Characterizing abiotic niches.
We used the SDMs described above to characterize abiotic niches of each species with respect to mean annual temperature (MAT), precipitation (MAP), and elevation, by parsing the fraction of the SDM attributable to an increment of that abiotic variable. Increments were 1°C for MAT and 250 mm for MAP. For elevation, we used the seven bioclimatic zones defined by the Global Mountain Biodiversity Assessment (GMBA; Körner et al., 2011, 2017; Körner and Paulsen, 2004; 2.5’ resolution). These zones integrate temperature, growing season length, and topographic information to provide a robust and biogeographically relevant assessment of the extent and type of montane and alpine habitat across the globe. They further permit an assignment of species distributions along elevational gradients despite latitudinal differences in what constitutes montane and alpine areas. Nevertheless, these bioclimatic categories are fundamentally defined by isothermal zones, and are not therefore an independent axis from MAT (Körner et al., 2011). However, our goal was not to define independent niche axes per se , but rather to characterize abiotic niches in a biologically meaningful way. The GMBA-defined zones provide the current best estimate of a biogeographically relevant classification of mountain systems (Körner et al., 2017). The seven GMBA bioclimatic zones used to define species’ elevation niche were: 1—nival; 2—upper alpine; 3—lower alpine; 4—upper montane; 5—lower montane; 6—mountain slope with frost; 7—mountain slope without frost or lowland. We chose to combine GMBA zone 7 with lowland areas for our study; thus, our assessment of lowland distributions was fundamentally linked with exposure to freezing temperatures.
This method yielded three vectors that jointly quantified the distribution of niche space potentially occupied by each species (e.g., 10% probability a species will be present at a site with MAP of 500 mm). Figure 1 plots some representative examples and provides a conceptual illustration of our characterization of abiotic niches and dataset assembly. In this way, we avoided having to arbitrarily bin species into habitat categories, such as alpine or lowland, and instead could assess niches as continuously valued distributions of occupied niche space. Parsing of SDMs was performed using custom Python scripts (available at: https://doi.org/10.5061/dryad.9cnp5hqgx).
To visualize the climatic (temp. or prec.) niche space occupied by species in our dataset, we integrated each climate vector to obtain a weighted average value that characterized the center of each species’ range (i.e., climatic conditions in locations where the SDM score was higher were given proportionally higher weight). The resulting values thus described the climatic conditions in regions of that species’ range where the species was expected to be more abundant, as opposed to characterizing the most extreme climatic values where a species is found. This realized climatic niche space was then contrasted among species occupying different elevations.
The elevation niche vector was used to assign species into categories such as ‘alpine’ or ‘specialist’. Though such distinctions can be arbitrary, when used to express where the center of a species’ range is expected to occur, they can be used to provide valuable biological insight (Körner, 2003). In this study, the ‘specialist’ category indicated the likely central habitat of a species’ range, and so we required specialists to have at least 60% of their SDM in that habitat type. For elevational categories, such as alpine or montane, we required a species to have at least 5% of their total SDM in that category, regardless of whether such habitat was contiguous or not. Thus, an alpine specialist in this study would have at least 60% of its SDM in alpine regions, whereas an alpine generalist could have anywhere from 5% to 59% of its modelled range in the alpine belt (Supplemental Fig. S1).