient noise tomography (Behr et al., 2010; Stehlyet al., 2009), can also yield information on Moho depth, althoughthis is the least reliable of the seismic methods.

Although they each carry inherent uncertainties and interpretationaldifficulties, these seismic methods provide generally robust knowledgeof the depth to the Moho, and indicate something of its character forthe portion of crust directly sampled by the study. However, in the crea-tion of extensive regional-scale or continent-scalemaps, sub-optimal cov-erage causes several problems. Firstly, where seismic data are sparse, themodel will be poorly constrained because the influence of local estimatesmust be extrapolated far beyond the sampled area. Secondly, irregularlyspaced coverage causes a requirement for a balance between good resolu-tion of smaller-scale featureswhere data are closely spaced, and adequatecoveragewhere data aremorewidely distributed. Thirdly, the vastmajor-ity of seismic observations are made on land, and imaging of the crust inoffshore regions is often poor. It is alsoworth noting that the various seis-mic methods are most sensitive to different aspects of the seismic struc-tures and so different Moho estimates need not correspond exactly.

Because the Moho is also a prominent density gradient, gravitydata have a role in understanding the geometry of the Moho, andcan help in interpreting its nature. Gravity data are often availableat higher resolution and with more regular coverage than seismicstudies. In particular, combined satellite-ground gravity models(e.g. EGM 2008 (Pavlis et al., 2008)) provide seamless global coverageat relatively high resolution. However, the usefulness of gravity datais limited by the lack of a reliable means to discriminate betweenthe signal from deep crustal density contrasts, the Moho, and densityvariations in the uppermost mantle.

This manuscript describes the application of a gravity inversionscheme tomodel the geometry of Australia'sMoho, using a recent compi-lation of seismicMoho depth estimates (Kennett et al, 2011). Our primaryaim is to generate a model of Australia's Moho that satisfies both seismicand gravity data. However, we also seek to understand the capabilitiesand limitations of the method applied, and the influence of the level ofseismic constraint on the resulting model.

2. Moho modelling using gravity data — a review

Gravity data have long been used to constrain models of crustalthickness (e.g. Vening-Meinesz, 1931). Throughout this time, a great

Please cite this article as: Aitken, A.R.A., et al., Australia's Moho: A test odepth, Tectonophysics (2012), doi:10.1016/j.tecto.2012.06.049

variety of approaches have been used, including forward modelling,inverse modelling and process-oriented modelling. Estimation ofMoho depth from the spectral content of the gravity data has alsobeen applied (e.g. using Euler deconvolution (Tedla et al., 2011) orpower spectrum analysis (Studinger et al., 1997)), however, the use-fulness of such methods has been questioned (Reid et al., 2012). Dueto the availability of more capable and robust methods, we do not dis-cuss parameter estimation here. Below we summarise the philoso-phies behind forward, inverse and process oriented modelling, andoutline several methods used within each.

2.1. Forward modelling

Forward modelling involves the generation of a model of theEarth's crust, which can then be modified until a satisfactory fit tothe gravity data is achieved. This approach has the key advantagethat model geometries can be explicitly controlled, and hence, thereare few problems in generating a believable Moho geometry. Forwardmodelling however has the problem that changes to the model aredriven by user-input, and are thus subjective, and are prone to errordue to misinterpretations on the part of the modeller. Severalmethods have been successfully used.

The least structure approach is based on the concept that the mostsimple model that adequately fits the data is the most reasonable, fol-lowing the concept of Occam's razor. At its simplest, this will involve alayered model with sedimentary basins, crystalline crust, perhapswith middle and/or lower crustal layers and the mantle, each withconstant density (e.g. Ebbing et al., 2001; Grobys et al., 2008), al-though additional blocks are often required to satisfy either the grav-ity data, or to include other known features (e.g. Ferraccioli et al.,2011; Hackney, 2004). With least structure, there is the advantagethat the number of variables is much reduced, and so the problem be-comes less under-determined, and solutions become less uncertain.Further advantages include reduced labour due to the relative sim-plicity of the modelling process and a more obvious link betweenmodel structure and the resulting gravity anomalies. Nevertheless,the least structure approach has several drawbacks. Firstly, theEarth's crust and mantle are not homogenous materials, and rep-resenting them as such is an oversimplification that can have seriousconsequences for the model. In particular, a modelled Moho geome-try using constant density contrast is an end-member of the rangeof possibilities, and the Moho could often be better represented by in-cluding relatively small changes in the densities of the crust and/ormantle. Secondly, the level of structural complexity that can be de-rived from gravity data is typically low, and is often much lowerthan would be expected in reality. Thus least structure models areoften at odds with the crustal structure derived from geological stud-ies or seismic reflection lines. Notwithstanding these drawbacks, thismethod is appropriate for many studies, and provides a relatively sta-ble and effective method where constraints are few. Continent-scalecoverage is possible using this approach (e.g. Grobys et al., 2008)but only with significant effort.

The structural–tectonic approach (e.g. Aitken et al., 2009; McLeanet al., 2010; Prezzi et al., 2009; Stewart and Betts, 2010; Williamset al., 2010) is based on the concept that incorporating knowledgegained from other studies (in particular near-surface detail) gener-ates a more geologically realistic result. For example, known struc-tures from geological cross sections may be included (McLean et al.,2010; Stewart and Betts, 2010), near-surface structure frommagneticmodelling may be modelled in detail to constrain upper-crustal struc-ture (Aitken et al., 2009), or the crustal structure imaged in seismicreflection profiles may be extrapolated to other areas (Williamset al., 2010). Where abundant data exist, these can all be combinedinto a detailed tectonic model, prior to modelling (e.g. Prezzi et al.,2009). So long as the densities of the surface rocks can be constrained,this approach has the advantage that, because the upper-crustal



3A.R.A. Aitken et al. / Tectonophysics xxx (2012) xxx–xxx

structure is better constrained, the geometry of the deeper crust is lessuncertain. However, the level of structural complexity is usuallyhigher than that required by the gravity data, and so the level of under-determinedness is increased. Hence, the scope for over-interpretationis increased and there is a greater risk that misinterpretations of thestructure of the upper-crust will lead to erroneous Moho geometries.Due to these drawbacks, this method is best used to resolveproblematic Moho geometry for relatively local areas where upper-crustal structure is well constrained by other means. 3D continent-scalecoverage using this approach is possible, but a major undertaking,given the labour-intensive nature of this approach, and the requirementfor in depth understanding of upper crustal structure.

2.2. Inverse modelling

Inverse modelling of the Moho surrenders the explicit control offorward modelling for an implicit approach, where a computer algo-rithm has control over the resulting geometry (Oldenburg, 1974;Zhdanov, 2002). Inversions have the advantage that the computerdoes most of the work, and large areas can be covered at high resolu-tion within a reasonable time. In addition inversions are repeatableand do not rely as much on subjective interpreter bias, although sig-nificant decisions must still be made to drive the inversion towardsa believable result. To avoid nonsensical results, the changes madeduring inversion must be controlled by some regularisation proce-dure, the exact choice of which can lead to dramatically differentmodel results (Zhdanov, 2002).

