Abstract
This study represents part of a research project focused on the Runaway Surface (MFS53), which is one of the reservoir levels in the Boonsville Field of the Fort Worth Basin in north central Texas, USA. This reservoir system is a subunit of the Bend Conglomerate section, which is a productive series of gas reservoirs deposited during the Middle Pennsylvanian fluvio-deltaic environment. This paper adopts an integrated approach to the seismic and well log data, using a combination of geostatistical and multi-attribute regression transform methods. The main focus of the research is to accurately predict the porosity distribution of the Runaway Formation away from the well locations. The input data consists of 32 wells (of which six wells contain porosity logs) together with a 3D seismic volume of Boonsville Field. The 3D seismic volume was inverted to obtain the acoustic impedance cube of the study area. Secondly, six multi-attribute data slices were extracted from both surface seismic and inverted acoustic impedance volumes. Subsequently, the porosity distribution of a selected reservoir level was estimated over the full extent of the study area using both acoustic impedance alone, and multi-attributes. Initially, the multi-attribute transform algorithm was trained using the well log data. The porosity at each well location was averaged over a particular depth zone of interest, and then compared with six extracted attribute slices averaged over the same depth window. In order to select the appropriate number of attributes for analysis, a cross-validation process was followed. The results of this cross-validation process and the training of the multi-attribute transforms were applied to the extracted attribute slices in order to produce the final porosity map of the Runaway Formation. The cross-plot between the seismically derived porosity and the well porosity values showed that accuracy of porosity prediction was increased from 75 %, when using a single attribute (acoustic impedance (AI)), to 90 % when multiple attributes are used. Additionally, when the actual well-derived porosities were overlaid onto the final predicted porosity maps from both techniques, a significant amount of mismatching was observed on the porosity map derived from AI alone, whereas the predicted porosity with the multi-attribute regression transform was a close match to the actual well-derived porosities. Beside this, the subsurface geology (i.e., karst collapse features) were not clear in the porosity map deduced from AI. Based on these cross-correlation results, the porosity map derived from multi-attribute regression transform was selected. The high level of correlation (90 %) with the actual and derived porosity indicates that the seismic multi-attributes were reliably transformed to the reservoir porosity log. The derived porosity map for the Runaway Formation indicates high lithological variation within the reservoir level with a porosity generally varying between 2 and 32 %. The western portion of the Runaway Formation is highly porous and can be considered for future exploration purposes. Although this study retains a certain level of uncertainty, which can be attributed to the well and seismic data used, due to data limitations, uncertainty analysis was not included in the current study, but this should be considered in future studies so as to improve the porosity prediction.