Ecological and Economic Predictors of Métiers in a Mixed Fishery

Marine ecosystem-based management requires the understanding of species interactions and what species are harvested together. This study combines two major questions: the first regarding what drives the probability that a m & eacute;tier (species assemblages, with spatial distribution and seasonality) will be observed as catch, and the second regarding the level of control fishers have over this catch mix. To address these questions, we analysed highly resolved logbook records of an Arctic and sub-Arctic industrial demersal fishery operating in Icelandic waters. The study employs a multi-class random forest model to identify predictors of m & eacute;tier occurrence and consistency of predictions using a dataset of >100 000 hauls over 4 years (2016-2019). The overall accuracy of the random forest model is 69-70%, indicating moderate predictability of catch mix based on known environmental, vessel, and company characteristics. We find that habitat-related variables (depth and temperature) are most important to predict catch mix. Still, company, trip, and vessel characteristics are also very important (e.g. vessel and trip length, distance to port). Beyond these more traditional bio-economic variables, important predictors include variables related to harvesting strategies, such as quota diversity and a vessel's mobility. These findings contribute to a fuller picture of fisher decision-making in mixed fisheries.