Enhanced Uniqueness and Linearity of Receiver Function Inversion

Inversion of receiver functions is widely regarded as both highly nonlinear and highly nonunique. In this article, we describe why this is not entirely accurate. We show that for synthetic data with or without noise, a quasi-continuous model stratification greatly reduces the nonuniqueness present for the typical coarser stratifications. However, convergence is slow when the model is parameterized in depth. If instead the quasi-continuous model is parameterized in delay time, convergence becomes very fast and stable. This indicates that the nonuniqueness and nonlinearity problems encountered by previous studies were mainly a product of an unfavorable model parameterization. To validate the quasi-continuous delay-time parameterization in receiver function inversion, we test this new approach on teleseismic data from two permanent broadband stations situated in well-studied and distinct geologic settings. Convergence of the linearized iterative inversions is fast, reliable, and practically independent of the starting model. No a priori constraints are imposed on the modeled S velocities. The inversion models obtained for each of the two stations agree very well with results of other seismic methods. Based on this general validation with observed data and synthetic tests, we propose the use of the quasi-continuous delay-time model parameterization to enhance linearity and uniqueness of receiver function inversion, regardless of the approach being linearized or Monte Carlo.