Should Hydraulic Tomography always rely on the Geostatistics Approach? A Laboratory Sandbox Investigation

Abstract:

Hydraulic tomography (HT) is an approach that has been designed to incorporate multiple hydraulic head records for model calibration. The promising performance of HT has been demonstrated through many synthetic, laboratory and field studies. In virtually all studies published to date, geostatistics forms the backbone of the algorithm. While geostatistics offers many advantages, there are some drawbacks. One disadvantage is its highly parameterized nature, which renders it computationally intensive for large-scale problems. This has resulted in the development of alternative HT methods (e.g., geostatistical reduced order, Krylov subspace, and principal component geostatistical methods). Another issue is that HT may produce overly smooth images of subsurface heterogeneity when there are few monitoring interval data. Therefore, one may question the utility of the geostatistics approach in certain situations and seek alternative approaches. For example, the simultaneous calibration of an effective parameter model may be sufficient for water supply studies, as detailed information on heterogeneity may not be necessary. Moreover, when data used for HT are sparse, the simultaneous calibration of a geological model may be more appropriate than a geostatistics approach. To investigate these issues, we conducted a HT study with different subsurface conceptualizations and parameter resolutions. The performance of different models (effective parameter, geological, and geostatistics models) is assessed through model calibration and validation. Calibration data consisted of data from eight pumping tests, while validation data consisted of 16 separate pumping tests not used in the calibration effort. Our ultimate goal is to conduct a field-based study for this comparison; however, we began our investigation in a synthetic, heterogeneous, sedimentary aquifer constructed in the laboratory under controlled conditions. Results show that the geostatistics approach performed the best among the approaches compared, although the geological model with perfect knowledge of the stratigraphy came a close second. This may suggest that the better knowledge of stratigraphy obtained via geophysics or other means may contribute to improved HT results, although further research is necessary.