Comparing Geochemical and Geodynamical Models of Plume and Ridge Mantle Source Composition

We use a multidisciplinary approach to investigate how the parameter space of mantle convection affects present-day mantle composition. We compare 22 forward geodynamical mantle circulation model simulations against 24 variants of a geochemical inversion model of the global radiogenic isotope data set of mantle-derived lavas. Both models are fully independent but able to output compositional parameters for the lower mantle sampled by upwelling mantle plumes and for the upper mantle sampled by mid-oceanic ridges. Geodynamical model results suggest an excess degree of peridotite melt-depletion ΔF_d = +0.4% ± 0.4% and an excess amount of recycled crust Δf_RC = +2.7% ± 3.1% in plumes compared to ridges, while the geochemical inversion returns ΔF_d = +0.4% ± 1.2% and Δf_RC = +1.5% ± 0.6%. Models are thus in quantitative agreement but with opposite sensitivities, allowing to restrict their respective parameter space. Geodynamical runs show best fits with the narrow geochemical Δf_RC for core-mantle boundary (CMB) temperatures of 3,400–3,800 K and a recycled crust buoyancy number of 0.44–0.66. A dense primordial layer at the CMB also leads to a better fit. Variants of our geochemical model show a best fit with the narrow geodynamical ΔF_d value when early mantle differentiation occurs in the garnet stability field. We also find that the formation of early compositional heterogeneities is needed to fully explain the isotope range of mantle melts. Our work emphasizes the need to correct isotopic data for the effects of non-magmatic processes in a quantitative geochemical model before extracting the parameters relevant to a comparison with geodynamical model results.