Themost straightforward approach tomodelling theMoho is to invertfor the geometry of a single layer-boundary with constant density con-trast (Moritz, 1990; Oldenburg, 1974; Parker, 1972; Vening-Meinesz,1931). This approach has the advantage that for a given density contrastit permits a unique solution, notwithstanding the imperfectness of theinput data (Parker, 1972). The application in the wavenumber domain,in either spatial or spherical coordinates (Moritz, 1990) is also exception-ally fast. Alternatively, a similar approach can be applied in the space do-main (Barbosa and Silva, 1994; Barbosa et al., 1999; Bott, 1960), althoughmore computationally intensive, space domain models are more readilyconstrained by known pierce-points (e.g. seismicMoho depth estimates)and geometrical changes can be constrained by weightings (Barbosaet al., 1997). These surfacemodellingmethods are limited by the inherentassumption that the geometry of theMoho is responsible for all observedgravity anomalies. Hence, for these inversions to produce a reasonableMoho geometry, great caremust be taken to filter the gravity data appro-priately. Typically, the gravity effect of topography is removed byBouguer reduction, and the gravity effect of known sedimentary basins,ice-sheets, oceans etc. are calculated and removed. In addition, wave-length filtering is applied in an attempt to remove short-wavelengthgravity anomalies that represent upper-crustal features, and alsovery-long wavelength anomalies that represent density structure withinthe mantle (e.g. Braitenberg et al., 2000; Ebbing et al., 2001; Shin et al.,2009; Steffen et al., 2011). In applying such wavelength filtering, thereis a risk that part of the Moho signal is inadvertently removed. There isalso the risk that significant anomalies from the crust or mantle densityvariations are not removed in an attempt to preserve the Moho signal.Additional modifications to this method include the inclusion of gravityanomalies due to thermally induced density variations in the upperman-tle (Alvey et al., 2008), and the spectral correlation of gravity anomalieswith topography (Braun et al., 2007).

An alternative conventional method for gravity inversion is toapply inversion for density within a 3D grid of small cuboidal cells(Li and Oldenburg, 1998). This approach has been extensively appliedto upper crustal problems, however, applications to the lower-crustor uppermost mantle are relatively few (although see Brandmayr etal., 2011; Welford and Hall, 2007; Welford et al., 2010). Despitebeing less popular than the surface-modelling approach, this ap-proach is not without utility, in that it can be used to simultaneously


model density variations within the crust and mantle, and by proxy,model the Moho geometry. Intelligent mesh design and sensibleregularisation procedures permit reasonable resolution of the Mohosurface, whilst also allowing density variations within the crust andmantle to be modelled as part of the same procedure (Welford andHall, 2007; Welford et al., 2010). The full 3D nature of this approach,and the requirement for many cells in the vertical direction, meansthat this method is more computationally intensive than the othermethods, and so applying it at very large scales requires significantcomputational resources and time.

The above approaches to the exploitation of gravity data each pro-vide a valid, although limited, approach. Recent software develop-ments have answered the need for flexible methods that can allowfor changes to both the geometries of model features, and the densitydistribution within them (Fullagar et al., 2008; Guillen et al., 2008;Schmidt et al., 2011) or at least allow the density contrast at this in-terface to be variable (Silva et al., 2006; Sjöberg and Bagherbandi,2011). The latter approach allows variable density contrast acrossthe Moho, as part of a joint inversion scheme. This reduces the sensi-tivity of the model to extreme Moho variations, and allows some in-terpretation of lower crust and uppermost mantle density structure(Sjöberg and Bagherbandi, 2011). In the former category, Aitken(2010) proposed a flexible Moho gravity inversion scheme usingVPmg™ software (Fullagar et al., 2008) that allows for the explicit in-corporation of seismic constraints, allows for precise definition of sur-faces and incorporates modelling of lateral variations in crust and/ormantle densities. Using Geomodeller™ software (Guillen et al.,2008), Schreiber et al (2010) applied a similar concept to modellingof the Moho in the southwestern Alps. The approach of Guillen et al.(2008) involves random changes to the density of cuboidal cells or al-ternatively, the lithologies of cells located at a lithological boundarycan be changed. Allowing the inversion to run for many iterations be-yond where an asymptotic fit to the gravity data was achieved pro-vides an estimate of the probability that an individual cell isoccupied by a particular lithology, and also allows 3D mapping ofthe most probable lithology (Guillen et al., 2008).

The combined density/geometry approach removes some of theoversimplifications present in other inversion schemes (e.g. the as-sumption of constant Moho density contrast), but introduces prob-lems of its own, in particular balancing the trade-off betweenchanges in crust (or mantle) density against changes in Moho depthas a means to reduce the gravity misfit.

2.3. Process oriented approaches

A third approach to gravity Moho modelling involves using aMoho surface that is considered a result of a physical process, or in-deed several. These modelling methods have several advantages.Firstly, the resulting Moho geometry is explicitly linked to a causalphysical process, and is inherently realistic (albeit subject to the nec-essary simplifications). Secondly, the inputs to the modelling processare usually much better known than the Moho geometry, for exam-ple, global topography is known in impressive detail, e.g. ETOPO1(Amante and Eakins, 2009), and the age of the oceanic crust (a prima-ry control on crustal thickness) has also been mapped in detail(Müller et al., 2008).

For the continental crust, the process most commonly used to con-strain the Moho geometry is the principle of crustal isostatic compen-sation either locally, or through regional flexure (e.g. Aitken et al.,2012; Jiang et al., 2004; Karner and Watts, 1983; Karner et al., 2005;Petit and Ebinger, 2000; Stern and Ten Brink, 1989; Vening-Meinesz,1931; Watts, 1994). This can be applied as a forward problem, inwhich case loads are applied to the lithosphere using either a knownflexural rigidity, or by applying a range of values in order to find theone that generates the closest fit to the gravity data (e.g. Aitken et al.,2012; Jiang et al., 2004; Stern and Ten Brink, 1989). Most commonly,



Table 1Initial density and permitted density bounds for each layer of the model.

Layer Initial density(g cm−3)

Minimum density(g cm−3)

Maximum density(g cm−3)

Seawater 1.03 NA NASedimentaryrocks 1

2.40 2.10 2.70

sedimentaryrocks 2

2.50 2.20 2.70

Sedimentaryrocks 3

2.60 2.45 2.70

Upper crust 2.75 2.55 3.00Lower/oceaniccrust

2.95 2.75 3.15

Eclogite 3.10 2.95 3.25Mantle From model NA NA


current topography is used as a load on an elastic plate with uniformflexural rigidity. This basic approach neglects the influence of mantlebuoyancy forces, uncompensated crustal loads, palaeotopographyand load evolution, and spatial and temporal variations in flexuralrigidity. These can be incorporated in a forward model if required(Ferraccioli et al., 2011; Karner et al., 2005), however, they mustfirst be known with reasonable confidence. The chief drawbackwith such an approach is the assumption that the crust is isostatical-ly equilibrated (notwithstanding applied mantle buoyancy forces),and the requirement that crustal isostatic compensation is observedwithin the studied area.

Oceanic regions, in general, do not display crustal isostatic com-pensation, with the thickness and density of the lithospheric mantleand buoyancy forces within the asthenosphere being the dominantisostatic controls (e.g. Zlotnik et al., 2008). Oceanic crustal thick-nesses are remarkably consistent, at approximately 7 km (YongshunJohn, 1992), except in regions where flexure is observed (e.g. Watts,1994), or the crust is otherwise anomalous, for example at oceanicplateaux, at fracture zones and at spreading ridges (Mooney et al.,1998). For these anomalous regions, gravity modelling has a role inunderstanding crustal thickness (Davy and Wood, 1994; Gladczenkoet al., 1997; Grobys et al., 2008; Pim et al., 2008; Watts, 1994). An ad-ditional consideration is that gravity anomalies in oceanic regions,and also rifted continental margins, contain a significant gravity con-tribution from thermal variations in the mantle. These thermal effectscan be explicitly removed from the gravity data prior to modelling(Alvey et al., 2008), however, a more sophisticated approach hasbeen developed that simultaneously inverts for crustal thickness,stretching factor and the mantle temperature field (Chappell andKusznir, 2008).

