SUPPLEMEMTARY INFORMATION
Supplementary methods.
Morans’s Index
As known, the Global Moran’s Index (IG) evaluates whether the pattern expressed by data is clustered or random (Moran 1950, Goodchild, 1986). Hence, the IG of canopy height was chosen to check whether spatial autocorrelation of this variable (i.e. pattern of data aggregation of canopy height) could affect plants richness.
Spatial autocorrelations strongly depends on the neighbourhood distance threshold adopted to run the test (for instance data may results randomly dispersed into a sampling of 10Km2 but strongly clustered into sampling of 100Km2 or vice versa) and on the relation of contiguity adopted. Contiguity was defined by a neighbourhood of four locations adjacent to each cell, whilst the threshold distance was set to 0.9°, consistently with the spatial resolution adopted in the present work.
IG ranges between 1 (total spatially autocorrelated data) and -1 (totally spatially dispersed data). Value 0 indicates a random distribution.
IG was computed for the canopy height within homogeneous DZ (thus, we carried out 10 values of IG of canopy height, one for each DZ). Significant spatial aggregation of canopy height (p<0.05) were found only within DZs ranging from 2 to 7:
DZ / IG / p-value1 / 0.4 / 0.1
2 / 0.62 / 0.05
3 / 0.96 / <0.05
4 / 0.92 / <0.05
5 / 0.73 / <0.05
6 / 0.68 / <0.05
7 / 0.84 / <0.05
8 / 0.14 / 0.75
9 / -0.04 / 0.99
10 / NaN / NaN
We then tested the correlation between significant IG and S through the non-parametric Kendall Rank test, which resulted non-meaningful (p>0.05).
References
Goodchild, M. F. (1986). Spatial Autocorrelation. Concepts and Techniques in Modern Geography 47. Norwich, UK: Geo Books.
Moran, P. A. (1950). Notes on continuous stochastic pheomena. Biometrika, 37(1/2), 17-23.
Supplementary Figures
Figure S1. Coefficient of Variation (Std/H) of Canopy Height projected using the same projection (i.e. Gall–Peters) and matching colour scale of figures 1a and 1b.
Figure S2. Forest land cover is expressed as Pixel density (dimension less 0-1, see Methods) of the resampled Canopy Height projected using the same projection (i.e. Gall–Peters) and matching colour scale of figure 1a and 1b.
Figure S3. Boxplots of the coefficient of variation (Std/H) within DZs. Symbols and colours as Figure 1c of the main text.
1