Stabilized resolution-density appraisal of regularized one-dimensional magnetotelluric inversion

Abstract A small data misfit can coexist with weakly constrained structure in a regularized magnetotelluric model. We examine a post-inversion appraisal workflow that combines resolution density, a residual-related model-variation scale and a numerical check of the regularized inverse operator. Synthetic one-dimensional layered models compare the effects of residual level, integrated sensitivity weighting, frequency coverage and regularization strength. Removing high-frequency data weakens shallow constraints. Removing low-frequency data leaves a larger depth range poorly constrained. In the resistive-layer example, one recovered profile contains a deep resistive feature below the true target layer. Forward-response tests compare selected model perturbations with the adopted error scale. The variation scale describes local propagation of the residual through the regularized operator, while the system residual measures agreement with the operator equations. For a fixed recovered model, integrated sensitivity weighting changes the appraisal profiles. Stronger inversion regularization suppresses the recovered conductive contrast and increases the final residual in these examples.