Finally, many geodynamic numerical modelling algorithms permitthe modelling of geodynamic scenarios which can be compared withreal-world observables such as topography, basin evolution,plate-motion, the observed stress field, gravity anomalies and thegeoid. Examples of such modelling includes the modelling of colli-sional orogens (Jiménez-Munt et al., 2005; Tang and Chemenda,2000), slab subduction processses (Hashimoto et al., 2008; Marottaet al., 2006), continental rifting processes (Elesin et al., 2010; Lesneet al., 2000), and the thickness of oceanic lithosphere (Zlotnik et al.,2008). These methods have the capability to generate models ofMoho depth that are fully consistent with a range of physical process-es, including those between which complex feedbacks are observed.Currently, these methods are somewhat limited in their abilities,and have not yet been applied to broad-scale Moho mapping in 3Dfor “real” examples. However, suchmethods may be the future of pro-cess oriented modelling to understand the nature and geometry ofthe Earth's Moho, and gravity data are likely to be a principal con-straint on such modelling.

The chief drawback to process oriented methods is that they areunable to realise Moho geometries that are inconsistent with the pro-cess model applied. Despite this, a process-oriented approach hasmuch utility in regions where Moho geometry is dominated by aphysical process for which the main parameters are known. This re-quirement for the processes to be both well understood and thesame throughout the area means that process-oriented modelling ismost useful at the regional-scale.

In summary, each of these techniques has particular situations towhich it is well suited, and other situations where it is probably un-suitable. All of these approaches suffer from the fundamental prob-lem that a gravity anomaly of appropriate wavelength to representthe Moho could, in fact, be due to anomalous density in the crust orthe mantle. Unless two of these parameters are known with reason-able certainty, resolving the other is a matter of high uncertainty.Hence, unless nature is simplified considerably, the models arehighly underdetermined and many solutions may be found thatwill fit the data.


2.4. Methods used in this study

This study largely follows the method of Aitken (2010), but withsome modifications to both the initial model and the method used.The initial model follows the same principles and uses similardatasets; however, some important differences are included. Firstly,the model covers a much larger area, so as to include more of the Aus-tralian continent and to avoid the influence of edge-effects, which canextend up to 500 km from the edge of the model (±0.5 mGal toler-ance). The topographic model is now ETOPO1 (Amante and Eakins,2009), which provides consistent and seamless coverage across thewhole region. The cell-size was increased to 20 km (from 15 km),however, because sensitivity to Moho geometry is on wavelengthsof ~100 km or more, this affects only the imaging of crustal densityvariations.

Sedimentary basins are a major limiting factor on the Moho inver-sion process. This is because they are close to the surface, and thuspossess high-amplitude gravity signals, but are often broad and so itis difficult to remove their gravity signal by wavelength filtering.The basin model used in the study of Aitken (2010) was based onthe continent-scale SEEBASE compilation of sedimentary basin thick-nesses (Frogtech pty ltd, 2005). We follow that approach here, addingsediment thicknesses from Laske and Masters (1997) in areas notcovered by SEEBASE.

Although it captures sedimentary basin thickness in reasonabledetail, the SEEBASE model does not include any density information.The constant density model used by Aitken (2010) was unrealisticin that it failed to capture the increasing density of sedimentaryrocks with depth, leading to irresolvable misfits over the deepest ba-sins. Ideally, density information for these basins would be derivedfrom a combination of density measurements, and seismic velocityinformation, as is routinely undertaken in well-sampled basins.Such data are not available for Australia at the continent scale, andso we seek to find a model that is simple but appropriate. Basins aresubdivided into three layers: an upper layer with a maximum thick-ness of 2 km, and an initial density of 2.4 g cm−3, a middle layerwith initial density of 2.5 g cm−3 has a maximum thickness of4 km. A lower layer with initial density of 2.6 g cm−3 occupies the re-mainder of the basin thickness. Initial, maximum and minimum den-sities (Table 1) are derived for each layer by converting typicalseismic velocities using Gardner's rule (Gardner et al., 1974; Telfordet al., 1990). The inclusion of sedimentary basin density as a variablein inversion means that the assumptions made (excepting the broadupper and lower limits on density) are not imposed too strongly,and variations form part of the uncertainty analysis of the inversionprocess.

The internal structure of the crystalline crust is also important, al-though less so than the sediment thickness. Of course, seismic studiesthat resolve the Moho must also resolve the crust, and this is oftenachieved in impressive detail. Unlike the Moho, which is a common




feature of all of the Earth's crust, intra-crustal structuring is highlyvariable at the regional scale, and a boundary within one tectonicblock need not necessarily have a correlative in an adjacent block.This regional variability, allied with the variability in resolvingpower from the different methods used, mean that it is difficult togenerate a continent-wide map of intra-crustal structure that is sim-ple enough to be included in this sort of analysis. Thus, we again seekto find a model that is simple but reasonable. A three-layer crystallinecrust is used, with an upper to lower crust boundary at half the crustalthickness, or at 20 km depth where Moho depth is greater than40 km. As indicated by Aitken (2010), an extremely dense “eclogite”layer is required below ca. 40 km depth to achieve a reasonable grav-ity fit for Australia. The effects of deviations of this assumed crustalstructure from reality are potentially significant. However, these aremitigated by the reduced sensitivity of gravity to these intra-crustallayers due to the smaller density contrasts compared to the Moho.As with the sedimentary basin density, the inclusion of the thicknessof these layers and their densities as inversion parameters allows acertain degree of freedom, and the assessment of variability underdifferent inversion conditions to be assessed.

The Moho surface is of course defined by much more seismic in-formation than in Aitken (2010), however the model extends well be-yond the limits of these data. Two areas were defined — a centralregion of “constrained” crust bounded by the outermost seismic con-straints, and a peripheral region of “unconstrained” crust. Moho ge-ometry in the latter was defined by a local Airy isostatic model(Eq. (1)) based on the ETOPO1 surface, represented as a grid with50 km cell size. This cell size captures the main topographic trendswithout being over-sensitive to local topography, and captures thewavelengths to which gravity inversion is sensitive (>100 km). In-side the constrained region, control nodes were manually added tothe seismic Moho estimates to provide more consistent resolution.The same isostatic model (Eq. (1)) was used to define Moho depthat these points. The resulting Moho surface is shown in Fig. 1a.

M ¼ M0−ρc−1ρm−ρc

� T1−T0ð Þ þ ρc

ρm−ρc� T2

� �ð1Þ

M is Moho depth, M0 is the reference Moho elevation, ρc and ρm arecrust and mantle densities respectively, T1 is bathymetric relief and T2is topographic relief T0 is the reference topographic elevation. In thiscase, M0=−14000, T0=−5000, ρc=2.8 g cm−3 and ρm=3.3 g cm−3. The crustal thickness returned at some ocean trenchesrby this method was less than 6 km. In these cases, crustal thicknesswas set at 6 km. Topography in the area is generally low and flat, andit was not deemed necessary to limit maximum crustal thicknesses ina similar way.

The “hard” Moho constraints used by Aitken (2010) are inappro-priate given the existence of seismic uncertainties and the often tran-sitional nature of the crust–mantle boundary. This resulted in severalartefacts around seismic estimates (Aitken, 2010). In this work, weapply a different approach, using the uncertainties in the seismicdata as estimated by Kennett et al (2011) to generate upper andlower bounds for the Moho surface in the vicinity of seismic esti-mates. In each case, the weighting was translated to an uncertaintyusing the formula U=12−10∗w, where U is uncertainty, and w isthe weighting. Weightings were assigned on the basis of the type ofdata (i.e. refraction, receiver function, reflection), and in the case ofreceiver functions and reflection lines, inversion quality and pickquality respectively (Kennett et al., 2011). Control nodes were notconstrained. This error was symmetrically applied to the Moho sur-face (M±U), thus permitting both upward and downward move-ments, and so takes account of random seismic uncertainties, butdoes not account for the transitional nature of the crust–mantleboundary, as the seismic picks of Kennett et al. (2011) were placed


at the base of the transition zone, and may be systematically deeperthan the Moho surface that is most consistent with the gravity data.

