Supplementary Materials - Chapter 3

Figure S1: Clay fruit models showing characteristic attack marks by different taxa. Fruit models numbered 1–27 display peck marks consistent with avian frugivory; models 28–30 show bite marks attributed to alien monkeys; models 31–34 show bite marks attributed to alien rats; and models 35–36 display damage consistent with insect attacks. Photo credit: Ghanishta Seeburrun. Figure S2: Correlation between acoustic indices and total bird detection rate. (a) Correlation between acoustic diversity index (ADI) and total bird detection rate, (b) correlation between acoustic complexity index (ACI) and total bird detection rate. Figure S3: DHARMa residual diagnostics for the generalised linear mixed model of frugivory. Quantile–quantile (QQ) plot of simulated residuals and residuals plotted against fitted values. Diagnostics indicate no significant deviations from model assumptions, with no evidence of overdispersion, outliers, or systematic patterns in residuals. Figure S4: Sensitivity analysis assessing site effects on drivers of frugivory. (a) Predicted frugivory rates from generalised linear mixed models are shown as a function of standardised Acoustic Complexity Index (ACI) for weeded (green) and invaded (yellow) forests at Brise Fer and Mare Longue. Shaded areas represent 95% confidence intervals; (b) Model comparison results for sensitivity analyses testing the inclusion of site and site-specific interactions. Models were compared using Akaike’s Information Criterion (AIC) and likelihood ratio tests (LRT). ΔAIC values are calculated relative to the base model. Non-significant LRTs and higher AIC values indicate that including site or site-specific interactions did not improve model fit. Table S1: Frugivory rates in weeded and invaded forests at Brise Fer and Mare Longue. (a) Mean monthly frugivory rates (± SE) calculated as the proportion of attacked fruits per tree, based on 11 months of sampling. (b) Wilcoxon rank-sum tests comparing frugivory rates between weeded and invaded forests for each site. W denotes the Wilcoxon test statistic. Table S2: Monthly differences in frugivory rates between weeded and invaded forests at each site. Wilcoxon rank-sum tests were used to compare monthly frugivory rates between forest types for each site. Analyses were based on 11 months of sampling. W denotes the Wilcoxon test statistic. Table S3: Origin and diet composition of bird species recorded in the study. Table S4: Mean monthly detection rates of bird species across weeded and invaded forests at Brise Fer and Mare Longue. Values represent mean detection rate ± standard error (SE), calculated as the proportion of 1-minute recordings in which a species was detected. Species marked with an asterisk (*) are non-native birds. Table S5: Species-specific differences in detection rates between weeded and invaded forests. Wilcoxon rank-sum tests were used to compare monthly detection rates of individual bird species between weeded and invaded forests, conducted separately for each site. P-values were adjusted for multiple testing using the false discovery rate (FDR) method. Significant differences are in bold. Species marked with an asterisk (*) are non-native bird species. Table S6: Effects of forest type on bird community composition at Brise Fer and Mare Longue. (a) Permutational multivariate analysis of variance (PERMANOVA) testing for differences in bird community composition between weeded and invaded forests based on Bray–Curtis dissimilarities (999 permutations). R² indicates the proportion of variance explained by forest type. (b) Permutation tests for homogeneity of multivariate dispersions (PERMDISP), showing no evidence that PERMANOVA results were driven by differences in within-group dispersion. Table S7: Species most strongly associated with community differences, based on permutation-based vector fitting (envfit). r² values indicate the strength of association between species and community structure. Asterisks* represent non-native bird species. r² values indicate the strength of association between species and community structure. Asterisks* represent non-native bird species. Table S8: Acoustic indices in relation to bird detection rate. (a) Spearman rank correlations between acoustic indices and total bird detection rate. Spearman’s rank correlation coefficients (ρ) and associated p-values are shown for associations between acoustic indices (ACI, ADI, BI, NDSI) and total bird detection rate across all recordings; (b) Spearman rank correlations between acoustic indices and total bird detection rate by forest type. Spearman’s rank correlation coefficients (ρ) are shown for associations between acoustic indices and total bird detection rate calculated separately for weeded and invaded forests. Table S9: Results of the final generalised linear mixed model (GLMM) examining drivers of frugivory. The model was fitted using a negative binomial error distribution with a log link function. The response variable was the number of attacked fruits per tree, with the log-transformed number of available fruits included as an offset. Month was included as a random intercept. Continuous predictors were centred and scaled prior to analysis. Effect sizes are presented as regression coefficients (β ± SE), incidence rate ratios (IRR = expβ), and 95% confidence intervals.