Lawrence Berkeley National Laboratory

We discuss field activities designed to characterize seepage into an underground opening at the potential site for geologic storage of high-level radioactive waste (HLRW) at Yucca Mountain, Nevada, and the use of these data for development and calibration of a model for predicting seepage into planned HLRW emplacement drifts. Air-injection tests were conducted to characterize the permeability of the fractured rock, and liquid-release tests (LRTs) were conducted and seepage monitored to characterize the seepage-relevant properties of the fractured rock. Both air-injection and liquid-release tests were performed in the same borehole intervals, located above the underground openings. For modeling, three-dimensional, heterogeneous permeability fields were generated, conditioned on the air-permeability data. The initial seepage data collected were used to calibrate the model and test the appropriateness of the modeling approach. A capillary-strength parameter and porosity were the model parameters selected for estimation by data inversion. However, due to the short-term nature of the initial data, the inversion process was unable to independently determine the capillary strength and porosity of the fractured rock. Subsequent seepage data collection focused on longer-term tests, a representative selection of which was used for data inversion. Field observations also played a key role by identifying factors such as evaporation and ceiling geometry that can enhance or reduce seepage. These observations help guide future test and model development by ensuring that relevant processes that influence seepage are identified, characterized, and incorporated into the model, thus increasing confidence in the parameter estimates. It is this iterative and collaborative approach to field testing and modeling, and the feedback mechanisms of field-test-methodology and model review and revision, that has been employed to continuously improve the scientific quality of the study. Initiation of modeling as soon as the first liquid-release data were available, review of the models with the field-testing team, and feedback of model results to the field-testing team proved to be important for optimizing both data collection and model development, resulting in increased confidence in the predictive models.