| Abstract |
Quantification of fluid distribution and flow in the Earth’s near-surface benefits from precise estimation of electrical properties of fluid-saturated rocks, such as resistivity estimated from inversion of electrical resistivity tomography (ERT) data. The predicted resistivity values are often uncertain due to two main types of uncertainties: epistemic uncertainty in the inversion process (e.g., inaccuracy in the physical models) and aleatoric uncertainty in the data (e.g., measurement errors). This work focuses on the quantification of aleatoric variability in the ERT measurements and its effect on the inverted resistivity models. We first investigate how measurement uncertainty, in the form of reciprocal error, correlates with the measured electrical contact resistance of electrodes with ground. Next, we apply a statistical approach based on the stochastic perturbation and inversion of multiple realizations of resistance data to study the uncertainty in the predicted resistivity tomograms. We then study the effect of data uncertainty on the inverted resistivity model for individual data sets. We finally quantify the effect of variation in data quality over time on the inverted time-lapse resistivity results. The results from 20 campaigns and two time-lapse ERT data sets show that reciprocal error is positively correlated with both contact resistance and ground’s apparent resistivity, confirming the significance of practicing lowering electrode contact resistance during ERT field campaigns. Additionally, our results show that uncertainty in the estimated resistivity model depends on both the ground’s resistivity and measurement error of the input data. The time-lapse results provide additional insight that model uncertainty is the highest at the driest and coldest months of the year, corresponding to the highest measured contact resistance and reciprocal error. |
, Andrew Parsekian 
