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

(‘complex’ and ‘barren’) as fixed factors, and ‘quadrant’ (Q1, Q2, Q3 and Q4) as the repeated measure. Data were square-root transformed and Greenhouse-Geisser correction was applied when epsilon was lower than 0.75 [33]. For significant sources of variation, the differences were highlighted by applying descriptive analysis [34]. Behavioral effects were evaluated using a multivariate approach. After the data on percent- age of behaviors were square-root transformed, they were used to build a resemblance matrix with Bray-Curtis distance. To test homogeneous dispersion, data were assessed using the PERMDISP procedure (permuted dispersion, which tests for homogeneity of dispersions). Following these procedures, PERMANOVA with 999 permutations [35] was carried out using the same model applied in univariate comparisons for swimming velocity and immobility time. For significant sources of variations, pairwise comparisons were performed and SIMPER analyses conducted to highlight the behaviors that most contributed to these differences. The univariate analyses (two-way and repeated measures ANOVA tests) were performed in the software Systat 12 and the multivariate procedures (PERMANOVA and SIMPER tests) in the software Primer 6 with PERMANOVA add-on. 3. Results Swimming velocity varied between habitat complexity depending on the temperature regime, both before and after mirror exposure (Table 1 and S1 Data). The differences between habitat complexity occurred only at 34 degrees. Fish from the complex habitat decreased velocity

before mirror exposure and those from the barren habitat increased it after the same exposure (Table 1 and Fig 2). For immobility time, we observed an effect only after mirror exposure, where fish from barren tanks showed a decrease at 34˚C (Table 1 and Fig 2). The time spent in different quadrants under the two temperature regimes depended on habitat complexity, both before and after mirror exposure (Table 2). Before exposure at 28˚C, the fishes spent more time in quadrant 4 (lower aggressiveness), with a greater difference observed when the habitat was barren (28C versus 28B). At 34˚C, although the fishes also spent more time in quadrant 4, the opposite was observed, with a larger difference when the Table 1. Two-way ANOVA to compare ‘Swimming velocity’ and ‘Immobility time’ between temperature (28 and 34˚C) and habitat structure (complex and barren) in the dusky damselfish before and after mirror exposure. Swimming velocity Before mirror (KS–p = 0.128; L–p = 0.718) After mirror (KS–p = 0.061; L–p = 0.493) Source DF MS F P MS F P Temp. 1 0.55 0.14 0.714 1.14 6.32 0.017 Compl. 1 2.14 0.53 0.471 2.39 13.20 0.001 T x C 1 33.60 8.34 0.007 2.81 15.55 < 0.001 Error 32 4.03 0.18 Before mirror (KS–p = 0.457; L–p = 0.141) After mirror (KS–p = 0.125; L–p = 0.373) Source DF MS F P MS F P Temp. 1 18.67 0.01 0.941 19.73 4.26 0.047 Compl. 1 4.08 0.00 0.973 24.43 5.27 0.028 T x C 1 12892.76 3.80 0.060 33.49 7.23 0.011 Error 32 108674.99 4.63

Square root-transformed data were used for swimming velocity after mirror exposure in order to achieve ANOVA assumptions. Bold p-values correspond to significant effects. KS–Kolmogorov-Smirnov test for normality; L–Levene test for homoscedasticity; Temp./T–temperature; Compl./C–complexity of the habitat. DF = degrees of freedom, MS = mean squared https://doi.org/10.1371/journal.pone.0235389.t001 PLOS ONE Temperature and habitat structure on damselfish behavior PLOS ONE | https://doi.org/10.1371/journal.pone.0235389 June 30, 2020 5 / 13