Lch t-tests. We conducted the same procedure to analyse patterns of phylogenetic diversity (PD) and its respective changes from wet to dry season. Our analyses are based on a time-calibrated phylogeny of mantellid species [50] as all tadpoles sampled in the streams belonged to this family [40]. We pruned from this tree all taxa not represented in our Ranomafana Win 63843 msds tadpole sampling using TreeEdit version v1.0a10. This reduced the number of species by retaining branch lengths. For a few so far undescribed species, genetic data were insufficient to include them in the tree. In these cases, other species that are known to be in the same clade (according to molecular data of Vieites et al. [34]) were used as replacement for the purpose of PD calculation. Similar to FD calculations [43, 44, 51], we extracted branch lengths from the tree for the assemblages [52] by using the function treedive included in the R package vegan. The sum of branch lengths needed to connect all species of an assemblage represents the assemblage’s PD. All analyses were run using the statistical software R 2.15.1 [53]. Packages used for FD calculations include car R package version 2.0?8 [54], gtools R package version 2.7.0. [55], cluster R package version 1.14.2 [56] and clue R package version 0.3?5 [57, 58]. For PD calculations we used the R packages ape version 3.0? [59, 60] and vegan version 2.0? [61]. Moran’s I was also calculated using the package ape.ResultsSummarizing data over wet and dry seasons, we found tadpoles of a total of 31 species in all twelve stream sections (= assemblages). In the wet season we found five to 15 species, in the dry season two to twelve species per assemblage. Mean SR in the dry season was about 27 lower than in the wet season (paired t-test, scan/nsw074 t = 3.44, df = 11, p = 0.006). Beside this decrease inPLOS ONE | DOI:10.1371/journal.pone.0151744 March 25,6 /Seasons Affect Functional and Phylogenetic DiversityFig 3. Change from the wet to the dry season: Deviation of observed loss of FD and PD from the null models. The dashed lines represent the null models for the change from the wet to the dry season, the box-whisker-plots show the observed deviations from the null models. Values below the dashed line show a smaller loss, values above the line show a higher loss of FD and PD than PG-1016548 chemical information predicted by the null models. jir.2012.0140 FD of tadpoles decreases significantly less than predicted whereas PD decreases significantly more than predicted. doi:10.1371/journal.pone.0151744.gSR (i.e., probably due to reduced reproductive activity), a species turnover from wet to dry season was observed in eight streams, i.e., an assemblage in the dry season harboured on average one or more species that were not present in the assemblage of the same stream during the wet season. However, no species was exclusively found in the dry season.Functional diversityAs observed for SR, FD decreased from the wet to the dry season. However, the observed relative loss of FD (8 ) differs from the loss predicted by the null models (17 ) based on the observed loss in SR (paired t-test, t = -2.34, df = 11, p = 0.04; Fig 3). This difference suggestsPLOS ONE | DOI:10.1371/journal.pone.0151744 March 25,7 /Seasons Affect Functional and Phylogenetic DiversityFig 4. Comparing null models and observed FD from tadpole assemblages from RNP. Given are observed (dark grey filled circles, continuous regression line) and null model FD (light grey circles with dark margins, dashed line) along the observed S.Lch t-tests. We conducted the same procedure to analyse patterns of phylogenetic diversity (PD) and its respective changes from wet to dry season. Our analyses are based on a time-calibrated phylogeny of mantellid species [50] as all tadpoles sampled in the streams belonged to this family [40]. We pruned from this tree all taxa not represented in our Ranomafana tadpole sampling using TreeEdit version v1.0a10. This reduced the number of species by retaining branch lengths. For a few so far undescribed species, genetic data were insufficient to include them in the tree. In these cases, other species that are known to be in the same clade (according to molecular data of Vieites et al. [34]) were used as replacement for the purpose of PD calculation. Similar to FD calculations [43, 44, 51], we extracted branch lengths from the tree for the assemblages [52] by using the function treedive included in the R package vegan. The sum of branch lengths needed to connect all species of an assemblage represents the assemblage’s PD. All analyses were run using the statistical software R 2.15.1 [53]. Packages used for FD calculations include car R package version 2.0?8 [54], gtools R package version 2.7.0. [55], cluster R package version 1.14.2 [56] and clue R package version 0.3?5 [57, 58]. For PD calculations we used the R packages ape version 3.0? [59, 60] and vegan version 2.0? [61]. Moran’s I was also calculated using the package ape.ResultsSummarizing data over wet and dry seasons, we found tadpoles of a total of 31 species in all twelve stream sections (= assemblages). In the wet season we found five to 15 species, in the dry season two to twelve species per assemblage. Mean SR in the dry season was about 27 lower than in the wet season (paired t-test, scan/nsw074 t = 3.44, df = 11, p = 0.006). Beside this decrease inPLOS ONE | DOI:10.1371/journal.pone.0151744 March 25,6 /Seasons Affect Functional and Phylogenetic DiversityFig 3. Change from the wet to the dry season: Deviation of observed loss of FD and PD from the null models. The dashed lines represent the null models for the change from the wet to the dry season, the box-whisker-plots show the observed deviations from the null models. Values below the dashed line show a smaller loss, values above the line show a higher loss of FD and PD than predicted by the null models. jir.2012.0140 FD of tadpoles decreases significantly less than predicted whereas PD decreases significantly more than predicted. doi:10.1371/journal.pone.0151744.gSR (i.e., probably due to reduced reproductive activity), a species turnover from wet to dry season was observed in eight streams, i.e., an assemblage in the dry season harboured on average one or more species that were not present in the assemblage of the same stream during the wet season. However, no species was exclusively found in the dry season.Functional diversityAs observed for SR, FD decreased from the wet to the dry season. However, the observed relative loss of FD (8 ) differs from the loss predicted by the null models (17 ) based on the observed loss in SR (paired t-test, t = -2.34, df = 11, p = 0.04; Fig 3). This difference suggestsPLOS ONE | DOI:10.1371/journal.pone.0151744 March 25,7 /Seasons Affect Functional and Phylogenetic DiversityFig 4. Comparing null models and observed FD from tadpole assemblages from RNP. Given are observed (dark grey filled circles, continuous regression line) and null model FD (light grey circles with dark margins, dashed line) along the observed S.