Analysis of the Aespoe LPT2 pumping test via simulation and inverse modelling with HYDRASTAR

A stochastic groundwater modelling study is presented for the Long-term Pumping Test (LPT2) conducted at the Aespoe Hard Rock Laboratory. It is a test case for HYDRASTAR, a stochastic continuum groundwater flow and transport model. Unlike previous modelling studies of LPT2, this study uses the inverse modelling capabilities of HYDRASTAR to condition the model results to the observed hydraulic heads. The purpose of this conditioning via inverse modelling is to improve the reliability of input hydraulic conductivity fields and thus minimise the uncertainty of the model predictions. Preliminary simulations evaluated the boundary conditions, grid extent and grid density, all of which were restricted by the computational demands of a larger domain and by code limitations. The preliminary simulations also indicated that the model is quite sensitive to changes in specific storativity and kriging parameters. The final calibrated model simulations were successful in representing the response of the rock mass to the LPT2. Specifically, the mean of the Monte Carlo realizations of simulated drawdowns generally reproduced the magnitude, timing, and shape of the observed drawdowns. The observed drawdowns generally were bracketed by an interval of plus or minus one standard deviation from the mean of the realizations. Some discrepancies in the magnitude and timing of the observed versus the simulated drawdowns were revealed at several locations, such as the upper sections of borehole KAS07. However, the ensemble of model realizations were centred on the observed drawdowns and bracketed the observed responses, supporting both the conceptual model and its representation by HYDRASTAR. 33 refs, figs