Damselfish face climate change: Impact of temperature and habitat structure on agonistic behavior

tank (no food was offered, thus it was recorded as a typical foraging behavior observed in nature). Locomotion was recorded from the overhead camera and analyzed using tracking software developed in MatLab [32]. The overhead view made it possible to divide the bottom of the tank into 4 quadrants: Q1 was the closest to the mirror, Q2 the side area where the mir- ror offered a more distant view of the image, and Q3 and Q4 were the back areas (Fig 1; see video at https://www.youtube.com/watch?v=wz_aOkiunOA&feature=youtu.be). The time spent in each quadrant and average swimming speed were measured. For speed, calculations are performed on a series of frames to produce quantified measurements of the animal behav- ior. It is known the position of the animals for each frame of the video, and the number of frames per second. Thus, the series of frames is analyzed, and the number of frames changed in a certain time is used to estimate the animal’s movement. 2.3 Statistical analysis All comparisons were performed through independent analyses before and after mirror expo- sure. We decided not to use ’time’ as a factor due to the dependence between two periods and because analysis of time spent in each quadrant (see details below) would result in two depen- dent factors in the same analysis (’time’ and ’quadrant’). The effects on mobility were evaluated using swimming velocity and immobility time. Each variable was compared separately between ‘temperature regime’ (fixed, ‘28˚C’ and ‘34˚C’) and

‘habitat complexity condition’ (fixed, ‘complex’ and ‘barren’), and the respective interaction, applying two-way ANOVA. Data were checked for normality (Komogorov-Smirnov test) and homoscedasticity (Levene test), and when a divergence was observed (swimming velocity and immobility time after mirror exposure) square-root transformation was applied. For signifi- cant sources of variation, posthoc pairwise comparisons were evaluated using the Student- Newman-Keuls test. The effects on aggressive behavior were assessed considering the time spent in each quad- rant, related to mirror position, as a proxy of aggressiveness level (see methods). Since the time spent in one quadrant is dependent on the others, repeated measures ANOVA was performed for each of them, with ‘temperature regime’ (‘28˚C’ and ‘34˚C’) and ‘habitat complexity’ Fig 1. Schematic overview of the tank (15L, 40x20x25 cm) used for the mirror test. Walls were covered with opaque white film, except the front and mirror walls. The wall that allowed mirror view was covered for the first 5 min of behavioral recording (basal behavior). The cover was then removed and the fish had access to its own image for 5 min. Behavior was recorded from the front to analyze aggression, and from above to assess fish distribution in the tank. Remaining longer in quadrant 1 (Q1) indicates close contact with the mirror image, followed by Q2, while lingering in Q3 and Q4 suggests greater distance from the mirror image. https://doi.org/10.1371/journal.pone.0235389.g001 PLOS ONE Temperature and habitat structure on damselfish behavior

PLOS ONE | https://doi.org/10.1371/journal.pone.0235389 June 30, 2020 4 / 13