Dry season assemblage as observed (six in our example). This was repeatedly done for each stream based on the species that we observed in the stream. For each stream and season we first calculated FD of the tadpole null assemblages in the seasons and then the change in FD from the wet to the dry season, also based on the null assemblages. To calculate the relative change of diversity from the wet season (“wet”) to the dry season (“dry”) we developed an approach using ??(1 ry/wet) resulting in positive values indicating an increase and negative values indicating a decrease in the respective diversity measure. FD was calculated following Petchey Gaston [43, 44]. This method applies a three-step dendrogram based classification function in which a species trait matrix is used to calculate a pair-wise species distance matrix. This matrix is used to construct dendrograms of specific species assemblages. The total branch length needed to connect all species in the assemblagePLOS ONE | DOI:10.1371/journal.pone.0151744 March 25,5 /Seasons Affect Functional and Phylogenetic Diversityrepresents the respective FD. Best distance measure (Gower’s distance) and cluster method (UPGMA) were identified following Mouchet et al. [45]. Our trait matrix consisted of categorical and continuous morphological trait variables (oral disc, body shape) of ecological relevance (i.e. feeding, microhabitat choice; [46]) for all species (for a list of traits used, see [25]). We calculated FD and the relative changes in FD for the observed assemblages, and for null assemblages (see above). We compared relative changes of observed and null assemblage FD using paired t-test. Accordingly, significant differences show that seasonal changes in FD are not random. Non randomness can be VP 63843 biological activity caused by differences in redundancy patterns, or by high or low FD in one or both seasons [24]. We therefore analysed data of the wet and the dry season separately. We applied polynomial regression with observed FD as dependent and SR (linear term) and SR^2 (polynomial term) as independent variables to reveal possible patterns of functional redundancy in each of both seasons (i.e., if nonlinearity is found). To test whether the full polynomial model or a simplified (i.e., linear) model performs better we used stepwise deletion (polynomial term first) and compared the models based on Akaike’s Information Criterion (AIC, [47, 48]) until the minimum adequate models were reached. Residuals were checked using diagnostic plots. We used Moran’s I autocorrelation coefficient [49] to prove that there is no spatial autocorrelation of the study sites regarding SR, FD, and PD, and their respective changes between wet and dry season. In this process no functional redundancy is indicated if these analyses show only a linear relationship between FD and SR (i.e., if the linear term in the polynomial model happens to be significant and the polynomial term to be non significant). Functional redundancy is indicated if the slope of the relationship between FD and SR decreases with increasing SR (i.e., if the polynomial term in the model is significant), caused by a stronger overlap of ecological traits with increasing number of AKB-6548 supplement syntopic tadpole species. We compared observed FD data with the respective null model data using paired t-tests, SART.S23503 separately for each season. This allowed identifying possible patterns of low or high FD in the tadpole assemblages if differences were significant. All t-test were two-sided We.Dry season assemblage as observed (six in our example). This was repeatedly done for each stream based on the species that we observed in the stream. For each stream and season we first calculated FD of the tadpole null assemblages in the seasons and then the change in FD from the wet to the dry season, also based on the null assemblages. To calculate the relative change of diversity from the wet season (“wet”) to the dry season (“dry”) we developed an approach using ??(1 ry/wet) resulting in positive values indicating an increase and negative values indicating a decrease in the respective diversity measure. FD was calculated following Petchey Gaston [43, 44]. This method applies a three-step dendrogram based classification function in which a species trait matrix is used to calculate a pair-wise species distance matrix. This matrix is used to construct dendrograms of specific species assemblages. The total branch length needed to connect all species in the assemblagePLOS ONE | DOI:10.1371/journal.pone.0151744 March 25,5 /Seasons Affect Functional and Phylogenetic Diversityrepresents the respective FD. Best distance measure (Gower’s distance) and cluster method (UPGMA) were identified following Mouchet et al. [45]. Our trait matrix consisted of categorical and continuous morphological trait variables (oral disc, body shape) of ecological relevance (i.e. feeding, microhabitat choice; [46]) for all species (for a list of traits used, see [25]). We calculated FD and the relative changes in FD for the observed assemblages, and for null assemblages (see above). We compared relative changes of observed and null assemblage FD using paired t-test. Accordingly, significant differences show that seasonal changes in FD are not random. Non randomness can be caused by differences in redundancy patterns, or by high or low FD in one or both seasons [24]. We therefore analysed data of the wet and the dry season separately. We applied polynomial regression with observed FD as dependent and SR (linear term) and SR^2 (polynomial term) as independent variables to reveal possible patterns of functional redundancy in each of both seasons (i.e., if nonlinearity is found). To test whether the full polynomial model or a simplified (i.e., linear) model performs better we used stepwise deletion (polynomial term first) and compared the models based on Akaike’s Information Criterion (AIC, [47, 48]) until the minimum adequate models were reached. Residuals were checked using diagnostic plots. We used Moran’s I autocorrelation coefficient [49] to prove that there is no spatial autocorrelation of the study sites regarding SR, FD, and PD, and their respective changes between wet and dry season. In this process no functional redundancy is indicated if these analyses show only a linear relationship between FD and SR (i.e., if the linear term in the polynomial model happens to be significant and the polynomial term to be non significant). Functional redundancy is indicated if the slope of the relationship between FD and SR decreases with increasing SR (i.e., if the polynomial term in the model is significant), caused by a stronger overlap of ecological traits with increasing number of syntopic tadpole species. We compared observed FD data with the respective null model data using paired t-tests, SART.S23503 separately for each season. This allowed identifying possible patterns of low or high FD in the tadpole assemblages if differences were significant. All t-test were two-sided We.
