Rare Earth Element Inversion for Melt Distribution: Sensitivity and Application

Inversion of rare earth element (REE) abundances in MgO-rich(MgO 6 wt.%) igneous rocks, described initially by McKenzie& O'Nions (1991), provides a useful means of constrainingthe distribution of melt fraction with depth, X(z), in the mantle.Here, we investigate the sensitivity of the REE inversion procedureto noise in measured REE abundances and to variation in parameterswhich must be specified a priori. Inverted melt distributionsare insensitive to realistic amounts of random noise but aresensitive to systematic noise. Melt distributions are also sensitiveto input parameters such as depth and range of melting, sourcecomposition, and mantle mineralogy. To determine unique meltdistributions, additional independent constraints on at leasttwo of the following input parameters are therefore required:depth and range of melting (i.e., lithospheric thickness); sourcecomposition; source mineralogy (dependent on mantle temperature).The successful replication of melt compositions where thesethree parameters are well constrained and the agreement betweeninverted melt distributions and melt distributions calculatedindependently assuming isentropic melting during adiabatic upwellingindicate that the scheme is reliable. REE inversion thereforeprovides a means for both routinely constraining X(z) and fordetermining unknown input parameters.