Analysis of the parameter identifiability of the in situ diffusion and retention (DR) experiments

Abstract

In situ diffusion experiments are performed at underground research laboratories to overcome the limitations of laboratory diffusion experiments. The diffusion and retention (DR) experiments are long-term, natural-scale, in situ experiments performed in the anisotropic Opalinus Clay Formation at Mont Terri, Switzerland. Dilution data are monitored at the injection interval and overcoring data will be measured at samples around the injection interval at the end of the experiment during overcoring. Interpretation of DR experiments is complicated by the non-ideal effects caused by the sintered filter, the gap between the filter and the borehole wall and the excavation disturbed zone (EdZ). Their impact on parameter estimates has been evaluated with numerical sensitivity analyses and synthetic experiments having the same geometry and parameters as the real DR experiments. Dimensionless sensitivities of tracer concentrations in the injection interval and along the overcoring profiles have been computed numerically. They have been used to identify the tracer parameters that can be estimated with the least uncertainty from tracer dilution and overcoring data. Sensitivities of tracer dilution data change with time and those of overcoring concentrations along the bedding are different from those along profiles normal to the bedding. Concentrations along overcoring profiles are sensitive to the effective diffusion normal to the bedding. Synthetic experiments generated with prescribed known parameters have been interpreted automatically with INVERSE-CORE 2D and used to evaluate the relevance of non-ideal effects and determine parameter identifiability in the presence of random errors for HTO and 22Na +. Model results show that it is difficult to estimate the parameters of the undisturbed clay when the tracer dilution data contain noise. The convergence of the estimation algorithm improves when the starting values are smaller than the true parameters. Although the parameters of the undisturbed clay and the EdZ cannot be estimated using tracer dilution data, their joint estimation from overcoring noisy data is possible for standard deviations of the noise up to 0.05. Large estimation errors in the parameters of the undisturbed clay and poor fits are obtained when the assumption about the existence of EdZ is incorrect. The effective diffusion of the filter is a key parameter for the interpretation of the experiments. Small errors in the volume of the circulation system do not affect the estimates of the component of the effective diffusion of undisturbed clay parallel to bedding. The results of the identifiability analysis clearly shows that: (1) The proper interpretation of in situ DR experiments requires accounting for the filter and the EdZ and (2) Overcoring data allow a more accurate estimation of the parameters of the undisturbed clay than the tracer dilution data.