The effective seismic attributes in porosity prediction for different rock types: Some implications from four case studies

Abstract

The present study evaluates the performance of PNN models for porosity prediction using seismic attributes. Four seismic datasets and more than 20 wells from different sedimentary basins located in Libya, Iraq, Egypt and USA are employed to characterize the effective attributes for porosity prediction. Verification and testing error analysis is adopted for evaluating the prediction performance. Results indicated that the porosity prediction models are primarily dependent to the propagation related attributes with frequency attributes as the most effective parameters in porosity prediction. In addition, the data quality and processing history strongly control the prediction model performance with relatively limited effects to dataset dimensionality (2D versus 3D) and the number of wells utilized in model construction. Such results are important to better understand and evaluate the performance of PNN porosity prediction models using various seismic attributes.