The mantle occupies the space from the base of the crust to thebase of the model at 75 km depth, and we treat the mantle as alayer with laterally variable density, but no vertical density variations.The mantle-density grid was updated from that of Aitken (2010)(Fig 1b). The same temperature calculation and model (Goes et al.,2005) was used, but the definition of lithospheric age was modifiedto better represent the age of the lithosphere at depth, rather thanthe surface, following the interpretation of Kennett et al (2011). Man-tle densities prior to the temperature perturbation were as follows:Archaean lithosphere (type 1) and Archaean to Palaeoproterozoiclithosphere (type 2) were assigned a density of 3.31 g cm−3;Palaeo-to-Mesoproterozoic lithosphere (type 3) and Neoproterozoiclithosphere (type 4) were assigned a density of 3.35 g cm−3, Phaner-ozoic lithosphere (type 5) and transitional lithosphere (type 6) wereassigned a density of 3.36 g cm−3, Oceanic lithosphere (type 7) wasassigned a density of 3.39 g cm−3. These initial densities were thenperturbed by the mantle temperature model (see Aitken, 2010 forthe method). There is some subjectivity in the assignment both of ini-tial densities and the age of the lithosphere, especially for regionswhere old lithospheric mantle may have been rejuvenated duringsome younger tectonic event. As with the other density informationin the model, errors in this model of mantle density will influencethe Moho geometry in the final result. Total dynamic range in mantledensity is less than 0.1 g cm−3. As this is a fixed parameter for the in-versions, errors in this grid do not form part of the uncertaintyanalysis.

These changes to the initial model were accompanied by changes tothe method. The combination of terrestrial/satellite derived free-airgravity data used in MoGGIE is replaced by the gravity disturbance de-rived from the EGM 2008 model. These data provide wider coverageand a consistent model across the whole modelled region. The modelused was calculated at the Earth surface on a 0.1° by 0.1° grid.Very-longwavelength gravity trends relating to deep-mantle processeswere removed by subtracting a separately calculated version of EGM2008 that is complete only to degree 10. This process is consideredequivalent to a high-pass filter with an approximate wavelength of2300 km. This high-pass filtered data was then re-gridded at 20 kmspacing and upward continued by 10 km, so as to match the resolutionof the model whilst avoiding short-wavelength content that resultsfrom the steep topographic gradients at the edges of cells (Fig 1c).

The most important change to the method is the management ofthe density/geometry trade-off. We recall that the method used em-ploys alternating iterations in which the density within prisms ismodified, then the geometry of layer boundaries, then density againand so on (Aitken, 2010). As well as the hard constraints listedabove, the degree to which changes to density and geometry are per-mitted are controlled by per-iteration constraints, which must be im-posed by the user. Previously a single combination was used (Aitken,2010); however for this work, we utilise a range of values for each pa-rameter, and so sample the range of solutions possible given the ini-tial conditions of the model.

From these results, the mean and standard deviation of the Mohogeometries can be calculated, giving an indication of the most likelymodel result and the variability of the results. It is important to notethat, although the variability is highly sensitive to the level of con-straint from seismic and gravity data, it is not exactly the same asan error estimate. Low-variability may indicate low-sensitivity, ratherthan low uncertainty. Furthermore, the variability is subject to thelimitations of the initial model, which may fail to include aspects ofthe true crustal structure (e.g. a thick mafic underplate) or mayhave errors in features that are fixed (e.g. basin thickness or mantledensity). These flaws in the initial model may drive the inversion to-wards a solution that does not represent true crustal structure, al-though variability may be low.



a

b c


Please cite this article as: Aitken, A.R.A., et al., Australia's Moho: A test of the usefulness of gravity modelling for the determination of Mohodepth, Tectonophysics (2012), doi:10.1016/j.tecto.2012.06.049

image of Fig.�1


0

10

20

30

40

50

60

70

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

density only

density - 0.025 g/cm-3

geometry - 25%

last good iteration

geometry only

all other models

RM

S M

isfit

(m

Gal

)

Iteration

Fig. 2. Misfit curves for each of the accepted models. Rejected models are not shown,except for the density only and geometry only models. The thicker black line indicatesthe model selected for the final density inversion. Dots indicate, for each inversion, thelast good iteration before the onset of instability.


3. Modelling results

In total, 23 inversions were run with a range of density and geom-etry constraints. Permitted per iteration (PPI) density changes rangedfrom 0 g cm−3 to 0.15 g cm−3 and PPI geometry changes rangedfrom 0% to 50%. Inversions were run for 20 iterations, however, anal-ysis of the misfit curves indicates initial reduction in the misfit, beforethe onset of instability, typically occurring around iteration 8–10(Fig. 2). The onset of instability was identified by large increases inthe misfit (>0.25 mGal RMS), and also the onset of high misfit vari-ability in the subsequent iterations. A single poor iteration was toler-ated, provided subsequent iterations were stable. The iteration priorto instability was considered the final result of the inversion. Of the23 inversion results, 12 were rejected. With a PPI density change of0.025 g cm−3, models with PPI geometry change of less than 5%were rejected on the basis that they did not generate an adequatefit to the data (RMS misfit>10 mGal). The model with PPI geometrychange of 50% was also rejected because it generated an identical re-sult to the model with 40% PPI geometry change. With 25% PPI geom-etry change, inversions with PPI density change of less than0.020 g cm−3 were rejected because they did not generate a believ-able Moho geometry. Models with PPI density change of greaterthan 0.06 g cm−3 were also rejected because the results were identi-cal to the result with a PPI density change of 0.06 g cm−3. Theremaining 11 models were used to calculate the mean Moho surfaceand its standard deviation.

The mean result (Fig 3) is considered to represent the most likelyMoho geometry given the initial conditions. This result shows thatthe initial model is largely validated, that dramatic changes are few,and that changes typically are fairly short-wavelength. Changesfrom the initial model are, naturally, focused within the less-well con-strained parts of the model, namely offshore regions and withincentral-western Australia.

The standard deviation of these results (Fig 4) shows that modelMoho variability is generally low offshore, and in Tasmania. These resultsindicate that the result in these regions is robust under different modelconditions, and indicates that seismicMoho depth estimates are compat-ible with the crustal density model. With some exceptions, seismic

Fig. 1. a) The initial Moho surface showing the location of seismic estimates (black dots) coMantle density. Black lines indicate the boundaries of differentmantle “age-types” labelled as fo4—Neoproterozoic, 5— Phanerozoic, 6— Transitional, 7—Oceanic. c) Gravity disturbance at 10to mitigate edge effects.


estimates merge smoothly with the gravity-modelling results in theselow-variability areas. Model variability is highly changeable onshore.The largest variability is observed in an ENE trending band in central Aus-tralia, roughly corresponding with the Albany-Fraser, Musgrave andArunta provinces. These regions possess thick to very thick crust(>45 km) but gravity anomalies are moderately low to very high, indi-cating high density crust (or perhaps mantle) that may not be well de-scribed by the initial model. Variability is much lower throughout mostof eastern and western Australia, although areas of high variabilityexist. This indicates that either only minor changes to the initial modelwere required to satisfy the gravity field, or that these solutions werewell constrained by either the seismic constraints, or by the gravitymodelling process. It is interesting to note that, despite low seismic esti-mate density, the area between longitude 135°E to 140°E and latitude−22°S to 30°S shows very low variability in comparisonwith similarly con-strained crust immediately to thewest (i.e. longitude 125°E to 135°E andlatitude −20°S to 30°S). This indicates that the robustness of modellingresults is not solely related to the density of seismic data, but also reflectsthe constraints inherent to the gravity modelling process.

