Dynamics of plumes in a compressible mantle: consequences for their seismic visibility

Hotspots are regions of excessive long-term volcanism whose position is independent of plate boundaries and which are not explained by the standard plate tectonic paradigm. Fluid dynamical considerations suggest their occurrence to result from the rising of hot material from the deep mantle in the form of mantle plumes [Morgan, 1971] due to the thermal boundary layer at the core-mantle boundary and the estimated convective vigor of the Earth's mantle. However, it has been contentious to recognize mantle plumes in seismic tomographic images of the mantle beneath hotspots given the relative poor image resolution. Here we present a further development of numerical models of thermal and thermochemical mantle plumes in an axisymmetric spherical shell geometry [e.g., Lin and van Keken, 2005]. Our new simulations consider viscous dissipation and work done against gravity as well as compressibility and a strongly temperature-dependent viscosity. Phase boundaries are incorporated at 410 km and 660 km depth. In addition, the efficiency of the model is improved by an automated mesh refinement using a series of meshes whose local resolution is matched to the respective position of the rising plume head. We use a fully consistent equation of state which incorporates variations in thermal expansivity, density and heat capacity. We compare the convective vigor and resulting plume shapes for a range of viscosity models (which influence the global Rayleigh number and local viscosity variations) and for variable thickness and density contrast of a deep dense layer from which the plume rises. Complex interaction at the phase changes is seen due to the compressible effects and the variable composition. From this series of models we choose the ones that have reasonable characteristics based in particular on the observed buoyancy flux. For this selection of models we determine seismic wave velocity variations using mineral physics constraints. These velocity models are then projected as tomographic images using the S40RTS resolution filter [Ritsema et al., 2010] to understand how the simulated plume head and tails are imaged tomographically.