Figure 3 : At the upper panel, description of the framework used to calculate plot’sfilling index es. Logic behind the filling index procedure, which tests whether the observed plot show a polygon-shaped pattern when compared to simulated line-shaped plots. A) Calculation of the observed plot filling index . First, the abundance-environment plot is rasterized, indicating whether there is one or more data points within the range of each cell (purple cells). Then, the upper boundary of the point cloud (purple line) is defined by connecting the upper cells of each raster column (dark purple cells). Finally, the rate of occupation is calculated below the upper boundary (i.e., sum of purple cells/sum of all cells bellow the purple line). Values of 1 indicate that the area bellow the ceiling is fully filled by data points. B) Generation of a null model based on 999 simulated line-shaped point clouds. The line-shaped simulations use the same environmental data and sample size as the observed data, and resemble the general trend of the observed plot. The filling index is calculated for each simulated distribution as described in A. C) Statistical significance of the observed filling index is different to the distribution of the simulated line-shaped patters. The probability of the observedfilling index (purple line) is greater than the simulated ones (blue distribution) and is calculated at a statistical significance of p<0.05. At the lower panel, exemplification of the performance of the filling index to identify polygon-shaped patterns in two simulated plots: polygon-shaped plot (D), and line-shaped plot (E). For more details on evaluation methods and results see Appendix S6.
Figure 4