Inversion of surface subsidence data to quantify reservoir compartmentalization: A field study

Abstract: Surface subsidence is in many cases an unwanted effect of subsurface exploitation. It can, however, also be used as a reservoir monitoring tool. The present paper reports a field study on the Roswinkel gas field, in which subsidence data were used to reduce the uncertainty about the reservoir architecture. The Roswinkel gas field in the northeast of the Netherlands was in production from 1980 to 2005. Located at about 2100m depth, it is a severely faulted anticlinal structure, constituting up to 30 reservoir compartments. As a result of its complexity there are large uncertainties about the fault transmissibilities and the strength of the connected aquifer. Consequently, it is possible there are undepleted compartments in the reservoir. Pressure depletion due to gas production causes the reservoir sandstones to compact, leading to surface subsidence. The gas production in Roswinkel has induced subsidence of approximately 17cm above the center of the field. The subsidence at any point on the surface is a result of compaction over a large area within the reservoir. We estimated the compaction in the reservoir using subsidence data in order to reduce the uncertainties about fault transmissibility and aquifer connection. The subsidence data had been obtained from leveling campaigns and satellite measurements (Persistent Scatterer InSAR). We used a previously developed Bayesian inversion method in which prior knowledge is combined with observations. The prior knowledge on the compaction and the associated uncertainties were generated by Monte Carlo simulations of the reservoir in which the fault transmissibilities and aquifer connectivity were varied. In the meantime, the geological reality was maintained and the production history was honored. The average prior compaction field was a relatively smooth field extending far into the aquifer and typically having an uncertainty of 40%. Our inversion results show a reservoir in which certain large faults divide the reservoir into compartments with different pressure histories. In addition, the aquifer activity appeared to be much weaker than the average prior knowledge suggested. The posterior uncertainty of the degree of compaction was reduced to about 10%. Our study demonstrates that a carefully executed inversion exercise can considerably reduce uncertainties, thus making it possible to identify possibly undepleted compartments in the reservoir. [Copyright &y& Elsevier]