Stochastic inversion of sequential hydraulic tests for transient and highly permeable unconfined aquifer systems

Abstract. A hydraulic tomography survey (HTS) is a conceptually improved technique that has been recognized to be efficient for estimating high-resolution aquifer parameters. Based on the concept of HTS, this study presents a modified stochastic inverse model for estimating hydraulic conductivity (K) and specific yield (Sy) in shallow and highly permeable unconfined aquifers. A well field with 15 fully screened wells was developed for the purpose of model implementations. In this study a synthetic example was first employed to assess the accuracy of the inverse model. We then implemented the model to field-scale, cross-hole injection tests in a shallow and highly permeable unconfined aquifer near the middle reach of the Wu River in central Taiwan. To assess the effect of constant head boundary conditions on the estimation results, two additional modeling domains were evaluated based on the same field data from the injection tests. Results for the synthetic example show that the modified inverse model can reproduce well the predefined geologic features of the unconfined aquifer. The inverse model can estimate accurately the ln K patterns and magnitudes. However, slightly fewer details of the ln Sy field are obtained due to the insensitivity of transient hydraulic stresses for specified sampling times. Model implementations of field-scale injection tests show that the model can estimate ln K and ln Sy fields with high spatial resolution. The estimated K and Sy values for the test site vary by one order of magnitude, indicating a relatively homogeneous aquifer for the tested well field. Results based on three different modeling domains show similar patterns and magnitudes of ln K and ln Sy near the well locations. This result suggests that the case with domain 40 m × 20 m should be sufficient for the injection tests at the well field.