Sensitivity studies applied to a two-dimensional resistivity model from the Central Andes

Long-period magnetotelluric (MT) data have been collected along an E-W profile across the Central Andes in Northern Chile and Southern Bolivia. A 2-D resistivity model explaining these data was found by inverse modelling. The model includes a prominent conductivity anomaly in the backarc of the South American subduction zone below the Bolivian Altiplano. Another, comparatively localized anomaly occurs beneath the Precordillera in the forearc. The enhanced conductivity in the backarc correlates with reflective zones, low seismic velocities and high seismic attenuation. In contrast, no correlation can be found in the forearc. A strong seismic reflector in context with the down-going slab, and another reflector in the middle crust below the Pre- and Western Cordillera do not coincide with anomalous conductivity. Other 2-D models may also be consistent with the data. We applied different techniques to find the range of relevant models. One simple but useful approach is the direct use of the sensitivity matrix containing the partial derivatives of the data with respect to the model parameters. The sensitivity of the model parameters to the data or data subsets can be visualized by calculating columnwise sums from the sensitivity matrix. These sensitivities may be used to indicate model parameters that are less resolved by the data and thus should not form part of an interpretation. Another approach is to perform systematic studies to establish the validity of geologically important features of the model. These studies comprise forward modelling and the use of a priori information. Combining all approaches generates a more complete assessment of the principal model features. We conclude that a minimum depth extent of the Altiplano conductor can be specified as approximately 50 km. Low sensitivity in the forearc beneath the Longitudinal Valley indicates a limited structure resolution, and therefore may explain the absence of a conductive slab in the model. The existence of the localized conductor below the Precordillera can be established by high sensitivity. Finally, we included the two seismic reflectors in the forearc as a priori information in the inversion process. As a result the absence of a conductive slab in the model is probably a result of low sensitivity, but cannot be totally excluded. A conductive zone coinciding with the strong reflector below the Pre- and Western Cordillera is not manifested by the MT data.