Fig 1.
Regional network of forest plots sampled as part of The Madidi Project, a floristic inventory of northwestern Bolivia.
The map shows the locations of 440 0.1-ha plots along a 4000-m elevational gradient in the Andes. A) Study region in Bolivia. B) and C) Distribution of plots along the eastern slopes of the Andes (∼250–4,350 m) in and around the multiple protected areas that form part of the Madidi region. Elevation data from WorldClim (www.worldclim.org); country borders from the GADM database (gadm.org).
Fig 2.
Elevational gradients in diversity at two contrasting spatial scales.
Small (among 0.01-ha subplots within a 0.1-ha plot; top row) and large (among 0.1-ha plots within an elevational band; bottom row). A) and D) Regional (γ-) and local (α-) diversity. B) and E) Observed β-diversity and mean null β-diversity. C) and F) β-deviations (standardized effect sizes of β-diversity). Null β-diversity and β-deviations were calculated based on two null models, one that randomizes the regional species abundance distribution (r-SAD) and one that fixes it to be identical to the one observed in the empirical data (f-SAD; see Methods). Diversity was partitioned following Jost [43] and by weighting each species proportionally by its abundance (i.e. diversity of order 1). All relationships were statistically significant (Table 1).
Table 1.
Relationships between diversity and elevation.
Fig 3.
Comparisons of the strength and shape of elevational gradients between scales and between observed β-diversity and β-deviations.
β-deviations were calculated using the random SAD (r-SAD) and fixed SAD (f-SAD) null models (see Methods). A) Strength of the gradients measured using adjusted R2 values (adj.R2) from cubic regressions between diversity and elevation. Black circles represent original adj.R2 estimates. Grey regions show the distribution of values based on 1,999 bootstrapped regressions. Black lines represent 99% confidence intervals. B) Shape of the gradients measured using standardized regression coefficients. Only the coefficients for elevation (b1) and elevation squared (b2) are presented. Other coefficients lead to similar conclusions. Black symbols represent original estimates. Black arrows show the change in coefficients between observed β-diversity and β-deviations at a given spatial scale. Black lines represent 99% data ellipses which define confidence regions. Other symbols show the distribution of values based on bootstrapped regressions.
Fig 4.
Variation in the overall magnitude of β-deviations between small and large spatial scales.
β-deviations were calculated using the random SAD (r-SAD) and fixed SAD (f-SAD) null models (see Methods). The horizontal grey line marks the value of no difference from null model expectations (i.e. β-deviation of zero). β-deviations above the line indicate higher β-diversity than expected by random sampling of individuals from observed species pools. Note that β-deviations are higher at large scales than at small scales (linear mixed-effects model: t276 < 38.97; p < 0.001). In addition, mean β-deviations are statistically different from zero for all combinations of spatial scale and null model (one sample t-tests: |t| > 4.77; p < 0.001).
Fig 5.
Distributions of 2,668 woody plant species along the elevational gradient.
Each vertical line represents the elevational range of a species in the Madidi region. Ranges are defined as the interval between the lowest and highest elevations at which a species was found within the full network of plots. The horizontal dashed line marks the elevation at which there seems to be a break in the continuous turnover in forest composition along the elevational gradient. Above 3,725 m, forests are composed only of 3 woody plant species: Gynoxys asterotricha, G. compressissima and Polylepis pepei.