Incorporating geological structures into hydraulic tomography based on geostatistical model

Abstract:

Hydraulic tomography can possibly obtain high-resolution hydraulic properties of the heterogeneous aquifer by integrating information collected from a series of cross-hole pumping tests. Typically, the collected data are interpreted by geostatistical models, in which the hydraulic parameters are treated as random fields with specified mean and correlation structures. Under the Bayesian framework, the inverse models start from an initial guess of coarse parameter structure and then obtain reasonable estimates that honor the observed data. Although hydraulic tomography can collect a lot of non-redundant information with a limited number of boreholes, the information may not be sufficient to characterize the complex heterogeneity in some geological formations, especially in fractured medium. In this situation, prior information on geological structures may be important to the quality of the estimates. Yet, conventional geostatistical inverse methods only consider the global structure of the parameter fields and specific geological knowledge available (e.g., geofacies data, fault location) is not fully appreciated. The proposed methodology can flexibly incorporate complex geological structure information and their uncertainties in geostatistical modelling, that are available usually only in a qualitative sense. An illustration example showed that, by integrating fault information from outcrop survey and borehole intersection data, the estimates obtained from hydraulic tomography with four pumping tests at a large-scale fractured granite aquifer can be improved using the proposed method, which lead to better flow prediction.