Impact of petrophysical uncertainty on Bayesian hydrogeophysical inversion and model selection

Abstract

Quantitative hydrogeophysical studies rely heavily on petrophysical relationships that link geophysical properties to hydrogeological properties and state variables. Coupled inversion studies are frequently based on the questionable assumption that these relationships are perfect (i.e., no scatter). Using synthetic examples and crosshole ground-penetrating radar (GPR) data from the South Oyster Bacterial Transport Site in Virginia, USA, we investigate the impact of spatially-correlated petrophysical uncertainty on inferred posterior porosity and hydraulic conductivity distributions and on Bayes factors used in Bayesian model selection. Our study shows that accounting for petrophysical uncertainty in the inversion (I) decreases bias of the inferred variance of hydrogeological subsurface properties, (II) provides more realistic uncertainty assessment and (III) reduces the overconfidence in the ability of geophysical data to falsify conceptual hydrogeological models.

Improved Characterization of DNAPL Source Zones via Sequential Hydrogeophysical Inversion of Hydraulic‐Head, Self‐Potential and Partitioning Tracer Data

“…But according to Ikard et al (2012), this salinity effect is rather modest (less than 1 order of magnitude) and can be removed by measuring the electrical conductivity of the pore water. Furthermore, one can consider this petrophysical uncertainty in the Bayesian framework by jointly estimating both the distribution of the unknown parameters (e.g., K eff ) and the distribution of petrophysical uncertainty (e.g., Brunetti & Linde, 2018). However, additional petrophysical uncertainty in the ERT and SP inversion will reduce the accuracy of the inversion, and collection of additional data (e.g., a larger amount of ERT and SP measurements) will be necessary.…”

Advancing hydrological process understanding from long‐term resistivity monitoring systems

“…An additional, and often overlooked, source of uncertainty in the interpretation of resistivity images is the petrophysical prediction uncertainty, whereby the equations (e.g., Equations (1)–(4)) are inherently uncertain, heterogeneous and characterized by limited predictive power (Brunetti & Linde, 2018). The most common approach remains to use calibrations of the petrophysical relationships (e.g., m and n in Equation (2)) to turn changes in imaged conductivity into changes into a hydrologic state variable.…”

Relative importance of conceptual and computational errors when delineating saltwater intrusion from resistivity inverse models in heterogeneous coastal aquifers

“…After data collection, the analysis of resistivity data may also benefit from the use of appropriate appraisal tools (Oldenburg and Li, 1999;Caterina et al, 2013). Additionally, hydrogeophysical SWI studies would be improved by performing uncertainty analysis (Linde et al, 2017), including the evaluation of uncertainty associated to the petrophysical model in the transfer of information between the hydrogeological and geophysical models (Brunetti and Linde, 2018). The results presented in this work suggest that, even for acceptable values of sensitivity, hydrogeological information derived from geophysics must be used with caution in the calibration of groundwater models, especially in heterogeneous aquifers.…”