Intra-crustal boundaries were analysed in a similar fashion. Theupper-crust/lower crust boundary shows generally quite low varia-tions, with a standard deviation of less than a kilometre even in theleast well constrained areas, and significantly less throughout muchof the model (Supplementary Fig. 1). This variability is most similarto the crustal density anomaly map (Fig 6), indicating that in thismodel the upper crust behaves differently to the lower crust. The var-iability of the top of the eclogite layer quite closely matches the Mohovariability in both magnitude and pattern, with the exception thatcentral Australia shows less variability (Supplementary Fig. 2).

Of all these models, a single model was selected with which torepresent crustal density variations. The model with permitted ge-ometry changes of 25% and density changes of 0.025 g cm−3 wasselected because it achieved the lowest misfit before instability(Fig. 2), and its geometry was close to the mean Moho, with an av-erage discrepancy of 275 m. After 40 further iterations,density-only inversion achieved a final RMS misfit of 3.31 mGal(Fig 5). Residual misfit is concentrated along the continent's south-ern margin and in the Bonaparte Gulf and Woodlark basins. The ex-istence of serpentinised peridotite has been inferred beneath thesouthern margin of Australia in the Great Australian Bight (e.g.Direen et al., 2011) but is not included in the model. The BonaparteGulf has a combination of thin crust and thick sediments that doesnot permit a fit to the gravity data. The residual anomaly in theWoodlark Basin likely reflects un-modelled gravity anomaliesfrom small-scale thermal anomalies in the mantle.

The crustal density anomaly is a vertically averaged record of thepercentage difference between observed density, and that of a modelwith the same geometry, but populated by initial density values(Table 1). This image shows that much of the continent is occupied byrelatively dense crust, much of it focused in areas of thick crust (Fig 6).In some cases, these crustal density anomaliesmirror themargins of tec-tonic boundaries (e.g. in eastern Australia) but in others, they transgressmajor boundaries (e.g. in central-western Australia). Oceanic crust isgenerally of reduced density relative to the initial model.

4. Discussion

4.1. Australia's Moho

This model generated few fundamental changes to the Moho ge-ometries generated in the work of Kennett et al. (2011), and

ntrol points (white dots) and the boundary of the “constrained” region (black line). b)llows— 1—Archaean, 2—Archaean to Palaeoproterozoic, 3— Palaeo-to-Mesoproterozoic,km above the earth's surface. The gravity disturbance grid is smaller than themodel area


image of Fig.�2


Fig. 3.MeanMoho elevation, taken to represent themost likely geometry of theMoho. KC—Kimberley Craton, NAC—North Australian Craton, CYP— Cape York Peninsula,WS—WillowraSuture, GSD—Great SandyDesert, PC— Pilbara Craton, CO— CapricornOrogen, YC—Yilgarn Craton,WMP—westMusgrave Province,MSZ—Mundrabilla Shear Zone, GC—Gawler Craton,CC— Curnamona Craton, LO — Lachlan Orogen, NEO— New England Orogen.


Salmon et al. (this issue), and readers are referred to those studies,and also Aitken (2010) for more in-depth discussion of the main fea-tures. Briefly, these Moho maps show moderately thin (~35 km)crust beneath the Archean Yilgarn and Pilbara cratons, except theSouthwest Zone of the Yilgarn Craton, where the crust is thicker(40–45 km). The intervening Capricorn Orogen has thickened crust.The Palaeoproterozoic North Australian Craton typically has thick tovery-thick crust (>40 km), with the exception of the Kimberley Cra-ton, where the crust is thinner (35–40 km). Central Australia is dom-inated by thick crust (>45 km) likely reflecting crustal thickeningduring intraplate orogenesis, which also caused localised Moho up-lifts of up to 15 km (Aitken et al., 2009; Goleby et al., 1989; Korschand Kositcin, 2010; Korsch et al., 1998). Magmatic underplating dueto theWarakurna Large Igneous Province is a further possibility to ex-plain the thick crust here (Aitken, 2010). The Gawler and CurnamonaCratons are characterised by moderately thick crust (40–45 km).There is a rapid transition to the much thinner crust of eastern Aus-tralia, which is generally between 25 and 40 km thick. Structuringin Eastern Australia follows the N-S tectonic fabric, with several dis-crete blocks (e.g. New England Orogen) defined. Tasmania and theBass Strait have the thinnest continental crust (20–25 km), reflectingcrustal thinning during Gondwana breakup.

Changes that have been made to the initial model generally in-volve the "sharpening" of Moho steps associated with major crustalboundaries — for example the boundary between the New EnglandOrogen and Lachlan Orogen is better defined, and resolution of theWillowra Suture is enhanced. In other areas, short wavelength fea-tures are introduced, for example the Mundrabilla Shear Zone insouthern Australia. In a few areas (for example the Cape York Penin-sula, and beneath the Great Sandy Desert), the gravity inversion pro-cess has generated significant changes across large areas with sparseseismic controls.


4.2. The capabilities and limitations of gravity inversion for Mohomodelling

As expected, the use of gravity inversion has increased the resolu-tion of short-wavelength Moho features, and has provided good cov-erage in offshore regions, and regions lacking seismic data. With thismodel, the changes are not so dramatic as with Aitken (2010), pri-marily due to the much denser network of seismic estimates, butalso the averaging process applied, which smoothes out any extremevalues from individual results.

The results of this work indicate that gravity inversion can providehighly consistent and reasonably accurate models of the Moho geom-etry, even where seismic constraints are relatively sparse. The major-ity of the model shows a standard deviation of less than 1.5 km,equivalent to an error (3σ) of less than 5 km. High-variability areashowever, have standard deviations upwards of 2.5–3 km, equivalentto an error of 7.5–9 km. These results are consistent with othermodels, where gravity estimates are usually within ~5 km of seismicestimates (e.g. Braun et al., 2007; Steffen et al., 2011) but are occa-sionally more than 10 km in error. Thus, gravity inversion methodscan have comparable accuracy to the less-reliable seismic methods,but with a greater risk of highly erroneous results.

Inmany instances, these high-error regions occur where the densitystructure of the crust is “anomalous” in that it is either thick and highdensity (e.g. due to a mafic underplate that thickens the crust) or thinand low density (e.g. due to deep sedimentary basins over thinnedcrust). Because the gravity responses from the Moho and the crustcancel-out to some degree, such crustal structures cannot be spontane-ously generated by gravity inversion. For a reliable model in such areas,a firm knowledge of either Moho geometry or crustal density is re-quired. Mantle densities of course are also influential, but less so dueto the narrower range of permissible densities. Forward modelling or


image of Fig.�3


Fig. 4. The standard deviation of the modelling results, indicating the variability of gravity inversion results from the 11 accepted models. AFP — Albany Fraser Province, MP —

Musgrave Province, AP — Arunta Province.


process-orientedmodellingmay provemore successful in these regionsdue to the greater degree of control on Moho geometry.

This leads to the twin requirements that uncertainty is reduced asfar as possible, and that the remaining uncertainty is characterised, soas to guide future efforts. Uncertainty can be directly reduced by intro-ducing new constraints from seismic Moho estimates. Uncertainty canalso be reduced by better resolving the density structure of the crustand mantle. Densities in the upper crust can be derived from samplingprograms or from higher resolution modelling, with appropriately fil-tered gravity data. Seismic velocities of the lower crust and uppermostmantle are harder to constrain, but can be obtained from models ofseismic-velocity or, less directly, from more robust models of theirage, composition and temperature.

Uncertainty will never be eliminated, and should be investigated byanalysing the sensitivity of the modelling to at least the assumptionsmade in the modelling process, but perhaps also to the uncertaintiesin a-priori constraints. The approach used here describes the sensitivityof results to a keymodelling parameter – the crustal density/crustal ge-ometry trade-off – and provides a means to identify regions where themodel requires either more data, or a more appropriate modelling ap-proach. The approach defined above does not include tests of the ro-bustness of assumptions made in the initial model, for example,assuming that mantle density and sedimentary basin thicknesses areknown could introduce erroneous (but low-variability) results wherethesemodels are in error. In an extension to the abovemethod, sensitiv-ity to such parameters could be included within the sensitivity analysiswith little difficulty, so long as the range of possibilities is known.

