Stronger tests of mechanisms underlying geographic gradients of biodiversity: insights from the dimensionality of biodiversity

Abstract

Inference involving diversity gradients typically is gathered by mechanistic tests involving single dimensions of biodiversity such as species richness. Nonetheless, because traits such as geographic range size, trophic status or phenotypic characteristics are tied to a particular species, mechanistic effects driving broad diversity patterns should manifest across numerous dimensions of biodiversity. We develop an approach of stronger inference based on numerous dimensions of biodiversity and apply it to evaluate one such putative mechanism: the mid-domain effect (MDE). Species composition of 10,000-km(2) grid cells was determined by overlaying geographic range maps of 133 noctilionoid bat taxa. We determined empirical diversity gradients in the Neotropics by calculating species richness and three indices each of phylogenetic, functional and phenetic diversity for each grid cell. We also created 1,000 simulated gradients of each examined metric of biodiversity based on a MDE model to estimate patterns expected if species distributions were randomly placed within the Neotropics. For each simulation run, we regressed the observed gradient onto the MDE-expected gradient. If a MDE drives empirical gradients, then coefficients of determination from such an analysis should be high, the intercept no different from zero and the slope no different than unity. Species richness gradients predicted by the MDE fit empirical patterns. The MDE produced strong spatially structured gradients of taxonomic, phylogenetic, functional and phenetic diversity. Nonetheless, expected values generated from the MDE for most dimensions of biodiversity exhibited poor fit to most empirical patterns. The MDE cannot account for most empirical patterns of biodiversity. Fuller understanding of latitudinal gradients will come from simultaneous examination of relative effects of random, environmental and historical mechanisms to better understand distribution and abundance of the current biota.

Figure 1. Pairwise associations among indices of biodiversity (A) and histogram describing the magnitudes of Spearman correlation coefficients (B).

Figure 2. Empirical (center) and MDE-generated (left and right) spatial variation in species richness of Noctilionoidea in the New World.

Left and right panels present unscaled and scaled spatial variation, respectively. Unscaled mean simulated gradients correspond to MDE generated variation that is scaled the same as for empirical patterns. Scaled mean simulated gradients correspond to MDE generated variation that ranges according to the magnitude of MDE results. Red shades depict areas of high species richness whereas blue shades represent areas of low species richness. Colors in the left and middle columns are directly comparable. Colors in the right column are not comparable to those of the two left columns because they are scaled differently. Areas in grey are those occurring outside the geographic distribution of the Noctilionoidea.

Figure 3. Spatial variation in phylogenetic (PD, PSV, PSC), functional (functional richness, functional diversity, functional evenness), and phenetic (morphological volume, minimum spanning tree SD, mean nearest neighbor distance) components of biodiversity of Noctilionoidea.

Each characteristic is represented by three panels. Unscaled mean simulated gradients correspond to MDE generated variation that is scaled the same as for empirical patterns (left). Empirical patterns are in the middle panel. Scaled mean simulated gradients correspond to MDE generated variation that ranges according to the magnitude of MDE results (right). Colors in the left and middle columns are directly comparable. Colors in the right column are not comparable to columns on the left because they are scaled differently. Areas in grey are those occurring outside the geographic distribution of the Noctilionoidea. Black areas are those for which diversity measures could not be estimated because they registered only one species.

Figure 4. Histograms describing goodness of fit based on coefficients of determination (r²) between empirical and MDE-generated variation in species richness across 1,000 runs of the MDE model.

Figure 5. Scatter plots indicating variation in combinations of slope and intercept for the relationship between empirical and simulated gradients in different measures of biodiversity generated from the MDE analysis.

Dots represent the combination of slope and intercept for each run. Large figures represent a subset of simulation runs that were close to the theoretical expectation of zero intercept and slope of unity. Inset represents relationship based on all 1,000 simulation runs. The two intersecting lines represent the theoretical expectation. If there is good fit between observed and expected values then the confidence envelope will overlap the point of zero intercept and slope of unity.