DI14A-01 INVITED
CitcomS v3.0 - A compressible thermo-chemical mantle convection code
The Earth's mantle extends ~3000 km from the surface to the core-mantle boundary. The vast pressure difference across the mantle changes the magnitudes of various physical properties of rock, e.g. density, thermal conductivity, and heat capacity. Here, we present CitcomS v3.0, a 3D spherical mantle convection code that incorporates compressibility, multi-component composition, phase changes, and non-Newtonian rheology. The governing equations are solved under the Truncated Anelastic Liquid Approximation. Two implementations of the Stokes equation solver, one using Conjugate Gradient (CG) method with an outer iteration and the other using Bi-CG Stabilized method, are discussed. The code has been benchmarked with a semi-analytical solution and has a good scaling on parallel computers. The code is open source and freely available and can be downloaded from the provided URL. http://www.geodynamics.org/cig/software/packages/mc/citcoms/
DI14A-02
Constraining the Bulk Major Element Composition and Thermal State of the Earth's Lower Mantle from a Joint Inversion of Electromagnetic Sounding, Seismic and Gravity Data
We jointly invert global seismic travel time data, mean mass, mean moment of inertia and deep electromagnetic sound data directly for the Earth's mantle composition and thermal state, using a, by us, recently developed method. The chemical composition of the silicate Earth is modeled within the system CaO-FeO-MgO-Al2O3-SiO2. Given these parameters, in addition to a geotherm, we calculate stable mineral modes, elastic properties, bulk density (ρ) at the prevailing physical conditions using Gibbs free energy minimisation. Voigt-Reuss-Hill averaging is subsequently employed to compute bulk seismic P and S wave velocity profiles (VP, VS). The mineral modes are combined with laboratory-based models for the conductivity of individual minerals to estimate the bulk Earth electrical conductivity structure (σ). Assuming shear attenuation to be a thermally activated process, we used our thermodynamic calculations to estimate radial bulk and shear attenuation profiles, and employed these to assess anelastic contributions to VP and VS. From these radial profiles, seismic travel times, mean mass, mean moment of inertia and electromagnetic responses at the surface of the Earth are calculated. Given this scheme, the data are jointly inverted using a Markov chain Monte Carlo algorithm, from which a range of compositions and temperatures fitting data within uncertainties are obtained. More specifically, we find a more chondritic lower mantle composition with Mg/Si~1.15, in comparison to 1.27 for a pyrolitic upper mantle, providing a sink in addition to the core for the missing Si from the upper mantle and at the same time making it possible to derive the Earth from cosmic abundances. The lower mantle geothermal gradient is superadiabatic, attaining a most probable value of around 0.56°C/km, with core mantle boundary temperatures of ~2900 °C. Our results also imply that the 660 discontinuity is most probably related to a change in chemistry, while evidence for deep chemical layering is not found. The calculated physical properties (σ, ρ, VP and VS) agree excellently with purely geophysically-derived models such as PREM, AK135 and electrical conductivity models obtained by various workers.
DI14A-03 INVITED
Predictive geodynamical modeling of large-scale mantle models
One of the longstanding challenges of solid-Earth geophysics is to discover the large-scale dynamical and chemical structure of Earth's mantle. Geophysical and geochemical observations, and geodynamical modeling have led to several dynamically-different conceptual mantle models which are currently being debated in the community. Key to distinguishing between these models is understanding the cause of the large, low-velocity seismic anomalies observed beneath Africa and the Pacific. Candidate mantle models to explain thse anomalies include isochemical convection with large clusters of thermal plumes, thermochemical piles, and thermochemical superplumes. We perform predictive geodynamical modeling combined with comparison to observations of seismic tomography and CMB topography to test the feasibility these conceptual models. Given the uncertainty in model parameters, particularly the initial condition, we also investigate the predictive ability of geodynamical models to differentiate between these models. Furthermore, we investigate the possible genesis of more-dense material in the thermochemical models.