4.3. The influence of seismic constraint

This model has used a much greater amount of seismic informa-tion (>1200 individual estimates compared to 230) than the modelof Aitken (2010), and provides a compelling opportunity to test theinfluence of this increase of seismic information on the model results.


We also can compare these models to a recent global gravity-onlyMoho model derived from GOCE data (Sampietro, 2011).

Both seismic-gravity models have imaged the overall geometry ofAustralia's Moho, including moderate crustal thickness beneath west-ern Australia's cratons, thick crust beneath the Proterozoic terranes ofcentral and northern Australia, and predominantly thin crust beneatheastern Australia. Each also defines several short-wavelengthMoho fea-tures, including major fault zones (e.g. Mundrabilla Shear Zone), Mohouplifts (e.g. central Australia), the boundaries between crustal blocks(e.g. New England Orogen) and the continental margins. The ampli-tudes of these features are typically smaller in the newmodel, reflectingthe tighter per-iteration constraints on the Moho surface. The averagediscrepancy is 2.4 km, with the greatest discrepancies (up to 25 km)in regions of thickened crust that were unconstrained in the Aitken(2010) model (Supplementary Fig. 3). Large discrepancies occur pri-marily where the crust is dense (e.g. the northern Gawler Craton, thewest Musgrave province).

Although the aims, resolution and scope to include local informationare different for the global gravity-only model, this model captures onlysome of the features of Australia's Moho. These include thicker crust inthewestern two-thirds of the continent, the structure of central Australia,and the difference between the oceans and the continents. However, itcontains many features that are known to be erroneous, such as thincrust (28 km) beneath the Mt Isa region, which is known to have thickcrust (50 km) from seismic refraction studies (Maccready et al., 1998).Overall the discrepancy between these models averages 3.5 km, andthis is generally low (b2 km ) in the oceanic regions and high(>6 km) on the continent (Supplementary Fig. 4).

The level of detail in each of the seismic-gravity models, and alsoKennett et al. (2011) and Salmon et al. (this issue) is similar, althoughthe newmodel has a more irregular surface due to the very high den-sity of seismic estimates in some areas (e.g. southeast Australia). Allof these models show greatly improved definition of Australia'sMoho over the model of Collins et al. (2003). This prompts a


image of Fig.�4


Fig. 6. Crustal density variations across the Australian region. Crustal density represents the vertically averaged percentage difference from the initial densities (Table 1) for thesame model geometry.

Fig. 5. The residual gravity disturbance of this final model. Positive anomalies indicate insufficient mass in the subsurface, while lows indicate excess mass. BG— Bonaparte Gulf, GAB—

Great Australian Bight, WB — Woodlark Basin.


Please cite this article as: Aitken, A.R.A., et al., Australia's Moho: A test of the usefulness of gravity modelling for the determination of Mohodepth, Tectonophysics (2012), doi:10.1016/j.tecto.2012.06.049

image of Fig�6

image of Fig.�5



discussion on the influence of the level of seismic constraint on theusefulness of gravity inversion to define Moho geometry.

Gravity inversion with low-levels of seismic constraint (e.g. centralAustralia) is useful, and may provide an understanding of likely Mohogeometry where otherwise there would be none. Although the resultsof gravity inversion generally compare quite well with seismic con-straints there is the potential for large errors, especially where thecrust or mantle density is anomalous. Unless the potential for error isquantified, this inhibits the usefulness of themethod. In such situations,it may be more effective to utilise a process-oriented method, or per-haps constrained forward modelling, to limit the range of potentialMoho geometries. Gravity inversionwithhigh levels of seismic constrainte.g. south-east Australia, post Kennett et al. (2011) also have limited use-fulness as the Moho surface is largely known, and there is little scope fornew structure to be imaged. In these situations, gravity inversion can stillprovide revealing insights into the nature of the lower-crust and upper-most mantle through modelling their densities.

The most useful areas in which to apply gravity inversion are thosewith a moderate level of well-spaced seismic estimates e.g. easternAustralia as in Collins et al. (2003), or areas where crustal structure ina low-data density region cannot be reasonably predicted from seismicdata in an adjacent well constrained region e.g. at continental margins.In each of these cases, gravity inversion can produce robust models ofthe Moho that genuinely add to our knowledge.

5. Conclusion

The studies discussed in this manuscript demonstrate the value ofgravity modelling to generate models of the Moho. In particular, theyshow that gravity inversion is capable of reasonable accuracy as ameans of extending coverage of Moho models beyond the limits ofseismic constraints and to interpolate between gaps in coverage.They also highlight the inability of gravity inversion to model crustalstructure reliably in areas to which the method is ill-suited (e.g. thick,dense crust or thin, low-density crust), unless the structure is definedby seismic methods. In such areas, more detailed gravity modellingusing strong a-priori constraints from non-seismic data may bemore effective.

Applying a broad range of inversion parameters to generate a var-iability estimate is effective in identifying areas where the result isnot well constrained by either seismic or gravity methods. Althoughthere is a general relationship with seismic coverage, the least con-strained areas are not necessarily those with the lowest seismic esti-mate density. Therefore, the variability map contributes to theoptimisation of future seismic efforts to map Moho geometry.

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.tecto.2012.06.049.

References

Aitken, A.R.A., 2010. Moho geometry gravity inversion experiment (MoGGIE): a refinedmodel of the Australian Moho, and its tectonic and isostatic implications. Earth andPlanetary Science Letters 297, 71–83.

Aitken, A.R.A., Betts, P.G., Weinberg, R.F., Gray, D., 2009. Constrained potential fieldmodeling of the crustal architecture of the Musgrave Province in central Australia:evidence for lithospheric strengthening due to crust–mantle boundary uplift. Jour-nal of Geophysical Research B: Solid Earth 114.

Aitken, A.R.A., Wilson, G.S., Jordan, T., Tinto, K., Blakemore, H., 2012. Flexural controlson late Neogene basin evolution in southern McMurdo Sound. Global and Plane-tary Change 80–81, 99–112.

Alvey, A., Gaina, C., Kusznir, N.J., Torsvik, T.H., 2008. Integrated crustal thickness map-ping and plate reconstructions for the high Arctic. Earth and Planetary Science Let-ters 274, 310–321.

Amante, C., Eakins, B.W., 2009. ETOPO1 1 arc-minute global relief model: proce-dures. Data Sources and Analysis. NOAA Technical Memorandum NESDISNGDC-21. (19 pp., March 2009).

Barbosa, V.C.F., Silva, J.B.C., 1994. Generalized compact gravity inversion. Geophysics59, 57–68.

Barbosa, V.C.F., Silva, J.B.C., Medeiros, W.E., 1997. Gravity inversion of basement reliefusing approximate equality constraints on depths. Geophysics 62, 1745–1757.


Barbosa, V.C.F., Silva, J.B.C., Medeiros, W.E., 1999. Stable inversion of gravity anomaliesof sedimentary basins with nonsmooth basement reliefs and arbitrary density con-trast variations. Geophysics 64, 754–764.

Beaumont, C., Kooi, H., Willett, S., 2000. Coupled tectonic-surface process models withapplications to rifted margins and collisional orogens. In: Summerfield, M.A. (Ed.),Geomorphology and Global Tectonics 29–55.

Begg, G.C., Hronsky, J.A.M., Arndt, N.T., Griffin, W.L., O'Reilly, S.Y., Hayward, N., 2010.Lithospheric, cratonic, and geodynamic setting of Ni–Cu–PGE sulfide deposits. Eco-nomic Geology 105, 1057–1070.

