Fig. 4.
From: Honeybee communication during collective defence is shaped by predation
Stinging response as a function of alarm pheromone concentration (first column) and performance (second column) in modelled colonies facing different environmental pressures. Triangles denote colonies trained for 80,000 trials with parameters: N=100,sth∈(16,40),k=1,tatt=0,rf=0 and Δtv=10. They are compared to colonies that a invest in guards to detect the predator earlier (tatt=60); b have higher false alarm rates (rf=0.3,0.6); c face only weak predators (sth∈(7,16)); d face a non-uniform (n-u) distribution of predators, in which weak predators (sth∈(16,26)) appear 4 times more often than strong ones (sth∈(27,40)); and e need more time to visually detect that the predator is escaping (Δtv=20). Only one parameter is varied in each comparison. Panel f compares colonies that defend small and large territory areas, which we model by setting tatt=Δtv=10 and tatt=Δtv=40, respectively. In all plots of the first column, percepts in which the probability of stinging remained as initialised (ps=0.5) because they were never reached are not included. Shaded areas indicate the sth range. Markers are at the end of each percept’s bin. Probabilities ps for percept vESC are given in Table 1. Average ± one standard deviation for 50 independently trained populations. In the second column, panels a’, b’, d’ and f’ display the percentage of live bees at the end of encounters, depending on the predator resistance sth. The upper bound indicates the optimal performance for the given scenario. Panels c’ and e’ show the total number of bees stinging as a function of predator resistance, and the dashed lines indicate the maximum number of times bees could sting before percept vESC is activated. Numbers below this boundary indicate self-limitation based on pheromone concentration. Average ± one standard deviation in the last 500 trials of 50 independently trained populations