Towards validating numerical simulations of drawdown in unconfined fractured rocks with field experiments: A comparative study at sub-seismic scale

Quantifying the effective permeability structure of fault and fracture zones is crucial for numerous geo-energy applications, especially at sub-seismic scales. However, the multi-scale presence of fractures and the structural heterogeneity of faults and host rocks often cause highly non-stationary and anisotropic hydraulic properties. This usually impedes the definition of representative elementary volumes and complicates the upscaling process. Thus, progressively integrating these multi-scale complexities into 3D numerical models of exploration targets has gained increasing scientific interest in recent years and is crucial to make predictions of flow and transport through reservoirs.Here, we aim to reproduce multiple drawdown curves obtained in analog field pumping tests with numerical models of fluid flow in order to develop a proof of concept for generating correct hydraulic representations of fault and fracture zones in numerical models with the final goal to upscale their effective permeabilities for numerical simulations above the sub-seismic scale.The test subject is a 30- by 30-meter-wide area in a quarry in the Franconian Alb, Germany, featuring an intensively deformed Upper Jurassic limestone formation, frequently explored for geothermal energy production in the southern German Molasse Basin. First, an initial 3D structural model of the main faults, fractures, and layer surfaces, based on multiple borehole logs and pavement traces is constructed with the GemPy software. In the next step, we employ a newly developed discretization method to convert the initial 3D GemPy model into various equivalent continuum models of the test field by parameterizing fracture, fault, and rock matrix permeabilities/porosities, resulting in high-resolution 3D voxel models with individual, anisotropic permeability tensors. Those serve as input for numerical simulations of a pumping test, where we solve for transient, unsaturated/saturated Darcy-flow using a newly developed parallel, 3D finite element code that utilizes a van Genuchten approximation for the arising non-linearities, i.e., relative permeability and water content. As a final step, we compare the drawdown curves logged in three observation wells in an analog constant-head hydraulic test in the field to the ones obtained from the numerical simulations by computing a cumulative misfit. While changing the parameters of the employed permeability-porosity parametrizations for faults, fractures, and rock matrix in a classical forward-approach manner, we can determine a range of best-fitting models. Preliminary results show that with some educated initial guesses on the hydraulic properties of the reservoir, we could reproduce the drawdown curves in two observation wells with a relative error below one percent after a couple of tens of simulations. The uniqueness of those results will be assessed during the discussion.