Behr, Y., Townend, J., Bannister, S., Savage, M.K., 2010. Shear velocity structure of theNorthland Peninsula,New Zealand, inferred from ambient noise correlations. Jour-nal of Geophysical Research B: Solid Earth 115.

Bierlein, F.P., Groves, D.I., Goldfarb, R.J., Dubé, B., 2006. Lithospheric controls on the formationof provinces hosting giant orogenic gold deposits. Mineralium Deposita 40, 874–886.

Bott, M.H.P., 1960. The use of rapid digital computing methods for direct gravity inter-pretation of sedimentary basins. Geophysical Journal of the Royal Astronomical So-ciety 3, 63–67.

Braitenberg, C., Zadro, M., Fang, J., Wang, Y., Hsu, H.T., 2000. Gravity inversion in Qing-hai–Tibet plateau. Physics and chemistry of the earth, part A. Solid Earth and Geod-esy 25, 381–386.

Brandmayr, E., Marson, I., Romanelli, F., Panza, G.F., 2011. Lithosphere density model inItaly: no hint for slab pull. Terra Nova 23, 292–299.

Braun, A., Kim, H.R., Csatho, B., von Frese, R.R.B., 2007. Gravity-inferred crustal thick-ness of Greenland. Earth and Planetary Science Letters 262, 138–158.

Chappell, A.R., Kusznir, N.J., 2008. Three-dimensional gravity inversion for Moho depthat rifted continental margins incorporating a lithosphere thermal gravity anomalycorrection. Geophysical Journal International 174, 1–13.

Collins, C.D.N., Drummond, B.J., Nicoll, M.G., 2003. Crustal thickness patterns in theAustralian continent. Geological Society of America Special Papers 372,121–128.

Davy, B., Wood, R., 1994. Gravity and magnetic modelling of the Hikurangi Plateau. Ma-rine Geology 118, 139–151.

Direen, N.G., Stagg, H.M.J., Symonds, P.A., Colwell, J.B., 2011. Dominant symmetry of aconjugate southern Australian and East Antarctic magma-poor rifted margin seg-ment. Geochemistry, Geophysics, Geosystems 12.

Ebbing, J., Braitenberg, C., Gotze, H.J., 2001. Forward and inverse modelling of gravityrevealing insight into crustal structures of the Eastern Alps. Tectonophysics 337,191–208.

Elesin, Y., Gerya, T., Artemieva, I.M., Thybo, H., 2010. Samovar: a thermomechanicalcode for modeling of geodynamic processes in the lithosphere-application tobasin evolution. Arabian Journal of Geosciences 3, 477–497.

Ferraccioli, F., Finn, C.A., Jordan, T.A., Bell, R.E., Anderson, L.M., Damaske, D., 2011. EastAntarctic rifting triggers uplift of the Gamburtsev Mountains. Nature 479,388–392.

Flesch, L.M., Haines, A.J., Holt, W.E., 2001. Dynamics of the India-Eurasia collision zone.Journal of Geophysical Research B: Solid Earth 106, 16435–16460.

Frogtech pty ltd, 2005. OZ SEEBASE™ Study 2005, Public Domain Report to Shell Devel-opment Australia by FrOG. Tech Pty Ltd.

Fullagar, P.K., Pears, G.A., McMonnies, B., 2008. Constrained inversion of geologic sur-faces — pushing the boundaries. The Leading Edge 27, 98–105.

Gardner, G.H.F., Gardner, L.W., Gregory, A.R., 1974. Formation velocity and density —

the diagnostic basis for stratigraphic traps. Geophysics 39, 770–780.Gladczenko, T.P., Coffin, M.F., Eldholm, O., 1997. Crustal structure of the Ontong Java

Plateau: modeling of new gravity and existing seismic data. Journal of GeophysicalResearch B: Solid Earth 102, 22711–22729.

Goes, S., Simons, F.J., Yoshizawa, K., 2005. Seismic constraints on temperature of theAustralian uppermost mantle. Earth and Planetary Science Letters 236, 227–237.

Goleby, B.R., Shaw, R.D., Wright, C., Kennett, B.L.N., Lambeck, K., 1989. Geophysical ev-idence for “thick-skinned” crustal deformation in central Australia. Nature 337,325–337.

Grobys, J.W.G., Gohl, K., Eagles, G., 2008. Quantitative tectonic reconstructions ofZealandia based on crustal thickness estimates. Geochemistry, Geophysics,Geosystems 9.

Guillen, A., Calcagno, P., Courrioux, G., Joly, A., Ledru, P., 2008. Geological modellingfrom field data and geological knowledge. Part II. Modelling validation using grav-ity and magnetic data inversion. Physics of the Earth and Planetary Interiors 171,158–169.

Hackney, R., 2004. Gravity anomalies, crustal structure and isostasy associated with theProterozoic Capricorn Orogen, Western Australia. Precambrian Research 128,219–236.

Hashimoto, C., Sato, T., Matsu'ura, M., 2008. 3-D simulation of steady plate subductionwith tectonic erosion: current crustal uplift and free-air gravity anomaly in north-east Japan. Pure and Applied Geophysics 165, 567–583.

Jiang, X., Jin, Y., McNutt, M.K., 2004. Lithospheric deformation beneath the Altyn Taghand West Kunlun faults from recent gravity surveys. Journal of Geophysical Re-search B: Solid Earth 109.

Jiménez-Munt, I., Garcia-Castellanos, D., Negredo, A.M., Platt, J.P., 2005. Gravitationaland tectonic forces controlling postcollisional deformation and the present-daystress field of the Alps: constraints from numerical modeling. Tectonics 24, 1–15.

Karner, G.D., Watts, A.B., 1983. Gravity anomalies and flexure of the lithosphere atmountain ranges. Journal of Geophysical Research 88, 10449–10477.

Karner, G.D., Studinger, M., Bell, R.E., 2005. Gravity anomalies of sedimentary basinsand their mechanical implications: application to the Ross Sea basins, West Antarc-tica. Earth and Planetary Science Letters 235, 577–596.

Kennett, B.L.N., Salmon, M., Saygin, E., Group, A.W., 2011. AusMoho: the variation ofMoho depth in Australia. Geophysical Journal International 187, 946–958.




Korsch, R.J., Kositcin, N., 2010. GOMA (Gawler Craton-Officer Basin-MusgraveProvince-Amadeus basin) seismic and MT workshop 2010. Geoscience AustraliaRecord 2010/39. Geoscience Australia, Canberra, p. 162.

Korsch, R.J., Goleby, B.R., Leven, J.H., Drummond, B.J., 1998. Crustal architecture of cen-tral Australia based on deep seismic reflection profiling. Tectonophysics 28, 57–69.

Langston, C.A., 1979. Structure under Mount Rainier, Washington, inferred fromteleseismic body waves. Jounal of Geophysical Research 84, 4749–4762.

Laske, G., Masters, G., 1997. A global digital map of sediment thickness. Eos Transac-tions AGU 78, F483.

Lesne, O., Calais, E., Deverchère, J., Chéry, J., Hassani, R., 2000. Dynamics ofintracontinental extension in the north Baikal rift from two-dimensional numeri-cal deformation modeling. Journal of Geophysical Research B: Solid Earth 105,21727–21744.

Li, Y., Oldenburg, D.W., 1998. 3-D inversion of gravity data. Geophysics 63, 109–119.Maccready, T., Goleby, B.R., Goncharov, A., Drummond, B.J., Lister, G.S., 1998. A frame-

work of overprinting orogens based on interpretation of the Mount Isa deep seis-mic transect. Economic Geology 93, 1422–1434.

Marotta, A.M., Spelta, E., Rizzetto, C., 2006. Gravity signature of crustal subductioninferred from numerical modelling. Geophysical Journal International 166,923–938.

