Thesis (M.Eng.)--Chulalongkorn University, 2004

Geostatistics offers a way to describe the spatial continuity of reservoir properties such as permeability and porosity. One of the most important parameters in reserve estimation is porosity; thus, it is chosen as a parameter of interest in this study. Realizations of spatial distribution of porosity can be generated from stochastic simulation at equal probability due to the inherent uncertainty in the data set. Anyway, it is impossible to use all realizations as input for reservoir simulation because it will consume too many resources. In this study, 60 realizations of porosity data set were generated from Sequential Gaussian Simulation. All realizations were denoised by multiresolution wavelet analysis, using Daubechies 4 wavelet. These denoised realizations were then compared and grouped based on correlation between each pair. The criterion used in the comparison and grouping is a cut-off value that is specified to select highly correlated realizations. The realization that has the largest number of similar realizations based on a certain cut-off value of correlation was chosen as the most representative realization and can be used as a porosity distribution reservoir model.