Figure 1.
A graphical example of potential ways in which polytomies in the phylogenetic tree may influence different metrics of community phylogenetic diversity.
The boxes indicate whether or not that species is found in the community. If the box is shaded grey, then the species is present in the community. If the box is not shaded, then the species is absent from the community. The left panel is a fully resolved phylogenetic tree and the three different measures of phylogenetic diversity (MPD, MNND, and FI) using four example assemblages. The MPD is the mean pair-wise phylogenetic distance between all taxa in the assemblage. The MNND is the mean nearest phylogenetic neighbor distance for all taxa in the assemblage. The FI (Faith's Index) quantifies the shared branch lengths between species in an assemblage as a proportion of the total branch lengths in the species pool phylogeny. The right panel shows the same phylogeny with one node now a polytomy and the same measures of phylogenetic diversity using this less resolved phylogeny with an increased total branch length. In all cases the FI measured is influenced as it represents a proportion of the total branch length. The MPD and MNND metrics are not influenced if the polytomy does not include species in the assemblage, but if it does include species in the assemblage these metrics may artificially increase or decrease.
Figure 2.
A graphical example how potential directional biases in phylogenetic dispersion produced by using phylogenetic trees containing polytomies were quantified in this study.
The value for a phylogenetic dispersion metric, in this example NRI, generated for a community using a phylogeny containing polytomies is regressed through the origin onto the NRI value generated from the same community using a fully resolved phylogeny (dashed lines). As in the above example, the expected relationship is a 1∶1 line through the origin (Solid Line). When the slope is greater than one (dashed line in the top panel) shows a bias towards higher phylogenetic overdispersion and phylogenetic clustering. In other words, a bias towards non-random phylogenetic structuring (False Positives; Type I Error). When the slope is less than one (dashed line in the bottom panel) this shows a bias towards lower phylogenetic overdispersion and phylogenetic clustering. In other words, a bias towards random phylogenetic structuring (False Negatives; Type II Error).
Figure 3.
A figure showing the power to predict NRI and NTI of an assemblage with the maximal possible phylogenetic diversity estimated using the Greedy Algorithm.
The slopes and r2 values from regressing the NRI and NTI values derived using a randomly ‘unresolved’ phylogeny onto the NRI and NTI values derived using a fully resolved phylogeny. The size of the phylogeny is represented by color and dashing of the lines. Specifically, the number of terminal taxa was 20 (finely dashed grey line), 40 (thickly dashed grey line), 80 (solid grey line), 160 (dashed black line), and 320 (solid black line). The percentage of nodes that were ‘unresolved’ is indicated by Rx on the x-axis. Slopes less than one show a bias towards under-predicting the phylogenetic diversity in an assemblage and vice versa for slopes greater than one (see Figure 1).
Figure 4.
A figure showing the power to predict NRI and NTI of an assemblage with the minimal possible phylogenetic diversity estimated using a dynamic programming algorithm implemented in PDA.
The slopes and r2 values from regressing the NRI and NTI values derived using a randomly ‘unresolved’ phylogeny onto the NRI and NTI values derived using a fully resolved phylogeny. The size of the phylogeny is represented by color and dashing of the lines. Specifically, the number of terminal taxa was 20 (finely dashed grey line), 40 (thickly dashed grey line), 80 (solid grey line), 160 (dashed black line), and 320 (solid black line). The percentage of nodes that were ‘unresolved’ is indicated by an R on the x-axis. Slopes less than one show a bias towards under-predicting the phylogenetic diversity in an assemblage and vice versa for slopes greater than one (see Figure 1).
Figure 5.
A figure showing the power to predict NRI and NTI of randomly generated assemblages.
The slopes and r2 values from regressing the NRI and NTI values derived using a randomly ‘unresolved’ phylogeny onto the NRI and NTI values derived using a fully resolved phylogeny. The size of the phylogeny is represented by color and dashing of the lines. Specifically, the number of terminal taxa was 20 (finely dashed grey line), 40 (thickly dashed grey line), 80 (solid grey line), 160 (dashed black line), and 320 (solid black line). The percentage of nodes that were ‘unresolved’ is indicated by Rx on the x-axis. Slopes less than one show a bias towards under-predicting the phylogenetic diversity in an assemblage and vice versa for slopes greater than one (see Figure 1).