DI14A-04 INVITED
Predicting Seismological and Geochemical Observations Using Global and Regional 3-D Spherical Convection Models Incorporating Self-Consistently Calculated Mineral Physics
The latest generation of the global 3-D spherical convection model yinyang-Stag3D allows the direct computation of a planet's thermo-chemical evolution, including self-consistently generated plate tectonics, chemical differentiation induced by melting, large viscosity variations, and a realistic treatment of phase diagrams and material properties. The latter has recently been added using free energy minimization to compute stable phases as a function of temperature, pressure, and composition as expressed by ratios of the five main oxides, and thus avoids the need for increasingly complicated and ad hoc parameterizations of phase transitions. Modern supercomputers and clusters also allow increasingly higher resolution, with up to 1.2 billion unknowns possible on only 32 dual-processor nodes of an opteron cluster. In ongoing research, results from such modeling efforts are compared to a wide range of seismological observations, ranging from statistical comparisons with global tomographic inversions (standard and probabilistic), and comparisons with regional models, for example of D" structure and heterogeneity, including structures that are sharp-sided and/or related to the post-perovskite phase transition. Such results also have implications for geochemistry including the possible location of "reservoirs". For high-resolution studies, a region of a sphere, instead of a whole sphere, can be modeled, and in this mode models of slab-CMB interaction are presented that show some of the small-scale thermo-compositional-phase structures that can be generated. Global models also allow the computation of planetary secular cooling, including prediction of how the core heat flux varies with time hence the evolution of the geodynamo, and possible transitions in plate tectonic mode. Thus, a suite of predictions can be made.
DI14A-05 INVITED
Global Mantle Flow Models Constrained by Observations of Long Term Sea Level Change
Geological and sedimentological observations of continental transgressions show that sea level has fallen 100- 200 meters during the past 65 Myr. ~60 m can be attributed to climatic cooling (ocean water contraction and ice sheet formation), and at least ~120 m is caused by an increase in basin volume associated with a shortening of the Pacific ridge system. Thus, between ~20 m of fall and ~80 m of rise were accomodated by other processes that affect sea level. We use this constraint to test the possibility that long- wavelength topography dynamically supported by viscous flow in the mantle can change the volume of the ocean basins, thus changing sea level. Using a global mantle flow model driven by tomographically-inferred mantle density heterogeneity, we produced a global map of dynamic topography that reproduces patterns of observed dynamic topography in some locations. In this model, the ocean basins, on average, are elevated ~150 m while the continents are depressed ~340 m. This dichotomy occurs because mantle slabs generate downwellings that produce negative surface topography and form preferentially beneath continental regions. When isostatically compensated, the elevated seafloor produces a sea level that is 105 m higher than it would be on a non-convecting Earth. We investigate the possibility that this dynamically-elevated sea level has changed with time to cause Cenozoic sea level change. Using the first time-derivative of our instantaneous mantle flow models, we show that changes in the dynamic topography field probably do produce sea level rise (at rates of 0.3-0.5 m/Myr), but that the motion of continents over this dynamic topography field also produces between -0.3 to +0.3 m/Myr of sea level change, depending on the plate motion reference frame. Thus, dynamic topography should produce 0 to 50 m of sea level rise during the Cenozoic, and could possibly be a significant contributor to eustatic variations, comparable in magnitude to the effects of continental fragmentation and climate change.
