A Bayesian network model for estimating the combined risk in Northeast Passage escort operations

Abstract

Escort and convoy operations are commonly employed and highly effective strategies in the Arctic, particularly when ship navigating independently becomes challenging due to adverse ice conditions. Nonetheless, these operations are also among the riskiest, given their potential to lead to collisions between ships and icebreakers, as well as ships besetting in ice. Consequently, a robust estimation of the combined risk involved in escort/convoy systems is of paramount importance. To this end, this paper introduces a Bayesian network model to address the combined risk in escort operations, which considers both ship besetting risk and ship-to-icebreaker collision (STIC) risk factors in a unified framework. This model considers technical, environmental, human, and organizational risk factors. The model's practical applicability is demonstrated through an analysis of a genuine voyage along the Northeast Passage in August 2015. Additionally, a comparison of model outputs with captains' judgments across 14 crafted scenarios has been performed. The proposed model suggests that the main factor causing ship besetting in ice is the ice concentration, while the distance between the icebreaker and the ship is the critical factor affecting STIC risk. According to experts, visibility outweighs both the ice channel radius and navigation experience in terms of STIC risk in escort operations.