Model-based explanation of uncertain observations [marine oil spill tracking]

Given a model for some process, the reliability of predicting the future behavior of the process through the model depends on how the past observations of the process were understood. Typically, the observational data is uncertain and incomplete while the model is crisp in the sense that the input and output of the model are precise. In explaining observations, the authors wish to construct a plausible history of the process that is consistent both with the observations and the model, in this paper, the authors propose a general architecture to explain uncertain observations through a model. The authors illustrate the validity of the approach by applying this architecture to marine oil spill tracking problem