Bayesian sunken oil tracking with SOSim v2: Inference from field and bathymetric data.

Sunken oil is often difficult to detect, and few oil spill models are designed to locate and track such oil. Therefore, the multi-modal Bayesian inferential sunken oil model, SOSim (Subsurface Oil Simulator), was expanded in this work for use during emergency response and damage assessment. Rather than requiring hydrodynamic data as input, SOSim v2 accepts available field concentration data, along with default or custom bathymetric data, for inference of the location and trajectory of sunken oil. Novel aspects include inference based on bathymetry and the Coriolis Effect, by constructing a prior likelihood function from sampled bathymetric data, scaled proportionally with field concentration data. SOSim v2 is demonstrated versus field data on the ITB DBL-152 oil spill in the Gulf of Mexico, with sensitivity analysis. Results suggest that the inferential approach presented can be effective for modeling relatively slow-moving pollutant masses such as sunken oil, when field concentration data are available.

A review on the sinking mechanisms for oil and successful response technologies.

A rise in the shipping of heavier hydrocarbon products increases the potential for an oil to sink after a spill. Further, sunken oil is difficult to locate and recover, and appropriate response technologies depend on the sinking mechanism. In this review, principal sinking mechanisms for oil are described and appropriate response technologies are suggested. Then, models appropriate for tracking sunken oil are compared. Oil may sink as burn residue, microscopic oil-particle aggregates (OPAs) or macroscopic oil-sediment mixtures (OSMs), marine oil snow during a MOSSFA event, or due to its high density. The most common mechanism is by sediment entrainment, and in such scenarios manual recovery has been reported as a successful response option. Among oil tracking models, trajectory models and Bayesian oil search models are compared for sunken oil capabilities. Many oil spill models require hydrodynamic inputs, whereas Bayesian models infer parameters based on available field concentration and bathymetric data.