We propose an application of the pattern-oriented modeling strategy to the elaboration of a spatially-explicit agent-based model of marine traffic in the context of marine ecosystem management. Our study area, the Saguenay/St. Lawrence Marine Park, is one of the best sites in the world for whale-watching cruises. During the tourist season, up to 50 cruise operators may have to cooperate to locate and observe whales. The sequences of their decisions at sea control the spatiotemporal dynamics of their excursions. Following interviews with cruise operators, several hypotheses and decision-making models were developed to reproduce their behavior during whale-watching excursions. Spatiotemporal features of excursions in the marine park are sampled on a yearly basis by the GREMM (Group for Research and Education on Marine Mammals) and Parks Canada. The whole dataset consists of sampled time series of excursion activities, including GPS records, the number of marine mammals observed as well as environmental conditions such as visibility. Excursions were characterized using patterns extracted from this dataset. Considering these patterns as observable footprints of cruise operators' underlying decisions in a given environmental setting (weather, whale abundance and spatial distribution...), we used them to select the fittest combination of hypotheses and decision-making models among a set of multiple alternatives.