Multiple-Point Stastistics

Multiple-point Statistics (MPS) technique is used to simulate categorical variables and continuous variables through the simulation of a pattern. It tends to reproduce the proportions, the relationships between facies and the geobodies, from the training image. The MPS can take into account conditioning data, local proportions, local homothety, local rotation and trend variables.

The method is based on the DeeSse algorithm from the University of Neuchâtel in Switzerland. The approach's principle is to mimic a reference or training image. This image could be an analog if the geological environment is known. You can also start from a simple theoretical image of a geological concept and deform it through scaling and rotation, or applying a trend to build a more elaborate scheme. This option is also very useful to "extrapolate" mineral properties from a characterized mine to a similar (but different one) for instance. You can also force some grid cells to connect each other, or each simulation realization to respect input anisotropies or facies proportions. All that makes the DeeSse implementation very flexible, enabling the modeling of complex relationships between facies. DeeSse also applies to multivariate issues.

The whole Multiple-point Statistics process can be run by providing information in five successive windows:

• Input: Definition of the Input File and Variables, the Output Grid and the MPS Options
• Non Stationarity Option: Definition of the Proportions, Trend Variables and Connectivity Variable
• Distortion Option: Definition of the Scaling and Rotation
• Search Pattern: Definition of the search pattern, MPS Parameters and Display of one realization
• Output: Definition of the Simulation Parameters and Output Variables