Pirate probabilities: On the importance of the dynamic coupling of meteorological and intelligence information for piracy interdiction
The heart of the effort is the construction of a model of pirate tactics, techniques, and procedures that take into account intelligence information (e.g. pirate CONOPS, equipment performance, base locations, numbers) and environmental information (e.g. winds, waves, and currents). The pirate model predicts the track of a single pirate skiff as a function of time. The uncertainty associated with the intelligence and environmental information is accounted for by running the pirate behavior model not once, but tens of thousands of times sampling from all available distributions, whether they be a subjective distribution of pirate base location or an objective distribution like a wave ensemble. This generates millions of possible pirate locations, and these millions of points can be interpreted as draws from a distribution function.
The resulting distribution function can be updated in real time should observations of pirate activity become available using particle filter techniques, it can also be combined with probabilistic information about commercial shipping activity to produce an estimate of the probability of a pirate attack.
The product is being operationalized at the Naval Oceanographic Office for dissemination to US and NATO forces off the Horn of Africa to aid in pirate interdiction efforts.