Abstract
This article describes a method for the quantitative analysis of fish behaviour relative to selection devices in trawl gears. Based on video observations, the method estimates probabilities for a given event to happen and establishes behavioural tree diagrams representing and quantifying behavioural patterns in relation to the selection device under assessment. Double bootstrapping is used to account for the uncertainty originating from a limited number of fish observations and the natural variation in fish behaviour. The method is used here to supplement standard analysis of catch data for the performance assessment of a flatfish excluder (FLEX). The Baltic Sea trawl fishery targeting cod (Gadus morhua) provides the pilot case. Results obtained by comparing catches with and without FLEX installed revealed that >75% of bycaught flatfish individuals escaped through the device, while no evidence was found that catches of cod in the targeted sizes were reduced. The behavioural analysis produced values of escape efficiency comparable to those obtained in the catch analysis. Furthermore, it revealed that ∼80% of the flatfish went calmly into the excluder, while most of the roundfish displayed avoidance swimming reactions. The method provides quantitative information of fish behaviour that can be relevant for developing and optimizing selection devices.