1
Faculty of Chemical, Petroleum and Gas Engineering, Department of Petroleum Engineering
2
Institute of Geophysics, University of Tehran
10.22050/ijogst.2026.604515.1796
Abstract
Total organic carbon (TOC) is a key parameter for evaluating source-rock potential, but laboratory Rock-Eval measurements are scarce and costly. Log-based proxies such as the ΔlogR method of Passey et al. (1990) are therefore widely used. In this study, we compare six tree-based regression models, AdaBoost, Random Forest, Extra Trees, Gradient Boosting, LightGBM, and XGBoost, for predicting TOC from seismic attributes in the Groningen gas field, the Netherlands. The target is ΔlogR-derived TOC from four Groningen wells (TJM-1, SCB-1, SDB-1, and SDB-10), not laboratory TOC. Negative proxy values were removed, leaving 98,861 samples. Test metrics measure agreement with this proxy over the Rotliegend and the Carboniferous source-rock interval. Feature selection retained seven seismic attributes. On an independent 20% hold-out set, Random Forest, Gradient Boosting, and XGBoost achieved the lowest test MSE (0.0315) and the highest test R² (0.8900–0.8901). These scores describe agreement with the ΔlogR proxy in the four training wells; they do not establish accuracy against laboratory TOC or spatial generalization beyond those wells.
Alhasan, R., Saberi, M. H. & Riahi, M. A. (2026). Total organic carbon prediction using seismic attributes and well log data: A comparative study of six tree-based machine-learning models. (e256343). Iranian Journal of Oil and Gas Science and Technology, (), e256343 https://doi.org/10.22050/ijogst.2026.604515.1796
MLA
Alhasan, R., Saberi, M. H., & Riahi, M. A. "Total organic carbon prediction using seismic attributes and well log data: A comparative study of six tree-based machine-learning models." .e256343 , Iranian Journal of Oil and Gas Science and Technology, , 2026, e256343. doi: 10.22050/ijogst.2026.604515.1796
HARVARD
Alhasan, R., Saberi, M. H., & Riahi, M. A. (2026). 'Total organic carbon prediction using seismic attributes and well log data: A comparative study of six tree-based machine-learning models.', Iranian Journal of Oil and Gas Science and Technology, (), e256343. doi: 10.22050/ijogst.2026.604515.1796
CHICAGO
Alhasan, R., Saberi, M. H. & Riahi, M. A., "Total organic carbon prediction using seismic attributes and well log data: A comparative study of six tree-based machine-learning models.." Iranian Journal of Oil and Gas Science and Technology, (2026): e256343, doi: 10.22050/ijogst.2026.604515.1796
VANCOUVER
Alhasan R., Saberi M. H., Riahi M. A. Total organic carbon prediction using seismic attributes and well log data: A comparative study of six tree-based machine-learning models. IJOGST. 2026;():e256343. doi: 10.22050/ijogst.2026.604515.1796
