Consistent selection towards low activity phenotypes when catchability depends on encounters among human predators and fish

Abstract

Together with life-history and underlying physiology, the behavioural variability among fish is one of the three main trait axes that determines the vulnerability to fishing. However, there are only a few studies that have systematically investigated the strength and direction of selection acting on behavioural traits. Using in situ fish behaviour revealed by telemetry techniques as input, we developed an individual-based model (IBM) that simulated the Lagrangian trajectory of prey (fish) moving within a confined home range (HR). Fishers exhibiting various prototypical fishing styles targeted these fish in the model. We initially hypothesised that more active and more explorative individuals would be systematically removed under all fished conditions, in turn creating negative selection differentials on low activity phenotypes and maybe on small HR. Our results partly supported these general predictions. Standardised selection differentials were, on average, more negative on HR than on activity. However, in many simulation runs, positive selection pressures on HR were also identified, which resulted from the stochastic properties of the fishes' movement and its interaction with the human predator. In contrast, there was a consistent negative selection on activity under all types of fishing styles. Therefore, in situations where catchability depends on spatial encounters between human predators and fish, we would predict a consistent selection towards low activity phenotypes and have less faith in the direction of the selection on HR size. Our study is the first theoretical investigation on the direction of fishery-induced selection of behaviour using passive fishing gears. The few empirical studies where catchability of fish was measured in relation to passive fishing techniques, such as gill-nets, traps or recreational fishing, support our predictions that fish in highly exploited situations are, on average, characterised by low swimming activity, stemming, in part, from negative selection on swimming activity.

Figure 1. Schematic daily squared displacement over time of a fish moving with a home range (HR).

Schematic representation derived from empirical observations using acoustic telemetry techniques in the marine small-bodied fishes Serranus scriba, Serranus cabrilla and Xyrichthys novacula , , . In all cases, the squared distances reached an asymptote as a characteristic of the general mechanism of the HR behaviour . Two general patterns can be derived from the within-population variability observed: there were some individuals that reached an asymptote at the same time (denoted by the vertical dashed bar) but had different HR sizes (approximated by the ratio ε/k, see methods section, panel B), and other individuals with the same HR size (denoted by the vertical dashed bar) but with a different way of exploring the whole HR (approximated by the parameter k, see methods section, panel C).

Figure 2. Spatial scenario considered for the IBM simulations.

Grey points represent the centre of the HRs of the 1,000 simulated fishes following the biased-random walk described in the methods section. The path (individual Lagrangian trajectory) for one fishing trip (12 time steps) of the fishers following the behaviours described in the M&M are shown: the fisher fixed in one spot (FP), the fisher moving randomly along the edge of the scenario (EM), the random search pattern (RSP) and the Lévy search pattern (LSP). A magnification of the RSP and LSP (at the same spatial scale) is provided to improve visualisation and to show the distribution of the distance travelled between two consecutive time steps during one simulation, presented on a log-scale (m).

Figure 3. Harvesting efficiency (approximated by the fishing-trips needed to finish the simulation) related to the fisher behaviour.

Box-plots of the harvesting efficiencies within 100 simulation runs until the virtual fish population collapsed to <10% of the initial size: one fixed spot (FP, green), edge-based random movement (EM, orange), random search pattern (RSP, purple) and Lévy search pattern (LSP, turquoise). The figure shows the box-plots of the days needed to overexploit the fish stock for each fisher style as a proxy of harvesting efficiency.

Figure 4. Distribution of mean-standardised selection differentials on the mathematical descriptors of movement.

Frequency distributions (n = 100) of the mean-standardised selection differentials (S_μ) calculated in each iteration for each movement characteristic (radius of the HR and the activity approximated by K) in the different fisher styles (left panels). Black dashed vertical lines mark the level of no selection (S_μ = 0), and grey dashed vertical lines mark the mean-S_μ calculated for the 100 iterations performed for the different fisher styles: the fisher fixed in one spot (FP), the fisher moving in the edge of the scenario (EM), the random search pattern (RSP) and the Lévy search pattern (LSP).

Figure 5. Lagrangian trajectories of the population-average fish in the original and exploited populations.

The path of the first 500 time steps of an individual with average movement characteristics (ratio ε/k and k) sampled from the un-exploited population (in blue) and from the exploited population by the different fisher styles: the fisher fixed in one spot (FP), the fisher moving in the edge of the scenario (EM), the random search pattern (RSP) and Lévy search pattern (LSP). Note the change in the tortuousity degree and a needing for more time steps to reach an asymptote (HR) in the spatial-utilisation in the exploited average fish.