Bayesian hierarchical approach and geophysical data sets for estimation of reactive facies over plume scales

A stochastic model is developed to integrate multiscale geophysical and point data sets for characterizing coupled subsurface physiochemical properties over plume-relevant scales, which is desired for parameterizing reactive transport models. We utilize the concept of reactive facies, which is based on the hypothesis that subsurface units can be identified that have distinct reactive-transport-property distributions. To estimate and spatially distribute reactive facies and their associated properties over plume-relevant scales, we need to (1) document the physiochemical controls on plume behavior and the correspondence between geochemical, hydrogeological, and geophysical measurements; and (2) integrate multisource, multiscale data sets in a consistent manner. To tackle these cross-scale challenges, we develop a hierarchical Bayesian model to jointly invert various wellbore and geophysical data sets that have different resolutions and spatial coverage. We use Markov-chain Monte-Carlo sampling methods to draw many samples from the joint posterior distribution and subsequently estimate the marginal posterior distribution of reactive-facies field and their associated reactive transport properties. Synthetic studies demonstrate that our method can successfully integrate different types of data sets. We tested the framework using the data sets collected at the uranium-contaminated Savannah River Site F-Area, including wellbore lithology, cone penetrometer testing, and crosshole and surface seismic data. Results show that the method can estimate the spatial distribution of reactive facies and their associated reactive-transport properties along a 300 m plume centerline traverse with high resolution (1.2 m by 0.305 m).