Predicting sand composition (QFR) delivered into Miocene GOM basin using basin catchments features and leveraging machine learning algorithm

Reservoir quality uncertainty can significantly impact original oil-in-place and ultimate recovery estimations for a given prospect/play. Exploration teams attempt to predict/constrain reservoir quality properties using available data and analogs; however, this can be challenging in frontier basins where rock data is limited. Knowing the mineral composition of the sediments entering the basin is helpful to assess and forecast reservoir quality in the absence of nearby well control. We present a model that leverages machine learning to predict the key framework components of sand delivered to a basin and we demonstrate and further validate the model using a Miocene Gulf of Mexico case study.