Receiver Functions in the San Fernando Valley, California: Graph-Regularized Bayesian Approach for Gravity-Informed Mapping

Abstract

The San Fernando Valley (SFV) in Southern California is a complex sedimentary basin whose shape strongly influences ground shaking. We develop a fully quantitative, probabilistic graph-regularized inference model that integrates both gravity and receiver function (RF) constraints and evaluate its ability to determine the basin's shape. The sediment-basement interface in single-station RFs is often difficult to interpret due to scattering and noise, which can render isolated stations unusable. By using RFs from a dense seismic array and incorporating gravity, we address the issue of non-uniqueness in converting the times of RF phases to layer thickness by comparing the predicted gravity to observations at each station. In areas where the density contrast may change, Bayesian inference with a graph Laplacian allows us to determine the effective density contrast by taking into account its neighbors' picks and densities. This method promotes spatial smoothness between neighboring stations, while preserving sharp contrasts in locations supported by the RF and gravity data. We applied this method to a dataset that was acquired in fall 2023, when 140 nodes were installed in the SFV. Our results show the deep Sylmar sub-basin, the San Fernando sub-basin, and the Leadwell high found in a previous study (Ju\'arez-Z\'u\~niga and Persaud, 2025), and our results also show good agreement with the industry seismic reflection profiles across the valley. This method demonstrates how to incorporate gravity with lateral density variations into receiver function interpretation to better map interfaces in the subsurface.