Machine Learning-Aided Seismic Mapping of Deepwater Turbidites in the Dangerous Grounds Region, Offshore Northwest Sabah, Malaysia

Summary This study utilizes 3D seismic and supervised machine learning to identify and classify deepwater turbidites in the frontier Dangerous Grounds region, offshore northwest Sabah. The time mapping of the seabed reveals prominent seafloor fairways and canyons that flow towards the deep sea. We identified two phases of unchannelized turbidites deposition based on stratigraphic position and attribute expression. The older buried turbidite shows a higher amplitude response and appears to have been affected by faulting. In contrast, the younger modern turbidite exhibits a lower amplitude response and has more continuous reflectors. We extracted 35 seismic attributes and labeled one inline of the 3D volume to predict turbidites and non-turbidites facies. Support vector machine (SVM), random forest (RF), and neural network (NN) classifiers were trained. All three classifiers showed excellent classification performance (above 99%), with the NN outperforming SVM and RF.