Direct updating of geostatistical reservoir models using iterative resampling with DISPAT

In many practical cases, a single legacy reservoir model generated through tedious manual construction/adjusting to match geological, geophysical and production history requires updating because of the availability of new production data. Current approaches such as assisted history matching, model updating using filtering technique and others are often not practical for two reasons: 1) they require geological modeling parameters as input and 2) they require an explicit parameterization of the model to generate stochastically updates through sampling or optimization. In many cases, the initial geological parameters such as variogram model, Boolean model may simply no longer be available (or have never been used in the first place) and hence need to be reconstructed. Parameterizations are desirable from a theoretical point of view but may require complex software coding since such parameterizations can be very case-specific. In this paper, we propose an alternative approach that, while being an approximation to the ideal case of parameterization/sampling, can provide a quick solution to the problem of model updating with a legacy reservoir model. Realizing that all real reservoir models require a non-stationary modeling approach, we use the current existing reservoir model as a training image in a non-stationary geostatistical algorithm termed dispat. Using iterative spatial resampling, we then update the current legacy model with the additional production data. In this paper, we perform some initial investigation on this idea and provide a proof of concept. Finally, we outline a plan for future work with this idea.