Abstract. The latitudinal diversity gradient is the largest scale, and longest known, pattern in ecology. We examined the applicability of three versions of the energy hypothesis, the habitat heterogeneity hypothesis, and historical contingency to the gradient of terrestrial birds. The productivity version of the energy hypothesis, tested using actual evapotranspiration, a water-energy variable closely associated with plant productivity, accounted for 72% of the variance in a model of global extent. An historical contingency model based on biogeographic region explained 58% of the variance. A combined climate-region model accounted for 78% of the variance, but 52% comprised the overlap between these effects. This suggests that further resolution of contemporary vs. historical processes at the global level will require the inclusion of phylogenetic information.Regional-extent regression models suggest a latitudinal shift in constraints on diversity; measures of ambient energy (potential evapotranspiration and mean annual temperature) best predicted the diversity gradient at high latitudes, whereas water-related variables (actual evapotranspiration and annual rainfall) best predicted richness in low-latitude, high-energy regions. Intraregional spatial autocorrelation analysis confirmed that climatic models adequately describe geographic richness patterns at all but the smallest spatial scales resolved by the analysis. We conclude that the ''water-energy dynamics'' hypothesis, originally developed for plant diversity gradients, offers a parsimonious explanation for bird diversity patterns as well, presumably operating via plant productivity. However, more refined tests of historical factors are needed to fully resolve their influences on the gradient.

Forecasts of species range shifts under climate change are fraught with uncertainties and ensemble forecasting may provide a framework to deal with such uncertainties. Here, a novel approach to partition the variance among modeled attributes, such as richness or turnover, and map sources of uncertainty in ensembles of forecasts is presented. We model the distributions of 3837 New World birds and project them into 2080. We then quantify and map the relative contribution of different sources of uncertainty from alternative methods for niche modeling, general circulation models (AOGCM), and emission scenarios. The greatest source of uncertainty in forecasts of species range shifts arises from using alternative methods for niche modeling, followed by AOGCM, and their interaction. Our results concur with previous studies that discovered that projections from alternative models can be extremely varied, but we provide a new analytical framework to examine uncertainties in models by quantifying their importance and mapping their patterns.

Abstract.-We propose a new method to estimate and correct for phylogenetic inertia in comparative data analysis. The method, called phylogenetic eigenvector regression (PVR) starts by performing a principal coordinate analysis on a pairwise phylogenetic distance matrix between species. Traits under analysis are regressed on eigenvectors retained by a broken-stick model in such a way that estimated values express phylogenetic trends in data and residuals express independent evolution of each species. This partitioning is similar to that realized by the spatial autoregressive method, but the method proposed here overcomes the problem of low statistical performance that occurs with autoregressive method when phylogenetic correlation is low or when sample size is too small to detect it. Also, PVR is easier to perform with large samples because it is based on well-known techniques of multivariate and regression analyses. We evaluated the performance of PVR and compared it with the autoregressive method using real datasets and simulations. A detailed worked example using body size evolution of Carnivora mammals indicated that phylogenetic inertia in this trait is elevated and similarly estimated by both methods. In this example, Type I error at a = 0.05 of PVR was equal to 0.048, but an increase in the number of eigenvectors used in the regression increases the error. Also, similarity between PVR and the autoregressive method, defined by correlation between their residuals, decreased by overestimating the number of eigenvalues necessary to express the phylogenetic distance matrix. To evaluate the influence of c1adogram topology on the distribution of eigenvalues extracted from the double-centered phylogenetic distance matrix, we analyzed 100 randomly generated c1adograms (up to 100 species). Multiple linear regression of log transformed variables indicated that the number of eigenvalues extracted by the broken-stick model can be fully explained by c1adogram topology. Therefore, the broken-stick model is an adequate criterion for determining the correct number of eigenvectors to be used by PVR. We also simulated distinct levels of phylogenetic inertia by producing a trend across 10, 25, and 50 species arranged in "comblike" c1adograms and then adding random vectors with increased residual variances around this trend. In doing so, we provide an evaluation of the performance of both methods with data generated under different evolutionary models than tested previously. The results showed that both PVR and autoregressive method are efficient in detecting inertia in data when sample size is relatively high (more than 25 species) and when phylogenetic inertia is high. However, PVR is more efficient at smaller sample sizes and when level of phylogenetic inertia is low. These conclusions were also supported by the analysis of 10 real datasets regarding body size evolution in different animal clades. We concluded that PVR can be a useful alternative to an autoregressive method in comparative data analysis.

