A Python Based Multi-Point Geostatistics by using Direct Sampling Algorithm

Abstract

Multi-Point Geostatistics (MPS) is a type of geostatistical method used to estimate the value of an unsampled location by utilizing several data points around it simultaneously. The MPS method estimates it by defining a model based on initial data in the form of a training image, which is a collection of data in the form of a geological conceptual model in the research area with the integration of geological and geophysical knowledge. The MPS method is currently starting to develop because it differs from conventional covariance-based geostatistical methods such as simple kriging and ordinary kriging, which only use a variogram based on the relationship between two points rapidly. In this study, we evaluated the use of the MPS method by using a direct sampling algorithm with Python that will directly sample the training image and then retrieve the data based on the sample data. A braided channel training image is used as the initial model to estimate the distribution of reservoir properties in lithology with sand and shale types. This study shows that MPS could reconstruct geological features better than kriging.