The interaction of subducted slabs and plume generation zones in geodynamic models

The interactions of surface and core-mantle boundary processes have been investigated in numerous studies in recent years. The interest in this topic was raised by observations like the correlation between hotspots, reconstructed large igneous provinces and erupted kimberlites, and core-mantle boundary structures. While these studies already prove the existence and reproduce single observations of these links, they fail in providing a convincing match to the observed hotspot distribution on Earth. In this work, models of global mantle convection with prescribed surface plate motions are developed. In particular, they are set up to test the possibility of creating a present-day state of the models that is comparable to present-day Earth observations. A special focus is lain on the interaction of subducted slabs and plume generation zones at the core-mantle boundary, which is investigated and compared to surface observations like published hotspot catalogues, seismic observations like global tomographies and constraints from temperature evolution and heat flow measurements. The preferred reference model develops an Earth-like distribution of plumes and zones of low seismic velocity at the core-mantle boundary. This similarity is proven to be statistically significant. The plumes are separated by a clustering algorithm into two groups around Africa and the South Pacific, which is nearly identical to the separation of observed hospots. Furthermore, the comparison to seismic tomography at the core-mantle boundary shows large congruency between the mantle convection model and the tomography models Savani and P07. An especially convincing agreement exists in the description of the seismologically observed "Perm" anomaly, which is reproduced for a wide range of model parameters, and has not been reproduced in previous convection models with similar setup. The model further matches thermal constraints and observations of the Earth like the secular cooling rate and the surface and core-mantle boundary heat flux. In a sensitivity analysis of the model results to assumed material properties, initial and boundary conditions, strong dependencies on the assumed viscosity formulation and surface plate reconstruction are shown, while the influence of the initial condition is minor. The investigated model results are insensitive to uncertain boundary conditions like the core-mantle boundary temperature, with a preference of the model towards a core-mantle boundary temperature jump of 1000 +/- 200 K. The second part of the thesis investigates the possibility to constrain a plate reconstruction on the basis of the fit between convection model, seismic tomography and hotspot positions. On the example of the Phoenix-Gondwana subduction zone around 200 Ma BP (before present), the influence of a regional change of subduction zone position, convergence velocity and convergence direction is tested. The models show a strong influence of the regional subduction zone setup on the present-day mantle state. The number and position of plumes, and the boundaries of the large-low velocity province beneath Africa are directly affected by the changes. The models are ranked according to their fit to present-day observations and suggest a closer subduction to Gondwana, or a flat-slab subduction scenario, with a possibly slightly oblique convergence direction. A subduction far out in the Panthalassa ocean does not fit observations in the African hemisphere, but improves the LLSVP position in the Pacific hemisphere, and might therefore be investigated in further studies as an additional intra-oceanic subduction zone. In summary, this thesis provides progress in the methods of mantle convection modelling by using a temperature-dependent viscosity formulation not investigated before. It shows that the resulting plume distribution is more Earth-like compared to other approaches, and a necessary consequence of the interaction of paleo-plate movements and processes at the core-mantle boundary. Additionally, it explores the possibility to constrain paleo-plate reconstructions by forward numerical modelling