Petrologically-constrained thermo-chemical modelling of continental upper mantle consistent with elevation, geoid, surface heat flow, seismic surface waves and MT data

摘要

The Earth comprises a single physio-chemical system that we interrogate from its surface or from spacernmaking observations related to various physical and chemical parameters. A change in one of thosernparameters affects many of the others; for example a change in velocity is almost always indicative of arnchange in density, which results in changes to elevation, gravity and geoid observations. Similarly, a changernin oxide chemistry affects almost all physical parameters to a greater or lesser extent. We have nowrndeveloped sophisticated tools to model/invert data in our individual disciplines to such an extent that we arernobtaining high resolution, robust models from our datasets. Thus, the challenges lie at the boundariesrnbetween the geoscientific sub-disciplines. However, in the vast majority of cases the different datasets arernmodelled/inverted independently of each other, and even without considering concomitant data in arnqualitative sense.rnI will present a framework for integrated inversion of geoscientific data to yield thermo-chemical models thatrnare petrologically consistent and constrained using the LitMod approach of Afonso and colleagues [Afonso etrnal., 2008; Fullea et al., 2009; Fullea et al., 2011] as the forward engine, and a delayed rejection adaptivernMetropolis (DRAM) Markov Chain Monte-Carlo (MCMC) algorithm to perform the stochastic inversionrn[Afonso et al., 2013a; Afonso et al., 2013b]. Input data can comprise any combination of elevation, geoid,rnsurface heat flow, Rayleigh and Love dispersion data, and MT data.rnThe basis of LitMod is characterization of the upper mantle in terms of five oxides in the CFMAS systemrn(CaO, FO, MgO, Al_2O_3 and SiO_2) and a thermal structure that is conductive to the lithosphere-asthenospherernboundary (LAB) and convective along the adiabat below the LAB to the 410 km discontinuity. Candidaternsolutions are chosen from four of the five oxides from prior distributions based on their inter-correlations,rnwhere the fifth oxide is given by the required sum to unity. For the crust, candidate solutions are chosen fromrndistributions of crustal layering, velocity and density parameters. (For MT however, currently the crustalrnsolution is fixed based on the high frequency MT data.) Those candidate solutions that fit the data withinrnprescribed error limits are kept, and are used to establish broad posterior distributions from which newrncandidate solutions are chosen.rnExamples will be shown of application of this approach fitting data from the Kaapvaal Craton in South Africa,rnthe Rae Craton in northern Canada, and central Tibet. I will show that the MT data are the mostrndiscriminatory, requiring many millions of candidate solutions to be tested in order to sufficiently establishrnposterior distributions. In particular, the MT data require layered lithosphere, whereas the other data can bernfit with a single lithosphere, and the MT data are particularly sensitive to the depth to the LAB.