| dc.description.abstract | Porosity is a key rock property that influences reservoir quality and plays an important role in petroleum exploration and production. The most common sources of information for reservoir characterization are well and seismic data. Well data, such as logs, provide appropriate vertical resolution, but the wells are separated by a wide distance. Three-dimensional seismic data, on the other hand, may be used to generate a more thorough reservoir characterization in the inter-well space. However, the resolution limitation intrinsic to seismic data is a key issue in the delineation of reservoir properties. Seismic inversion is a vital approach for determining indirectly attributes such as P-impedance and porosity from seismic data. There are various types of seismic inversion approaches, with model-based inversion receiving the most prominence. In this thesis, I analyze the porosity of an deltaic sandstone reservoir from offshore Netherlands, using the public F3 subsurface dataset. The first goal of this study is to predict porosity through seismic inversion followed by collocated co-kriging. To estimate acoustic impedance and porosity in the inter-well zone of the F-3 block, the model-based inversion has been applied. Two low frequency models based on different approaches are used to understand the influence of their accuracy on the porosity result. The second part of the thesis compares the porosity estimation from the previous approach with the porosity cube estimated using a machine learning, neural network approach. Both porosity cubes are based on the same acoustic impedance cube which is a part of the published F3 data set. The main result of this comparison is that machine learning / neural network does not provide a significant different result compared to than the geostatistical approach. | |