Machine learning integration with seismic attributes for lithofacies prediction and distribution in Abu Madi Reservoir, onshore Faraskour Gas Field, East Nile Delta, Egypt

The complex lithological heterogeneity of Abu Madi Formation introduces substantial challenges for accurate reservoir evaluation, furthermore the lack of core data, due to the expensive nature of core acquisition and its sparse coverage, undermines the reliability of conventional interpretation methods. In this study, an integrated workflow that merges supervised machine learning and two-dimensional seismic attributes was applied to delineate lithofacies distribution and evaluation of Abu Madi Formation within Faraskour gas Field in the onshore Nile Delta. Five supervised machine learning algorithms, including Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Neural Network (NN), were implemented and compared to choose the most effective approach for accurate and reliable lithofacies prediction. Conventional well logs such as gamma ray, density, neutron, and resistivity logs, which are more accessible and cost-effective alternative, were utilized as input features, while lithofacies interpretations derived from the limited available core data served as ground truth labels. The models were applied using data from four wells that fully penetrated the Upper Abu Madi (UAM), Lower Abu Madi 1 (LAM1), and Lower Abu Madi 2 (LAM2) units of the formation. Two wells (SF-3 and SF-4), that contained both well log data and corresponding core-based lithofacies labels, were used for the models training and validation. The RF model outperformed the others, achieving 79% cross validation accuracy and 84% blind test accuracy, supported by high F1-scores and confusion matrix performance across all classes, which was subsequently used for the lithofacies prediction of the un-cored intervals in the available wells. The lithofacies prediction revealed four distinct lithofacies associations, sandstone, siltstone, shale and marl, associated with the fluvial to estuarine depositional environments. Seismic attributes such as relative acoustic impedance, RMS amplitude, envelope and instantaneous frequency were extracted to validate the distribution of predicted lithofacies and identify zones of hydrocarbon potential. The proposed methodology proved its effectiveness in lithofacies classification and characterization of Abu Madi Formation, providing a valuable tool for guiding exploration and development decisions in data-constrained settings.