The geostatistical method of kriging, conditioned with geological information, has been used to indicate potential methods for characterising permeability of reservoir domains. In this method, the hard permeability data (eg. core data) obtained from drilled wells needs to be supplemented with "soft" data between wells. This paper examines the way in which soft data may be inferred from statistical distributions and how geological controls are used to develop soft data for enhancing the hard data.
A minipermeameter was used to develop examples of permeability data at different scales. Data from outcrops and in the laboratory on rock slabs were used for the evaluation of geostatistical methods for representation of permeability. Monte Carlo methods were also used to generate data of different statistical realizations to mimic some probability distribution functions of permeability which might derive from different depositional environments.
This study has demonstrated how use of geological controlled soft permeability representation from appropriate data sets leads to improved reservoir simulation models and performance prediction. It has also shown that compared to the conventional geostatistical method of kriging, the ‘conditional kriging’ approach is much superior.