Identifying Field-scale Bioremediation Status from Geochemical and Geophysical Data Using Dynamic Linear Models with Switching: Development and Application at a Uranium Contaminated Aquifer

Many field bioremediation experiments have been carried out at the uranium-contaminated Rifle Integrated Field Research Center (IFRC) site in Rifle, Colorado. The experiments include continuously injecting acetate and bromide for a period of 1~2 months and subsequently collecting multiple geochemical samples from downstream monitoring wells. Surface spectral induced polarization data along several two-dimensional (2D) profiles have also been collected to obtain information on the spatial distribution of biogeochemical transformations induced by bioremediation. The biogeochemical reactions vary over space and time during the contaminated aquifer transitions from iron to sulfate reduction following introduction of the electron donor. Developing methods to identify the onset and distribution of these transitions could improve our ability to assess remediation efficacy and sustainability. In this study, we develop a dynamic linear model with switching to identify bioremediation transitions using time-lapse aqueous geochemical data (such as Fe(II), sulfate, sulfide, acetate, uranium, chloride, and bromide concentrations) and spectral induced polarization data. We consider the multivariate geochemical concentrations as hidden random processes (observed at borehole locations but unknown at other locations) and the time-lapse geophysical data as observations at each location along the 2D profiles. The connection between the geophysical observations and geochemical time-series is determined by design matrices, which vary depending upon redox status. We describe the unknown biogeochemical events as categorical random variables. We take a Bayesian approach to estimate unknown parameters by first assigning suitable priors to the unknowns and then drawing many samples from their joint posterior distribution using Markov Chain Monte Carlo methods. The developed approach can provide us a wide range of information on bioremediation for evaluating the effectiveness of bioremediation, including: (1) the probability of being in each redox stage over time and space following biostimulation; (2) diagnostic parameters as functions of aqueous geochemical concentrations and geophysical attributes.