Joint inversion of Magnetotelluric Data and Surface-Wave Dispersion Curves using Correspondence Maps

We use a correspondence map approach to jointly invert surface-wave dispersion curves and magnetotelluric data for subsurface shear velocity and resistivity but also for a possible relationship between them. Our first experiments consist of inversions of synthetic data computed from models linked by linear and second-order polynomial relationships. Our methodology produces joint inversion model-pairs (resistivity-shear velocity) from where 80% fit the 'observed' parameter relationship within a 5% error vs only 1%  of the separate inversion model-pairs for the linear relationship experiment. For the non-linear relationship synthetic test, 85% of the joint inversion model-pairs fit the 'observed' relationship within a 5% error while just 40% of the separate inversion model-pairs. This reduces the number of acceptable models without compromising the data fit ('reduction of non uniqueness'). Using the non-linear synthetic data we show how to select an appropriate polynomial degree for joint inversion. Having validated the approach with synthetic cases, we applied our methodology to field data from the ECOGI and EstOF surveys in North Alsace, France. We compare separate and joint inversions and we find that the 1D subsurface models obtained from joint inversions are more similar to previous models documented in the area than the separate inversion models. We are currently extending this work to higher dimensions. At the spatial scale of our problem, sensitivity analysis suggests that shear velocity models can benefit from the lateral sensitivity of the magnetotelluric data.