Individual processes shaping geographical patterns of biodiversity are increasingly understood, but their complex interactions on broad spatial and temporal scales remain beyond the reach of analytical models and traditional experiments. To meet this challenge, we built a spatially explicit, mechanistic simulation model implementing adaptation, range shifts, fragmentation, speciation, dispersal, competition, and extinction, driven by modeled climates of the past 800,000 years in South America. Experimental topographic smoothing confirmed the impact of climate heterogeneity on diversification. The simulations identified regions and episodes of speciation (cradles), persistence (museums), and extinction (graves). Although the simulations had no target pattern and were not parameterized with empirical data, emerging richness maps closely resembled contemporary maps for major taxa, confirming powerful roles for evolution and diversification driven by topography and climate.

The latitudinal diversity gradient is one of the most important problems in ecology and biogeography, but there is little consensus on its causes. Broad-scale species richness gradients are usually correlated with contemporary climate, with the highest number of species in warm, wet areas (Wright et al., 1993;Hawkins et al., 2003a). Even so, such associations do not ABSTRACTAim The aim of this study was to test a variant of the evolutionary time hypothesis for the bird latitudinal diversity gradient derived from the effects of niche conservatism in the face of global climate change over evolutionary time.Location The Western Hemisphere.Methods We used digitized range maps of breeding birds to estimate the species richness at two grain sizes, 756 and 12,100 km 2 . We then used molecular phylogenies resolved to family to quantify the root distance (RD) of each species as a measure of its level of evolutionary development. Birds were classified as 'basal' or 'derived' based on the RD of their family, and richness patterns were contrasted for the most basal and most derived 30% of species. We also generated temperature estimates for the Palaeogene across the Western Hemisphere to examine how spatial covariation between past and present climates might make it difficult to distinguish between ecological and evolutionary hypotheses for the current richness gradient.Results The warm, wet tropics support many species from basal bird clades, whereas the northern temperate zone and cool or dry tropics are dominated by species from more recent, evolutionarily derived clades. Furthermore, crucial to evaluating how niche conservatism among birds may drive the hemispherical richness gradient, the spatial structure of the richness gradient for basal groups is statistically indistinguishable from the overall gradient, whereas the richness gradient for derived groups is much shallower than the overall gradient. Finally, modern temperatures and the pattern of climate cooling since the Eocene are indistinguishable as predictors of bird species richness.Main conclusions Differences in the richness gradients of basal vs. derived clades suggest that the hemispherical gradient has been strongly influenced by the differential extirpation of species in older, warm-adapted clades from parts of the world that have become cooler in the present. We propose that niche conservatism and global-scale climate change over evolutionary time provide a parsimonious explanation for the contemporary bird latitudinal diversity gradient in the New World, although dispersal limitation of some highly derived clades probably plays a secondary role.