McLean, M.A., Morand, V.J., Cayley, R.A., 2010. Gravity and magnetic modelling of crust-al structure in central Victoria: what lies under the Melbourne Zone? AustralianJournal of Earth Sciences 57, 153–173.

Mooney, W.D., Laske, G., Masters, T.G., 1998. CRUST 5.1: a global crustal model at 5°×5.Journal of Geophysical Research B: Solid Earth 103, 727–747.

Moritz, H., 1990. The figure of the earth. Theoretical Geodesy and the Earth's Interior.Wichmann, Karlsruhe.

Müller, R.D., Sdrolias, M., Gaina, C., Roest, W.R., 2008. Age, spreading rates, and spreadingasymmetry of the world's ocean crust. Geochemistry, Geophysics, Geosystems 9.

Naliboff, J.B., Lithgow-Bertelloni, C., Ruff, L.J., de Koker, N., 2012. The effects of litho-spheric thickness and density structure on Earth's stress field. Geophysical JournalInternational 188, 1–17.

Oldenburg, D.W., 1974. Inversion and interpretation of gravity anomalies. Geophysics39, 526–536.

Parker, R.L., 1972. The rapid calculation of potential anomalies. Geophysical Journal ofthe Royal Astronomical Society 31, 447–455.

Pavlis, N.K., Holmes, S.A., Kenyon, S.C., Factor, J.K., 2008. An Earth gravitational modelto Degree 2160: EGM2008. EGU General Assembly, 2008, Vienna, Austria.

Petit, C., Ebinger, C., 2000. Flexure and mechanical behavior of cratonic lithosphere:gravity models of the East African and Baikal rifts. Journal of Geophysical ResearchB: Solid Earth 105, 19151–19162.

Pfiffner, O.A., Ellis, S., Beaumont, C., 2000. Collision tectonics in the Swiss Alps: insightfor geodynamic modeling. Tectonics 19, 1065–1094.

Pim, J., Peirce, C., Watts, A.B., Grevemeyer, I., Krabbenhoeft, A., 2008. Crustal structureand origin of the Cape Verde Rise. Earth and Planetary Science Letters 272,422–428.

Prezzi, C.B., Götze, H.J., Schmidt, S., 2009. 3D density model of the Central Andes. Phys-ics of the Earth and Planetary Interiors 177, 217–234.

Reid, A.B., Ebbing, J., Webb, S.J., 2012. Comment on ‘A crustal thickness map of Africaderived from a global gravity field model using Euler deconvolution’ by GetachewE. Tedla, M. van der Meijde, A. A. Nyblade and F. D. van der Meer. Geophysical Jour-nal International 189, 1217–1222.

Salmon, M.L., Stern, T.A., Savage, M.K., 2011. A major step in the continental Moho andits geodynamic consequences: the Taranaki-Ruapehu line, New Zealand. Geophys-ical Journal International 186, 32–44.

Salmon, M.L., Kennett, B.L.N., Stern, T.A., Aitken, A., this issue. The Moho in Australiaand New Zealand. Tectonophysics.


Sampietro, D., 2011. http://geomatica.como.polimi.it/elab/gemma/global.php.Schmidt, S., Plonka, C., Götze, H.J., Lahmeyer, B., 2011. Hybrid modelling of gravity,

gravity gradients and magnetic fields. Geophysical Prospecting 59, 1046–1051.Schreiber, D., Lardeaux, J.M., Martelet, G., Courrioux, G., Guillen, A., 2010. 3-Dmodelling of

AlpineMohos in Southwestern Alps. Geophysical Journal International 180, 961–975.Shin, Y.H., Shum, C.K., Braitenberg, C., Lee, S.M., Xu, H., Choi, K.S., Baek, J.H., Park, J.U.,

2009. Three-dimensional fold structure of the Tibetan Moho from GRACE gravitydata. Geophysical Research Letters 36.

Silva, J.B.C., Costa, D.C.L., Barbosa, V.C.F., 2006. Gravity inversion of basement relief andestimation of density contrast variation with depth. Geophysics 71, J51–J58.

Sjöberg, L.E., Bagherbandi, M., 2011. A method of estimating the Moho density contrastwith a tentative application of EGM08 and CRUST2.0. Acta Geophysica 59, 502–525.

Steffen, R., Steffen, H., Jentzsch, G., 2011. A three-dimensional Moho depth model forthe Tien Shan from EGM2008 gravity data. Tectonics 30, TC5019.

Stehly, L., Fry, B., Campillo, M., Shapiro, N.M., Guilbert, J., Boschi, L., Giardini, D., 2009.Tomography of the Alpine region from observations of seismic ambient noise. Geo-physical Journal International 178, 338–350.

Stern, T.A., Ten Brink, U.S., 1989. Flexural uplift of the Transantarctic Mountains. Jour-nal of Geophysical Research 94 (10,315-310,330).

Stewart, J.R., Betts, P.G., 2010. Implications for Proterozoic plate margin evolution fromgeophysical analysis and crustal-scale modeling within the western Gawler Craton,Australia. Tectonophysics 483, 151–177.

Studinger, M., Kurinin, R.G., Aleshkova, N.D., Mlller, H., 1997. Power spectra analysis ofgravity data from theWeddell Sea embayment and adjacent areas. Terra Antarctica4, 23–26.

Tang, J.C., Chemenda, A.I., 2000. Numerical modelling of arc-continent collision: appli-cation to Taiwan. Tectonophysics 325, 23–42.

Tedla, G.E., van der Meijde, M., Nyblade, A.A., Meer, F.D., 2011. A crustal thickness mapof Africa derived from a global gravity field model using Euler deconvolution. Geo-physical Journal International 187, 1–9.

Telford, W.M., Geldart, L.P., Sheriff, R.E., 1990. Applied Geophysics Second Edition.Cambridge University Press.

Valera, J.L., Negredo, A.M., Jiménez-Munt, I., 2011. Deep and near-surface consequences ofroot removal by asymmetric continental delamination. Tectonophysics 502, 257–265.

Vening-Meinesz, F.A., 1931. Une Nouvelle Méthode Pour la Réduction IsostatiqueRégionale de L'intensité de la Pesanteur. Bulletin Géodésique 29, 33–51.

Watts, A.B., 1994. Crustal structure, gravity anomalies and flexure of the lithosphere inthe vicinity of the Canary Islands. Geophysical Journal International 119, 648–666.

Welford, J.K., Hall, J., 2007. Crustal structure of the Newfoundland rifted continentalmargin from constrained 3-D gravity inversion. Geophysical Journal International171, 890–908.

Welford, J.K., Shannon, P.M., O'Reilly, B.M., Hall, J., 2010. Lithospheric density variationsand Moho structure of the Irish Atlantic continental margin from constrained 3-Dgravity inversion. Geophysical Journal International 183, 79–95.

Williams, H.A., Betts, P.G., Ailleres, L., Burtt, A., 2010. Characterization of a proposedPalaeoproterozoic suture in the crust beneath the Curnamona Province, Australia.Tectonophysics 485, 122–140.

Yongshun John, C., 1992. Oceanic crustal thickness versus spreading rate. GeophysicalResearch Letters 19, 753–756.

Zhdanov, M.S., 2002. Geophysical Inverse Theory and Regularization Problems.Elsevier, Amsterdam.

Zlotnik, S., Afonso, J.C., Díez, P., Fernández, M., 2008. Small-scale gravitational instabil-ities under the oceans: Implications for the evolution of oceanic lithosphere and itsexpression in geophysical observables. Philosophical Magazine 88, 3197–3217.


http://geomatica.como.polimi.it/elab/gemma/global.php


Documents

Australia's Moho: A test of the usefulness of gravity ...rses.anu.edu.au/~brian/PDF-reprints/Proofs/TECTO-125531.pdf · 2011), resource exploration (e.g. Begg et al., 2010; Bierlein