Interactions between all pairs of neighboring trees in 16 forests worldwide reveal details of unique ecological processes in each forest, and provide windows into their evolutionary histories
Fig 1
An overview of a typical EAA analysis.
At top is a diagram of a large focal tree in the Lambir (Malaysia) FDP, surrounded by 10 concentric annuli each of area 50 m2. For simplicity, the trees shown in the diagram are only a sample of some of the annular surviving trees (S), recruits (R), and trees that die (D). In the diagram the trees shown are sampled from among the trees that lie at zero or at 185 Ma (mega-annum) phylogenetic distances from their LCA with the focal tree, although of course all the trees in the annuli are used in the entire EAA analysis. Generalized Additive Model (GAM) fits to patterns of clustering of surviving annular trees, using data from the closest and the furthest annulus, are shown in the two-dimensional graphs on the right of the diagram. In this analysis, the observed annular clustering is presented as deviations (z-values) from a null model expectation for four quantiles of focal tree diameter. The null model is generated by repeated shuffling of the focal tree diameters within species, so that any positive z-values for some of the focal-annular size quantiles must be balanced by negative values for others. Such positive-negative balances are expected in analyses of recruitment, clustering and mortality, but not in analyses of growth (Materials and Methods). The 95% confidence intervals of the GAM curves are shown in gray. Brown horizontal lines show the 95% confidence intervals around zero z-values. To help orient the viewer, gray arrows connect some of the closely-related and distantly-related survivors in the diagram to the places at which their data contributes to the largest-quantile focal tree lines (red) on the two-dimensional graphs. Each data point in the 2D graphs represents the difference between the actual and the null-model data for all focal trees in a given census period that have annular trees within a specific phylogenetic range. The null-model data have been generated by repeated shuffling of focal tree properties (size or growth rate) within species. A new point is generated for each of the ten replicates of the actual-null comparisons and for each of the census periods at the FDP. The points in the graphs form clusters because, with each replicate, species pairs separated by similar phylogenetic distances are shuffled at random in order to fill each of the phylogenetic distance quantiles. Thus, each of the gray arrows that shows the contribution of an individual tree simply shows where the tiny amount of information that is contributed by that tree’s focal-annular interactions ends up in the data points in the graphs. The three-dimensional graphs show GAM fits of the largest-quantile (graph A) and smallest-quantile (graph B) focal tree size data across all ten annuli. Regions of the surfaces that lie within the range of non-significant z-values along the z-axis are gray; those that lie outside this range, and that therefore represent significant z-values, are colored. The colors start with green and shade through blue as the significance of the positive or negative z-values increases. The orientations of the three-dimensional graphs presented here sometimes differ, in order to reveal details of the surfaces. As with the lines on the two-dimensional graphs, the 3D surfaces themselves have confidence intervals, but the confidence intervals are not shown here. Typical confidence intervals on the 3D surfaces, which tend to be small, are visualized more easily if these three-dimensional graphs can be rotated by the viewer. A sampling of such rotatable graphs is presented as html files in S3–S11 Figs.