DI14A-06
Dynamics of Thermal Plumes: Comparison of Laboratory and Numerical Models
Plumes originating from point sources have been widely studied by analytical, numerical and experimental techniques to better understand mantle plume dynamics using fluid mechanical constraints. However, a detailed comparison is made difficult due to the widely different contexts (injection of hot fluids or conductive heating, use of different heating modes) and assumptions of fluid dynamical parameters. This has led to a wide range of different scaling laws, diverse ideas about plume shapes and sizes, and diverging thoughts about entrainment in the plume head and tail. It is essential to investigate how the technical differences and underlying assumptions influence our understanding of plume dynamics, in order to figure out which configuration should apply to mantle plumes in the Earth. To facilitate this study we compare laboratory experiments of a plume growing in a viscous fluid from a heated patch with numerical models that attempt to reproduce the laboratory conditions as closely as possible. A new method of visualization set up in the IPG laboratory allows to visualize in situ the thermal and dynamical structures of the convection patterns on a 2-D section of the tank, without interfering with the flow. The numerical simulations are axisymmetric finite element simulations of starting plumes where we use the measured properties of the laboratory fluids under the assumptions of infinite Prandtl number and laminar flow. We find excellent quantitative agreement between the two fully independent approaches, in both temperature and detailed velocity field. That suggests that the laboratory simulation can be accurately described by laminar Boussinesq low at infinite Prandtl number, at least for finite size boxes. We use the numerical models to understand the origin of remaining minor quantitative differences, especially to estimate the effects of the boundary conditions and the influence of the weak temperature-dependence of viscosity. We also quantify the power leakage from the heater in the laboratory setup by measuring the heat flux through the plume stem in the numerical models. The combination of these numerical analyses allows to find reasonable agreement with independently derived scaling laws for the conduit velocity and the head dynamics versus power for thermal plumes, depending on the geometry of the box.
DI14A-07
Continuously closing plates: A new Paleogeographic concept and application to geodynamic models
Global plate tectonic reconstructions are inadequate for geodynamics, either as information to be assimilated into a model or as the basis to map a prediction into the geological record. Published reconstructions are often crudely spaced in time, have large swaths of the surface ambiguously defined, and/or have plate margin evolution inconsistent with plate motions. We have overcome these limitations with the formulation and implementation of a new method to represent plate tectonic reconstructions. Referred to as either "continuously closing" or "dynamically closing" plate polygons, the new method has been implemented using the new plate tectonic modeling package GPlates, global reconstruction have been developed with the method, and then reconstructions have been assimilated into forward and adjoint mantle convection models. Essentially, a plate is defined as a polygon that is made up of a finite set of plate boundaries. Each plate boundary is associated with its own set of finite rotations in an absolute reference system. These plate boundaries are continuously rotated and an algorithm finds the intersection of adjacent plate boundaries. Two adjacent plates always share a boundary. Using this method in GPlates, we have developed several global plate reconstructions from 140 Ma to the present. Since plate closure is continuous in time, reconstructions can exist at any granularity of time. Our present model has been output at 1 Myr time intervals. Subduction zones and their polarity are continuously tracked. The present plate reconstructions are self-consistent with a set of oceanic paleo age grids. We will illustrate the use of the new reconstructions in several applications drawn from our recent work: (1) regional subduction models; (2) global models of thermo-chemical convection in the lower mantle; (3) inverse and adjoint models of the descent of the Farallon slab; and (4) instantaneous models of global plate motions.
DI14A-08
Mantle Convection and the Recent Geological Evolution of the Southwestern United States
The present-day dynamic topography in the Southwestern US and its corresponding rate of change are predicted with a high-resolution mantle convection simulation based on a new global seismic tomography model (Simmons et al. 2007) which incorporates mineral, physical and surface geodynamical constraints. The 2° horizontal resolution in the global tomography model enables us to perform flow calculations with much greater spatial resolution than previously possible. These convection simulations also incorporate a viscosity profile derived from joint inversions (Mitrovica and Forte, 2004) of convection-related surface observables (surface gravity anomalies, topography, divergence of tectonic plate motions, excess ellipticity of the core-mantle boundary) and data associated with the response of the Earth to ice-age surface mass loading (decay times inferred from post-glacial sea-level histories in Hudson Bay and Fennoscandia, and the Fennoscandian relaxation spectrum). A fundamentally important aspect of the mantle convection model is the use of a 3-D mantle density distribution that explicitly accounts for both thermal and compositional heterogeneity. The predicted near-surface convective flow velocities and associated surface topography thereby include the stabilizing effect of compositional buoyancy in the continental tectosphere and deep lower mantle. We will present the implications of this new convection model for the time-dependent topography of the Colorado Plateau where we find a focused dynamic topography high in the Western US which overlies a regional mantle upwelling in the asthenosphere. Additionally, we will consider the implications of our convection model on the late Cenozoic evolution of the Colorado Plateau, the second stage of rifting in the Rio Grande valley and the associated magmatic activity along the Jemez lineament.