Bayesian submerged oil tracking with SOSim: Inference from field reconnaissance data and fate-transport model output

Abstract

When spilled oil collects at depth, questions as to where and when to dispatch response equipment become daunting, because such oil may be invisible by air, and underwater sensing technology is limited in coverage and by underwater visibility. Further, trajectory modeling based on previously recorded flow field data may show mixed results. In this work, the Bayesian model, SOSim, is modified to locate and forecast the movement of submerged oil, with confidence bound, by inferring model parameters based on any available field concentration data and the output of one or more deterministic trajectory models. Novel aspects include specification of a prior likelihood function, and generation of results in 3-D from data in the 2-D density space of the isopycnal layer containing oil. The model is demonstrated versus data collected following the Deepwater Horizon spill. This new inferential modeling approach appears complimentary to deterministic methods when field concentration data are available.