BAYESIAN SEISMIC REFRACTION INVERSION FOR CRITICAL ZONE SCIENCE AND NEAR-SURFACE APPLICATIONS

The critical zone (CZ) is the region of the Earth’s surface that extends from the bottom of the weathered bedrock to the tree canopy and is important because of its ability to store water and support ecosystems. A growing number of studies use active source shallow seismic refraction to explore and define the size and structure of the CZ across landscapes. However, conventional approaches to retrieve seismic velocity structure do not extensively explore the uncertainty and tradeoffs in the inverted velocity models, nor do they assess the effect of horizontal and vertical model smoothing assumptions. These limitations hinder interpretation, particularly of deeper structures. To reliably resolve seismic velocity with depth, we develop a Transdimensional Hierarchical Bayesian (THB) framework with reversible-jump Markov Chain Monte Carlo (rjMCMC) to generate samples from the posterior distribution of velocity structures. This approach allows us to evaluate measurement noise as well as model resolution along distance and depth. With sufficient number of model iterations, this approach can also eliminate influence from the initial velocity model. We perform 2D synthetic tests with imposed measurement noise to fully explore the capability and limitation of imaging subsurface structure using seismic refraction. Despite inability to precisely recover input model layers, this method can capture the overall velocity pattern along hillslopes and estimate model uncertainty with depth. It also distinguishes subtle differences of velocity at shallow depths (< 10m). Based on a 24-geophone surveying system, we explore the velocity structure in a series of ridges and valleys of laterally homogeneous Cretaceous lithology in northern California. The posterior velocity model shows an increasing thickness of low velocity material from channels to ridgetops along a transect parallel to bedding strike, implying that the near-surface velocity structure is more strongly influenced by the weathering process than by variation in bedrock lithology. The THB rjMCMC method strengthens the ability to reliably image and interpret CZ structure. It also has additional applications for other near-surface studies, especially in the presence of significant surface topography.