In order to efficiently model fluids flows in heterogeneous reservoirs, one must inevitably face the problem of correctly estimating parameters for the equations governing the flow. Reaching this objective requires to use different types of measurements or interpretations, acquired at different scales and through time, driven as much as possible by auxiliary data, such as geological concept and seismic imaging. However, the access to all those data in the same location and at the same time is usually not possible, leading to a high uncertainty for the numerous flow parameters. Observatories in groundwater reservoirs are a new way to use analogues in the petroleum industry, as : 1) they offer an easier accessibility to the reservoir than buried oil fields; 2) they allow simplifications of this problem, thanks to single phase flows.
On the Céreirède experimental site (Lattes, Montpellier, South of France), a unique playground was offered to face the question of quantifying the permeability anisotropy in a fluvial depositional analogue. 14 wells were drilled in a a 50 x 50 m area (about the surface of one gridblock in a petroleum reservoir model),with an inter-well distance ranging from 2 to 20 m. The shallow water-bearing reservoir is composed of 2 meters of unconsolidated and very permeable gravels belonging to quaternary alluvial sediments.
Well tests and interference tests were conducted on all 14 wells. Pressure Transient Analysis (PTA) of reservoir pressure behaviour provided permeability and other related information on reservoir architecture (such as no flow boundaries or lateral permeability changes between wells). A trend in permeability was observed from South-East to North-West of the site, and anomalies of the pressure evolution showed an anisotropic behaviour. An electrical survey was also conducted, showing an abrupt change of resistivity from South to North.
The integration of all these data through a 3D geological model of the Céreirède site was carried out. At a first step, a simplified transmissivity map was painted in the model from ground observation, to initialise the dynamical simulations. Then a direct inverse method, the DSM “Differential System Method” was applied. This technique allows the computation of the transmissivity field in a confined aquifer when several “independent” data sets are available, (Parravicini et al. (1995) and Giudici et al.(1995)). Following calibration and sensitivity analyses, the model was tested by matching the pressure history using a pressure solver. The workflow based on a loop process delivered the updating of the permeability model. At the very end, a very good match between simulated and actual water table evolution was obtained, with a permeability field very consistent with the depositional scheme of the sediments.
The authors thank TOTAL for funding this R&D project and giving permission to publish this paper.