RESULTS
Dataset construction and description of abiotic niches .
Our data cleaning and filtering methods produced a dataset of 50,002,722 georeferenced occurrence records spanning 72,372 seed plant species, of which 68,241 could be matched back to our phylogeny (totaling 5397 genera or ~18% of total estimated seed plant lineages; Stevens, 2001; Smith and Brown, 2018), with broad geographic coverage of the Americas, including all mountains (Fig. 2a). There were notable geographic areas of poor sampling, including the Amazon basin and extreme southern and northern latitudes. From these occurrence records, we built species distribution models (SDMs) to characterize mean annual temperature (MAT), precipitation (MAP), and elevation niches. Importantly, these niches were characterized as continuously valued distributions and not as scalar values. Figure 1 illustrates an example of this niche characterization for a single species, Astragalus alpinus L. (‘alpine milkvetch’), whose composite niche is centered within montane habitat (i.e., >60% of the modelled range falls within montane area) and is bounded between approximately -17° and +10° C, and generally below 1000 mm of precipitation. This type of niche characterization allowed us to define the fraction of each species’ range occurring within different combinations of abiotic conditions.
Species richness in different elevational categories .
Comparing species richness across different elevational belts revealed contrasting patterns of diversity for species occupying lowland (Nlowland=36,420), montane (Nmontane=33,015), and alpine (Nalpine=2937) habitats. Across all seed plants, we observed a traditional latitudinal diversity gradient, with richness peaking near the equator (Fig. 2a). This pattern was nearly identical when considering solely montane communities (i.e., species with at least 5% of their range in the montane belt; Fig. 2b). For alpine communities, however, species richness was latitudinally bimodal, peaking in both the western North American cordillera and the central Andes (Fig. 2c).
Climatic niche space of lowland, montane, and alpine communities .
The temperature and precipitation niche space occupied across all species in our dataset (Fig. 3a) generally mirrored that defining the climate space of the regional biomes, indicating that our modelled distributions recapitulated the likely climatic niche space of the Americas overall. Lowland and montane species occupied similar regions of climatic space as seed plants generally (Fig. 3a-c), though lowland taxa exhibited a greater density at higher precipitation values, and very few lowland species had ranges centered in habitats with mean annual temperature below freezing (though certainly many had ranges extending into these regions). Indeed, montane species seemed to occupy a greater total niche volume than lowland taxa here. In contrast, the climatic niche space of alpine species (Fig. 3d) was drastically reduced and shifted toward colder and drier habitats. Alpine communities were more uniform in their occupied precipitation niche space, but somewhat bimodal in the occupied temperature space, with a small, dense cluster of species having ranges centered on areas with -10° C MAT.
Climatic niche breadth across elevational gradients .
The distributions of average, standardized, niche breadths (B ) across all American seed plants (Fig. 4a) were generally narrow (BTEMP = 0.130 ± 0.08; BPREC = 0.099 ± 0.08), but with long tails, indicating much greater niche breadth for only some species as relatively few species had both largeBTEMP and BPREC . The niche breadths of montane species were generally similar to those of seed plants overall (Fig. 4c), though montane species had increasedBTEMP . Alpine species, however, occupied a reduced total niche breadth space, characterized by broadBTEMP and narrow BPREC (Fig. 4d).
Generalist species within each elevational category had maximalBTEMP (Fig. 5, top row ), with an abrupt increase in BTEMP observed for alpine generalists. BPREC , in contrast, was generally narrow regardless of elevation or specialization (Fig. 5, bottom row ). Across our dataset, genera with alpine specialists had a greater fraction of both alpine generalists and montanespecialists (Supplemental Fig. S2; Figueroa et al., pers. comm. ). Spatial variation in community-averaged niche breadth (Supplemental Fig. S3) indicated higher BTEMP at greater absolute latitudes (particularly in the Northern Hemisphere) and in mountains compared to surrounding lowlands. In contrast,BPREC was largest near the equator, regardless of elevation. Most mountains showed lower BPREC than nearby lowlands, exceptions being some Central American mountains, the Northern Andes, and the southernmost Andes. As a result, the Northern Andes are notable here for having greater community-averaged niche breadth for both temperature and precipitation.
Climatically Similar Non-Alpine (CSNA) species .
Climatically Similar Non-Alpine (CSNA) species, defined here as non-alpine species whose ranges otherwise overlapped with the climatic conditions of alpine taxa (Fig. 2d and 6) tended to cluster geographically in the southwestern US and Mexico (Fig. 2d, 6e), and therefore also departed from a traditional latitudinal diversity gradient as their greatest richness occurred at ~20° N. Although regions with higher CSNA richness harbored the lowest alpine diversity (Fig. 2d), richness between these groups at these areas was strongly and positively correlated (R2=0.68, p<0.001; Fig. 6e). CSNA species ranges (Fig. 6d) contained a significantly greater fraction of frost-exposed mountain foothills (p<0.01) and lower fraction of frost-free lowland (p<0.01), compared to other non-alpine species. To a lesser extent, they also had a greater proportion of their ranges in the lower montane belt than other non-alpine taxa. At the phylogenetic scale of the American seed plant flora, there was no clear separation between alpine and CSNA species (Fig. 6a). However, there were substantial differences in the taxonomic composition of these groups at finer phylogenetic scales (Table 1 and online supplement), with several orders containing CSNA but not alpine species, and only ~28% overlap in genera between CSNA and alpines (Nalpine_genera=717; NCSNA_genera=1421; Nshared_genera=395).