We propose a new method to estimate and correct for phylogenetic inertia in comparative data analysis. The method, called phylogenetic eigenvector regression (PVR) starts by performing a principal coordinate analysis on a pairwise phylogenetic distance matrix between species. Traits under analysis are regressed on eigenvectors retained by a broken-stick model in such a way that estimated values express phylogenetic trends in data and residuals express independent evolution of each species. This partitioning is similar to that realized by the spatial autoregressive method, but the method proposed here overcomes the problem of low statistical performance that occurs with autoregressive method when phylogenetic correlation is low or when sample size is too small to detect it. Also, PVR is easier to perform with large samples because it is based on well-known techniques of multivariate and regression analyses. We evaluated the performance of PVR and compared it with the autoregressive method using real datasets and simulations. A detailed worked example using body size evolution of Carnivora mammals indicated that phylogenetic inertia in this trait is elevated and similarly estimated by both methods. In this example, Type I error at a = 0.05 of PVR was equal to 0.048, but an increase in the number of eigenvectors used in the regression increases the error. Also, similarity between PVR and the autoregressive method, defined by correlation between their residuals, decreased by overestimating the number of eigenvalues necessary to express the phylogenetic distance matrix. To evaluate the influence of c1adogram topology on the distribution of eigenvalues extracted from the double-centered phylogenetic distance matrix, we analyzed 100 randomly generated c1adograms (up to 100 species). Multiple linear regression of log transformed variables indicated that the number of eigenvalues extracted by the broken-stick model can be fully explained by c1adogram topology. Therefore, the broken-stick model is an adequate criterion for determining the correct number of eigenvectors to be used by PVR. We also simulated distinct levels of phylogenetic inertia by producing a trend across 10, 25, and 50 species arranged in "comblike" c1adograms and then adding random vectors with increased residual variances around this trend. In doing so, we provide an evaluation of the performance of both methods with data generated under different evolutionary models than tested previously. The results showed that both PVR and autoregressive method are efficient in detecting inertia in data when sample size is relatively high (more than 25 species) and when phylogenetic inertia is high. However, PVR is more efficient at smaller sample sizes and when level of phylogenetic inertia is low. These conclusions were also supported by the analysis of 10 real datasets regarding body size evolution in different animal clades. We concluded that PVR can be a useful alternative to an autoregressive method in comparative data analysis.

Knowledge about biodiversity remains inadequate because most species living on Earth were still not formally described (the Linnean shortfall) and because geographical distributions of most species are poorly understood and usually contain many gaps (the Wallacean shortfall). In this paper, we developed models to infer the size and placement of geographical ranges of hypothetical non-described species, based on the range size frequency distribution of anurans recently described in the Cerrado Biome, on the level of knowledge (number of inventories) and on surrogates for habitat suitability. The rationale for these models is as follow: (1) the range size frequency distribution of these species should be similar to the range-restricted species, which have been most recently described in the Cerrado Biome; (2) the probability of new discoveries will increase in areas with low biodiversity knowledge, mainly in suitable areas, and (3) assuming range continuity, new species should occupy adjacent cells only if the level of knowledge is low enough to allow the existence of undiscovered species. We ran a model based on the number of inventories only, and two models combining effects of number of inventories and two different estimates of habitat suitability, for a total of 100 replicates each. Finally, we performed a complementary analysis using simulated annealing to solve the set-covering problem for each simulation (i.e. finding the smallest number of cells so that all species are represented at least once), using extents of occurrence of 160 species (131 real anuran species plus 29 new simulated species). The revised reserve system that included information about unknown or poorly sampled taxa significantly shifted northwards, when compared to a system based on currently known species. This main result can be explained by the paucity of biodiversity data in this part of the biome, associated with its relatively high habitat suitability. As a precautionary measure, weighted by the inferred distribution data, the prioritization of a system of reserves in the north part of the biome appears to be defensible.

We tested the proposition that there are more species in the tropics because basal clades adapted to warm paleoclimates have been lost in regions now experiencing cool climates. Molecular phylogenies were used to classify species as "basal" and "derived" based on their family, and their richness patterns were contrasted. Path models also evaluated environmental predictors of richness patterns. As predicted, basal clades are more diverse in the lowland tropics, whereas derived clades are more diverse in the extratropics and high-altitude tropics. Seventy-four percent of the variation in bird richness was explained by environmental variables, but models differed for basal and derived groups. The overall gradient is described by the spatial pattern of basal clades, although there are differences in the Old and New Worlds. We conclude that in ecological time, the global richness gradient reflects birds' responses to climatic gradients, partially operating via plants. Over evolutionary time, the gradient primarily reflects the extirpation of species in older clades from parts of the world that have become cooler in the present. A strong secondary effect arises from dispersal of clades from centers of origin and subsequent radiations. Overall, the diversity gradient is well explained by niche conservatism and the "time-for-speciation" hypothesis.